Cortical locations of maximal spindle activity: magnetoencephalography (MEG) study

Authors


Valentina Gumenyuk, PhD, Henry Ford Hospital, Department of Neurology/Sleep center, Clara Ford Pavilion room 75, 2799 West Grand Boulevard, Detroit, MI 48202, USA. Tel.: +1 313 916 1075; fax: +1 313 916 0526; e-mail: vgumenyu@neurnis.neuro.hfh.edu

Summary

The aim of this study was to determine the main cortical regions related to maximal spindle activity of sleep stage 2 in healthy individual subjects during a brief morning nap using magnetoencephalography (MEG). Eight volunteers (mean age: 26.1 ± 8.7, six women) all right handed, free of any medical psychiatric or sleep disorders were studied. Whole-head 148-channel MEG and a conventional polysomnography montage (EEG; C3, C4, O1 and O2 scalp electrodes and EOG, EMG and ECG electrodes) were used for data collection. Sleep MEG/EEG spindles were visually identified during 15 min of stage 2 sleep for each participant. The distribution of brain activity corresponding to each spindle was calculated using a combination of independent component analysis and a current source density technique superimposed upon individual MRIs. The absolute maximum of spindle activation was localized to frontal, temporal and parietal lobes. However, the most common cortical regions for maximal source spindle activity were precentral and/or postcentral areas across all individuals. The present study suggests that maximal spindle activity localized to these two regions may represent a single event for two types of spindle frequency: slow (at 12 Hz) and fast (at 14 Hz) within global thalamocortical coherence.

Introduction

Sleep spindles are a characteristic of human stage 2 sleep. Based on electroencephalographic (EEG) studies, sleep spindles represent waxing and waning waves at a specific frequency band between 11 and 14 Hz (Steriade and Amzica, 1998). Duration of the spindle can vary between 0.5 and 3 s and the spindle pattern waveform mixed with slow rhythm (0.1–0.3 Hz) may periodically reoccur every 3–10 s (Contreras et al., 1996).

Findings in animal research have demonstrated that the cerebral cortex and thalamus are closely integrated structures in which the depolarizing component of the cortically generated slow oscillation drives thalamic reticular and thalamocortical cells to produce spindles (Steriade and Amzica, 1998). Thus, both the cortex and thalamus are important for the generation of spindles as demonstrated by studies of thalamectomy and experiments using decorticate cats (Amzica and Steriade, 1995; Timofeev and Steriade, 1996). It was shown that the slow oscillation is reflected in the thalamus by thalamic reticular and thalamocortical cells that produce sleep spindles (Amzica and Steriade, 1995). These fundamental studies demonstrate that, while spindle activity is generated in the thalamic reticular nucleus, the activation is synchronized by corticothalamic projections (Steriade, 2000), suggesting a significant role for the cortex in the production of sleep spindle oscillations.

Multiple widespread generators for sleep spindles were suggested by human studies using whole-head EEG (Anderer et al., 2001; Zeitlhofer et al., 1997; Zygierewicz et al., 1999) and magnetoencephalography (MEG) (Ishii et al., 2003; Lu et al., 1992; Manshanden et al., 2002; Shih et al., 2000; Urakami, 2008). In humans, attempts to localize spindle activation were performed and studies have found common brain regions that are active during the occurrence of sleep spindles (Brazier, 1968; Montplaisir et al., 1981) and that the amplitude of sleep spindles is the highest in the frontal–central regions (Caderas et al., 1982). The reason for such widespread cortical distributions of sleep spindles is not clearly understood.

The localization of the spindle generators becomes even more complex when one considers the frequency-dependent nature of their special cortical distribution. Scalp-recorded sleep spindles in humans involve both slow and fast frequency oscillations that have distinct cortical distributions (Anderer et al., 2001; Broughton and Hasan, 1995; Gibbs and Gibbs, 1950; Zeitlhofer et al., 1997; Zygierewicz et al., 1999). The relatively low-frequency spectra centered at 12 Hz is more pronounced over frontal areas, whereas higher frequency spectra centered at around 14 Hz are distributed more posteriorily (Gibbs and Gibbs, 1950; Zygierewicz et al., 1999). These studies clearly show that spindle activity is complex and in order to understand the biological function of that complexity a source localization method might be helpful to track the neuronal activity underlying these types of spindles.

