Advanced age is associated with an overall decrease in the effectiveness of specialized communication at the global network level (i.e., integration) and loss of functional specialization at the local level of specific brain regions (i.e., segregation). That is, neuronal networks become less distinct with advanced age (Foo et al., 2021).

Age-related differences in the reorganization of functional connectivity and cognitive abilities may also differ by sex. In adulthood, sex differences in brain structure and function can be observed (Cosgrove et al., 2007) and, similarly, sex differences in resting state networks (RSNs) have been reported. For example, studies of age differences have observed that males show increasing connectivity between-networks when compared to females (Allen et al., 2011; Filippi et al., 2013; Goldstone et al., 2016). Males also exhibited more marked changes in default mode network (DMN) connectivity, especially in the posterior cingulate cortex (PCC), but showed smaller differences (and possibly increases) in connectivity to the lateral prefrontal regions of the fronto-parietal network (FPN) relative to females (Allen et al., 2011; Filippi et al., 2013). Females on the other hand showed smaller differences in DMN connectivity but showed greater decreases in FPN connectivity when compared to males.

Sex differences in network architecture have also been reported (Zhang et al., 2016). Zhang et al. (2016) observed that female functional networks have significantly more connected nodes than males suggest an increase in network homogeneity in female brains. They also observed that the cerebellar nodes have a higher clustering coefficient and local efficiency for females (Zhang et al., 2016). This adds evidence to their findings that the clustering coefficient and local efficiency in males are higher, while female connections are diffuse across lobes and the network is less modular. These results jointly support the notion that networks in female brains, compared to those in males, are more spatially distributed but with lower correlation strengths (Scheinost et al., 2015; Zhang et al., 2016). Overall, males and females showed differential patterns in connectivity in RSNs during normal aging, and from early adulthood to late middle age (Allen et al., 2011; Filippi et al., 2013; Scheinost et al., 2015). However, these differences vary between studies with respect to regions and networks that are impacted.

Just as there are mixed results regarding sex differences with age in brain networks, there are also disagreements in the literature regarding differences in resting-state functional connectivity in the context of hormonal differences between biological sexes. To this point, there have been a limited number of studies that have investigated sex steroid hormones on brain networks. Further, there is currently disagreement among the rapidly expanding number of studies on the possible neuroprotective effects of sex hormones on cognitive and brain function more generally (Moffat, 2005; Sundström Poromaa and Gingnell, 2014; Toffoletto et al., 2014). Given that RSNs appear to be differentially impacted in males in females in later life, further exploration of the impact of sex steroid hormone levels on network architecture across adulthood stands to improve our understanding of underlying factors contributing to these differences. Specifically, it may be that sex hormone levels are related to the integration and segregation of neuronal networks (Peper et al., 2011).

In the brain, hormone receptors can be found across several regions. Post-mortem studies have found estrogen receptors in the hippocampus, claustrum, cerebral cortex, amygdala, hypothalamus, subthalamic nucleus, and thalamus (Osterlund et al., 2000; Weiser et al., 2008; Syan et al., 2017). As for progesterone, a post-mortem study reported high concentrations of its receptors in the amygdala, hypothalamus, and cerebellum (Bixo et al., 1997; Syan et al., 2017). Testosterone exerts an early organizational effect on the development of the hypothalamus (Jacobson et al., 1981), the cerebral cortex (Diamond, 1991), and the hippocampus (Roof and Havens, 1992) in addition to other brain structures. With respect to testosterone, after conversion to estradiol, it can also interact with estrogen receptors (Moffat, 2005). The relationship between these two hormones (testosterone and estrogen) have been shown to affect vascular health directly (Aggarwal et al., 2018; Raparelli et al., 2022). The ratio of testosterone to estradiol is important for vascular function suggesting that these hormones may not act independently (van Koeverden et al., 2019; Raparelli et al., 2022). Therefore, it is also important to investigate the interactions that may occur between sex hormones and whether together they may be involved related to aging processes, as well as differences in neuronal networks (Syan et al., 2017). Testosterone, progesterone, and estrogens are present in both males and females, but their levels and production vary, mainly with respect to sex and age, though fluctuations in estrogens and progesterone occur across the female menstrual cycle as well (Sundström Poromaa and Gingnell, 2014).

