Fifty-six healthy adult participants between 18 and 45 years of age (mean age = 30.8 ± 8.1 years, 29 males) participated in this study. Detailed age-wise (18–45 years) (A) and age-range (Group 1: 18–25, Group 2: 26–35, and Group 3: 36–45 years) (B) distributions over number of participants are shown in Figure 1. Participants were screened via a comprehensive telephone interview and were excluded for any history of psychiatric, neurological, or significant medical problems (e.g., heart problems, diabetes), including head injury with loss of consciousness longer than 30 min, sleep disorders, or current use of psychotropic medications that could affect neuroimaging. Participants were also excluded for any history of drug or alcohol treatment or current use of illicit substances. Current alcohol use was required to be lower than the Center for Disease Control criteria for excessive alcohol use³. Written informed consent was obtained from each participant before the experiment and the study protocol was approved by the Institutional Review Boards of McLean Hospital and Partners Healthcare, as well as by the United States Army Human Research Protections Office. Other unrelated behavioral data from this sample have been reported elsewhere (Killgore et al., 2013), but the cortical thickness and cortical surface area measures reported here are novel and have not been previously reported. We recorded high-resolution anatomical magnetic resonance imaging (MRI) data using a 3-Tesla Siemens whole brain MR scanner located at the McLean Hospital Imaging Center. Each participant was instructed to rest, relax and try his/her best to stay motionless during the entire scan. Anatomical data for each participant was acquired using a 3D magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence which consisted of 128 sagittal slices (thickness = 1.33 mm, voxel resolution = 1 × 1 × 1.33 mm, field of view (FOV) = 256 mm) with TR/TE/FA/inversion time of 2100 ms/2.25ms/12°/1100 ms.

We used the “recon-all” pipeline in FreeSurfer (version 6.0)⁴ installed on a Debian 8 LINUX machine to process anatomical images for all the participants. This processing involved motion-correction, brain extraction (i.e., removal of skull, skin, neck and eyes), automated transformation to Talairach co-ordinate system, intensity correction, volumetric segmentation, and whole brain parcellation into seven functional networks (visual network, somatomotor network, dorsal attention network, ventral attention network, limbic network, fronto-parietal network, and default mode network) using the Yeo atlas (Yeo et al., 2011). These seven networks and their corresponding regions/components are summarized in Table 1. The naming of all the regions constituting each network was performed by overlaying annotation file (split components) of 7 networks over 7-network parcellation in FreeView. The measures of mean CT, mean CV and mean CSA were evaluated individually for the left and the right hemisphere to determine any significant differences in these measures across different age groups. We also estimated the correlations between age and each of these measures averaged over each hemisphere and ICV. For network level analysis, the correlation analysis between all three cortical measures and age as well as within cortical measures was performed. All the correlation analysis was performed after extracting subject-wise CT, CV and CSA measures from all seven functional networks for each hemisphere. For multiple comparisons, the Holm–Bonferroni technique (Holm, 1979) was used to adjust the p-value of 0.05 across all correlations of CT, CSA, and CV with age and for all correlations within cortical measures, such as between CV and CT, CV and CSA, and CT and CSA.

Further, we evaluated the association among all three (CV, CT, and CSA) cortical measures.

Previous research has identified age-related changes in cortical measures such as thickness, surface area and gray matter volume across different regions of interest using voxel-based and surface based measures (Salat et al., 2004; Lemaitre et al., 2012). In a study of 70 healthy men between the ages of 19 and 76 years, significant age-related loss in gray matter volume in the frontal and temporal lobes was reported (Bartzokis et al., 2001). It was suggested that a reduction in the number of large neurons and an increase in the proportion of small neurons contributed to an overall reduction in cortical volume (Terry et al., 1987; Bartzokis et al., 2001). However, network level changes in cortical characteristics have rarely been investigated in prior work. In addition, most of the previous studies reported inconsistent trend effects of aging on surface-based cortical measures, especially cortical surface area (Salat et al., 2004; Fjell et al., 2006; Rettmann et al., 2006; Dickerson et al., 2009). In this study, we found a nearly consistent global reduction in mean CT and unchanged mean CSA of each hemisphere as well as for all the functional networks with the exception of the limbic network. Consistent with a study by Grieve and colleagues (Grieve et al., 2005), we found a preserved limbic network in terms of non-significant changes in magnitude of CT and CSA with age. In that study, significant preservation, relative stability, and sparing of the medial temporal lobe including the entorhinal and parahippocampal cortices across the entire age range were reported. In addition, in younger and older adults, robust functional activation during detection of emotional faces was found in the amygdala, suggesting that the amygdala remains relatively preserved with aging (St Jacques et al., 2010; Campbell et al., 2012; Hampson et al., 2012). Significant stability of the hippocampal region until the 7th decade followed by a sharp decline after the 8th decade was also reported (Allen et al., 2005). Several other studies support the notion that regions associated with the limbic network remain relatively preserved in aging humans (Raz et al., 1997), aging rats (Rapp and Gallagher, 1996), and aging primates (Keuker et al., 2003).