Despite this complexity, it is clinically and scientifically important to understand what cortical brain regions are involved in maximal spindle activity and what type of spindles are contributing to such activation. It was shown that different types of spindles appear to reflect neuronal networks that react differentially to age-related changes, developmental maturation changes, circadian factors, pharmacological agents and homeostatic changes (Dijk and Czeisler, 1995; Dijk et al., 1995; Landolt et al., 1996; Shinomiya et al., 1999).

To understand the biological function of widespread distribution of the spindles EEG studies were performed on topographic and physiologic aspects of spindle activity; however, the EEG source localization analyses have limitations. One limitation of scalp-recorded EEG is that the result is a summation of all simultaneously active neuronal generators in the given time period. This makes separation and localization of individual spindle generators more difficult. Unlike the EEG, the magnetic fields are less distorted by the resistive properties of the skull and scalp, which results in better spatial resolution of the MEG. In addition, MEG with its high temporal resolution, on the order of milliseconds, when compared with other neuroimaging techniques, is well suited for localizing sources of brain activity, particularly cortical ones, as its spatial resolution is more accurate (5 mm or less) compared with that of the EEG (Tepley, 2005).

The MEG method has been successfully used for localization of sleep spindles (see Table 1); however, cortical regions related to the maximal source of spindle amplitude activity have received limited research attention. Urakami (2008) showed that fast EEG/MEG spindles are more frequently observed in the postcentral areas, whereas slow spindles are localized more to precentral than to postcentral areas. From that study and others (Ishii et al., 2003; Lu et al., 1992; Manshanden et al., 2002; Shih et al., 2000), we may hypothesize that despite multisource generators of the spindle activity, there are possibly unifying cortical networks representing a common neural basis of both types of spindles, as at least slow spindles were localized to both pre- and postcentral areas (see Urakami, 2008).

Table 1.   Source localization of sleep spindles across studies using whole-head magnetoencephalographic system
StudyMethodology of source localization performanceNumber of subjectsNumber of spindles processed for source localizationArea of activation during spindle activity
Shih et al. (2000)Principal component analysis (PCA) and equivalent dipole source model410 spindles from each subjectSleep spindles were localized to all four cerebral lobes, but the frontal and parietal lobes showed more frequent involvement in spindle activity
Manshanden et al. (2002)Principal component analysis (PCA) and equivalent dipole source model450–99 among the subjectsCentro-parietal region
Ishii et al. (2003)Beam formers, spatial filtering technique (SAM)830–77 spindle epochs among the subjectsFrontal and parietal lobes, including vicinity of the primary sensorimotor areas
Urakami (2008)Equivalent current dipole model710 (13–14 Hz) and 5 spindles (11–12 Hz) for each subjectPre- and postcentral areas of the frontal and parietal lobes

The aim of the present study was to localize the maximal source spindle amplitude activity and examine the frequency and amplitude differences between two types of spindles contributing to this maximal activity within the region of activation. The present study computed MEG localizations using MR-FOCUSS, a current source density technique (Moran et al., 2005) which utilizes a wavelet statistical operator; so, spatial resolution can be chosen appropriately for extended sources such as sleep spindle activity. This technique is capable of localizing spontaneous multisource brain activity and determining the scale of the activation using the values of the source amplitude.

Materials and methods

Subjects

The data set used for the present study is a subset of a larger data set from an ongoing project that is aimed to determine if a general trait of vulnerability to sleep disturbances exists. From a group of 19 subjects who participated in this larger study, 13 subjects reached stage 2 sleep; however, only nine subjects (mean age: 26.1 ± 8.7; six women, all subjects are right handed) met the criterion for the number of spindle per 15 min of stage 2 sleep. The subject inclusion criteria for the present study were based on the literature review (see Table 1). Across four MEG studies, the average number of spindles submitted for the source localization algorithms was approximately 30 spindles per subject. Therefore, at least 20 spindles per subject were set as the subject’s inclusion criterion in the present study. This was regarded as evidence of stage 2 sleep recorded during 15 min.

