Skip to main content
NCBI home page
The Journal of Neuroscience logoLink to The Journal of Neuroscience
. 2025 Apr 11;45(21):e1138242025. doi: 10.1523/JNEUROSCI.1138-24.2025

Joint Heritability of Sleep EEG Spindle Activity and Thalamic Volume in Early Adolescence

Andjela Markovic 1,2,3,4,✉, Duco Veen 5,6, Christoph Hamann 7, Kristina Adorjan 4, Michael Kaess 1,8, Ruth Tuura O’Gorman 9,10,11, Leila Tarokh 1,4
PMCID: PMC12096044  PMID: 40216547

Abstract

Sleep spindles, transient bursts of rhythmic activity during non-rapid eye movement sleep, are generated by the thalamocortical network through an intricate interplay between the thalamus and the cortex. Emerging research has shed light on the role of sleep spindles in cognitive function, memory consolidation, and overall brain health. Using a behavioral genetics approach in female and male adolescent humans, this study examined the degree to which sleep spindles (measured via high-density sleep electroencephalography) and thalamic volume (measured via magnetic resonance imaging) are driven by common genetic and environmental factors. Here we show a strong correlation between thalamic volume and sleep spindle amplitude and density. Bayesian structural equation modeling estimated that over posterior regions, genetic factors accounted for approximately half of the covariance between sleep spindle activity and thalamic volume. Our findings demonstrate that genetic factors play a role in shaping the structural and functional integrity of the thalamocortical network, with implications for understanding how these processes contribute to neurodevelopmental outcomes.

Keywords: Bayesian inference, electroencephalography, MRI, thalamus, twin design

Significance Statement

Sleep spindles, oscillatory activity generated in the thalamus, are crucial for cognitive functions and brain health. This study investigated the joint genetic and environmental influences on sleep spindles and thalamic volume in adolescents. Our findings suggest a significant overlap in genetic factors influencing thalamic volume and spindle amplitude over posterior brain regions. Given that sleep spindle activity is altered in several brain disorders involving the thalamocortical system, this work not only enhances our understanding of the biological phenomena underlying the neuroanatomical substrates of the sleep electroencephalography but also offers crucial insights for developing targeted interventions in neurodevelopmental disorders.

Introduction

Sleep spindles are bursts of activity during non-rapid eye movement (NREM) sleep which can be recorded on the scalp using electroencephalography (EEG). Sleep spindles are generated in the thalamus and projected to the cortex via extensive thalamocortical circuits (Crunelli et al., 2018). As such, sleep spindles provide a unique in vivo measure of the integrity of the thalamocortical system that can be measured noninvasively in vulnerable populations (e.g., children and patient populations). Interest in the thalamocortical system has arisen because this network plays a pivotal role in various cognitive functions and is crucial for understanding the complex interplay between sensory information processing and higher-order brain functions (Hwang et al., 2017). While our knowledge about the circuitry involved in the generation and synchronization of sleep spindles largely originates from animal models, studies in humans have identified an association between thalamic activation and sleep spindle activity (Schabus et al., 2007).

Sleep spindle activity is altered in several disorders of the brain in which the thalamocortical system is implicated. For example, in schizophrenia, a link between reduced sleep spindle activity and thalamic volume/connectivity has been reported (Ferrarelli et al., 2010; Wamsley et al., 2012; Manoach et al., 2014), with reduced left mediodorsal thalamic volumes correlating with spindle density over frontal areas (Buchmann et al., 2014). Moreover, the degree of spindle deficits is tied to the severity of hallucinatory experiences (Markovic et al., 2020a), suggesting a role of the thalamocortical network in generating hallucinations. Similarly, sleep spindle activity has been proposed as a biomarker of depressive symptomatology (Baena et al., 2024). Investigating the relationship between spindle activity and the morphology of the thalamocortical network may thus provide crucial insights into the etiology of such brain-based disorders.

While emerging evidence suggests an association between sleep spindle activity and thalamic structure, it remains unclear if this is because the same genes code for both features. Disentangling genetic and environmental influences is important for understanding individual differences in sleep neurophysiology and brain morphology. A strong genetic contribution to sleep spindle activity has been found in adults (Adamczyk et al., 2015) as well as adolescents (Rusterholz et al., 2018), however, with large regional variability. On the other hand, thalamic volume is largely shaped by genetic factors during childhood and early adolescence (Swagerman et al., 2014) to become predominantly determined by the environment in adulthood (Strelnikov et al., 2022). A recent study identified shared genes for thalamic volume and cerebral cortical areas, along with shared genes for thalamic volume and 10 psychiatric and neurological disorders confirming the crucial role of the thalamus in cortical functioning and brain disorders (Elvsåshagen et al., 2021). Taken together, these findings underscore the complex interplay and developmental shifts in genetic and environmental influences on each of the two measures.

