Circadian rhythms are physiological processes accompanied by behavioral changes that follow a ˜24-hour period in most living organisms. Humans are normally diurnal creatures, and biological clocks endogenously control their activity–rest patterns. Notably, these cycles have profound effects on fundamental molecular and behavioral processes (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019).
The expression of circadian preference, often referred to as chronotype, is the natural phenomena accounting for the variation in an individual’s preference for early or late sleep and activity driven by sleep–wake cycles due to internal circadian rhythms (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). Furthermore, chronotype defines whether an individual is a morning person, implying a preference for going to bed and waking up early, or an evening person, meaning someone who prefers later sleep and waking up times. A considerable proportion of chronotype variance is due to factors such as age, gender and the exposure to different levels of light. However, genetic variation also accounts for chronotype differences, and twin studies have estimated its heritability at ∼50% (Allebrandt et al., Reference Allebrandt, Teder-Laving, Kantermann, Peters, Campbell, Rudan and Roenneberg2014; Duffy & Czeisler, Reference Duffy and Czeisler2002; Kalmbach et al., Reference Kalmbach, Schneider, Cheung, Bertrand, Kariharan, Pack and Gehrman2017; Leocadio-Miguel et al., Reference Leocadio-Miguel, Louzada, Duarte, Areas, Alam, Freire and Pedrazzoli2017).
Adequate sleep is vital for healthy brain function and physiological systems (Medic et al., Reference Medic, Wille and Hemels2017). The restoration theory states that sleep is essential to replete and restore the cellular components necessary for biological functions (Brinkman & Sharma, Reference Brinkman and Sharma2020). In contrast, the brain plasticity theory describes the vital neural reorganization and development of the brain’s diverse structures and functions during the restorative state (Brinkman & Sharma, Reference Brinkman and Sharma2020). Recently, circadian parameters such as chronotype, average mid-sleep and social jetlag have been described as predictors of sleep quality (Harfmann et al., Reference Harfmann, Swalve, Mitrzyk and Montoye2020).
Brain white matter constitutes half of the brain, being a fundamental element of neural networks responsible for neurobehavioral activity, including sleep duration and REM sleep (Altendahl et al., Reference Altendahl, Cotter, Staffaroni, Wolf, Mumford, Cobigo and Walsh2020; Filley & Fields, Reference Filley and Fields2016; Yaffe et al., Reference Yaffe, Nasrallah, Hoang, Lauderdale, Knutson, Carnethon and Sidney2016). Diffusion-weighted imaging is a noninvasive neuroimaging technique that can be used to characterize the brain’s white matter integrity in acute sleep deprivation, insomnia and chronotype (Lu et al., Reference Lu, Dai, Couto, Liu, Chen, Lu and Yuan2017; Wei et al., Reference Wei, Bresser, Wassing, Stoffers, Van Someren and Foster-Dingley2019). In particular, diffusion tensor imaging (DTI) refers to the implementation of a tensor model to explore the main directions (eigenvectors of the model) and magnitude (eigenvalues) of water diffusivity (Basser et al., Reference Basser, Mattiello and LeBihan1994), which are sensitive to the underlying microstructure (Beaulieu, Reference Beaulieu2002). Water diffusivity parameters, measured through DTI, include axial diffusivity (AD), radial diffusivity (RD), mean diffusivity (MD), fractional anisotropy (FA) and mode of anisotropy (MO).
DTI, being a useful tool to explore brain regions with single fiber orientation, is used to infer histological changes in normal development and pathology. For instance, AD is associated with axonal integrity, while RD is related to axonal density, myelin integrity, axonal diameter and fiber coherence (Beaulieu, Reference Beaulieu2002).