All participants were free of sleep complaints, as assessed by a board-certified sleep specialist, and had no objective signs of sleep disturbance determined from a standard overnight 8-h polysomnogram (PSG) including respiratory measurement and anterior tibialis electromyograph (apnea–hypopnea index <10 per hour and periodic limb movements <10 per hour; see Table 2). One subject was excluded due to abnormal sleep parameters (sleep efficiency <50%). Thus, analyses were performed on eight normally sleeping subjects.

Table 2.   PSG, sleep parameters obtained from each individual subject during the sleep screen night
SubjectsSOL (min)WASO (min)TST (min)TIB (min)SE (%)Stage 1 (%)Stage 2 (%)Stage 3 (%)REM (%)
  1. SOL, sleep onset latency; WASO, wakefulness after sleep onset; TST, total sleep time; TIB, time in bed; SE, sleep efficiency; PSG, polysomnograph; REM, rapid eye movement sleep; SD, standard deviations (in min or %).

18.511461480.595.96.451.123.119.4
2420458480.595.31.681.72.614.1
335.528417480.586.86.756.22116.1
48.533440480.591.57.764.917.113.3
51411454.548094.72802.615.4
61816.5434.5480.590.42.252.12619.7
74.536433486.591.4460.516.219.3
82513447.5486.5920.753.32818
Mean ± SD15 ± 1121 ± 10443 ± 15482 ± 392 ± 34 ± 2.662 ± 1217 ± 9.717 ± 2.5

The subjects had no history of any medical or psychiatric disorder as determined by physical evaluation, blood chemistries and a scheduled diagnostic interview for mental disorders-IV (SCID). Subjects were free of all central nervous system-acting medications at least 2 weeks prior to the laboratory assessment. All subjects were free of alcohol and nicotine at least 24 h prior to the study. Each participant kept a sleep diary 2 weeks prior to the MEG study. All subjects were paid for participation and the study was approved by the Internal Review Board of Henry Ford Hospital, Detroit, MI, USA.

Procedure

Both the MEG (148 MEG sensors, 4D Neuroimaging WH2500 Neuromagnetomer System, BTI, San Diego, CA, USA, see Fig. 1a) and EEG (NeuroScan, Charlotte, NC, USA) activity of sleep spindles were recorded simultaneously. Four EEG electrodes were applied to the subject’s scalp (C3, C4, O1 and O2); two EOG electrodes (electrooculogram, vertical/horizontal) were used for eye movement recordings and EMG electrode was placed on the chin for electromyogram activity recording. The electrocardiogram (ECG) was recorded using a standard V5 leads. EEG, EOG and EMG data were recorded with the left earlobe electrode as a reference.

Figure 1.

 (a) Magnetoencephalography (MEG) sensor position. Circled MEG channels were selected for visualizing sleep spindles during sample procedure. (b) Electroencephalogram (EEG) and MEG data segment of sleep spindle in individual subjects.

Magnetoencephalography recording was performed from 07:30 to 08:30 hours. The neuromagnetometer detector array was placed over the head of the subject while in supine position. After a short (1–2 min) biocalibration (eye’s and muscle’s movement), subjects were asked to ‘try to fall asleep’ and data acquisition began. In order to maintain accurate head positioning during the MEG recording, the recording session was divided into separate 15-min segments. Each subject performed from 30 to 45 min of MEG recording (two or three segments) with no awakening between runs. Data acquisition was monitored online, and an experienced PSG technician (CJ) detected sleep stages in real time on the computer screen.

Prior to data acquisition, the locations of the nasion, left and right preauricular points and additional points to delineate the location and head shape of each subject were acquired. Five signal coils (three coils were attached to the forehead and two coils were placed in front of preauricular notch of left and right ear) were used to track any potential head movement during the MEG session. The head position was measured online before and after each 15-min segment with no awakening of the subject. During the MEG recording, subjects slept comfortably on a bed in supine position inside the magnetically and sound-shielded room with lights off. Total sleep time during all 30–45 min of MEG recording across eight subjects was from 18 to 30 min (see Table 3). The data were acquired using an analog filter from 0.1 to 100 Hz with a sampling rate of 508.69 Hz, and stored on a disk for later analysis. Any potential ECG and EOG artifacts were visually identified and removed from continuous MEG data.

