Normal brain aging is accompanied by a decline of brain region volumes (Anderton, 2002) and cognition such as conceptual reasoning, executive function, and memory (Harada et al., 2013; Kirova et al., 2015). As the brain ages, many age-related diseases emerge, such as Alzheimer’s disease (AD) (Amaducci and Tesco, 1994; Ferri et al., 2005). As the fifth leading cause of death in people over the age of 65 years (Kirova et al., 2015), AD burdens the society heavily. The risk of developing AD increases exponentially with age (Plassman et al., 2007). Thus, revealing the mechanism of the normal brain age is the key to understanding age-related diseases (Raji et al., 2009). Recent studies combining neuroimaging and machine learning methods predict brain age successfully and found that the chronological age is not exactly equal to brain age in both normal and pathological subjects such as patients with schizophrenia, mild cognitive impairments, and depression (Gaser et al., 2013; Habes and Janowitz, 2016; Hajek et al., 2019; Han et al., 2021; He et al., 2020). This discordance between brain age and chronological age helps explain individual differences in brain aging (Jylhävä et al., 2017). However, the underlying biological basis of brain age is not well elaborated.

Extensive efforts have been made to identify reliable indictors of biological age (Wagner et al., 2016). In recent years, the brain age method identifying normal aging pattern has turned out to be an informative biomarker of healthy brain aging at the individual level (Cole and Franke, 2017; Franke et al., 2010). For example, Vishnu et al. accurately predicted MRI-derived brain age, helping to identify various brain diseases (Bashyam et al., 2020). Using this framework, studies have uncovered accelerated brain aging in several neurological diseases using the brain-predicted age difference (brain-PAD) scores, defined as the discordance between the predicted brain age and the chronological age (Gaser et al., 2013; Habes and Janowitz, 2016; Hajek et al., 2019; Han et al., 2021; He et al., 2020). The brain age method outperforms other state-of-the-art biomarkers, with accuracy rates reaching 81% in identifying mild cognitive impairments (Gaser et al., 2013). Despite these remarkable findings, these studies have failed to elucidate the underlying biological basis of brain age, limiting our understanding of the biological mechanism of brain age and its application.

It is widely accepted that genetic factors play important roles in normal brain aging (Lin et al., 2020). For example, the expressions of genes playing roles in synaptic functional and neuronal plasticity in the frontal cortex are reduced with aging (Sikora et al., 2021). However, the relation between genetic factors and brain age derived from neuroimaging remains unknown. Advances in comprehensive brain-wide gene expression atlases make possible linking the spatial variations in gene expressions to macroscopic neuroimaging phenotypes (Fornito et al., 2019; Zhu et al., 2021). For example, Reardon et al. found that the genetic spatial expression is tied with cortical scaling gradients (Reardon and Seidlitz, 2018). Resting-state intrinsic brain synchronization is also supported by related gene expression (Richiardi et al., 2015). Combing neuroimaging and gene transcripts provides insights into how disease-related aberrance at the microscale architecture drives macroscale brain abnormalities in mental disorders such as depression and schizophrenia (Romero-Garcia et al., 2020; Li and Seidlitz, 2021). The details of the underlying transcriptional mechanisms of brain age remain unknown.

The aims of the current study were twofold. Firstly, we investigated the importance of brain networks in brain age prediction. The Southwest University Adult Lifespan Dataset (N = 492) was used in the current study. For each subject, the gray matter volumes (GMV) quantified by voxel-based morphometry (VBM) of brain regions were treated as features to predict an individual’s brain age. In the prediction model, Gaussian process regression (GPR) was chosen for its superior performance compared to existing methods (Han et al., 2021). The importance of a distinct brain network was determined by computational lesion analysis (Feng et al., 2018). Secondly, genetic annotation of the brain networks playing decisive roles in brain age prediction was generated by employing the Brain Annotation Toolbox (BAT) (Liu et al., 2019) followed by functional enrichment analysis to infer the ontological pathways of the brain age-related genes.

In this study, we investigated the importance of brain networks contributing to brain age prediction and the underlying molecular mechanisms of brain age. As a result, the central executive network turned out to be a vital network in predicting brain age due to the performance of the prediction model being significantly degraded (Steiger’s Z = 2.114, p = 0.035) when it was excluded from the model. The genes associated with the central executive network were ontologically enriched in clusters such as regulation of ion transmembrane transport, synaptic signaling, synapse organization, retrograde endocannabinoid signaling (e.g., dopaminergic synapse), behavior (e.g., memory and associative learning), and so on. In all, these results reveal that the GMV of the central executive network played a vital role in predicting brain age and bridged the gap between transcriptome and neuroimaging promoting an integrative understanding of the pathophysiology of brain age.