The influence of sex hormones on functional networks is vital to our understanding of brain function and organization during periods of endogenous hormonal fluctuation, such as that which occurs in females during midlife with the transition to menopause when levels of estrogens and progesterone drop markedly (Toffoletto et al., 2014; Foo et al., 2021). Our study here aims to investigate how sex steroid hormones relate to age-network relationships in both males and females, with a focus on network segregation. As such, we predicted that there would be associations between network segregation and hormone levels, as well as interactions between hormones that may be affecting or enhancing the segregation of RSNs.

This study investigated network segregation in aging females in the context of sex steroid hormones. Primarily, we were interested in network segregation in adult females, as sex steroid hormone levels—which undergo dramatic changes in females during mid and later life—may impact brain network properties. Understanding differences in network segregation in females in the context of aging and hormone levels stands to provide a greater understanding around factors contributing to functional differences in aging, particularly given that older females are at greater risk for negative outcomes in later life (Gao et al., 1998; Burger, 2008; Lahousse et al., 2014; Alzheimer’s Association, 2022). Somewhat surprisingly, we found no significant relationships between sex steroid hormone levels and network segregation levels in adult females. To our knowledge, this is the first study to directly investigate endogenous hormone levels with network segregation and patterns of aging.

In the context of both endogenous hormone levels and exogenous sex hormone treatments, sex hormones have displayed impacts on brain structure and function (e.g., cortical connectivity, subcortical connectivity, and within-network coherence) (Peper et al., 2011; Taylor et al., 2019; Pritschet et al., 2020). Given the notable vacillations in hormone levels during distinct reproductive stages (Diedrichsen, 2006) and cognitive inefficiencies associated with certain reproductive stages (Greendale et al., 2011; Epperson et al., 2013; Weber et al., 2014; Rentz et al., 2017; Taylor et al., 2019; Pritschet et al., 2020), we expected to find associations between network segregation and hormone levels. However, females did not exhibit significant associations between progesterone, 17β-estradiol, or testosterone and network segregation. These findings are inconsistent with a recent study of network segregation from our group (Ballard et al., 2022c) which demonstrated some differences with reproductive stage, suggesting hormones may play impact network segregation. That is, female reproductive aging is associated with declines in 17β-estradiol and progesterone (Burger, 2008; Harlow et al., 2012; Sundström Poromaa and Gingnell, 2014) and Ballard et al. (2022c) demonstrated significant differences between distinct reproductive stages and network segregation (i.e., COTC, DMN, DA, FPTC, and Sa) in females (Ballard et al., 2022c). Of note, Ballard et al.’s (2022c) findings may be driven by age as they also found linear relationships between age and network segregation in the aforementioned networks. Pritschet et al. (2020) demonstrated that estradiol is associated with increasing global efficiency in the DMN and DA networks, whereas progesterone was associated with reduced coherence throughout the brain (Pritschet et al., 2020). While Pritschet et al.’s (2020) findings suggested influences of estradiol and progesterone on network dynamics, it is difficult to directly compare their findings to ours as their study was based on dense sampling in a young female across a menstrual cycle and evaluated a different aspect of network function. However, it broadly demonstrates the purported relationship between brain network organization and sex steroid hormones in the female brain. Although we cannot directly compare our findings with this and other studies, network segregation is a proxy for measuring functional organization in the brain and may loosely be interpreted as such.

Lastly, the combined impact (interaction) of hormone levels (i.e., estradiol, progesterone, and testosterone) on network segregation did not reveal a relationship with network segregation in females. We found interactions between estradiol and testosterone in SSH and SSM networks for raw p-values, but these findings did not survive FDR correction. However, this finding agrees with the study by Moffat (2005), in which they note that testosterone can interact not only with androgen receptors, but also with estradiol receptors, and therefore, its administration can, in some cases, parallel the effects of estradiol on the entire nervous system. An important element in understanding the effects of testosterone on the nervous system is that many of its behavioral and anatomical effects occur after it has been converted to its metabolically active derivatives–estradiol or dihydrotestosterone (Moffat, 2005).