Our findings did not show how the observed age-structure associations correspond to changes in actual cognitive performance. However, reduced cortical thickness is considered one of the possible mechanisms responsible for significant deviation in global cognition resulting in lower cognitive reserve (Ferreira et al., 2017), as measured by the Wechsler Adult Intelligence Scale (WAIS). To some extent, higher cognitive reserve restricts the clinical symptoms following brain pathology (Stern, 2009), and shields and compensates for the effect of cortical thinning (Ferreira et al., 2016). It is also possible that cognitive reserve mechanisms may protect against the changes in brain structure within the limbic network. The present findings reported in our study, as well as those of previous studies showing a protective role of cognitive reserve (Freret et al., 2015; Ferreira et al., 2017) and memory decline with age (Peters, 2006), suggest that the basis for age-related decline in cognitive skills is not entirely due to structural loss of a specific brain region within the limbic network. It is intriguing to speculate that the relative preservation of the limbic network with age in our sample and prior evidence of spared amygdala activation in older adults may suggest a greater reliance upon emotional encoding/retrieval strategies with increasing age.

In this study, significant reductions in CT and CV as a function of age for six functional networks reflect a widespread pattern of age-related cortical changes. This widespread pattern of age-related structural changes was also evident from lower CT, CSA, and CV in older individuals versus younger individuals. Significant negative associations between age and averaged CT and CV over each hemisphere as well as between age and ICV were also observed. There is abundant evidence showing age-related declines in brain activation and cognitive performance but limited evidence showing age-related structural changes in multiple regions of the human brain (Trollor and Valenzuela, 2001; Scahill et al., 2003). Age-related functional changes, however, could be a possible direct/indirect indicator of associated age-related structural modifications. Such changes in brain morphology have been extensively investigated in postmortem as well as in vivo magnetic resonance imaging studies (Kemper, 1994; Raz, 2000), with most showing age related weakening of the brain structure (i.e., decline in weight, volume, and white matter integrity) and global decline of cognitive functions.

Below we discuss our findings showing instability of each of the six networks with age:

It should be noted that most of the previous studies on human aging concentrated on brain function or functional connectivity for middle-aged and/or old-aged individuals. Very few studies have focused on age-related structural changes in the brain. As such, there was a lack of reported matrices (i.e., cortical thickness, cortical surface area, and cortical volume) sensitive to the identification of early structural brain changes for healthy younger adults, especially in various pre-defined functional networks. Previously, in a study of 207 healthy adults between 23 and 87 years of age, Storsve and colleagues reported age related annual decrease in thickness, volume, and area in most brain areas (Storsve et al., 2014). In another study of 974 individuals (aged between 4.1 and 88.5 years), Fjell and colleagues reported that genetic factors, genetic organization and maturation of brain areas affect cortical changes throughout the life span (Fjell et al., 2015). To our knowledge, our study is the first to focus on changes in all three morphometric measures in individual functional brain networks of young healthy adults. Although the underlying mechanisms of reduction in functional and structural measures have not been explored in detail, it is suggested that the basis of this significant reduction could be due to non-pathological processes such as transition of neurons to neurofibrillary tangles and reduction in dendritic arborization or pruning (Morrison and Hof, 1997). On the other hand, it is also possible that such declines could be due to pathological processes such as amyloid deposition or unwanted chemical changes such as disorders in glucose metabolism over time (Walhovd et al., 2014). A blend of decreased neural specificity and white matter integrity could be another possible explanation of functional and structural loss with aging (Park et al., 2004). Since post mortem studies and studies on non-human primates found relatively comparable neuron count between older and younger participants, neuron death has not been suggested as a cause for reduction in magnitude of cortical measures (Morrison and Hof, 1997; Peters et al., 1998).

Consistent with a study by Lemaitre et al. (2012), we did not find a functional network with pronounced reduction in CSA with aging. This suggests that CSA is less sensitive to morphometric changes with aging, which makes CT and CV relatively more informative measures to identify age-related morphometric changes across the brain (Lemaitre et al., 2012). Lemaitre and colleagues make an analogy to a dry apple, stating “When an apple dries, its flesh and thickness reduce and its skin shrivels keeping the surface area relatively constant but gaining spatial complexity from a flat to an uneven surface” helps to better understand the possible mechanisms underlying these cortical changes (Lemaitre et al., 2012, p. 617.e7). This underlines an idea that with age, an atrophied brain might display an increase in gyral complexity without a decrease in surface area. Consistent with this hypothesis, Rettmann and colleagues also observed age-related reduction in thickness and volume but no change in surface area among eight sulcal regions (Rettmann et al., 2006). Since cortical volume, simply put, is the product of the thickness and surface area, so changes in volume account for changes in thickness as well as surface area, which could be one of the possible reasons for an inconsistent pattern of age-related changes in CV across functional networks, especially before and after considering the effect of ‘gender.’ In the field of imaging genetics, it has been suggested that methods providing measures of gray matter volume could be less sensitive for gene identification than the methods providing CT and/or CSA (Winkler et al., 2010). In addition, we did not observe significant inter-dependence between CT and CSA, reflecting the idea that the sensitivity of thickness and area to normal aging could be entirely different depending on aging and/or type of neuropsychiatric disorder (Lemaitre et al., 2012) and/or type brain structure under study. Therefore, non-uniformity in age-related changes in CT, CSA, and CV but uniformity in age-related changes in CT, dependence of CV on CT and CSA and independence of CT and CSA in the present set of functional networks for relatively younger participants reflects the importance of analyzing multiple aspects of age-related cortical changes. For instance, individual measures of CT, CSA, and CV in isolation are not enough to distinguish the age-related susceptibility of functional brain networks to completely understand the effect of normal aging on cortical measures across the whole brain. While each of these measures contributes unique and important information regarding specific age-related histological changes in brain, when used in combination, these measures provide substantially greater precision in characterizing these associations.