Genome-wide association studies (GWAS) of neuroimaging measures have uncovered 213 independent significant genetic variants associated with 90 DTI parameters (Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019) and 351 loci with chronotype (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). Also, neuroimaging studies have identified white matter differences in evening people (Rosenberg et al., Reference Rosenberg, Maximov, Reske, Grinberg and Shah2014). Studies investigating the disruption of circadian cycles show that a day of wakefulness is associated with increases in white matter FA due to RD reductions (Elvsåshagen et al., Reference Elvsåshagen, Norbom, Pedersen, Quraishi, Bjørnerud, Malt and Westlye2015; Voldsbekk et al., Reference Voldsbekk, Maximov, Zak, Roelfs, Geier, Due-Tønnessen and Groote2020), thus suggesting that wakefulness-related effects may regulate white matter (Voldsbekk et al., Reference Voldsbekk, Maximov, Zak, Roelfs, Geier, Due-Tønnessen and Groote2020).
It is unclear whether the association between white matter and sleep is due to genetic factors, and if so, whether pleiotropic or causal effects drive the association. In the present study, we address this question by leveraging GWAS summary statistics data sets to estimate whole-genome genetic correlations using LD score regression. Also, we assess causality between DTI measures for white matter and circadian preference using Mendelian randomization analysis. Our results provide new insights into the relationship between chronotype and white matter structure and contribute to our understanding of the genetic and neurological factors underlying chronotype.
Materials and Methods
Discovery GWAS for Circadian Preference
This study used summary statistics from GWAS for self-reported chronotype (N = 449,734). Summary statistics include allele frequency, effect size, standard error, as well as the p value of every genetic variant that was tested on the trait of interest. These GWAS summary statistics were obtained from samples with European ancestry from the repository of its corresponding publication (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). Chronotype was assessed as a single self-reported measure (Field-ID: 1180): ‘Do you consider yourself to be?’ with six possible answers including ‘Definitely a morning person’, ‘More a morning than evening person’, ‘More an evening than a morning person’, ‘Definitely an evening person’, ‘Do not know’ or ‘Prefer not to answer’ (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). Here, chronotype was assessed as a dichotomous variable with two levels, including being a morning or an evening person. Chronotype GWAS summary statistics were adjusted for age, sex, study center and genotyping array (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). A full description of GWAS summary statistics used in this study is available in its corresponding publication (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019).
Discovery GWAS for DTI Brain Measures
The present study used summary statistics from GWAS with European ancestry for white matter microstructure, particularly DTI, objectively measured with magnetic resonance imaging (MRI) and has units of mm2 s−1. The GWAS summary statistics used here were obtained from samples with European ancestry (N = 17,706) from a public repository as reported in its corresponding publication (Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019). For instance, UK Biobank imaging data were processed according to the automated processing and quality control pipeline described in previous studies (Alfaro-Almagro et al., Reference Alfaro-Almagro, Jenkinson, Bangerter, Andersson, Griffanti, Douaud and Smith2018; Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019), and white matter tracts were labeled following the ENIGMA-DTI pipeline (Jahanshad et al., Reference Jahanshad, Kochunov, Sprooten, Mandl, Nichols, Almasy and Glahn2013; Kochunov et al., Reference Kochunov, Jahanshad, Sprooten, Nichols, Mandl, Almasy and Glahn2014; Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019). White matter microstructure parameters included mean diffusivities (MD), axial diffusivities (AD), radial diffusivities (RD) mode of anisotropies (MO) and FA. These five parameters were measured in 21 brain regions and all regions combined (Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019). GWAS were adjusted for age, age-squared, sex, age*sex interaction, age-squared*sex interaction, and the top 10 genetic principal components (Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019). A full description of the DTI preprocessing and the DTI analysis and imaging quality control procedures is available in the supplementary information of its corresponding publication (Zhao et al., Reference Zhao, Zhang, Ibrahim, Luo, Santelli, Li and Zhu2019).
