Brain-derived neurotrophic factor gene polymorphism predicts interindividual variation in the sleep electroencephalogram
Abstract
Previous studies have suggested that brain-derived neurotrophic factor (BDNF) participates in the homeostatic regulation of sleep. The objective of this study was to investigate the influence of the Val66Met functional polymorphism of the BDNF gene on sleep and sleep EEG parameters in a large population-based sample. In total 337 individuals participating in the São Paulo Epidemiologic Sleep Study were selected for analysis. None of the participants had indications of a sleep disorder, as measured by full-night polysomnography and questionnaire. Spectral analysis of the EEG was carried out in all individuals using fast Fourier transformation of the oscillatory signals for each EEG electrode. Sleep and sleep EEG parameters in individuals with the Val/Val genotype were compared with those in Met carriers (Val/Met and Met/Met genotypes). After correction for multiple comparisons and for potential confounding factors, Met carriers showed decreased spectral power in the alpha band in stage one and decreased theta power in stages two and three of nonrapid-eye-movement sleep, at the central recording electrode. No significant influence on sleep macrostructure was observed among the genotype groups. Thus, the Val66Met polymorphism seems to modulate the electrical activity of the brain, predicting interindividual variation of sleep EEG parameters. Further studies of this and other polymorphic variants in potential candidate genes will help the characterization of the molecular basis of sleep. © 2014 Wiley Periodicals, Inc.
The electroencephalogram (EEG) during sleep and wakefulness is considered to be under strong genetic control, with heritability estimated as high as 96% in the 8–16-Hz EEG frequency band of nonrapid-eye-movement (non-REM) sleep (De Gennaro et al., 2008). Studies in healthy young individuals have shown that sleep EEG presents pronounced interindividual variability but is remarkably stable over time within subjects (De Gennaro et al., 2005; Tucker et al., 2007). Genetic polymorphisms in a number of candidate genes have been shown to alter sleep stage proportions and EEG activity in different frequency ranges during wakefulness, REM sleep, and non-REM sleep (Retey et al., 2005; Viola et al., 2007; Bodenmann et al., 2009; Mazzotti et al., 2011, 2012).
Brain-derived neurotrophic factor (BDNF) is a key molecule involved in the growth, development, and modulation of the nervous system (Greenberg et al., 2009). Previous studies have suggested that BDNF may participate in the homeostatic regulation of sleep, since its expression has consistently been found to be upregulated in the brain of awake and sleep-deprived animals (Faraguna et al., 2008; Cirelli, 2009; Guindalini et al., 2009). A common nonsynonymous single nucleotide polymorphism (rs6265) in the BDNF gene, which leads to a valine (Val)-to-methionine (Met) substitution at amino acid 66 (Val66Met) of the protein (Pruunsild et al., 2007), has been associated with reduced performance on different types of cognitive tasks (Egan et al., 2003; Frielingsdorf et al., 2011) and with susceptibility to neuropsychiatric disorders (Levinson, 2006; Rybakowski et al., 2009; Terracciano et al., 2010). The Met variant has been shown to alter intracellular tracking and packing of pro-BDNF negatively, affecting the activity-dependent secretion of the mature peptide (Egan et al., 2003; Lang et al., 2009).
Gatt et al. (2008) reported that Met homozygosity of the BDNF gene predicted an increase of relative theta and delta power and a reduction of alpha power in both eyes-open and eyes-closed conditions. In a recent study, Bachmann and colleagues (2012) found reductions of delta/theta range and an increase of alpha and sigma EEG activity during non-REM sleep in Met allele carriers undergoing 40 hr of prolonged wakefulness. In REM sleep, alpha, theta, and sigma activities were lower in Val/Met genotype carriers, suggesting that this polymorphism affects functional brain oscillations and may contribute to the regulation of sleep homeostasis. The objective of the present study was to investigate the influence of the BDNF gene polymorphism on sleep and sleep EEG parameters in a large population-based sample of individuals of different ages without sleep disorders from São Paulo, Brazil.
