Due to immense developments in medical technologies and treatments in recent decades, the oldest old (individuals 85+ years of age) have become the fastest growing segment of the U.S. population (National Institute on Aging and U.S. Department of Health and Human Services [DHHS], 2020). However, accompanying this enhanced longevity is an increase in the prevalence of conditions such as Alzheimer’s Disease, osteoporosis, and cardiovascular disease, among others (Jaul and Barron, 2017). Even among relatively healthy adults, the normal aging process is characterized by declines in cognitive function, especially in the domain of executive functioning (EF), and changes to emotional and motivational processing (Scheibe and Carstensen, 2010; Murman, 2015). Such changes often lead to decreased quality of life, as well as an increased susceptibility to fraud, scams, and financial distress due to impaired decision making (Denburg et al., 2007; Gamble et al., 2015).

With the utilization of functional neuroimaging techniques such as resting state functional connectivity magnetic resonance imaging (rs-fcMRI), alterations in functional coordination between brain regions, in the absence of a task, have become increasingly more apparent and robust across various populations (Zhang and Raichle, 2010). An examination of the current literature suggests at least two major resting state networks critical to executive, socioemotional, and economic functioning, namely the frontoparietal control network (FPCN) and the default mode network (DMN). The FPCN, which includes the dorsolateral prefrontal cortex (dLPFC), anterior prefrontal cortex, portions of the anterior cingulate (ACC), and dorsal precuneus, has been consistently implicated in general control processes across a wide variety of domains including socioemotional and neuroeconomic processes, as well as emotion regulation (Vincent et al., 2008; Depue et al., 2010; Ochsner et al., 2012), the selection and manipulation of information in working memory (Barnes et al., 2016), and processes related to the prioritization of information, especially under conditions of conflict (Zaki et al., 2010). Moreover, altered function of the dLPFC among clinical groups is associated with impulsive economic and risk-seeking behaviors (Meade et al., 2020).

The DMN generally comprises the medial prefrontal cortex (PFC), posterior cingulate cortex, and bilateral inferior parietal lobe, among other brain regions. It has been suggested to play an important role in introspective, self-generated processes (e.g., acting on internally generated information and evaluation of the self in reference to others) that may also influence socioemotional and neuroeconomic processes (Peters and Büchel, 2011; Andrews-Hanna, 2012). Of note, the FPCN and DMN often act in an anti-correlated manner to one another such that when individuals are engaged in cognitive control there is decreased activity in the DMN (Fox et al., 2005), although under conditions in which executive resources are needed to support internally directed cognition, these two systems may act in concert (Spreng et al., 2010).

In general, previous research suggests a marked impact of aging on structural and functional connectivity within each of these networks, but some discrepancies do exist. For example, some studies have found thickening of the ACC with age (e.g., Dotson et al., 2016) while other longitudinal analyses report significant thinning (e.g., Fjell et al., 2014). Similar discrepant results have been reported for regions of the DMN; notably, Salat et al. (2004) and Abe et al. (2008) both report age-related reductions in cortical gray matter in most anterior structures of the brain, though Salat et al. (2004) also found thickening of the medial orbitofrontal cortex with age. Findings for age-related alterations in functional connectivity are more consistent, with most studies reporting decreases in resting state connectivity within the FPCN (e.g., Geerligs et al., 2015) and DMN (e.g., Damoiseaux et al., 2008) among older adults relative to young adults (see also, review from Sala-Llonch et al., 2015). However, there are still discrepancies across studies with regard to the magnitude of age-related effects.

In a recent review paper, Andrews-Hanna et al. (2019) provide input on potential sources of variation in structural and functional connectivity findings by highlighting several challenges that often arise while conducting aging research. Broadly, they argue that previously reported age-related changes may be driven, to some degree, by motion/physiological, contextual, or motivational factors. For example, motion artifacts may partially account for discrepant results from structural studies, as older adults tend to show more variability in head motion during scanning (Savalia et al., 2017). Additionally, young and older adults may differ in their motivation for scanner-related tasks (e.g., staring at a fixation cross). Finally, though most connectivity studies exclude participants diagnosed with Alzheimer’s disease, many fail to measure the associated risk factor of Mild Cognitive Impairment (Boyle et al., 2006). Thus, age-related changes reported by previous studies may be driven in part by the presence of undetected preclinical Alzheimer’s disease in clinically normal adults.