Here, we examine the joint heritability of sleep spindles and thalamic volume in healthy adolescent twins. By considering both genetic and environmental determinants of the covariance of these two measures, we enhance our comprehension of the shared pathways that may influence both spindle activity and thalamic structure. Mapping these shared influences provides a foundation for future research to identify at-risk individuals and to develop more targeted early interventions for disorders involving thalamocortical disruption.

Materials and Methods

Participants

Forty-three participants aged 10–14 years were included in the presented analyses. They were recruited for the purposes of a twin study investigating the heritability of the sleep EEG including 14 monozygotic (MZ; n = 28; mean age, 12.57; SD, 1.16; 12 females) and 8 dizygotic (DZ; n = 16; mean age, 13.25; SD, 0.71; 4 females) twin pairs (Inderkum and Tarokh, 2018; Markovic et al., 2018, 2020b, 2022; Rusterholz et al., 2018). A questionnaire (95% accurate) was administered to the parents to determine zygosity (Goldsmith, 1991). No difference with regard to the distribution of gender (χ2 = 0.7; p = 0.4) or age (t = −1.5; p = 0.15) was observed between the groups. A set of triplets that included a MZ and a DZ twin pair was included in both MZ and DZ analyses. Participants were born after 30 weeks of gestational age and healthy at the time of the assessments. All procedures of the study were performed according to the Declaration of Helsinki and approved by the local ethics committee. Consent was obtained from participants and their parents.

Sleep EEG

All participants underwent whole-night sleep EEG recordings during two consecutive nights (adaptation and baseline). Only data from the second night (baseline) were used for analyses. Sleep EEG recordings were performed at participants’ homes via a Geodesics system (Electrical Geodesic) with 64 channels (58 EEG, two electrooculogram, two electromyogram, and two electrocardiogram channels) and a sampling rate of 1,000 Hz (downsampled to 250 Hz for analysis). All participants were on a sleep schedule for at least 5 d prior to the sleep EEG recordings with 9.5–10 h of time in bed to ensure an adequate amount of sleep. Compliance to this schedule was confirmed through actigraphy and sleep diaries. Standard criteria (Rechtschaffen and Kales, 1968) were applied to score each 30 s epoch of sleep data. Power density spectra were calculated in 30 s epochs (5 s windows; Hanning window; no overlap). After visually inspecting spectrograms, channels of insufficient signal quality were excluded. The remaining signals were re-referenced to the average of all derivations with good signal quality (i.e., average reference). A semiautomated procedure was applied to detect and exclude epochs with artifacts based on the level of power in low (0.8–4.6 Hz) and high (20–40 Hz) frequencies, whenever this level exceeded a threshold (Buckelmüller et al., 2006). Sleep EEG power during NREM sleep in the sigma band (11–16 Hz) was examined. In addition to power, we characterized sleep spindle activity by means of an algorithm detecting individual spindle events in the frequency range between 10 and 16 Hz as described by Rusterholz et al. (2018). A spindle event was identified based on the envelope of the bandpass-filtered signal from the Hilbert transform whenever criteria regarding the duration and amplitude of the signal were fulfilled. These criteria included a minimum duration of 0.5 s, a minimum amplitude of 1.25 times the mean of the rectified filtered signal, and a maximum amplitude of five times the mean of the rectified filtered signal. We then calculated the mean sleep spindle amplitude, duration, and frequency as well as the number of spindle events per second (i.e., spindle density) for each recording. Additionally, we ran a series of sensitivity analyses for spindle amplitude by modulating the criteria for inclusion of detected spindle events. First, we separately analyzed slow (10–12 Hz) and fast (12–16 Hz) spindles, as distinct mechanisms of generation and function have been proposed for these two spindle classes in adults (Werth et al., 1997). Second, we separately analyzed spindles from the first half of the sleep episode and those from the second half of the sleep episode to account for potential overnight dynamics in sleep spindle activity. Third, we adjusted the lower amplitude threshold to twice the mean signal and the upper amplitude threshold to six as well as eight times the mean signal, to ensure that our findings are not dependent on the chosen amplitude thresholds. All thresholds we used have been employed in prior studies (Ferrarelli et al., 2007; Warby et al., 2014), providing a basis for comparison.

Structural MRI

Within one month after the sleep EEG assessment, all participants underwent structural magnetic resonance imaging (MRI) scanning on a 3 T GE Medical Systems MR750 scanner. A T1-weighted three–dimensional inversion recovery-prepared spoiled gradient recalled (IR-SPGR) sequence was obtained using an eight-channel receive–only head coil, with the following parameters: inversion time, 600 ms; echo time, 4.25 ms; repetition time, 11.4 ms; flip angle, 8°; and an isotropic voxel resolution of 1 mm. Thalamic volume was obtained from segmentation of the anatomical image using FreeSurfer (Version 5.3).