Genetic Correlations Analyses
We performed LD score regression (Bulik-Sullivan, Loh et al., Reference Bulik-Sullivan, Loh, Finucane, Ripke, Yang and Neale2015) using GWAS summary statistics as implemented in the Complex Trait Genomics Virtual Lab (CTG-VL, http://genoma.io; Cuéllar-Partida et al., Reference Cuéllar-Partida, Lundberg, Kho, D’Urso, Gutiérrez-Mondragón, Ngo and Hwang2019) to estimate genetic correlations between chronotype and white matter microstructures. Briefly, CTG-VL uses the analysis software PLINK to clump and estimate linkage-disequilibrium between SNPs (Cuéllar-Partida et al., Reference Cuéllar-Partida, Lundberg, Kho, D’Urso, Gutiérrez-Mondragón, Ngo and Hwang2019). Only samples with European ancestry were used to avoid potential biases regarding population differences in linkage-disequilibrium and allele frequencies. Benjamini-Hochberg’s false discovery rate (FDR < 5%) was used to account for multiple testing.
Mendelian Randomization
Mendelian randomization (MR) is a method used in epidemiology to assess causal relationships between phenotypes, environmental exposures or disease outcomes (Davies et al., Reference Davies, Holmes and Davey Smith2018; Swerdlow et al., Reference Swerdlow, Kuchenbaecker, Shah, Sofat, Holmes, White and Hingorani2016; Zhu et al., Reference Zhu, Zheng, Zhang, Wu, Trzaskowski, Maier and Yang2018). We used two-sample generalized summary-data-based Mendelian randomization (GSMR; Zhu et al., Reference Zhu, Zheng, Zhang, Wu, Trzaskowski, Maier and Yang2018), a method that uses genetic instruments strongly associated (i.e., genome-wide significant SNPs) with the outcome, to perform bidirectional analyses and assess possible causal relationships between chronotype and 29 DTI parameters that displayed a significant genetic correlation. P values were corrected for multiple testing using Benjamini–Hochberg’s FDR (FDR < 5%). Both corrected and uncorrected p values are reported.
Results
Genetic Correlations
From the 110 DTI parameters that were analyzed, we identified 29 significant negative genetic correlations (FDR < 5%) between chronotype and DTI parameters (Table 1). Negative correlations with chronotype imply an increase in eveningness, that is, a preference for late sleep patterns and activity. The average MD, AD and RD for all tracts showed significant negative genetic correlations with chronotype. When exploring the individual tracts (Figure 1), 10 out of 26 genetic correlations involved MD, 10 included AD, four implicated RD and two involved MO. No significant genetic correlations were identified with FA.
Table 1. White matter microstructure measures genetically correlated with chronotype
Note: This table shows all the diffusion tensor imaging (DTI) measures with a significant false discovery rate (FDR < 5%) genetic correlation (r G) with chronotype. A negative genetic correlation (r G < 0) indicates an association with eveningness, whereas a positive genetic correlation (r G > 0) reflects an association with morningness. No significant positive correlations were identified. Results for all genetic correlations are reported in Supplementary File 1.
Fig. 1. Significant genetic correlations between chronotype and DTI parameters classified by individual tracts including mean diffusivity (MD), mode of anisotropy (MO), axial diffusivity (AD) and radial diffusivity (RD). All genetic correlations are negative. Negative correlations with chronotype imply a tendency to be an evening person.
For MD, the strongest genetic correlations with chronotype were observed for the internal capsule, particularly in its posterior limb (r G = −.21, p value = 4.76 × 10–04) and the retrolenticular region (r G = −.20, p value = 3.00 × 10–03). Further, the most significant genetic correlation was identified in the splenium of the corpus callosum (r G = −.19, p value = 4.31 × 10–06), but the genu (r G = −.13, p value = 8.92 × 10–04) and body (r G = −.14, p value = 1.05 × 10–03) of the corpus callosum were also genetically correlated with being an evening person. In addition, the posterior (r G = −.13, p value = 4.11 × 10–03), superior (r G = −.12, p value = 7.20 × 10–03) and anterior (r G = −.10, p value = 1.31 × 10–02) regions of the corona radiata were also genetically correlated with chronotype. A similar pattern was observed for the sagittal stratum and the superior fronto-occipital fasciculus.
