Introduction

The detrimental effects of aging on cognition are widely recognized [1]. While these changes sometimes reflect cerebrovascular disease or early stages of a neurodegenerative process such as Alzheimer’s disease, normal, healthy aging is also associated with declines in processing speed, attention and memory, even in the absence of any known vascular disease, obvious cortical insult, or proteinopathy such as misfolding of amyloid or tau [2–4].

In contrast to cognitive decline in older adults, many aspects of emotional functioning are maintained or even improved as part of healthy aging [5–8]. Despite shrinking social networks, decreasing cognitive agility, and declining physical health, older adults demonstrate an emotional well-being that is equal to or better than their younger counterparts. High levels of happiness exist even among centenarians [9], and older adults maintain the ability to generate and experience emotions and empathic responses to others [10, 11].

The retention and in some cases improvement of emotional functioning in the face of well-described decline in general health, social networks and cognitive function has been dubbed the emotion paradox of aging [6–8]. There have been several proposed explanations for this preservation of emotional function. Emotional preservation likely results from a combination of adaptation and circumstance. Older adults may adapt to the ravages of time with a learned ability to make use of attentional strategies. Laboratory-based studies have found that while older adults are more reactive to emotional stimuli and may recover more slowly from both positive and negative emotional events [10, 12], older adults are better able to regulate their emotions through situation selection and reappraisal tactics that enable them to view even negative situations in a positive light [5, 13].

More circumstantial explanations propose that the emotion paradox of aging is a biological consequence of the aging process. Within the brain, relative emotional preservation could be an incidental result of different cerebral regional rates of atrophy over time. For example, some studies have suggested that the anterior cingulate cortex is relatively preserved [14, 15], though such preservation is not universally described [16]. If true, the ACC could contribute to emotional health due to its involvement with selective attention and reappraisal [17]. Another circumstantial explanation for preserved emotional positivity in aging is that positive material may be less cognitively demanding than negative information [18]. As a result, systemic efficiency demands that energy be spent on less exhausting positive information, driving an active regulation of sensory input based on recognized cognitive resources. Such a theory suggests that structures involved with appraisal of information as positive or negative, such as the amygdala [19], or structures involved with emotion-related information processing and learning, such as the nucleus accumbens [20], could be especially important in maintaining a positive outlook during healthy aging. While some have suggested different rates of volume loss within these structures as a potential cause of emotional preservation with aging [6], referencing existing longitudinal volumetric analysis of healthy aging brains [14, 15], direct comparisons of volume loss in brain regions associated with cognition versus those of emotion have not been made.

To further investigate the neuroanatomical foundations of emotional preservation in aging, we compared brain regions historically associated with cognition to those associated with emotion. We applied a cross-sectional analysis to compare volumes of cognitive regions and emotional regions as a function of age, and also performed a longitudinal analysis of regional volumes as a function of time. We hypothesized that in healthy aging, the collective group of regions associated with emotional processing would lose volume more slowly than brain areas associated with cognitive functioning. A slower rate of volume loss in brain structures associated with emotion could potentially explain retained emotional acumen despite cognitive deterioration over time.

Materials and Methods

Regional Selection

While recognizing evidence that the right hemisphere is more involved with emotional processing generally, the functionality between the left and right hemispheres remains complex and poorly understood [23, 24]. Consequently, regions associated with cognition or emotion were selected bilaterally. The following were included as regions predominantly influencing cognition: the middle frontal and lateral frontal gyri, the parietal cortex, and the medial and lateral gyri of the temporal lobes [25–29]. The following were included as regions predominantly influencing emotion: amygdala, rostral anterior cingulate cortex, insula, orbitofrontal cortex (medial and lateral orbitofrontal cortex) and nucleus accumbens [30–35]. When we could not reasonably decide whether a region should be considered cognitive or emotional in nature, we excluded that region from the analysis. For example, the inferior temporal gyrus was not included due to its prominent involvement in semantic dementia, which has pronounced cognitive and emotional changes [36].

