Structural complexity is negatively associated with brain activity: a novel multimodal test of compensation theories of aging

Highlights

• Fractal dimensionality (FD) was measured in cortical and subcortical structures.

• We provided a novel test of the atrophy-compensation hypothesis.

• Cortical FD was negatively associated with brain activity during memory retrieval.

• Lower FD and greater brain activity were associated with poorer cognition.

• Findings argue against the modal perspective of the atrophy-compensation hypothesis.

Abstract

Fractal dimensionality (FD) measures the complexity within the folds and ridges of cortical and subcortical structures. We tested the degree that FD might provide a new perspective on the atrophy-compensation hypothesis: age or disease-related atrophy causes a compensatory neural response in the form of increased brain activity in the prefrontal cortex to maintain cognition. Brain structural and functional data were collected from 63 middle-aged and older adults and 18 young-adult controls. Two distinct patterns of FD were found that separated cortical from subcortical structures. Subcortical FD was more strongly negatively correlated with age than cortical FD, and cortical FD was negatively associated with brain activity during memory retrieval in medial and lateral parietal cortices uniquely in middle-aged and older adults. Multivariate analyses revealed that the lower FD/higher brain activity pattern was associated with poorer cognition—patterns not present in young adults, consistent with compensation. Bayesian analyses provide further evidence against the modal interpretation of the atrophy-compensation hypothesis in the prefrontal cortex—a key principle found in some neurocognitive theories of aging.

Introduction

As the world's aging population grows, so too does the risk for cognitive decline and age-related neurodegenerative disorders such as Alzheimer's disease (AD). Understanding and preventing such declines in cognition is critical to increase the length of time that one leads a happy and independent life (Rowe and Kahn, 1987). Over the past 3 decades, neuroimaging has been used as a tool to understand how age differences in the brain might inform the causes and consequences of cognitive aging (for a brief review, see Park and McDonough, 2013). This literature has generally revealed declining brain structures (Fjell et al., 2014) alongside patterns of both higher and lower brain activity (for meta-analyses, see Li et al., 2015; Spreng et al., 2010). In the present study, we took a multimodal approach using a relatively novel measure of brain structure, fractal dimensionality (FD), to investigate the relationship between brain structure and function in the aging brain.

A prominent hypothesis arising from the past 15 years of neuroimaging research in aging is that brain atrophy, due to advanced aging or disease, causes an increased neural response in an attempt to maintain cognition (for review, see (Cabeza et al., 2018). Although lower brain activity with aging also has been widely found, these patterns have been intuitively linked to a lack of brain maintenance (Nyberg et al., 2012), neural inefficiency (Logan et al., 2002), or decline in neural distinctiveness (Li et al., 2001). More counterintuitive and controversial is the notion of compensation in which atrophy in aging adults, largely in the prefrontal cortex (PFC), is offset by an increase in brain activity in nearby or contralateral PFC, and some research has extended this notion to the lateral parietal cortex (LPC) (Cabeza, 2002, Greenwood, 2007, Park and Reuter-Lorenz, 2009, Reuter-Lorenz and Cappell, 2008). The PFC often is implicated in executive control processes to flexibly adapt to ongoing task demands across many tasks (e.g., Vincent et al., 208; Power and Petersen, 2013). One of the first studies to document higher age-related PFC activity was in an episodic memory task (e.g., Cabeza et al., 2002), and subsequent research investigating compensatory activity also has used episodic memory tasks during either encoding or retrieval (Brassen et al., 2009; Cabeza et al., 2002; Düzel et al., 2011; Persson et al., 2012; Pudas et al., 2013; Rajah et al., 2011). Supporting these single studies, quantitative meta-analyses have confirmed reliably higher brain activity in the PFC in older relative to younger adults at memory encoding (Li et al., 2015, Spreng et al., 2010) and retrieval (Li et al., 2015, Spreng et al., 2010). Similar findings also have been found in AD (Schwindt and Black, 2009). Together, these studies support part of the atrophy-compensation hypothesis.