One consideration beyond the scope of the current study is how stages within a female menstrual cycle or menopausal stage could impact brain connectivity. Syan et al. (2017) examined endogenous estradiol, progesterone, and a neuroactive metabolite of progesterone (allopregnanolone) at different menstrual phases and demonstrated an impact of hormone levels in the late luteal phase on resting state connectivity in both cortical and subcortical regions for reproductive aged females (Syan et al., 2017). As mentioned earlier, Pritschet et al. (2020) exhibited an impact of sex hormones on network architecture in tandem with the cycle of a reproductive aged female. Thus, sex steroid hormones have shown a relationship with brain connectivity as it relates to regular menstrual cycle fluctuation. Individual variation exists both in a regular menstrual cycle and menopausal stages (Fehring et al., 2006; Harlow et al., 2012; Boker et al., 2014). Notably, hormone levels are not the sole indicator for identifying distinct reproductive stages. Seminal work in categorizing stages in reproductive aging described principal criteria for determining reproductive stage by changes in cycle regularity and days to years since last cycle (Harlow et al., 2012). Endocrine information such as follicle stimulating hormones, antimullerian hormone, and inhibin-B are used to support categorization by changes in cycle (Harlow et al., 2012). We did not examine menstrual cycle, the above specified endocrine information, menopausal stage, or within individual hormone variance in this investigation. As such, the nature of our analyses and scope of our study may not capture the complexity of female network segregation.

In the context of aging, females did display a significant quadratic relationship between age and network segregation for the CBBG and Sa networks. In both cases, segregation was still increasing through adulthood and highest in midlife with a downturn thereafter. Patterns of connectivity between the cerebellum and basal ganglia have been positively linked to cognitive performance and self-reported balance confidence in older adults (Power et al., 2011; Bernard and Seidler, 2013). Further, in their review Diedrichsen et al. (2009) (Diedrichsen, 2006) implicated the relationship between the cerebellum and basal ganglia as critical for modulating cortical functions such as cognition and relying on subcortical processes. We can see in Figure 3, that both females and males demonstrate an inverted “U-shaped” decline in CBBG network segregation across the span of aging adults. Functionally this pattern in females may be related to the drop in cognition and decreased balance that females experience above and beyond males in aging (Burger, 2008; Pritschet et al., 2020), though we should note that this is speculative as we did not include any behavioral analyses here.

The quadratic age relationship with Sa network segregation in females was somewhat consistent with work from Chan et al. (2014), where they showed a significant quadratic age relationship across female and male participants in “association systems” which included the Sa network, but was aggregated with DMN, FPTC, VA, COTC, and DA (Chan et al., 2014). The Sa network has been associated with “conscious integration of autonomic feedback and responses with internal goals and environmental demands” (Weis et al., 2008). This network has also been conceptualized as an “integral hub” for facilitating communication between the DMN and central executive networks in a triple network model (Van Goozen et al., 1995). Notably, dysfunction in the Sa network has been linked to reduced cognitive performance in older adults (Dimech et al., 2019) and increased connectivity in this network has been linked to Alzheimer’s disease (Morrison et al., 2006). The quadratic relationship seen in females, but not males (see Figure 3), could similarly be highlighting cognitive inefficiencies that have been demonstrated during advanced aging in females (Pritschet et al., 2020).