In cases where these regions were not predefined from the atlas, regions were combined as needed. The orbitofrontal cortex measurement was obtained by summing the medial and lateral orbitofrontal cortex volumes. The middle frontal measurement summed the caudal middle frontal and rostral middle frontal volumes. The lateral frontal measurement summed the pars opercularis, pars orbitalis, and pars triangularis volumes. The parietal cortex summed the inferior parietal, superior parietal, precuneus, and supramarginal volumes. The lateral temporal region summed the superior and middle temporal gyrus volumes. The medial temporal region summed the hippocampal, entorhinal and parahippocampal volumes. (Fig 1, Table 1)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Regions included in cognitive and emotional collectives.

For a listing of how these regions were developed, see Table 1.

https://doi.org/10.1371/journal.pone.0158187.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of regions used and source atlases.

https://doi.org/10.1371/journal.pone.0158187.t001

Results

Subject characteristics for cross-sectional and longitudinal analyses are outlined in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline Demographic Data of Healthy Participants by Analysis Type.

https://doi.org/10.1371/journal.pone.0158187.t002

Cross-sectional Analysis

Age was associated with decreased brain volume both in regions of cognition ( = -0.05, R² = 0.61, p < 0.001, CI [-0.06, -0.04]) and in regions of emotion ( = -0.04, R² = 0.64, p < 0.001, CI [-0.05, -0.03]) (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Cross-Sectional Analysis of Regions of Emotion and Cognition.

Absolute brain volume in collective regions associated with cognition compared to those associated with emotion as a function of healthy aging research participants’ age in years. There was an estimated 0.51% per year absolute volume loss in regions of emotion and 0.60% per year absolute volume loss in regions of cognition. Note that while absolute volumes are shown here for clear visualization of the two groups, for statistical comparison of the association between age and volume between regions of emotion and cognition, volumes were converted to Z-scores in order to normalize the baseline difference in absolute volume between groups. No significant proportional difference was found between the cognitive or emotional regions in volume loss as a function of age.

https://doi.org/10.1371/journal.pone.0158187.g002

Cross-sectional analysis showed no significant interaction between age and region type ( = 0.01, R² = 0.62, p = 0.096, CI [-.002, 0.03]) after adjusting for intracranial volume and gender. While a p value < 0.1 might be considered a trend, any effect size for the interaction is very small (eta² = 0.005). Education, vascular risk factors and white matter hyperintensities did not contribute significantly to the model (all p > 0.20).

On subgroup analysis, the rostral ACC had the smallest association between volume and age, with the most significant interaction coefficient found in the medial temporal lobe (Table 3). Comparisons between groups in the interaction between age and volume suggest the ACC has a significantly smaller association between age and volume when compared with the amygdala ( = 0.04, p<0.001, CI[0.02, 0.06]), nucleus accumbens ( = 0.06, p<0.001, CI [-0.08, -0.04]), parietal lobes ( = 0.03, p = 0.001, CI {-0.05, 0.01]), medial temporal lobes ( = = -0.04, p<0.001, CI[-0.07, -0.03]), and lateral temporal lobes ( = = -0.04, p<0.001, CI[-0.06, -0.02), after adjusting for multiple comparisons. In contrast to the relative preservation of the ACC, the nucleus accumbens had a stronger relationship between age and decreasing volume when compared to the insula ( = = 0.05, p<0.001, CI [0.03, 0.07]), the orbitofrontal cortex ( = = 0.04, p<0.001, CI[0.02, 0.06]), the middle frontal cortex ( = = 0.04, p<0.001, CI[0.02, 0.06]), as well as the ACC. The medial temporal lobe differed significantly from the insula ( = = -0.04, p<0.001, CI[-0.06, -0.02]) and middle frontal cortex ( = = -0.04, p<0.001, CI[-0.06, -0.02]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Regression Coefficients for Sub-Regions by Analysis Type.