MATERIALS AND METHODS
Subjects
The study was conducted on individuals participating in the São Paulo Epidemiologic Sleep Study, which was a population-based survey adopting a probabilistic three-stage cluster sample from São Paulo to represent the population according to gender, age (20–80 years), and socioeconomic status. The study was carried out in 2007 to establish the epidemiological profile of sleep disorders in the adult population of São Paulo. Questionnaires, polysomnography (PSG), and blood samples were collected to investigate associations between sleep patterns and sociodemographic status; physical activity habits; mood disturbances; memory complaints; sexual dysfunction in males; drug addiction; genetic markers; and anthropometric, clinical, biochemical, hematological, endocrine, immunological, and inflammatory indicators. More details on the design, sampling, and procedures of the São Paulo Epidemiologic Sleep Study are provided elsewhere (Santos-Silva et al., 2009). Sleep disorders were ruled out using PSG combined with the questionnaires listed below.
The study was approved by the Ethics Committee for Research of the Universidade Federal de São Paulo (CEP 0593/06) and was registered with ClinicalTrials.gov (NCT00596713, Epidemiology of sleep disturbances among adult population of the São Paulo City). All volunteers read and signed an informed consent form.
PSG and Clinical Assessment
A complete full-night PSG was performed in the sleep laboratory of the Sleep Institute (São Paulo, Brazil) using a digital system (EMBLA S7000; Embla Systems, Broomfield, CO). The recording sensors were attached to the patient in a noninvasive manner using tape or elastic bands, and physiological variables were monitored simultaneously and continuously. All PSG and sleep-stage scoring were performed according to standardized criteria (Rechtschaffen and Kales, 1968). EEG arousals, sleep-related respiratory events, and leg movements during sleep were scored according to the criteria established by the American Academy of Sleep Medicine Manual for scoring sleep and associated events (Iber et al., 2007). In accordance with the International Classification of Sleep Disorders (ICSD-2; American Academy of Sleep Medicine, 2005), subjects were considered to have obstructive sleep apnea syndrome (OSAS) if they had an apnea-hypopnea index (AHI) between 5 and 14.9 and presented at least one of the following complaints: loud snoring, daytime sleepiness, fatigue, or breathing interruptions during sleep. Subjects with an AHI of 15 were also considered to have OSAS regardless of whether they had any of the aforementioned complaints. Participants were classified as insomniacs or noninsomniacs using subjective insomnia measures. Items were extracted from the questionnaires and combined as an algorithm of the DSM-IV (American Psychological Association, 1994) general criteria to define subjective insomnia as 1) difficulty initiating sleep (sleep-onset latency >30 min) or difficulty maintaining sleep (early-morning awakenings); 2) insomnia occurring 3 or more days per week, persisting for at least 6 months, and occurring in the past 30 days; and 3) relevant daytime impairment, interfering “much” or “extremely” with daily functioning. To be diagnosed with DSM-IV Insomnia (DSM-IS), individuals must have met all three criteria (Castro et al., 2013). A complete description of the clinical assessment and the outcomes are presented elsewhere (Santos-Silva et al., 2009).
The following questions based on standard diagnostic criteria were used to assess restless leg syndrome (RLS) status (Allen et al., 2003, 2005; Walters et al. 2003). 1) When you try to relax in the evening or go to sleep at night, do you ever have unpleasant, restless feelings in your legs that can be relieved by walking or movement? (This question was used to assess whether the person was susceptible to RLS.) 2) Are the feelings due to muscle cramps in the legs or feet? (This question was used to exclude false positives resulting from muscle cramps.) 3) Will simply changing leg position by itself once without continuing to move usually relieve these feelings? (This question was used to determine positional discomfort.) 4) In the past 12 months, how often did you experience these feelings in your legs? (This question was used to determine frequency.) All subjects diagnosed with a sleep disorder were excluded from the analysis.