It is important to emphasize that we did not find significant quadratic age relationships in any additional networks for females, nor any networks in males in our exploratory analyses. These findings are consistent with Lee et al. (2016) study that examined rs-fMRI graph network analysis in the DMN, Sa, and Central Executive network and a quadratic age relationship in both groups of “good” and “poor” cognitive performers (Lee et al., 2016). While Lee et al. (2016) assessed components of both network segregation and network integration (i.e., global efficiency, local efficiency, betweenness centrality, connectivity strength, and nodal degree), their analyses notably varied from ours in that they controlled for gender (Lee et al., 2016). That is, the relationship between quadratic age and segregation in certain networks may be driven by inherent hormonal differences between sexes, but as gender was controlled for in their study that relationship was left unexplored (Lee et al., 2016). Thus, the methodological differences may explain the differences in outcomes relative to what we report here.

Comparisons of model fit demonstrated the quadratic model as a substantially better fit for Sa network segregation and age in females; however, no additional quadratic age models displayed a significantly better fit in our analyses. To our knowledge, associations between quadratic age and network segregation have not been otherwise evaluated or reported. We would suggest that this may be a useful area of investigation in future work. We did not, however, find linear relationships between age and network segregation when examining females or males after FDR correction. Similarly, Grady et al. (2016), did not see linear relationships with age and network segregation, although their examination was specific to the DMN, DA, and FPTC networks (Grady et al., 2016). As stated, earlier Chan et al. (2014) found a significant quadratic age relationship in association systems (Chan et al., 2014); however, these systems were also significantly linearly associated with age. The same study also revealed linear age relationships with network segregation in “sensory-motor systems” which aggregated Hand somato-motor, Visual, Mouth somato-motor, and Auditory within-network segregations (Hausmann, 2005). While we did not replicate Chan et al.’ (2014) findings, we also did not aggregate networks for analysis in the same fashion. Further there are also differences in the data used for analyses with respect to both sample size and length of scans. That is, their sample size was twice as large as our study and scan length was about 5 min per participant. Of note, our approach to data collection (guided by recent scientific advancement and literature) often produces more reliable data (Pannunzi et al., 2017) as we collected 24 total minutes of resting state scans, alternating from anterior to posterior slice collection. Thus, Chan et al.’ (2014) investigation of network segregation in sensory-motor systems may not be appropriate for drawing direct comparisons to our data given methodological advancements in recent years resulting in differences in data collection parameters. Additional studies evaluating linear and quadratic relationships by sex are warranted, particularly with larger samples that include longer resting state acquisition times such as that used here.

Overall, we found quadratic age relationships for females in CBBG and Sa network segregation. We did not find significant impacts of individual or combined hormone levels on within-network segregation in the networks of interest. Future research would benefit from examining inter-individual differences over time to gain more insight into subtle influences of endogenous hormones. Another relevant area of research would be examining individual differences of network segregation over time in those taking hormonal contraceptives or receiving hormone replacement therapy to better understand the dynamic interplay between sex steroid hormones and brain network organization.

This study demonstrated a quadratic relationship between aging and network segregation for the CBBG and Sa networks in females. In both cases, segregation was still increasing through adulthood and highest in midlife with a downturn thereafter. These networks are functionally related to cognitive performance, balance, and integrating autonomic feedback in response to environmental demands (Bernard and Seidler, 2013; Chan et al., 2014; Hausman et al., 2020).

Prior research has shown that the sex hormones can regulate neurogenesis, inflammatory processes, impact network segregation, and may also play a role in regulating cognitive and affective processes, mainly in the aging process (Syan et al., 2017; Foo et al., 2021). Furthermore, the notable variability in hormone levels in certain reproductive stages and cognitive deficits associated with specific reproductive stages did not have an impact on network segregation as was expected (Greendale et al., 2011; Epperson et al., 2013; Weber et al., 2014; Rentz et al., 2017; Taylor et al., 2019; Pritschet et al., 2020).

Future studies could focus on examining participants longitudinally and pairing these types of data with behavioral outcomes.

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 Texas A&M Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

TH, HB, TJ, and SC completed the data collection. TH, TM, HB, and TJ conducted the data analyses under the guidance and supervision of JAB. TH and TM drafted the manuscript with input from all authors and again under the guidance of JAB. All authors contributed to the conceptualization of this project.