Introduction
The brain structure aspect alters throughout life. In particular, grey and white matter volumes are known to shrink with age in diseased and normal brains (Ge et al.
2002; Fjell and Walhovd
2010; Lockhart and DeCarli
2014). Numerous cross-sectional and longitudinal studies have confirmed this trend by either studying grey matter volume changes (Lemaitre et al.
2012) or cortical sulci widening (Kochunov et al.
2005; Shen et al.
2018). Importantly, the magnitude of brain shrinkage varies across regions and individuals, and increases with age (Raz et al.
2005). Multiple factors related to the environment or to genetics likely play a role in these changes. Such genetic effect is characterized in the hippocampus atrophy with the apolipoprotein E, ε4 allele (ApoE-ε4), which is also associated with an increased risk for developing late onset Alzheimer’s disease (AD) (Moffat et al.
2000). However, the genetic underpinnings of the brain sulcal features have not been investigated, except a few studies that were interested in the heritability of sulcal depth in extended pedigrees of young adults (Le Guen et al.
2018) or non-human primates (Kochunov et al.
2010). To the best of our knowledge, no genome wide association studies (GWAS) in imaging genetics with a sample size above 10,000 subjects were conducted on the brain sulcal features. In imaging genetics, previous GWAS with such sample sizes have looked into the hippocampal and intracranial volumes (Stein et al.
2012), or the human subcortical brain structures (Hibar et al.
2015). These studies traditionally used meta-analyses which pooled subjects scanned in different centers with various scanners and from different age ranges. Such a meta-analysis initiative is best exemplified by the ENIGMA (Thompson et al.
2014) and CHARGE (Psaty et al.
2009) consortia. The UK Biobank project (Allen et al.
2012) offers a remarkable opportunity to address these issues by gathering data from a fairly homogenous population of subjects, and acquiring magnetic resonance images (MRI) on identical scanners operated at the same location. Additionally, it enables researchers to directly access a cohort with numerous participants while alleviating the uncertainty of meta-analyses.
In this paper, we consider ten prominent brain sulci that are automatically extracted and labelled using the Brainvisa cortical sulci recognition pipeline (Rivière et al.
2009; Perrot et al.
2011). These sulci are the central, the anterior and posterior cingulate, the inferior and superior temporal, the intraparietal, the subparietal, the superior and inferior frontal sulci and the Sylvian fissure (Mangin et al.
2015; Shen et al.
2018). Even though the subparietal sulcus is not as prominent as the others, it was included in our analysis because it lies in the precuneus which is a major target for atrophy in AD (Karas et al.
2007; Bailly et al.
2015). First, we replicated the known trends of cortical shrinking with age in a large sample of individuals from the UK Biobank, considering the grey matter thickness and sulcal opening. Second, we estimated the genetic influence on these features with the genome-wide complex trait analysis (GCTA) (Yang et al.
2011). In this method the genetic relationship (kinship) matrix between subjects is computed to estimate the variance of an observed phenotype, which can be explained by the single nucleotide polymorphisms (SNPs), referred to as the heritability. Finally, we performed a genome-wide association study (GWAS) of the phenotypes with the genotyped variants using PLINK (Purcell et al.
2007). A functional annotation of the phenotype-associated variants was then performed using the gene expression level published by the Genotype-Tissue Expression (GTEx) consortium (GTEx Consortium
2017), allowing the identification of expression quantitative trait loci (eQTLs).
Discussion
The bulk of participants in the UK Biobank are older than 50 years old and results have to be discussed within this interval. To scrutinize certain genetic effects on brain features, it is important to understand the effect of age. Indeed, some genetic influences may only be revealed in a relatively aged cohort such as the UK Biobank. In this study, we emphasized a steady age effect on the sulcal opening and GM thickness between 45 and 75 years of age. Furthermore, we underlined that these brain features are heritable, thus their variance across individuals is significantly regulated by gene additive effects. Finally, we highlighted several causal genetic variants whose allelic configuration directly contributes to the phenotypic variance.
It is well known that with age, the GM and white matter (WM) volumes decreases (Raz et al.
2005), while the amount of cerebrospinal fluid (CSF) in the cortical folds increases (Good et al.
2001). In our analysis, we confirmed the well-established results regarding the GM thickness decrease and sulcal opening increase with age in the UK Biobank cohort. The assessment of cortical sulci opening is well described and correlates with neurocognitive decline in mild cognitive impairment and dementia disorders (Bastos Leite et al.