Sleep EEG Spectral Analysis
A specific syntax in R (version 2.10.1) was used for spectral analysis of the sleep EEG. Waves from C3-A2, C4-A1, O1-A2, and O2-A1 derivations were decomposed into delta (<4 Hz), theta (4–7.9 Hz), alpha 1 (8–9.9 Hz), alpha 2 (10–12.9 Hz), beta 1 (13–17.9 Hz), beta 2 (18–29.9 Hz), and gamma (=30 Hz) frequency bands using fast Fourier transformation with a sampling rate of 200 Hz using epochs of 20 sec. The filter settings used were according to standard criteria of sleep EEG data acquisition (low-frequency filter of 0.3 Hz, high-frequency filter of 35 Hz, time constant of 0.3 sec, and notch filter of 60 Hz). Artifacts were excluded as follows: the descriptive data (mean, standard deviation, median, and interquartile range) from each 20-sec window were calculated, and the 5% of time windows with the highest signal amplitude (maximum–minimum) at each sleep stage were considered outliers and excluded from analysis. For validation of this procedure, an experienced polysomnographist subjectively identified and excluded artifacts in 30 polysomnographies. Results of the visual analysis and the R syntax were compared using a kappa test (κ = 0.79, P = 0.002), showing good agreement between the two methods. Results are represented by mean spectral power (μV2/Hz) ± standard error of the mean.
Sample Collection and Genetic Analysis
Approximately 5 ml of venous blood was collected in vacuum tubes, and DNA was extracted from peripheral whole blood using a standard protocol (Miller et al., 1988). The Val66Met polymorphism (rs6265; http://www.ncbi.nlm.nih.gov/SNP/snp_ref.cgi?rs=rs6265) of the BDNF gene was selected for analysis. This polymorphism has different prevalence across populations, and association studies in admixed populations such as the Brazilians are prone to population stratification bias. To correct for the presence of population stratification in our sample, we selected a total of 31 ancestrally informative markers, e.g., genetic polymorphisms that exhibit large allelic frequency differences among the three main Brazilian ancestral populations (Europeans, Africans, and Native Americans; for more details see Guindalini et al. [2010]). By using the genotypic data, the number of ancestral populations (K) in the sample and the individual proportions of alleles were estimated using the Bayesian Markov chain Monte Carlo method implemented in the program STRUCTURE (Pritchard et al., 2000; Falush et al., 2003). The genotyping of all selected polymorphisms for this study was performed blinded to sample status under contract by Prevention Genetics (Marshfield, WI) using an allele-specific PCR assay protocol, which has been previously described (Myakishev et al., 2001).
Statistical Analysis
Hardy-Weinberg equilibrium was tested by calculating a χ2 statistic with one degree of freedom for the whole sample. Genotype and allele frequencies and risk factor prevalence were compared using the Fisher's exact test. Student's t-tests were used to compare quantitative variables among groups. Correction for multiple comparisons was performed when appropriate, using the Benjamini-Hochberg correction for false discovery rate. A general linear model was used to verify the effect of potential covariates on the association between genotypes and sleep power spectra. P < 0.05 was considered statistically significant.
RESULTS
In total 337 of the 1,042 individuals participating in the epidemiological assessment were eligible for the present study. Subjects who presented OSAS, periodic leg movements > 5/hr, RLS, or DSM-IV IS (N = 640) and individuals taking medication that could influence sleep (e.g., antidepressants, narcoleptics; N = 93) were excluded from the analysis. Genotyping success rate was 95%. The final sample was 44.5% male, with a mean age of 37.5 ± 13.2 years. The overall genotypic frequencies were 69.7% for the Val/Val genotype, 26.7% for the Val/Met genotype, and 3.6% for the Met/Met genotype. BDNF genotypes were in Hardy-Weinberg equilibrium for the sample as a whole (χ2 = 0.84, P = 0.36). Because of the low frequency of the Met/Met genotype and its previously reported functional effects, the Val/Met and Met/Met genotypes were grouped for analysis (“Met carriers”). No significant differences were found among the genotype groups in terms of sex (58.3% vs. 49.0% female gender in Val/Val genotype vs. Met allele carriers, respectively; P > 0.05); age; BMI; or estimated European, African, or Native American genetic ancestries (P > 0.05; Table 1). The genotype groups also did not differ in terms of daily caffeine consumption, caffeine intake on the day prior to PSG, anxiety, depression, morningness–eveningness, or sleepiness scores, as measured by validated questionnaires (P > 0.05, data not shown; Santos-Silva et al., 2009).