Regarding AD, robust genetic correlations were identified in the corona radiata, with the most significant one in the superior region (r G = −.21, p value = 6.57 × 10–05). The anterior region of the corona radiata was also genetically correlated with chronotype (r G = −.13, p value = 6.32 × 10–03). Moreover, genetic correlates were observed for the retrolenticular regions (r G = −.22, p value = 6.42 × 10–04) and the posterior limb of the internal capsule (r G = −.19, p value =3.82 × 10–04), followed by the splenium (r G = −.16, p value = 9.43 × 10–04) and genu (r G = −.13, p value = 7.36 × 10–03) of the corpus callosum. More genetic correlations were uncovered for the superior fronto-occipital fasciculus, the superior longitudinal fasciculus, the sagittal stratum and the corticospinal tract.
Genetic correlations identified between chronotype and RD included the splenium (r G = −.13, p value = 6.56 × 10–03) and body (r G = −.13, p value = 8.62 × 10–03) of the corpus callosum, the posterior region of the corona radiata (r G = −.12, p value = 8.62 × 10–03) and the sagittal stratum (r G = −.15, p value = 6.61 × 10–03). In contrast, MO displayed moderate genetic correlates with chronotype in the superior region of the corona radiata (r G = −.18, p value = 9.48 × 10–04) and the retrolenticular section of the internal capsule (r G = −.15, p value = 7.11 × 10–03).
Mendelian Randomization
To further explore the associations between chronotype and DTI, we performed a two-sample GSMR (Zhu et al., Reference Zhu, Zheng, Zhang, Wu, Trzaskowski, Maier and Yang2018) and assessed the potential of the 29 genetic correlations we identified to be explained by a causal association (Supplementary File 2).
We observed 13 one-way causal relationships in which an evening chronotype influences increases in mean, axial and radial diffusivities (Table 2). GSMR results suggest that eveningness exerts a causal effect on higher MD for the superior fronto-occipital fasciculus (p value = .005) and the reticular region and posterior limb (p value = .006) of the internal capsule. Further, increases in mean and axial diffusivities for the posterior corona radiata and the splenium of the corpus callosum were found to be causally influenced by an evening chronotype. A similar pattern was observed for mean and radial diffusivities for the superior and anterior regions of the corona radiata. Also, a one-way causal relationship was identified for eveningness increasing AD for the superior longitudinal fasciculus (p value = .027).
Table 2. Mendelian randomization results for chronotype and DTI
Note: DTI, diffusion tensor imaging. One-way causal relationships between chronotype (exposure) and white matter microstructure parameters (outcome) are shown here. Negative effect estimates with chronotype imply a tendency to be an evening person. This means that eveningness is causally associated with increases in radial, axial and mean diffusivities. Corrected p values (false discovery rate < 5%) are referred to as pval corrected. Results for all 29 Mendelian randomization tests are reported in Supplementary File 2.
Discussion
This study provides insights into the relation between chronotype and white matter structure. We identified 29 pairwise significant genetic correlations between chronotype and DTI parameters, including MD, AD, RD and MO. Of these, 13 were found to be indicative of a causal relationship. These results show that changes in white matter microstructural organization associated with an evening chronotype are explained by variation in MD, AD, RD and MO tracts.
Recently, there has been a growing interest in the genetic architecture of chronotype and its role in health and disease (García-Marín et al., Reference García-Marín, Campos, Martin, Cuellar-Partida and Renteria2021; Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). For instance, being a morning person has been proven to be genetically correlated with metabolic traits such as body mass index (BMI), type 2 diabetes and fasting insulin (Jones et al., Reference Jones, Lane, Wood, van Hees, Tyrrell, Beaumont and Weedon2019). Further, phenotypes involving depression, such as ever had prolonged feelings of sadness or depression and ever depressed for a whole week (García-Marín et al., Reference García-Marín, Campos, Martin, Cuellar-Partida and Renteria2021), have shown negative genetic correlations with morningness, suggesting that individuals leaning toward an evening chronotype share genetic variants with depression phenotypes. In contrast, diseases of the musculoskeletal system and connective tissue have shown a positive genetic correlation with being a morning person (García-Marín et al., Reference García-Marín, Campos, Martin, Cuellar-Partida and Renteria2021).