Statistical analysis

The association between thalamic volume and sleep EEG variables (i.e., NREM sigma power, sleep spindle amplitude, density, duration, and frequency) was examined by means of Spearman's correlation. We used a permutation test to compare the observed rho value to a distribution of rho values generated by shuffling the original data randomly in 1,000 iterations. In each iteration, we randomly shuffled the order of entries for one of the measures while keeping the other measure fixed. This ensured that the data entries no longer corresponded to the original pairings, effectively removing any meaningful correlation between the two measures. By running the correlation analysis between these shuffled datasets in each iteration, we generated a distribution of rho values. The actual rho value was considered statistically significant if it was larger than 95% of the corresponding distribution. The shuffling order was kept consistent across derivations in each iteration to control the error rate. Throughout the manuscript, we report p values derived from this permutation test. Only the sleep EEG parameters that were significantly associated with thalamic volume across multiple electrodes were further analyzed by means of Bayesian structural equation modeling (SEM). Bayesian methods allow for the integration of prior knowledge, thereby enhancing the robustness of model estimates, particularly when sample sizes are small (van de Schoot et al., 2021). The influence of three latent factors on the observed variables was estimated: genes (A), environmental impact shared among twins (C), and environmental impact unique to each twin and measurement error (E). A, C, and E range between 0 and 1 depicting the amount of the variance explained by each of the factors. Typically, twin studies apply SEM based on the assumption that MZ twins have the same genetic material (100% genetically concordant), while DZ twins have half of their genes in common (Rijsdijk and Sham, 2002). Both groups of twins were raised together and, thus, have the same shared environment (100% correlated) and uncorrelated unique environments and measurement error. Under these assumptions, we constructed models to estimate the contribution of A, C, and E to (1) only thalamic volume, (2) each sleep EEG variable separately, and (3) thalamic volume and each sleep EEG variable together in a joint model. The joint model extended the aforementioned assumptions to include two observed variables (MRI and EEG) and estimate the contribution of A, C, and E to their covariance (Fig. 1). For this purpose, we used the package blavaan (Merkle et al., 2021) in R 4.2.2 (R Core Team, 2022) within RStudio 2022.12.0 (Posit Software, PBC, Boston MA, USA). We used the default priors of blavaan which were seen as suitable after standardization of our data. These priors are weakly informative given the scale of the data, limiting their influence on the posterior but enhancing computation for the model with a limited sample size. All other computations including data visualization were conducted in MATLAB R2022b (MathWorks, Natick MA, USA).

Figure 1.

Figure 1.

Open in a new tab

Illustration of the structural equation model applied to estimate the joint contribution of genetic (A), shared environmental (C), and unique environmental (E) factors to EEG features (i.e., different measures of sleep spindle activity) and MRI data (i.e., thalamic volume). The relevant assumptions for correlations between genetic and environmental factors among MZ and DZ twins are shown. The remaining correlations were set to 0. A, C, and E estimates were forced to be equal across the observed variables. One such model was employed for each of the analyzed EEG features at each of the 58 derivations. In each of these models, the thalamic volume comprised the MRI measure.

Results

Sleep stage parameters were in the expected range for this age group and not significantly different between MZ and DZ twins (Table 1). In general, our participants were good sleepers with an average sleep efficiency of 92 ± 5% and an average sleep onset latency of 19 ± 10 min. Furthermore, the groups did not differ with regard to their cognitive abilities as reflected in their scores on the verbal learning and memory task, self-ordered pointing task (Cragg and Nation, 2007), digit symbol substitution task (Jaeger, 2018), and trail making test (Reitan, 1958; Table 2).

Table 1.

Sleep architecture

Sleep parameter MZ DZ z-statistic
Total sleep time (min) 514.54 (±67.26) 535.3 (±32.2) −0.31 (p = 0.76)
Wake after sleep onset (min) 25.63 (±26.33) 25.7 (±29.36) 0.13 (p = 0.89)
Sleep latency (min) 18.73 (±10.36) 18.4 (±9.85) −0.13 (p = 0.89)
Sleep efficiency (%) 91.9 (±4.73) 92.18 (±4.44) −0.13 (p = 0.89)
REM latency (min) 114.96 (±49.28) 90.92 (±37.99) 1.7 (p = 0.09)
Stage 2 (%) 45.21 (±10.44) 44.09 (±9.15) −0.02 (p = 0.98)
Slow wave sleep (%) 28.3 (±10.27) 27.48 (±8.5) −0.19 (p = 0.85)
Stage REM (%) 26.15 (±5.94) 27.88 (±6.68) −0.76 (p = 0.44)
Open in a new tab

The mean and standard deviation (in parentheses) of sleep parameters are shown in the second column for MZ (n = 28) and in the third column for DZ (n = 16) twins. The percentage values were calculated with respect to total sleep time. Sleep latency is defined as the first occurrence of Stage 2 sleep following lights out. The last column shows the results from a Wilcoxon rank sum test comparing the two groups with regard to each parameter (z values; p values in parentheses).

Table 2.