However, it is also the case that aging is associated with reductions in physical activity (Sarkisian et al., 2005), voluntary exercise (Meisner et al., 2013), and cardiovascular fitness (Sandbakk et al., 2016). There is a growing body of research that suggests exercise in general (Li et al., 2017) and cardiovascular fitness in particular (Voss et al., 2016) have beneficial effects on brain function and connectivity. Much of this work has supported this relationship within older populations and in the domain of EF (Heyn et al., 2004). In the majority of studies comparing functional network connectivity in older vs. younger samples, the relative fitness/physical activity status of participants is not measured nor is there an explicit attempt to match participants on these variables. It is likely, however, that owing to general demographic trends the younger participants in these studies are both more active and more fit than the older participants. If true, it may be that the “age-related” network connectivity differences reported in previous studies have been overestimated and may be due—at least in part—to the effects of age-related exercise and fitness declines on connectivity rather than solely age.

The current analysis explores functional connectivity in core resting state networks strongly associated with cognition and EF in both young and older adults who were not currently meeting physical activity guidelines from the American College of Sports Medicine (American College of Sports Medicine [ACSM], 2014) at the time recruitment began for the study. The goals of this research are to replicate previous studies showing age-related differences in functional connectivity, specifically in the FPCN and DMN, and to investigate whether fitness may partially account for these differences in functional connectivity between young and older adults. Data were collected as part of a large-scale randomized controlled trial of a 16-week supervised exercise intervention designed to improve fitness, as well as cognitive, socioemotional, and economic function among sedentary adults 60 years of age and older. Given past research on age-related differences in functional and structural activity (reviewed in Sala-Llonch et al., 2015), we hypothesized that older adults would show decreased FPCN and DMN connectivity relative to young adults. Drawing from research on the neuroprotective effects of exercise (Colcombe et al., 2006), we further hypothesized that accounting for baseline fitness should result in attenuated differences in functional connectivity between young and older adults.

The fitness and neurocognitive function data herein were collected as part of a larger intervention to increase physical activity among older adults (NCT02068612; 5R01 AG043452-05) we called Fitness, Older Adults, and Resting State Connectivity Enhancement (FORCE). Briefly, the aims of the FORCE study were to characterize associations between fitness and various domains of functioning (e.g., cognitive, emotional), as well as to examine the effects of regular exercise participation on such associations. Older adults 60 years of age and older were recruited and completed extensive assessments including cardiovascular fitness (VO₂ peak), self-reported health and functioning measures, objective cognitive functioning measures, and a functional magnetic resonance imaging (fMRI) session. They were then randomly assigned to either low- or moderate-intensity exercise as part of a 16-week supervised exercise paradigm (results not reported here). A smaller group of young adults was recruited as a comparison for baseline fitness and neurocognitive function. These young adults completed all of the same baseline assessments but did not engage in the exercise intervention.

In the present investigation, we sought to replicate findings from previous studies reporting age-related differences in functional connectivity and associations between fitness and connectivity, and then determine whether age-related declines in fitness might partially account for observed age differences in connectivity. In regression analyses examining associations between age, fitness, and average functional connectivity, we found somewhat perplexing results. When comparing across age groups, average connectivity in both the FPCN and DMN did not differ between young and older adults. This result is inconsistent with past research (e.g., Andrews-Hanna et al., 2007). Interestingly, in the pairwise ROI-to-ROI rs-fcMRI analyses, the strength of many of the functional correlations between specific ROIs differed between young and older adults in this sample, with older adults generally exhibiting reduced connectivity. It is therefore possible that the observed weaker functional correlations between specific regions among older adults were “washed out” when examining average connectivity across nodes in a particular network.

In contrast to the group analyses, when age was treated continuously, we observed a significant negative relationship between age and FPCN connectivity among older adults. If fitness is indeed neuroprotective (Colcombe et al., 2006), we might expect this relationship to be weaker after accounting for fitness. Instead, though fitness shared a marginal negative association with FPCN connectivity, the effect of age remained significant (and even became a bit stronger). In fairness, this is likely a statistical suppression effect given the observed correlations between age and fitness (see Table 3). More specifically, among older adults, age was significantly negatively correlated with FPCN connectivity and fitness, but FPCN connectivity and fitness themselves were uncorrelated. This is clearly problematic, given that a potential covariate should have a statistical relationship with the outcome. It also further supports our previous conclusion, namely, that using a measure of average connectivity in the context of regression analyses may not be the best method to assess the effects of fitness on age-related differences in functional connectivity.