In the present study, no positive genetic correlations were identified, thus no associations with a tendency to be a morning person are reported. Nonetheless, MD, AD, RD and MO showed significant negative genetic correlations with chronotype. Specifically, negative correlations with chronotype imply that DTI parameters corresponding to white matter structures are correlated with an increase in eveningness. In the present study, the eveningness chronotype is associated with higher average MD, AD and RD for all tracts, but with tract-specific distributions for each DTI parameter. Our results are highly consistent with previous findings describing white matter microstructural differences in several brain regions, including the corpus callosum, corona radiata, internal capsule and frontal and temporal lobes, when comparing evening to morning and intermediate chronotypes (Rosenberg et al., Reference Rosenberg, Maximov, Reske, Grinberg and Shah2014).
We identified genetic correlations between an evening chronotype and MD in the entire corpus callosum and corona radiata, as with the posterior segments of the internal capsule. Increased MD may be associated with decreased tissue density, altered properties of the myelin, axonal and neuronal membrane, as well as altered tissue organization and shape of glia and neurons (Sagi et al., Reference Sagi, Tavor, Hofstetter, Tzur-Moryosef, Blumenfeld-Katzir and Assaf2012; Beaulieu, Reference Beaulieu2002). Further, in our study, the posterior region of the corona radiata also showed increased RD, but no associations were detected with AD. An increase in AD, but not in RD, was identified in the posterior portion of the internal capsule for the evening chronotype. Similarly, the anterior part of the corona radiata showed genetic correlations between evening chronotype and AD, but not with RD. AD is usually associated with axonal membranes or axonal density (Kumar et al., Reference Kumar, Nguyen, Macey, Woo and Harper2012), while RD is typically associated with myelin density (Kumar et al., Reference Kumar, Nguyen, Macey, Woo and Harper2012). However, both measures could be affected by multiple tissue factors, including the presence of glial tissue, edema or multiple crossing fiber tracts (Winklewski et al., Reference Winklewski, Sabisz, Naumczyk, Jodzio, Szurowska and Szarmach2018).
Recently, it has been shown that glial cells are involved in the regulation of circadian rhythms and maintenance processes of a healthy brain (Chi-Castañeda & Ortega, Reference Chi-Castañeda and Ortega2017), and DTI parameters such as AD and RD can be affected by interactions between neuron and glial cells (Winklewski et al., Reference Winklewski, Sabisz, Naumczyk, Jodzio, Szurowska and Szarmach2018). We speculate that the role of glial cells in the regulation of circadian rhythms may go further than supporting and participating in metabolic functions. Possibly, abnormalities in glial tissue affecting AD and RD could also contribute to a preference for a morning or evening chronotype and, for evening people, increments in AD, RD or both could influence higher MD. However, more research is needed to disentangle the intricate relationship between AD, RD and MD.
Higher MO is associated with a more tubular shape of the diffusion tensor, consistent with the presence of a predominant fiber population. Our results showed that higher MO was associated with a genetic predisposition to the evening chronotype in regions with expected crossing fibers. Although it is expected to observe crossing fibers in superior corona radiata and the retrolenticular portion of the internal capsule, such an increase in MO may be attributed to decreased quality (less densely packed fibers and axonal density) of a portion of the crossing fibers, resulting in a predominant fiber population, also consistent with the increase of AD and MD in the same regions (Concha, Reference Concha2014; Douaud et al., Reference Douaud, Jbabdi, Behrens, Menke, Gass, Monsch and Smith2011).