Cognitive ability

Cognitive parameter MZ DZ z-statistic
Verbal learning and memory 58 (±8.72) 59.8 (±6.67) −0.35 (p = 0.72)
Self-ordered pointing task 6.31 (±4.17) 5.8 (±3.75) 0.42 (p = 0.67)
Digit symbol substitution task 61.46 (±11.98) 63.6 (±11.37) −0.18 (p = 0.86)
Trail making test A 16.77 (±17.78) 14.5 (±3.81) −1.32 (p = 0.19)
Trail making test B 26 (±8.09) 25.95 (±9.06) 0.1 (p = 0.92)
Open in a new tab

The mean and standard deviation (in parentheses) of cognitive parameters are shown in the second column for MZ (n = 28) and in the third column for DZ (n = 16) twins. Values represent the sum of words recalled across the learning trials in the verbal learning and memory task (first row), the number of errors in the self-ordered pointing task (second row), the number of correct symbols in the digit symbol substitution task (third row), and performance in the trail making test measured in seconds (fourth and fifth row). The last column shows the results from a Wilcoxon rank sum test comparing the two groups with regard to each parameter (z values; p values in parentheses).

Association between thalamic volume and sleep spindle activity

We observed positive correlations (Fig. 2) between thalamic volume and NREM sigma power (0.33 ≤ rho ≤ 0.51; 0.002 ≤ p ≤ 0.033), spindle amplitude (0.29 ≤ rho ≤ 0.48; 0.001 ≤ p ≤ 0.048), and spindle density (0.33 ≤ rho ≤ 0.48; 0.001 ≤ p ≤ 0.031). In contrast, spindle duration and frequency did not show robust associations with thalamic volume. Therefore, only NREM sigma power, spindle amplitude, and spindle density were used for further analysis.

Figure 2.

Figure 2.

Open in a new tab

Topographic distribution of Spearman's correlation coefficients (rho) between thalamic volume and the sleep EEG variables (i.e., NREM sigma power, sleep spindle amplitude, density, duration and frequency). Statistically significant (corrected p < 0.05) derivations are depicted with white circles.

Heritability of thalamic volume and sleep spindle activity

As shown in Figure 3, thalamic volume was largely impacted by shared environmental factors (51%) with a weaker influence of genes (34%) and a negligible influence of unique environmental factors (14%). With regard to sleep spindle activity, we observed a genetic impact with a contribution >50% at 29 derivations for NREM sigma power (maximum of 74%) and at 40 derivations for spindle amplitude (maximum of 74%), primarily over posterior regions (Fig. 3; second and fourth column). In contrast, shared environmental factors explained >50% of the variance at 10 derivations over frontal regions for sigma power (maximum of 63%). No derivations with a shared environmental contribution >50% were found for spindle amplitude.

Figure 3.

Figure 3.

Open in a new tab

Topographic distribution of genetic (A), shared environmental (C), and unique environmental impact (E) on thalamic volume, different measures of sleep spindle activity (i.e., NREM sigma power, spindle amplitude, and spindle density) and both. The lower half shows the corresponding fit measures, i.e., the BCFI, the BTLI and the BRMSEA. Analogous results for spindle amplitude, using sensitivity analyses with variations in the spindle detection and selection process, are shown in Extended Data Figures 3-1–3-3.

The joint models resulted in similar topographic patterns as those observed for sleep EEG spindle parameters, albeit with regionally less widespread heritability (Fig. 3; third and fifth column). Specifically, genes accounted for >50% of the joint variance between thalamic volume and sigma power (26 derivations) and sleep spindle amplitude (32 derivations). On the other hand, environmental factors shared between the twins explained >50% of the joint variance of sigma power and thalamic volume (12 derivations). Similarly, spindle amplitude (6 derivations) and thalamic volume were jointly influenced by environmental factors shared among the twins.

The impact of unique environmental factors was overall negligible. The fit measures were acceptable with a Bayesian comparative fit index (BCFI) and a Bayesian Tucker–Lewis index (BTLI) close to 1 and a Bayesian root mean square error of approximation (BRMSEA) below 0.1 (MacCallum et al., 1996) for thalamic volume, sigma power, and spindle amplitude, with the exception of a confined frontal area (Fig. 3, second to fourth column). However, the models involving spindle density performed worse and did not reach the recommended levels for the three fit measures, given that only half of electrodes exhibited a BRMSEA below 0.1 (Fig. 3, sixth and seventh column). We thus refrain from further discussing spindle density. Exemplary posterior distributions of all parameter estimates are shown in Figure 4.

Figure 4.

Figure 4.

Open in a new tab

Posterior distributions of all parameter estimates (A, C, and E) for the three EEG measures (NREM sigma power, spindle amplitude, and spindle density) at a parietal derivation (P4). The posterior distribution, shown on the y-axis, represents the probability density for each parameter value across its range on the x-axis, derived from the iterative sampling of the Bayesian analysis. For every parameter estimate, three distributions are shown: (1) from the model including only EEG data, (2) from the model including only MRI data (thalamic volume), and (3) from the model including both EEG and MRI data. The width of the distribution reflects the uncertainty in the parameter estimates, with wider distributions indicating a broader range of plausible values given the data and prior information.