| Val/Val carriers | Met carriers | |||||||
|---|---|---|---|---|---|---|---|---|
| Sample characteristics | N | Mean | SD | N | Mean | SD | t | P |
| Age | 235 | 37.46 | 13.13 | 102 | 37.18 | 13.34 | −0.184 | 0.854 |
| Body mass index | 235 | 25.67 | 4.70 | 102 | 25.44 | 4.23 | −0.428 | 0.669 |
| European | 234 | 0.74 | 0.18 | 101 | 0.74 | 0.19 | 0.266 | 0.791 |
| African | 234 | 0.20 | 0.17 | 101 | 0.18 | 0.16 | −1.085 | 0.279 |
| Native American | 234 | 0.07 | 0.10 | 101 | 0.08 | 0.13 | 1.014 | 0.312 |
- a N, number of individuals; SD, standard deviation; Met carriers, Val/Met + Met/Met genotype carriers.
Analysis of the sleep structure parameters indicated that the two BDNF genotype groups had similar polysomnographic characteristics. No significant differences were found between the Val/Val and the Met carriers with regard to sleep latency and REM sleep latency; total sleep time; percentage of time spent in stages 1, 2, 3 + 4, and REM sleep; sleep efficiency; minutes awake; and arousals per hour index (P > 0.05; Table 2).
| Val/Val carriers | Met carriers | |||||||
|---|---|---|---|---|---|---|---|---|
| Polysomnography measurements | N | Mean | SD | N | Mean | SD | t | P |
| Sleep latency (min) | 235 | 15.91 | 18.93 | 102 | 19.30 | 33.20 | 0.964 | 0.337 |
| REM sleep latency (min) | 235 | 95.00 | 44.43 | 102 | 88.67 | 38.59 | −1.249 | 0.213 |
| Total sleep time (min) | 235 | 347.74 | 75.27 | 102 | 352.62 | 78.77 | 0.540 | 0.590 |
| Sleep efficiency (%) | 235 | 83.65 | 11.76 | 102 | 82.65 | 13.02 | −0.694 | 0.488 |
| Stage 1 (%) | 235 | 4.23 | 3.29 | 102 | 4.19 | 3.48 | −0.088 | 0.930 |
| Stage 2 (%) | 235 | 53.99 | 8.03 | 102 | 53.08 | 8.64 | −0.937 | 0.349 |
| Stages 3 + 4 (%) | 235 | 22.62 | 7.59 | 102 | 22.79 | 7.75 | 0.192 | 0.848 |
| REM sleep (%) | 235 | 19.16 | 6.41 | 102 | 19.94 | 6.22 | 1.032 | 0.303 |
| Minutes awake | 235 | 51.87 | 40.71 | 102 | 54.52 | 38.63 | 0.558 | 0.577 |
| Arousals per hour index | 235 | 11.23 | 7.05 | 102 | 11.07 | 5.95 | −0.202 | 0.840 |
- a REM, rapid eye movement; N, number of individuals; SD, standard deviation; Met carriers, Val/Met + Met/Met genotype carriers.