Evening chronotype has been associated with lower sleep quality (Taillard et al., Reference Taillard, Philip, Coste, Sagaspe and Bioulac2003; Vitale et al., Reference Vitale, Roveda, Montaruli, Galasso, Weydahl, Caumo and Carandente2015), which in turn has been related to increased MD in the hippocampus, basal ganglia, anterior corpus callosum and prefrontal white matter (Khalsa et al., Reference Khalsa, Hale, Goldstone, Wilson, Mayhew, Bagary and Bagshaw2017; Takeuchi et al., Reference Takeuchi, Taki, Nouchi, Yokoyama, Kotozaki, Nakagawa and Kawashima2018). Our findings show similar patterns of white matter integrity in evening chronotype and further contribute to identifying an underlying genetic component between chronotype and MD regions. We hypothesize that the relationship between eveningness and higher MD could be mediated by a deficit in sleep quality.
Limitations of the present study must be recognized. Our analyses only used data from participants with European ancestry from the UK Biobank cohort. Since previous studies have pointed out racial differences in circadian rhythms (Egan et al., Reference Egan, Knutson, Pereira and von Schantz2017; Malone et al., Reference Malone, Patterson, Lu, Lozano and Hanlon2016), the generalizability of our results may be limited and should be addressed with caution. Furthermore, MR methods rely heavily on the statistical power of the samples, which in turn is tied to the availability of sufficient genetic instruments (i.e., genome-wide significant loci; García-Marín et al., Reference García-Marín, Campos, Martin, Cuellar-Partida and Renteria2021; Pierce et al., Reference Pierce, Ahsan and Vanderweele2011). In the present study, individual DTI parameters showed scarce genome-wide significant loci ranging from none to 12, lessening the statistical power of GSMR. Also, MR analyses were performed in the same cohort population (UK Biobank); therefore, the requirement for independent samples could not be met and a possible sample overlap between chronotype and DTI parameters may bias MR estimates (Burgess et al., Reference Burgess, Davies and Thompson2016). Nonetheless, LD score regression is unbiased by sample overlap, thus the estimates for genetic correlations remain unaffected (Bulik-Sullivan, Finucane et al., Reference Bulik-Sullivan, Finucane, Anttila, Gusev, Day, Loh and Neale2015). Despite reduced statistical power, our findings still provide evidence for causal associations suggesting eveningness is causal for increases in mean, axial and radial diffusivities and thus uncover meaningful insights and hypotheses for future studies to confirm when more powerful genetic statistical instruments and techniques become available.
In summary, we provide evidence for 29 genetic correlations between chronotype and brain white matter structure, of which 13 were found to be indicative of a causal relationship. Our results confirmed findings from previous studies and uncovered insights into the relationship between DTI parameters and chronotype. For instance, we reveal that increments in MD in the entire corona radiata and corpus callosum are associated with an evening chronotype. Further, we identified associations between eveningness and increments in the anterior and posterior parts of the corona radiata in RD and AD, respectively. Also, we found the retrolenticular part of the internal capsule to have higher MD, AD and MO among individuals leaning toward an evening chronotype. Altogether, our results contribute to a better understanding of the relationship between the brain and circadian preference.
Financial support
AIC and LMGC are supported by UQ Research Training Scholarships from The University of Queensland (UQ). MER thanks the support of the NHMRC and Australian Research Council (ARC) through a Research Fellowship (GNT1102821). PFK is supported by an Australian Government Research Training Program Scholarship from Queensland University of Technology (QUT). The research funders did not participate in the study design, data analyses, interpretation of results or writing of the manuscript.
Conflict of interest
GC-P contributed to this study while employed at The University of Queensland. He is now an employee of 23andMe Inc., and he may hold stock or stock options. All other authors declare having no conflicts of interest.
Ethical standards
This study was approved by the Human Research Ethics Committee of the QIMR Berghofer Medical Research Institute.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/thg.2021.4.