Our sensitivity analyses confirm these findings with a pattern of shared genetic contribution to thalamic volume and spindle amplitude, particularly over posterior regions, and a small impact of shared environmental factors over frontal areas, regardless of the differences applied in the spindle detection and selection process (Extended Data). Specifically, we found no difference in joint heritability when distinguishing between slow and fast spindles (Extended Data Fig. 3-1). Similarly, the results were highly comparable when separately analyzing the first and the second half of the sleep episode (Extended Data Fig. 3-2). Finally, differences were negligible when comparing different amplitude thresholds for spindle detection (Extended Data Fig. 3-3). Of note, the fit measures of all models calculated as part of sensitivity analyses were worse compared with our main analysis. It is possible that the variations in signal-to-noise ratio, influenced by both the number of detected spindles and the duration of the sleep episode considered, could affect the robustness of these findings. We thus conclude that the parameters chosen in our main analysis produce the most reliable results and limit our discussion to these.

Figure 3-1

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of slow (10 to 12  Hz; first column) as well as fast sleep spindles (12 to 16  Hz; second column). The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-1, TIF file (2.6MB, tif) .

Figure 3-2

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of sleep spindles detected in the first half of the sleep episode (first column) and the second half of the sleep episode (second column). The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-2, TIF file (2.3MB, tif) .

Figure 3-3

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of sleep spindles detected with different amplitude thresholds. The first column shows our results for a lower threshold of 2 (i.e., twice the mean sigma amplitude) and an upper threshold of 6 (i.e., six the mean sigma amplitude), while the second column shows our results for a lower threshold of 2 and an upper threshold of 8, in line with previously published procedures. The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-3, TIF file (2.8MB, tif) .

Discussion

This study explores the neuroanatomical substrates of high-density sleep EEG using structural MRI in the same subjects by estimating the contributions of genetic and environmental factors to the covariance between sleep spindle measures and thalamic volume. Our results suggest an overlap in genetic factors accounting for the variance in thalamic volume and spindle amplitude over posterior brain regions, while shared environmental factors were found to account for the variance between the two measures over the frontal regions.

We found a correlation between thalamic volume and certain aspects of sleep spindle activity, specifically spindle amplitude and density, along with NREM sigma power serving as a broader indicator. Similar to the observations of Saletin et al. (2013), we found no association of thalamic volume with spindle frequency or duration suggesting that variations in thalamic volume have a selective influence and may not impact these characteristics of spindle activity.

To shed light on the nature of the association between these measures, we disentangled genetic from environmental influences. In the context of previous work (Swagerman et al., 2014; Strelnikov et al., 2022), the observed moderate heritability of thalamic volume confirms a shift toward environmental influence during adolescence. Spindle activity, on the other hand, when only considering the regions exhibiting a sufficient model fit, is largely driven by the genes, in line with our previous report (Rusterholz et al., 2018).

Examining the two measures together, we find a regional pattern in the overlap between factors contributing to the covariance between spindle activity and thalamic volume with shared genetic influences over posterior regions and shared environmental influences over frontal regions. The substantial genetic influence over posterior region suggests that there may be a common set of genetic factors that contribute to both sleep spindle activity and thalamic volume in this specific area. Interestingly, in healthy adults, the thalamus was identified as the only subcortical structure that correlated with intelligence, and this correlation was influenced by a common genetic factor between the two (Bohlken et al., 2013). The authors propose that thalamocortical connections, which develop early during fetal stages (Taymourtash et al., 2022), influence the maturation of cortical areas and thereby intellectual ability. As the thalamocortical network is responsible for the generation of sleep spindles, the same genetic factors may underlie the already established link between sleep spindle activity and intelligence (Fogel and Smith, 2011).

One potential genetic mechanism linking spindle activity with thalamic volume is the schizophrenia risk gene CACNA1I. Knocking out this gene in mice causes spindle deficits (Astori et al., 2011), suggesting it may be a major contributor to diminished spindle activity observed in schizophrenia (Manoach et al., 2016). Given that CACNA1I has also been linked to thalamic dysfunction in schizophrenia (reviewed in Jiang et al., 2021), it may indeed be one common genetic denominator of sleep spindle activity and thalamic volume.

In contrast, the presence of shared environmental factors explaining the covariance between thalamic volume and sigma power over the frontal regions observed in our study indicates that environmental factors that affect both measures may be more prominent in this area. This finding emphasizes the potential for preventative measures against conditions involving thalamic dysfunction by targeting the relevant environmental factors that have yet to be identified. Alternatively, the poor fit measures over the frontal regions suggest that the corresponding estimates may not be reliable. Our findings may thus also imply that distinct factors drive thalamic volume and frontal sigma power. Therefore, we cannot conclude that there is an overlap in shared environmental factors accounting for the covariance between spindle activity and thalamic volume.