Table 3.   Total sleep time for each individual obtained during MEG (30–45 min) session
SubjectsTST (min)Stage 1 (min)Stage 2 (min)
  1. Stage 1 and stage 2 sleep obtained from 15-min segment were used for MEG source localization.

  2. MEG, magnetoencephalography; TST, total sleep time.

13015
2245.59.5
32115
4281.513.5
526.5114
623212
718310
829.5213

A combination of second- and fourth-order independent component analysis (ICA) method was used to remove heart artifacts from continuous MEG data. A power density operator/template was constructed using the ECG signal. The ECG can be a template for constructing a power density operator, for extracting the heart artifacts in the MEG data. MEG heart artifact is very similar to the ECG, recorded from the chest. The heart artifact was removed from the MEG data by using this template (for details see Moran et al., 2004, 2005).

Identification of sleep spindles

Visual sleep scoring of the EEG and detection of spindles were based on 30-s segments and scored according to the criteria of Rechtschaffen and Kales, using either C3 or C4 leads (Rechtschaffen and Kales, 1968). Simultaneous MEG spindles were detected in frontal, bilateral temporal, central and parietal MEG channels, for illustration see Fig. 1. For each subject, the MEG segment (duration of 15 min) containing the most stage 2 sleep was selected for further analyses (see Table 3). Criteria for the spindle selection were: (1) 10–15 Hz frequency EEG/MEG waves; duration between 0.7 and 2.5 s, (2) appearing only at non-rapid eye movement (NREM) stage 2 sleep, (3) not associated with a vertex wave or with a K-complex. All spindles that matched these criteria were marked for source localization analysis.

Prior to further localization analyses, all spindles were visually checked for any artifact-related activity (eye and muscle movements), if these types of artifact were present the respective spindles were rejected from further source localization analyses. Thus, two to four spindles per subject were excluded due to EOG artifacts in the waveform. The occurrence of spindles in MEG only or EEG only was found. However, because the identification of spindles was performed using both the EEG and MEG, and localization was performed using MEG, analyses were restricted to those spindles that were detected in both EEG and MEG data (see Table 5). For visual sleep scoring, continuous EEG and MEG data were filtered off-line using a band pass filter from 6 to 30 Hz. Thus, 20–26 EEG/MEG spindles per subject were selected for source localization analyses. Furthermore, in order to isolate spindle-related activity, a band pass filter from 10 to 15 Hz was applied to selected MEG sleep spindle data prior to localization of the spindle’s activity and averaging these source imaging results across spindle events.

Table 5.   Maximal source spindle amplitude (nA m) for slow and fast spindles measured in precentral and postcentral gyri
SubjectSpindles for each subjectPrecentral gyrusPostcentral gyrus
Slow spindlesFast spindlesSlow spindlesFast spindles
  1. Values are given as mean ± SD. Values in parentheses indicate the number of individual spindles.

1260.28 ± 0.02 (10)0.405 ± 0.03 (16)0.237 ± 0.02 (12)0.500 ± 0.1 (14)
2220.73 ± 0.1 (12)0.28 ± 0.02 (10)
3201.25 ± 0.03 (5)1.05 ± 0.07 (15)0.45 ± 0.03 (6)1.90 ± 0.14 (14)
4200.66 ± 0.04 (11)0.59 ± 0.02 (9)0.39 ± 0.04 (11)0.57 ± 0.07 (9)
5202.42 ± 1.4 (11)0.62 ± 0.06 (9)
6211.01 ± 0.11 (13)1.6 ± 0.2 (8)1.73 ± 0.16 (8)1.88 ± 0.12 (13)
7202.5 ± 0.23 (12)1.06 ± 0.08 (8)
8200. 95 ± 0.05(6)1.62 ± 0.09 (14)0.99 ± 0.06 (10)0.87 ± 0.05 (10)

Magnetoencephalographic data processing

The distribution of brain activity corresponding to MEG spindle data was separately calculated using a combination of ICA and current source imaging technique: MR-FOCUSS (Moran et al., 2004, 2005). MEG data were co-registered with subject’s anatomical magnetic resonance imaging (MRI) image.