Our results hinted that the GMV of the central executive network is a potential biomarker of brain age. Normal brain aging is associated with GM volume loss (Allen et al., 2005; Walhovd et al., 2005), including in the parietal lobe, temporal cortex, and especially in the frontal lobe (Matsuda, 2013; Van Petten et al., 2004). Along with losses of GMV, normal aging is characterized by a gradual decline in cognitive processes such as executive function, episodic memory, working memory, and processing speed (Lee et al., 2016). Consistent with these studies, our results presented that the GMV of all networks correlated with brain age significantly. In addition, we found that only when the central executive network was excluded did the performance of the prediction model significantly degrade (Steiger’s Z = 2.114, p = 0.035). These results hinted that the central executive network could be a potential biomarker of brain age. The reason might be that the effect of brain aging on the central executive network was more consistent across different populations than regions like the amygdala, hippocampus, and thalamus (Matsuda, 2013). Individuals exhibiting age-related decline tended to show impairments of executive functions first, suggesting that this network might be particularly vulnerable during normal aging (Sorel and Pennequin, 2008). In addition, a linear volume reduction of the central executive network with increasing age even occurred during the earlier stages of adulthood (Terribilli et al., 2011). As a supplement to these studies, our results revealed that the GMV of the central executive network played a decisive role in predicting brain age.

We further investigated the transcriptional signatures of the brain age-related networks. Although brain age was employed in abnormal aging trajectories in various diseases (Gaser et al., 2013; Habes and Janowitz, 2016; Hajek et al., 2019; Han et al., 2021; He et al., 2020), studies investigating the underlying biological foundation of brain age are scarce. To the best of our knowledge, only one study linked polygenic risk score and accelerated brain aging in AD (Habes and Janowitz, 2016). For the first time, we found that brain age-related genes were enriched in GO processes/KEGG pathways clustered into a number of groups such as regulation of iron/calcium transmembrane transport, synaptic signaling, synapse organization, retrograde endocannabinoid signaling (e.g., dopaminergic synapse), behavior (e.g., memory and associative learning), neurotransmitter secretion, and dendrite development. Calcium-dependent signals were key triggers of the molecular mechanisms underlying learning and memory; dysregulation of its homeostasis in the aging brain was hypothesized to underlie aging-related cognitive decline (Oliveira and Bading, 2011). In the brain, iron was involved in many fundamental biological processes, including neurotransmitter synthesis and metabolism; its homoeostasis played an important role in maintaining normal function (Ward et al., 2014). Normal brain aging is accompanied by selective accumulation of iron. Greater accumulation of iron was observed in neurodegenerative diseases associated with oxidative stress and cellular damage (Zecca et al., 2004). In addition, both the density and morphology of dendritic trees mainly possessed by pyramidal neurons underwent progressive regression in the neocortex (Dickstein et al., 2013) without neuronal death (Morrison and Hof, 1997). Consistent with the notion that no single mechanism explains the aging process (Kyng et al., 2003), we identified a number of GO processes/KEGG pathways underlying brain age.

Several limitations should be considered when understanding our results. Firstly, factors such as educational level could also affect the GMV. For example, greater GMV in the superior temporal gyrus, insula, and anterior cingulate cortex were found in more educated individuals (Arenaza-Urquijo et al., 2013). As this information was not included in the dataset used in the current study, future studies might explore its effect on brain age. Secondly, the gene expression data and neuroimaging data did not come from the same subjects. Considering the high degree of conservation in overall gene expression across human populations (Stranger et al., 2007; Zhu et al., 2021), the expressions of brain age-related genes could be believable.

As a supplement to previous studies exploring brain age, our results reveal a decisive role of the GMV of the central executive network in brain age prediction. In addition, the present study investigated the underlying transcriptional profiling of the central executive network. As a result, we found that brain age-related genes were enriched in GO processes/KEGG pathways clustered into a number of aging-related mechanisms such as regulation of iron/calcium transmembrane transport and dendrite development. In all, these results reveal that the GMV of the central executive network played a vital role in predicting brain age and bridged the gap between transcriptome and neuroimaging promoting an integrative understanding of the pathophysiology of brain age.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by the Research Ethics Committee of the Brain Imaging Center of Southwest University, in accordance with the Declaration of Helsinki. The patients/participants provided their written informed consent to participate in this study.

KF analyzed the data and wrote the manuscript. SH designed the research, analyzed the data, and wrote the manuscript. YL and JD searched the literature. JW modified the language. WZ directed the research program and provided guidance and suggestions for the study. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.