The individual ROI-to-ROI rs-fcMRI analyses were more consistent with findings from previous studies (e.g., Andrews-Hanna et al., 2007; Damoiseaux et al., 2008), indicating weaker functional correlations for older adults between several regions in both the FPCN and DMN. Furthermore, we found that, consistent with Voss et al. (2016) controlling for fitness generally resulted in some attenuation in the differences in functional connectivity in these particular ROI-to-ROI connections between the age groups. As one example to illustrate this point, the strength of the correlation between the right MPFC and the right superior temporal sulcus no longer differed between young and older adults once fitness was accounted for. This pattern held for other regions in the DMN, as well as for several functional correlations between regions in the FPCN. This suggests that “age-related” differences in network connectivity reported by previous studies may have been overestimated because many of those studies failed to control for the effects of fitness on connectivity. Bearing in mind that fitness was unrelated to average FPCN and DMN connectivity in both age groups, these findings also suggest that the effects of fitness on functional connectivity do not seem to be global. In other words, fitness can account for some of the age-related differences in functional correlations between specific brain regions, but it may not be useful for explaining age-related differences in average connectivity. Future work is needed to confirm these findings and to elucidate whether there is a consistent pattern to those connections that seem to be influenced by fitness changes and those that do not. It may also be fruitful to explore this global vs. specific distinction in the extent to which other factors, such as engagement in activities that are intellectually stimulating, might explain differences in functional connectivity between young and older adults (Park et al., 2007).

Future research could also examine how connectivity between large-scale brain networks, in addition to alterations within large-scale networks, are related to both age and fitness. We primarily focused on within-network coupling in the FPCN and DMN given prior studies reporting sensitivity of those networks to age (reviewed in Sala-Llonch et al., 2015; Andrews-Hanna et al., 2019) and cardiorespiratory fitness (e.g., Voss et al., 2016), and in an attempt to reduce the number of overall statistical tests. However, older adults tend to show increased functional coupling between the FPCN and the DMN compared to young adults, as reflected in the Default-to-Executive Coupling hypothesis of aging (Turner and Spreng, 2015; Spreng and Turner, 2019). Since the FPCN is considered a brain-wide network “hub” that alters its coupling patterns with the DMN depending on one’s current goals (Spreng et al., 2010), we reasoned that between-network in the absence of an explicit task at hand (i.e., “rest”) might be especially sensitive to the co-occurring mental state of the participant, making findings difficult to interpret. Some participants may have viewed the “rest” period as an effortful externally oriented task (i.e., “remain focused on the fixation crosshair”) whereas others may have viewed the rest period as an opportunity to advance their internally oriented goals (i.e., “use this time to plan the rest of my day”). Nevertheless, a more complete analysis of relationships between fitness and brain aging, and of the relation between resting state connectivity and ongoing thought patterns, would be of broad interest in future work.

As is always the case, some limitations need to be addressed. First, it is entirely possible that those participating in the FORCE study represent an especially healthy sample of older adults, in terms of both physiology and cognitive function. Participants were recruited from the Boulder-Denver area, two cities consistently cited as being among the healthiest and most active in the United States (Warren, 2020). In addition, our inclusion criteria required participants to be engaging in fewer than 80 min of moderate-to-vigorous exercise each week, rather than being completely sedentary. Thus, the older adults in our sample may have been healthier and fitter than what is typical. A related limitation is that most adults in the current sample reported being white, educated, and of middle socioeconomic status, further restricting the generalizability of the findings. The size of the young adult sample was also relatively small. Finally, our MRI scanner was upgraded part-way through the larger study, which past studies have shown can introduce bias and therefore reduce the integrity of the results (Chen et al., 2014). Efforts were made to reduce this bias by equating critical parameters across the systems and including a covariate for scanner in our analyses.

To conclude, this investigation employed a combination of analysis techniques to explore whether fitness partially accounts for important age-related differences in functional connectivity. Our initial approach involving a series of regressions using an average measure of connectivity showed that FPCN and DMN connectivity did not differ between young and older adults, and our measure of fitness was uncorrelated with average connectivity in both age groups. Seed-based rs-fcMRI analyses were more consistent with findings from previous studies (e.g., Andrews-Hanna et al., 2007), indicating weaker functional correlations between several regions in the FPCN and DMN among older adults as compared to younger adults. Critically, many of these differences were attenuated when fitness was accounted for. Taken together, our findings suggest that fitness exerts regional rather than global effects on network connectivity, which can account for some of the differences in functional connectivity across age.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the University of Colorado Boulder Institutional Review Board, Office of Research Integrity. The patients/participants provided their written informed consent to participate in this study.

DS, MB, JA-H, KH, and AB contributed to conception and design of the study. AB was responsible for funding acquisition and project administration. CG and EM performed the statistical analyses. CG, EM, and AB wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

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.