However, few studies have provided specific evidence for the association between smaller brain structures (e.g., gray matter density or volume loss) and higher brain activity in the PFC or LPC among aging adults. For example, Kalpouzos et al. (2012) found that the higher brain activity in PFC and LPC in older than younger adults during memory retrieval was eliminated after controlling for gray matter density in those regions, but not in other frontal and parietal regions (i.e., the left dorsomedial PFC and right LPC). This study provides both support for and against the general notion that atrophy should be associated with elevated brain activity. Similarly, Tyler et al. (2010) found that lower gray matter density in the left PFC and left temporal cortex was associated with higher activation in the right homologous regions during a language comprehension task in older adults. However, because gray matter density in each of those regions also was negatively correlated with brain activity in other frontal and temporal regions, these relationships appeared to be global rather than specific to nearby or contralateral regions. Other studies purportedly providing evidence for the atrophy-compensation hypothesis (1) assessed structure-function relationships only indirectly (Colcombe et al., 2005, Düzel et al., 2011, Persson et al., 2012, Thomsen et al., 2004), (2) found positive (not negative) structure-function relationships (Brassen et al., 2009; Rajah et al., 2011), (3) did not find a relationship between brain structure and function (Pudas et al., 2013), or (4) found relationships between brain function and white matter pathways using DTI (Daselaar et al., 2013; Persson et al., 2006), the latter of which can be difficult to classify as nearby or contralateral given the long distance of the fiber bundles. The lack of empirical support for the atrophy-compensation hypothesis is surprising given its wide-spread influence across multiple neurocognitive aging theories (Cabeza, 2002, Greenwood, 2007, Park and Reuter-Lorenz, 2009, Reuter-Lorenz and Cappell, 2008).

One possibility is that traditional measures of brain structure (e.g., gray matter density or volume loss) may not capture the type of atrophy that serves as a catalyst for neural compensation. FD, a measure of structural complexity, has been shown to be highly correlated with chronological age and a dementia diagnosis. FD quantifies fractal patterns, or irregularities, of cortical or subcortical surfaces similar to calculating the complexity of continental coastlines (Mandelbrot, 1967). One of the earliest studies using this method demonstrated that FD can provide better sensitivity to dementia-related differences in brain structure than thickness and gyrification (King et al., 2009, 2010). King et al. (2010) additionally found that FD was more strongly correlated with global cognition than other structural measures. In a series of studies, Madan and Kensinger extended this method to cognitively normal adults across the adult lifespan and also found higher correlations with chronological age and are more reliable in test-retest assessments than conventional measures (Madan and Kensinger, 2016, 2017b; see also Liu et al., 2020). Subsequent studies showed that these benefits can be found with improved precision in more localized regions, including cortical parcellations (Madan and Kensinger, 2018) and subcortical regions (Madan, 2019; Madan and Kensinger, 2017a).

The present study had 3 primary goals. The first goal was to provide a novel test for the basic premise of the atrophy-compensation hypothesis that lower brain integrity is associated with higher brain activity in a sample of middle-aged and older adults. We used FD as a proxy for brain integrity and task-related functional magnetic resonance imaging (fMRI) during encoding and retrieval in a paired association task to assess potential negative associations with brain activity. In accordance with the atrophy-compensation hypothesis, lower FD should be associated with higher task-related brain activity in the PFC and LPC. We also predicted that any such negative associations would be stronger at retrieval than encoding because of the greater task demands during this phase (Mandzia et al., 2004; McDonough et al., 2013). All fMRI analyses attempted to minimize vascular confounds to the blood oxygen level dependent (BOLD) signal by using a participant-specific hemodynamic response function (Handwerker et al., 2004; Huettel et al., 2001) and scaling the contrasts using resting state fluctuation analyses (Kalcher et al., 2013; Kannurpatti et al., 2011). We conducted additional control analyses to ensure that potential negative structure-function relationships did not also occur in a sample of healthy younger adults—a sample in which atrophy does not yet occur. The second goal was to test whether individual differences in FD would be associated with higher risk for dementia using a cumulative dementia risk score in middle-aged and older adults (McDonough et al., 2019). Given the previous associations with FD and AD (King et al., 2009, 2010), we predicted that higher dementia risk would be associated with lower FD in the medial temporal lobe (MTL) and regions within the default mode network, consistent with previously found patterns of atrophy in AD (e.g., Buckner et al., 2005). The third goal was to link the structure-function patterns to cognition. Although finding a positive correlation between higher brain activity and better cognitive performance intuitively captures the notion of successful compensation (Cabeza et al., 2018), some perspectives suggest between-subject correlations are not necessary, especially in the cases of attempted rather than successful compensation (Dennis and Cabeza, 2012). To the extent that such relationships with cognition are indeed compensatory, then similar relationships should not also be present in a healthy young adult sample—a possibility that we tested.