The final sample group for the sleep EEG spectral power consisted of 281 individuals for whom valid spectral data were available. After correction for multiple comparisons for each EEG bandwidth and sleep stage, Met carriers showed lower alpha 1 power in the C3-A2 derivation in stage 1 (2.9 ± 1.1 μV2/Hz) compared with Val/Val homozygotes (3.4 ± 1.8 μV2/Hz; Table 3; adjusted P = 0.046). In addition, Met carriers showed lower theta power during stage 2 in both C3-A2 (3.0 ± 0.9 μV2/Hz) and C4-A1 (2.7 ± 0.7 μV2/Hz) derivations compared with individuals with the Val/Val genotype (3.3 ± 1.0 μV2/Hz for C3-A2 and 3.0 ± 0.8 μV2/Hz for C4-A1; Table 3; adjusted P = 0.022). Theta power was also reduced during sleep-stage 3 in Met carriers (2.1 ± 0.6 μV2/Hz for C3-A2 and 1.9 ± 0.6 μV2/Hz for C4-A1) compared with Val/Val homozygotes (2.3 ± 0.8 μV2/Hz for C3-A2 and 2.1 ± 0.8 μV2/Hz for C4-A1; Table 3; adjusted P = 0.019, 0.031, and 0.014, respectively). A general linear model was applied to control for effects of possible factors that might influence the observed results. After adjustment for age, sex, BMI, and European ancestry, the association remained significant (data not shown).
| EEG spectral power | Val/Val carriers | Met carriers | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Derivation | Bandwidth | Sleep stage | N | Mean | SEM | N | Mean | SEM | t | P | Adjusted P |
| C3-A2 | Alpha 1 | 1 | 175 | 3.4 | 1.8 | 80 | 2.9 | 1.1 | −2.727 | 0.007 | 0.046 |
| C3-A2 | Theta | 2 | 190 | 3.3 | 1.0 | 91 | 3.0 | 0.9 | −2.597 | 0.010 | 0.022 |
| C4-A1 | Theta | 2 | 190 | 3.0 | 0.8 | 91 | 2.7 | 0.7 | −2.715 | 0.007 | 0.019 |
| C3-A2 | Theta | 3 | 189 | 2.3 | 0.8 | 91 | 2.1 | 0.6 | −2.483 | 0.014 | 0.031 |
| C4-A1 | Theta | 3 | 189 | 2.1 | 0.8 | 91 | 1.9 | 0.6 | −2.680 | 0.008 | 0.014 |
- a N, number of individuals; SEM, standard error of the mean.
DISCUSSION
This is the first study to evaluate the relationship between the Val66Met polymorphism of the BDNF gene and sleep physiology in a large population sample. We have shown that this polymorphism modulates EEG spectral power during sleep with no significant influence on sleep macrostructure. More specifically, Met carriers showed significantly decreased EEG power in the theta and alpha bands compared with Val/Val individuals in stages 1, 2, and 3 + 4 of non-REM sleep.
A growing body of behavioral and molecular genetics literature suggests that individual sleep profiles and sleep disorders are a result of complex interactions between genetic and environmental factors (Dauvilliers and Tafti, 2008). In fact, studies in monozygotic and dizygotic twins have provided consistent evidence that EEG patterns in non-REM and REM sleep are among the most heritable traits in humans (De Gennaro et al., 2008). Repeated recordings in the same individual and comparison among different subjects under highly controlled conditions suggest that EEG profiles show high interindividual variability and intraindividual stability (De Gennaro et al., 2005; Tucker et al., 2007). Although our knowledge is still incipient, studies evaluating genetic variants in candidate genes have contributed to substantial progress in our understanding of how genetic factors influence sleep physiology (Retey et al., 2005; Viola et al., 2007; Bodenmann et al., 2009; Bachmann et al., 2012; Mazzotti et al., 2012). Consistent with our findings, Bachmann et al. (2012) and Gatt et al. (2008) demonstrated that the Val66Met BDNF polymorphism induced frequency-specific alterations in EEG activity in wakefulness and sleep. Gatt et al. (2008) demonstrated reduced relative alpha power and elevated relative theta and delta power in Met/Met genotype carriers compared with Val/Met and Val/Val carriers across different brain regions, under both eyes-open and eyes-closed conditions. In contrast, Bachmann et al. (2012) reported that EEG power in the entire delta/theta range in non-REM sleep was consistently reduced and alpha power was increased in Val/Met heterozygotes compared with Val/Val genotype carriers. Met carriers also showed reduced alpha, theta, and sigma activity during REM sleep. The discrepancies between those studies and the present study may be explained by differences in experimental parameters, subjects, conditions, and analytical methods. For example, both previous studies used relative power rather than absolute power for analysis of EEG frequencies. Nevertheless, all three studies agree that the BDNF polymorphism influences functional brain oscillations.