To localize cortical source activation of sleep spindle amplitude, a model of gray matter was constructed using a T1-weighted high-resolution MRI volumetric image from each individual. The realistic head model consisted of X-, Y- and Z-oriented dipoles at approximately 4000 locations distributed such that each location represented the same amount of gray matter identified in the subject’s volumetric MRI image. For each of the brain-modeled locations the X, Y and Z amplitudes were combined to create total amplitude of activity; averaged over the time interval of spindle activation and across all spindles for each subject. The amplitude of imaged brain activity is displayed in units of nanoampere meters (nA m).

Forty per cent of the maximal source spindle amplitude was localized. The region of activation, related to maximal activity, was determined based on 50% or greater activity within this region. The center of spindle source activity in each individual region of activation was transformed to Talairach space and Montreal Neurological Institute (MNI) space.

To test the statistical significance of maximal spindle source amplitude between slow and fast spindles, as well as between hemispheres, the mean across time window of the spindle of maximal source amplitude was normalized and statistically compared. For all statistical comparisons, the Wilcoxon-matched pairs test was applied. Significance was accepted at the 0.05 level.

Results

Table 2 summarizes the mean values and standard deviations of sleep parameters across all individuals. From 2-week sleep diaries collected from each subject, the mean total sleep time over 2 weeks was 8.6 ± 1.15 h; sleep efficiency was 93% ± 1.8%; and habitual time in bed was 8.8 ± 1.25 h.

The total number of EEG/MEG spindles identified for MEG analysis ranged from 20 to 26, depending on which subject was selected. The duration of the spindles varied from 0.7 to 2.5 s. Fig. 2 demonstrates the multisource locations of spindle amplitude for all spindles from a representative individual subject (no. 8). Note that central and temporal and parietal regions are most active for individual spindles.

Figure 2.

 Individual spindle’s source localization for the selected subject (no. 8) (on the left). The example of the selected individual spindle across 148 magnetoencephalography sensors (on the right).

In order to identify the main cortical regions for maximal source spindle amplitude activity the top 40% of source activity was imaged to each individual’s MRI. The main region of activations corresponding to maximal source activity is depicted in Fig. 3 across all eight subjects. The maximal source sleep spindle amplitude was localized to frontal, temporal and parietal lobes in each subject. Although the data clearly show inter-subject variability in locations, the precentral and postcentral gyri are most frequently observed in localization of the maximal source spindle activity. Table 4 summarizes the MEG locations for the maximal source amplitude across the region of activation in Talairach, MNI spaces and named in Broadman’s areas. The results of further examination of the region of activation within precentral and postcentral areas for each location showed that a mixture of activities related to slow and fast spindles were present in these regions of activation.

Figure 3.

 Maximal source spindle amplitude locations depicted on the coronal view of the individual MRIs. Note: precentral and postcentral gyri are most frequently observed in maximal spindle activation across all subjects.

Table 4.   Talairach and MNI coordinates for maximal source spindle amplitude across regions of activation (given in Broadman’s area)
SubjectsBroadman's areaTalairach locationsMNI locations
XYZXYZ
  1. MNI, Montreal Neurological Institute.

1BA 4, R. superior frontal gyrus11−196612−2072
BA 4, R. precentral gyrus33−255437−2558
2BA 1, L. postcentral gyrus−43−3052−45−3158
BA 2, R. postcentral gyrus39−385741−3961
3BA 4, L. precentral gyrus−33−2158−35−2165
BA 4, L. superior frontal gyrus−10−2566−11−2673
BA 4, R. precentral gyrus51−163956−1543
4BA 1, R. postcentral gyrus47−215150−2155
BA 1, L. postcentral gyrus−48−1948−51−1954
5BA 6, R. middle frontal gyrus2655828663
BA 44, R. precentral gyrus5133456536
6BA 6, L. superior frontal gyrus−13−1062−14−1069
BA 4, R. precentral gyrus37−135041−1356
BA 1, R. postcentral gyrus59−152164−1422
7BA 40, L. postcentral gyrus−50−3846−53−3550
BA 42, R. supramarginal gyrus61−332264−3323
8BA 4, R. precentral gyrus31−276133−2767
BA 4. R. superior frontal gyrus11−266412−2771
BA 40, R. postcentral gyrus63−221867−2218

Maximal source spindle activity from the right hemisphere was compared with maximal source spindle activity in the left hemisphere within each individual subject. Statistical comparison yielded larger source amplitude spindle activation over the right hemisphere compared with that over the left hemisphere in four subjects (see Fig. 4). One subject showed a tendency for this pattern (subject 7), but it did not reach statistical significance (P < 0.056). Despite a left hemisphere dominance pattern of activation in subject 2, the result of a comparison did not show significant differences (P < 0.33) between hemispheres in maximal source spindle activity for this subject.