The modest sample size represents a limitation in our study. Nonetheless, we have implemented Bayesian inference in our statistical modeling to mitigate the potential for drawing inaccurate conclusions from the small sample. This approach provides not only point estimates but also confidence intervals, offering a deeper understanding of uncertainty in our findings. Consequently, we restricted our discussion to findings that meet these criteria. Of note, since the models are forced to provide estimates, the resulting genetic and environmental contributions should always be interpreted in relative terms as a means of comparison among the factors. Exemplary of this limitation are the poor model fits with regard to the joint heritability of thalamic volume and sleep spindle density, a metric often associated with neurodevelopmental and psychiatric disorders (Markovic et al., 2020a; Herrera and Tarokh, 2024). Nonetheless, the Bayesian framework allows for the posterior distributions to accurately reflect the underlying uncertainty given our modest sample size, as exemplified in Figure 4. By leveraging the strength of Bayesian inference, we maximize the informational value of our data, providing insights that are as robust as possible within the constraints of our study's scale.

In summary, our findings highlight the complex interplay between genes and environment in shaping the relationship between sleep spindle activity and thalamic volume. Once more, the observed regional variability highlights the importance of taking brain topography into account when studying sleep neurophysiology (Markovic et al., 2018). Furthermore, model fit measures reveal crucial information that must be considered before drawing any conclusions. Finally, our study demonstrates that while thalamic volume affects EEG-measured sleep spindle activity, the sleep EEG provides additional insights into brain function beyond structural characteristics. The corresponding genetic and environmental determinants may foster our understanding of biological phenomena underlying the neuroanatomical substrates of the sleep EEG, as our findings highlight shared influences on sleep spindles and thalamic structure. These results also facilitate translational research efforts by laying the groundwork for exploring how these shared pathways manifest in clinical populations, bridging the gap between basic and clinical research. This knowledge extends beyond recognizing alterations in spindle activity or thalamic volume individually, as it helps distinguish whether these deficits arise from common mechanisms or independent pathways, providing critical insights into the etiology and potential targeted treatment of neurodevelopmental disorders. Importantly, not all individuals with a particular neurodevelopmental disorder exhibit deficits in the thalamocortical system, which suggests that treatments aimed at modulating thalamocortical function may not be universally applicable. However, our findings suggest that sleep spindle activity can serve as a proxy for genetic vulnerabilities in the thalamus, which is often involved in neurodevelopmental disorders (Herrera and Tarokh, 2024). This approach offers a noninvasive alternative to MRI measurements, which are challenging in young children with neurodevelopmental disorders due to difficulties associated with staying stationary, heightened anxiety in enclosed environments, and sensory sensitivities to the loud noises of scanners. In contrast, sleep EEG recordings are more feasible, even in the most sensitive and vulnerable children. This accessibility allows for early and precise identification of those with diminished spindle activity, indicative of potential thalamocortical dysfunction, paving the way for personalized interventions that could ultimately improve outcomes.

Data Availability

The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request. The code necessary for conducting the presented analysis is available at https://github.com/Andjela-M/joint-heritability.