2004). The cortical sulci widening with age is likely related to the reduction of gyral thickness resulting in the dilatation of the sulci (Magnotta
1999; Symonds et al.
1999), but could also account for neurodegenerative processes occurring in the underlying white matter (Gunning-Dixon et al.
2009). While, the GM thickness and sulcal widening are correlated (Kochunov et al.
2008; Liu et al.
2013), the robustness of the two measures might differ. We obtained less significant variant-phenotype associations in GM thickness, because the sulcal opening might be more consistently measured across individuals than the GM thickness. This might be due to the fact that the MRI contrast between GM and CSF remains more stable across the lifespan than the GM / WM contrast (Kochunov et al.
2005). Thus, sulcal widening is commonly used by radiologists as a surrogate of cortical atrophy in clinical settings (Shen et al.
2018). It could also reflect a higher sensitivity to the primal effects of aging because of the consequences of grey and white matter atrophy. Note that the opening of larger sulci like the Sylvian valley could also be impacted by global mechanical compensation for aging processes.
The main finding of this paper is that the locus upstream of the
KCNK2 transcription start site influences the sulcal opening and GM thickness. Additionally, the tissue specific gene expression (eQTL) analysis of the GTEx consortium emphasizes that overall (meta-analysis of all tissues), and particularly in the brain, this DNA region regulates the expression of
KCNK2 (GTEx Consortium
2017). Thus, we can legitimately assume a link between the regulation of
KCNK2 expression and the amplitude of sulcal opening. In other words, depending on the allelic configurations in the region upstream of
KCNK2, an individual will have his sulci comparatively enlarged. Because sulcal widening is a marker of cortical atrophy as we pinpointed in the previous paragraph, there is a potential link between
KCNK2 expression level and brain atrophy. The
KCNK2 gene, also known as
TREK1, is a member of the two-pore-domain potassium channel family which is expressed predominantly in the brain (Hervieu et al.
2001). Previous literature emphasized several functions for
KCNK2 gene in the brain. First, the
KCNK2 regulates the blood–brain barrier function and inflammation in the brain of mice (Bittner et al.
2013) and humans (Bittner et al.
2014). The inhibition or deletion of
KCNK2 facilitates lymphocytes migration into the central nervous system (CNS) and promotes autoimmune CNS inflammation (Bittner et al.
2013). Second, in mice, the knockdown of
KCNK2 gene impairs the neuronal migration of late-born cortical excitatory neurons, which are precursors of Layer II/III neurons (Bando et al.
2014). Third, in rat hippocampal astrocytes, the increase of
KCNK2 expression mediates neuroprotection during ischemia (Banerjee et al.
2016). The mechanism might involve
KCNK2 blockade, inhibiting neuronal apoptosis and protecting the brain from cerebral ischemic injury (Wang et al.
2018). Finally,
KCNK2 over expression was shown to exacerbate memory impairment in middle-age mice (Cai et al.
2017). To summarize,
KCNK2 controls several major cellular responses involved in memory formation and is believed to participate in neuroinflammation, cerebral ischemia and blood–brain barrier dysfunction (Bittner et al.
2014; Cai et al.
2017; Wang et al.
2018).
The first role suggests the most promising direction of future work, because previous studies have proposed that neuroinflammation is involved in cognitive decline in midlife (Marsland et al.
2015) and implicated in pathological age-related changes and AD (McGeer and McGeer
2001). Throughout life, stress, recurrent inflammation and subclinical cerebrovascular events potentially contribute to brain aging (Raz and Rodrigue
2006). The link between our findings and inflammation indicates a potential mediation role for
KCNK2. Finally, it is difficult to disentangle whether or not brain inflammation has a deleterious role on cognitive functions, since there is no clear consensus. A recent study however emphasizes a slower progression of AD in patients with early neuroinflammation (Hamelin et al.
2016).
One limitation of this study is that the genetic associations were found in a British sample of elderly individuals. Thus, this association might not replicate in a sample composed of individuals with different ancestry or a younger cohort for which the brain features under scrutiny have not sufficiently evolved with age or with potentially recurrent neuroinflammatory events.
In conclusion, in a sample of 15,597 subjects representative of the general population of British ancestry, we have shown that an eQTL of KCNK2 influences sulcal widening. This appears coherent with the role of KCNK2, which affects the regulation of inflammation response in the brain.