Figure 4.

 Normalized source spindle amplitude measured from right and left hemispheres across subjects.

Region of activation for maximal spindle activity

The region of activation for maximal spindle activity was determined on 50% or greater overall activity within the active region. For example, for subject 3, the maximal source amplitude activity was observed across seven locations (see Fig. 3), but only three locations showed predominant activation of 66%, 70% and 74% within left precentral, left superior frontal and right precentral gyrus, respectively, whereas other regions showed 1% (right superior temporal gyrus), 12% (middle temporal gyrus), 32% (left postcentral gyrus) and 5% (right postcentral gyrus). The regions of activation ≥ 50% for maximal source spindle amplitude are shown in Table 4 for all subjects.

Further examination of the spindles within this region of activation showed that subject 1 had a total of 26 spindles during 15-min segment of sleep. Ten of these spindles were of slow type (frequency at 12 Hz) and 16 were of fast type (frequency at 14 Hz) in the precentral gyrus; and 12 (slow)/14 (fast) in the postcentral gyrus (see Table 5) contributing to the maximal source activity. Subjects 2 and 7 did not show the precentral gyrus as a region of activation in the maximal source spindle activity, but they had activation in the postcentral region: 10 slow and 12 fast, 12 slow and 8 fast, respectively. Five subjects displayed activation in both regions (see subjects 1, 3, 4, 6 and 8 in Table 5). Subject 5 did not have maximal source activity in the postcentral gyrus.

Statistical comparison of normalized source spindle amplitude activity between slow and fast spindles within and between region(s) of activation centered in precentral and postcentral gyri did not show a statistical difference (P > 0.05), suggesting that activity of the source amplitudes between these two regions of activation may reflect a unifying network underlying these spindles over the central area of the brain.

Discussion

The present interdisciplinary study had a number of shortcomings related to ‘time-of-night’ sleep. However, our findings show similarities to studies using different design, time of sleep and even method (e.g. EEG). The present study used whole-head MEG to examine the cortical regions involved in maxima of spindle amplitude activity in healthy, normal sleeping individual subjects during a morning nap. Despite the fact that in the present study spindles were obtained from stage 2 sleep during morning nap, our results show that sleep spindles have multiple cortical sources that are seen in frontal, temporal and parietal brain regions that correspond to the previous EEG and/or MEG findings (Anderer et al., 2001; Broughton and Hasan, 1995; Gibbs and Gibbs, 1950; Ishii et al., 2003; Lu et al., 1992; Manshanden et al., 2002; Shih et al., 2000; Urakami, 2008; Zeitlhofer et al., 1997; Zygierewicz et al., 1999). In addition, a new approach for spindle localization used in this study, such as averaging spindle amplitude and localizing the source of the maximal activity, showed specific locations for the maximal spindle activity centered at precentral and postcentral gyri. Our findings also show that this maximal activity probably reflects a unifying network for both slow and fast spindles within these regions.

The complex nature of the spindle activity, especially slow- and fast-frequency spindle activity was revealed by topographic EEG studies (Anderer et al., 2001; Zeitlhofer et al., 1997; Zygierewicz et al., 1999; for a review, De Gennaro and Ferrara, 2003). The findings from these studies show that the neuronal network underlying these two types of spindles is topographically distinct. Slow-frequency spindles are more frontally distributed, whereas fast spindles are distributed more parietally. In our study, we did not find a topographical distinction for the source amplitudes of slow and fast spindles that were measured in the precentral and postcentral gyri. This may possibly be due to the common neuronal proximity between these regions or perhaps, that the sleep segment in our study did not contain enough spindles related to the beginning of stage 2 sleep and deeper stage 2 as well as nocturnal sleep due to the time limitation aspects of the MEG experiment. Thus, we were also not able to compare spindles from distinctive portions of the sleep cycle, because we collected a portion of the light sleep stage 2 during the morning nap. It is well known that fast and slow spindle density changes from light to deep sleep across sleep cycles (for a review, De Gennaro and Ferrara, 2003). In our study, the spindles that we localized are elicited during stage 2 sleep during morning nap time. Taking into account the time of sleep and duration, there are potential limitations in neuronal sources underlying spindle activity that relate to sleep cycles and time of sleep. Further studies are needed to understand the difference in source localization of the sleep spindles obtained from the full range of all sleep cycles.