References

  1. Adamczyk M, Genzel L, Dresler M, Steiger A, Friess E (2015) Automatic sleep spindle detection and genetic influence estimation using continuous wavelet transform. Front Hum Neurosci 9:624. 10.3389/fnhum.2015.00624 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Astori S, Wimmer RD, Prosser HM, Corti C, Corsi M, Liaudet N, Volterra A, Franken P, Adelman JP, Lüthi A (2011) The Ca(V)3.3 calcium channel is the major sleep spindle pacemaker in thalamus. Proc Natl Acad Sci U S A 108:13823–13828. 10.1073/pnas.1105115108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baena D, et al. (2024) Sleep spindles in adolescents with major depressive disorder. J Affect Disord 344:535–545. 10.1016/j.jad.2023.10.039 [DOI] [PubMed] [Google Scholar]
  4. Bohlken MM, Brouwer RM, Mandl RCW, van Haren NEM, Brans RGH, van Baal GCM, de Geus EJC, Boomsma DI, Kahn RS, Hulshoff Pol HE (2013) Genes contributing to subcortical volumes and intellectual ability implicate the thalamus. Hum Brain Mapp 35:2632–2642. 10.1002/hbm.22356 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Buchmann A, Dentico D, Peterson MJ, Riedner BA, Sarasso S, Massimini M, Tononi G, Ferrarelli F (2014) Reduced mediodorsal thalamic volume and prefrontal cortical spindle activity in schizophrenia. Neuroimage 102:540–547. 10.1016/j.neuroimage.2014.08.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Buckelmüller J, Landolt H-P, Stassen HH, Achermann P (2006) Trait-like individual differences in the human sleep electroencephalogram. Neuroscience 138:351–356. 10.1016/j.neuroscience.2005.11.005 [DOI] [PubMed] [Google Scholar]
  7. Cragg L, Nation K (2007) Self-ordered pointing as a test of working memory in typically developing children. Memory 15:526–535. 10.1080/09658210701390750 [DOI] [PubMed] [Google Scholar]
  8. Crunelli V, Lőrincz ML, Connelly WM, David F, Hughes SW, Lambert RC, Leresche N, Errington AC (2018) Dual function of thalamic low-vigilance state oscillations: rhythm-regulation and plasticity. Nat Rev Neurosci 19:107–118. 10.1038/nrn.2017.151 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Elvsåshagen T, Shadrin A, Frei O, van der Meer D, Bahrami S, Kumar VJ, Smeland O, Westlye LT, Andreassen OA, Kaufmann T (2021) The genetic architecture of the human thalamus and its overlap with ten common brain disorders. Nat Commun 12:2909. 10.1038/s41467-021-23175-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Ferrarelli F, Huber R, Peterson MJ, Massimini M, Murphy M, Riedner BA, Watson A, Bria P, Tononi G (2007) Reduced sleep spindle activity in schizophrenia patients. Am J Psychiatry 164:483–492. 10.1176/ajp.2007.164.3.483 [DOI] [PubMed] [Google Scholar]
  11. Ferrarelli F, Peterson MJ, Sarasso S, Riedner BA, Murphy MJ, Benca RM, Bria P, Kalin NH, Tononi G (2010) Thalamic dysfunction in schizophrenia suggested by whole-night deficits in slow and fast spindles. Am J Psychiatry 167:1339–1348. 10.1176/appi.ajp.2010.09121731 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fogel SM, Smith CT (2011) The function of the sleep spindle: a physiological index of intelligence and a mechanism for sleep-dependent memory consolidation. Neurosci Biobehav Rev 35:1154–1165. 10.1016/j.neubiorev.2010.12.003 [DOI] [PubMed] [Google Scholar]
  13. Goldsmith HH (1991) A zygosity questionnaire for young twins: a research note. Behav Genet 21:257–269. 10.1007/BF01065819 [DOI] [PubMed] [Google Scholar]
  14. Herrera CG, Tarokh L (2024) A thalamocortical perspective on sleep spindle alterations in neurodevelopmental disorders. Curr Sports Med Rep 10:103–118. 10.1007/s40675-024-00284-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hwang K, Bertolero MA, Liu WB, D’Esposito M (2017) The human thalamus is an integrative hub for functional brain networks. J Neurosci 37:5594–5607. 10.1523/JNEUROSCI.0067-17.2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Inderkum AP, Tarokh L (2018) High heritability of adolescent sleep-wake behavior on free, but not school days: a long-term twin study. Sleep 41:zsy004. 10.1093/sleep/zsy004 [DOI] [PubMed] [Google Scholar]
  17. Jaeger J (2018) Digit symbol substitution test. J Clin Psychopharmacol 38:513–519. 10.1097/JCP.0000000000000941 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Jiang Y, Patton MH, Zakharenko SS (2021) A case for thalamic mechanisms of schizophrenia: perspective from modeling 22q11.2 deletion syndrome. Front Neural Circuits 15:769969. 10.3389/fncir.2021.769969 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. MacCallum RC, Browne MW, Sugawara HM (1996) Power analysis and determination of sample size for covariance structure modeling. Psychol Methods 1:130–149. 10.1037/1082-989X.1.2.130 [DOI] [Google Scholar]
  20. Manoach DS, Demanuele C, Wamsley EJ, Vangel M, Montrose DM, Miewald J, Kupfer D, Buysse D, Stickgold R, Keshavan MS (2014) Sleep spindle deficits in antipsychotic-naïve early course schizophrenia and in non-psychotic first-degree relatives. Front Hum Neurosci 8:762. 10.3389/fnhum.2014.00762 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Manoach DS, Pan JQ, Purcell SM, Stickgold R (2016) Reduced sleep spindles in schizophrenia: a treatable endophenotype that links risk genes to impaired cognition? Biol Psychiatry 80:599–608. 10.1016/j.biopsych.2015.10.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Markovic A, Achermann P, Rusterholz T, Tarokh L (2018) Heritability of sleep EEG topography in adolescence: results from a longitudinal twin study. Sci Rep 8:7334. 10.1038/s41598-018-25590-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Markovic A, Buckley A, Driver DI, Dillard-Broadnax D, Gochman PA, Hoedlmoser K, Rapoport JL, Tarokh L (2020a) Sleep spindle activity in childhood onset schizophrenia: diminished and associated with clinical symptoms. Schizophr Res 223:327–336. 10.1016/j.schres.2020.08.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Markovic A, Kaess M, Tarokh L (2020b) Environmental factors shape sleep EEG connectivity during early adolescence. Cereb Cortex 30:5780–5791. 10.1093/cercor/bhaa151 [DOI] [PubMed] [Google Scholar]
  25. Markovic A, Kaess M, Tarokh L (2022) Heritability of REM sleep neurophysiology in adolescence. Transl Psychiatry 12:399. 10.1038/s41398-022-02106-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Merkle EC, Fitzsimmons E, Uanhoro J, Goodrich B (2021) Efficient Bayesian structural equation modeling in Stan. J Stat Softw 100:1–22. 10.18637/jss.v100.i06 [DOI] [Google Scholar]
  27. R Core Team (2022) R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: https://www.R-project.org/
  28. Rechtschaffen A, Kales A (1968) A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Los Angeles: UCLA Brain Information Service/Brain Research Institute. [Google Scholar]
  29. Reitan RM (1958) Validity of the trail making test as an indicator of organic brain damage. Percept Mot Skills 8:271–276. 10.2466/pms.1958.8.3.271 [DOI] [Google Scholar]
  30. Rijsdijk FV, Sham PC (2002) Analytic approaches to twin data using structural equation models. Brief Bioinform 3:119–133. 10.1093/bib/3.2.119 [DOI] [PubMed] [Google Scholar]
  31. Rusterholz T, Hamann C, Markovic A, Schmidt SJ, Achermann P, Tarokh L (2018) Nature and nurture: brain region specific inheritance of sleep neurophysiology in adolescence. J Neurosci 38:9275–9285. 10.1523/JNEUROSCI.0945-18.2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Saletin JM, van der Helm E, Walker MP (2013) Structural brain correlates of human sleep oscillations. Neuroimage 83:658–668. 10.1016/j.neuroimage.2013.06.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Schabus M, et al. (2007) Hemodynamic cerebral correlates of sleep spindles during human non-rapid eye movement sleep. Proc Natl Acad Sci U S A 104:13164–13169. 10.1073/pnas.0703084104 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Strelnikov D, et al. (2022) Heritability of subcortical grey matter structures. Medicina 58:1687. 10.3390/medicina58111687 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Swagerman SC, Brouwer RM, de Geus EJC, Hulshoff Pol HE, Boomsma DI (2014) Development and heritability of subcortical brain volumes at ages 9 and 12. Genes Brain Behav 13:733–742. 10.1111/gbb.12182 [DOI] [PubMed] [Google Scholar]
  36. Taymourtash A, et al. (2022) Fetal development of functional thalamocortical and cortico–cortical connectivity. Cereb Cortex 33:5613–5624. 10.1093/cercor/bhac446 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. van de Schoot R, et al. (2021) Bayesian statistics and modelling. Nat Rev Methods Primers 1:1–26. 10.1038/s43586-020-00001-2 [DOI] [Google Scholar]
  38. Wamsley EJ, Tucker MA, Shinn AK, Ono KE, McKinley SK, Ely AV, Goff DC, Stickgold R, Manoach DS (2012) Reduced sleep spindles and spindle coherence in schizophrenia: mechanisms of impaired memory consolidation? Biol Psychiatry 71:154–161. 10.1016/j.biopsych.2011.08.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Warby SC, Wendt SL, Welinder P, Munk EGS, Carrillo O, Sorensen HBD, Jennum P, Peppard PE, Perona P, Mignot E (2014) Sleep-spindle detection: crowdsourcing and evaluating performance of experts, non-experts and automated methods. Nat Methods 11:385–392. 10.1038/nmeth.2855 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Werth E, Achermann P, Dijk DJ, Borbély AA (1997) Spindle frequency activity in the sleep EEG: individual differences and topographic distribution. Electroencephalogr Clin Neurophysiol 103:535–542. 10.1016/S0013-4694(97)00070-9 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figure 3-1