BDNF is a neurotrophin that seems to play an important role in synaptic plasticity and connectivity regulation (Greenberg et al., 2009) as well as in sleep homeostasis (Cirelli and Tononi, 2000; Conti et al., 2007; Cirelli, 2009; Guindalini et al., 2009). The Met genetic variant of the BDNF gene has been shown to have functional effects and to modulate features that have also been associated with sleep and sleep deprivation, such as cognitive function, memory performance, and symptoms of depression (Hariri et al., 2003; Pezawas et al., 2004; van Wingen et al., 2010). In this sense, it is reasonable to suggest that differences in EEG activity may be associated with the mechanisms promoting the phenotypic variations in Met carriers. The reduction in alpha activity observed in Met carriers, which was also demonstrated by Gatt et al. (2008), may be of particular importance. The authors evaluated 305 healthy individuals and demonstrated that the reduced EEG alpha activity mediated the relationship between the Met/Met genotype and trait depression. Studies of patients with degenerative neurological disorders have shown that reductions of EEG fast frequencies during sleep occur earlier in the course of the disease than EEG changes during wakefulness (Petit et al., 2004).
A genetic influence on sleep duration and depth has recently been shown, with some individuals tending to sleep less with a higher density of slow-wave sleep (He et al., 2009). We did not find any differences in sleep stages among the genotype groups, but the significant changes in EEG spectral power suggest that the genetic influence on overall sleep structure is complex. We hypothesize that sleep might be characterized under many phenotypic levels, including the macro- and microstructure and the spectral proprieties of EEG, and at a more complex level in chemical and electrical changes between neurons and within synapses. One would expect that relatively common genetic variants might have less pronounced effects on phenotype at a broader level. However, when we look more closely at the molecular mechanism that is supposed to generate the phenotype, the variant effect tends to be higher. In this sense, although we did not find differences in a broader sleep phenotype (sleep macrostructure), we did find an effect on a more specific phenotype (sleep EEG spectral power), which supports previous findings that the BDNF gene polymorphism plays a role in the modulation of EEG activity. Our findings also extend previous findings to an ethnically diverse population with a large range of ages. If this allele is associated with impairments in cognitive performance and with anxiety and depression, future studies may address the possible relationship between such phenotypes and sleep EEG features.
It is worth pointing out, however, that this study has some limitations that might be fundamental and should be addressed in future studies. Unfortunately, we could not measure serum BDNF levels, which would enhance the results and provide further biological evidence for the described associations. Also, the relatively small sample size of Met homozygous did not allow us to evaluate its effect separately. Indeed, when performing pairwise comparisons among the three genotypes, we observed that the statistical signal came mainly from the Val/Val vs. Val/Met genotype group comparison. Therefore, because of the limited sample of Met/Met carriers, a potentially enhanced effect of this genotype on EEG spectral power might have been masked. In this sense, although the present study supports the dominant model effect of Met allele on the measured outcomes, further studies with larger samples, especially of Met/Met carriers, would help to characterize better the inheritance model of this polymorphism.
This study provides the first comprehensive evaluation of this important polymorphism in a large and ethnically diverse population. The findings call for further studies of this and other polymorphic variants in candidate genes, with the overarching goal of characterizing the molecular basis of sleep.
ACKNOWLEDGMENTS
All the efforts of the AFIP staff, in particular Roberta Siuffi and Diva Maria Lima, are deeply appreciated. The authors have no conflicts of interest.