Two previous MEG studies (Manshanden et al., 2002; Urakami, 2008) demonstrated the ability of the MEG to distinguish the neuronal sources underlying spindle activity. Thus, Manshanden et al. (2002), using equivalent dipole source techniques were able to localize sleep spindles and distinct spindle activity source from alpha activity source. In that study, authors showed the common centro-parietal distribution of spindles across four individuals.

Further, Urakami (2008), using the same method (equivalent dipole source techniques), was able to show the existence of slow and fast spindle-neural networks that propagate to the common cortical level of activation. Our findings support this result by showing that this common cortical neural circuit consists of precentral and postcentral regions, as the maximal source of the spindle activity was localized to these regions, although some inter-subject variability was observed.

Using a current density imaging technique in the present study, we were able to localize both distributed and focal sources associated with spindle activity (for reviews on source reconstruction methods, see Fuchs et al., 1999; Lopes da Silva, 2004; Moran et al., 2005; Yao and Dewald, 2005); we also found that the right hemisphere is more active in maximal activation of the source amplitude in 50% of our subjects. This laterality pattern should be studied more thoroughly, but an effect appears to be present.

Although we applied different methods and a real head model (reconstructed using individual MRIs) for MEG source reconstruction to spindle amplitude, several aspects of our findings are consistent with studies using EEG (Anderer et al., 2001; Zeitlhofer et al., 1997; Zygierewicz et al., 1999). Similar to their findings, our results show frontal and parietal lobe activation (precentral and postcentral gyri), although some other regions (superior frontal gyrus and middle frontal gyrus) of activation for maximal source amplitude spindle activity were observed. Moreover, as our results are derived from MEG recordings, it is more likely that neuronal generators contributing to spindle amplitude and measured in the present study originate from cortical areas, because the signal rapidly attenuates with an increasing distance between the generator and the detector (Lopes da Silva, 2004).

Steriade et al. (1993) suggested that slow- and fast-frequency spindles may be attributable to a single event that depends on the duration of the hyperpolarization-rebound sequence in thalamocorticial neurons. They hypothesized that if the duration of hyperpolarization is about 70 ms, the frequency of the spindle would be 14–15 Hz, whereas longer hyperpolarization would be associated with a slower spindle frequency. Because we did not find amplitude and location differences between maximal slow and fast spindle activity within pre- and postcentral areas, our data are consistent with both slow and fast spindle activity as a single event in global thalamocortical coherence.

Limitations of the study

There are some limitations of the present study. A relatively short duration of sleep recording using MEG methods that led to: (i) a relatively small number of spindles used for localization. (ii) The nature of sleep spindles studied might be different from spindles measured during nocturnal sleep carried out by using EEG methods (for a review see De Gennaro and Ferrara, 2003). (iii) The lack of simultaneous MEG and EEG recordings may reflect the limitation of some spindle activity that MEG does not detect (e.g. deep neuronal sources or sources that have a radial orientation). Further MEG studies of source reconstruction in spindle activity should take into account these limitations as well as others related to spontaneous data localization techniques.

In summary, the present study suggests that precentral and postcentral gyri are main cortical regions that are associated with maximal portion of cortical contribution to thalamocortical loop of spindle oscillations elicited during morning nap. These two regions may also mediate synchronization of cortical spindle across cortex, as both regions are involved in slow and fast spindle network as well as in maximal spindle activity.

Acknowledgements

This study was supported by NIMH Grant 068372 to CLD and NINDS Grant R01 NS30914 to NT. The authors thank Mrs Karen Mason, Mr Sylvester Parker and Ms Rhonda Morris for their contribution in the data collection. We thank Dr GL Barkley for useful comments to the previous version of the manuscript.

Disclosure statement

This was not an industry-supported study. All authors have indicated no financial conflicts of interest.

Ancillary