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of slow (10 to 12  Hz; first column) as well as fast sleep spindles (12 to 16  Hz; second column). The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-1, TIF file (2.6MB, tif) .

Figure 3-2

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of sleep spindles detected in the first half of the sleep episode (first column) and the second half of the sleep episode (second column). The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-2, TIF file (2.3MB, tif) .

Figure 3-3

Topographic distribution of genetic (A), shared environmental (C) and unique environmental impact (E) on thalamic volume and amplitude of sleep spindles detected with different amplitude thresholds. The first column shows our results for a lower threshold of 2 (i.e., twice the mean sigma amplitude) and an upper threshold of 6 (i.e., six the mean sigma amplitude), while the second column shows our results for a lower threshold of 2 and an upper threshold of 8, in line with previously published procedures. The lower half shows the corresponding fit measures, i.e., the Bayesian comparative fit index (BCFI), the Bayesian Tucker-Lewis index (BTLI) and the Bayesian root mean square error of approximation (BRMSEA). Download Figure 3-3, TIF file (2.8MB, tif) .

Data Availability Statement

The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request. The code necessary for conducting the presented analysis is available at https://github.com/Andjela-M/joint-heritability.

Articles from The Journal of Neuroscience are provided here courtesy of Society for Neuroscience

RESOURCES