https://doi.org/10.1371/journal.pone.0158187.t003

Longitudinal Analysis

Preliminary data inspection showed that a longer period of time between scans was positively correlated with volume in the amygdala ( = 1.12, p = 0.002, CI [0.42, 1.82]) as well as the nucleus accumbens ( = 1.35, p < 0.001, CI [0.65, 2.1]) at time 1. Baseline volume was therefore included as a covariate in further longitudinal analyses. Again, education, vascular risk factors and white matter hyperintensities did not contribute significantly to the model (all p > 0.20). Mixed effects modeling confirmed the non-independence of longitudinal volumetric data between individuals ( = 0.95, p < 0.001, CI [0.8, 1.1]).

Significant volume loss was found over time in regions of emotion ( = -0.05, p = 0.003, CI [-0.08, -0.03]) as well as cognition ( = -0.07, p < 0.001, CI [-0.09, -0.02]) (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Longitudinal Analysis of Brain Regions of Emotion and Cognition.

Absolute volume of collective cognitive and emotional regions in healthy aging participants who were scanned over time. While volumes were converted to Z-scores for rate comparisons analysis in order to adjust for baseline differences in absolute volume between regions of cognition and emotion, absolute brain volumes are here depicted for visualization of individual slopes. In absolute volume, regions of cognition lost approximately 0.65% per year, and regions of emotion lost approximately 0.46% per year. No statistically significant proportional difference was found between brain regions of cognition and emotion in rates of volume loss over time.

https://doi.org/10.1371/journal.pone.0158187.g003

When comparing regression slopes, no significant difference in the time/volume interaction term was found between cognitive and emotional collectives ( = 0.01, p = 0.70, CI [-0.04, 0.06]). As with the cross-sectional studies, sub-regions were investigated as well to assess their potential contribution to the observed trend. In this analysis, two regions of emotion, the nucleus accumbens and the amygdala, demonstrated statistically significant positive volume changes with time. Including baseline volumes as a covariate did not alter this trend. Among regions of cognition, the medial temporal lobe did not show significant age-related volume loss.

Discussion

The purpose of this study was to examine whether differing rates of volume loss between brain regions commonly associated with emotion as opposed to cognition might contribute to preserved emotional health with aging. We used two different study designs (cross-sectional and longitudinal) to approach the question of how cognitive and emotional brain regions change with time. Our primary analyses found no difference between emotional and cognitive brain volumes as a function of age or time’s passage.

One of the strengths of this study was our use of both cross-sectional and longitudinal approaches. The cross-sectional design included a larger sample size. The longitudinal approach examines fewer individuals, but has the benefit of allowing better causal inference between volume and time, rather than relying on age as a proxy for the passage of time.

Subgroup analyses within both the longitudinal and cross-sectional studies support previous reports of relative preservation of the rostral anterior cingulate cortex, with smaller rates of decline compared to many other cortical regions (Table 3). Relative preservation of the ACC has been previously connected with the emotion preservation of aging [6], and with the positivity effect in particular [17]. The term “positivity effect” describes the tendency of elders to preferentially attend to positive over negative stimuli in a variety of contexts, including facial expressions, word lists [41], and emotional images [42–44]. Elders tend to recall more positive than negative stimuli in working, short-term, and autobiographical memories [43, 45, 46]. The use of such selective attention to avoid of negative stimuli another common emotional regulatory strategy used by elderly adults, which may relate to this positivity effect [6, 15, 17, 47]. Brassen and colleagues have correlated the anterior cingulate with such selective attention using functional connectivity. While the region identified by Brassen and colleagues is at the dorsal-most part of the ACC included in our partitioning, in an area many would associate with cognitive processes, the entire cingulate, both ventral and dorsal, has been shown to be relatively preserved in previous imaging studies [14, 15]. Furthermore, a more ventral portion has been associated with the positivity effect by Ritchey et al during a task demanding concentrated focus, akin to positive reappraisal [48].

The cross-sectional analysis also showed a significantly greater negative association between age and volume in the nucleus accumbens compared to other brain regions such as the rostral ACC, insula, orbitofrontal cortex, and the middle frontal lobe. Previous studies have also demonstrated age-related volume loss in the nucleus accumbens [49]. Unlike the ACC, little has been written on the accumbens and preservation of emotional functioning in aging, with an emphasis instead placed on decision-making [50, 51]. While tempted to speculate on how loss of this structure might place a greater processing burden on cortex, thereby driving a preference for positive emotion [18], the nucleus accumbens is a complex structure, and consequences of its loss are also likely to be complex. For example, while the nucleus accumbens has been traditionally tied with reward systems, it also mediates aversive reactions [52]. Further research will be needed to correlate degeneration of the nucleus accumbens with cognitive and emotional changes in healthy aging.

Volume in brain regions of cognition and emotion declined as a function of age in the cross-sectional study, and as a function of time in the longitudinal study, to a degree similar to overall cortical volume loss demonstrated in prior studies [53]. While the collective trajectories of cognitive versus emotional brain volumes are consistent between our cross-sectional and longitudinal studies, some sub-regions appear discordant on sub-regional analysis. Chiefly, the medial temporal lobe, amygdala and nucleus accumbens appear to gain volume over time. Previous studies have shown conflicting evidence of volume changes in these regions over time [14, 54]. On the other hand, Freesurfer has been previously noted to have questionable scan-rescan reliability in the parcellation of small, subcortical structures [55]. The apparent sub-regional volume gain may reflect a greater amount of variability in volumetric analysis of relatively small or medial structures using our techniques, which may be especially important with the smaller sample size found in the longitudinal group. In the comparison between cognitive and emotional regions collectively, any consequent bias present in the emotion regions would likely contribute to apparent emotional preservation. Failure to reject the null hypothesis in spite of such potential bias bolsters the negative finding between cognitive and emotional volume changes collectively. We consider the cross-sectional data to be more reliable within our study, and therefore limited between-subgroup analysis to the cross-sectional portion.

Other potential weaknesses to this study design include the subjective nature of regions selected as “emotional” or “cognitive.” As we know of no widely accepted battery of emotional functioning, this study instead relies on established correlations found in scientific literature. Because cognitive and emotional changes with age, as well as their neuroanatomical underpinnings, are complex and interwoven [56], selection of regions that could be congregated reliably into either emotional or cognitive group was a particular challenge of this research. While we recognize the imperfection of our regional groupings, absolute parcellation of cognitive and emotional functions is impossible, necessitating reliance on best judgment. Another potential limitation is the possibility that some of our healthy population in fact has histopathological changes associated with neurodegenerative disease. Our definition of this population as healthy agers must be recognized as strictly clinical.

In addition to our collective approach and the use of two different study designs, strengths of this research include a large sample size and relative consistency of magnetic resonance imaging and analysis techniques within a healthy aging population. Future studies could examine differential rates of volume loss within neuroanatomical regions over more visits to allow for imputation, as well as correlate active measures of emotion and cognition to confirm corresponding neuropsychological performance. Functional connectivity or other techniques could also be explored.

Conclusions

In summary, our results suggest that during healthy aging, brain regions commonly associated with emotional processing collectively undergo a rate of decline proportional to those of cognitive regions. Clustering all regions of emotion versus those of cognition does not provide sufficient specificity to provide a biological explanation for emotional preservation with aging. More detailed cross-sectional comparisons between subregions suggests that some regions of emotion such as the anterior cingulate cortex may be truly spared, whereas others such as the nucleus accumbens may be particularly susceptible to age-dependent volume loss. Further studies may further explore how these patterns of volume loss contribute to emotional functioning by correlation with dedicated cognitive and emotional tests.

Supporting Information

Acknowledgments

We thank the staff and faculty at the Memory and Aging Center at the University of California, San Francisco as well as all of the research participants who volunteered for this study. Special thanks to John Neuhaus, PhD and Michael Edge, PhD for statistical guidance through this research