Background
Aging is becoming an increasingly pressing issue, with estimates suggesting that by 2050, 16% of the world’s population will be over 60 years old [
1]. With age comes a decline in cognitive functions such as episodic memory (EM) [
2,
3], which can significantly impact the quality of life of older individuals. It is, therefore, of paramount importance to understand the biological basis of EM by investigating the association between EM and brain aging and to use such evidence for the early detection of neurodegenerative diseases.
EM involves a wide range of brain regions. The most consistent findings indicate a positive correlation between EM and the gray matter volume (GMV) or the thickness of gray matter in the medial temporal lobe (MTL) [
4‐
8]. The age-related decline in EM has been associated with decreased GMV in specific regions of MTL, including the hippocampus [
5,
9], especially the right hippocampus [
10,
11]; the middle and inferior temporal cortex [
12]; entorhinal cortex [
7]; and amygdala [
12,
13]. In addition to MTL, other implicated regions include left superior temporal cortex [
10] and lateral prefrontal cortex (lPFC) [
14].
Three factors have been found to moderate the relation between EM and the brain: type of EM, age, and sex. First, two types of EM have been commonly studied: verbal (VEM) and spatial (SEM). They have been found to be associated with different brain regions. Suri, Topiwala [
15] found that VEM was correlated with functional connectivity between temporal and frontal nodes, whereas SEM was correlated with functional connectivity between hippocampal and parietal regions. The differential neural bases for VEM and SEM may be linked to two neural systems—the anterior temporal (AT) system and the posterior medial (PM) system—that interact with the hippocampus to process different types of information [
16]. The AT system is essential for familiarity recognition, emotional processing, social cognition, and semantic representation, whereas the PM system plays a crucial role in context memory retrieval and spatial navigation. Consequently, both systems are expected to be involved in EM, with the AT system being more relevant to VEM and the PM system being more relevant to SEM. However, there is limited empirical research exploring the relationship between these two types of EM and their corresponding neural systems in older adults.
Second, as mentioned earlier, EM declines with age, so it is likely that the neural bases of EM also vary by age. Some longitudinal studies have revealed that a higher rate of EM decline is associated with greater volume loss in the MTL [
17,
18] and occipital lobes [
12], and with the annual atrophy rate of the hippocampus [
6,
9,
10]. These studies have focused on the covariation of age-related EM and brain structural changes, but they have not investigated whether this relationship varies across different age groups. However, one longitudinal study [
6] has focused on the role of age in the association between GMV of the hippocampus and EM, and found that the association was significant in older adults (aged 65–80), but not in middle-aged adults (aged 55–60). Results from cross-sectional studies have also revealed similar effects. One study found that maintaining memory function throughout the adult lifespan may mitigate the typical age-related volume loss in the hippocampus. Furthermore, an age-by-EM interaction term was found to better predict GMV, suggesting that the predictive effects of EM on GMV are modulated by age across different age groups [
19]. The broader literature on the association between gray matter structure and cognition has shown the pattern that the association is stronger in older adults than in younger individuals [
20].More research is needed to confirm and understand the role of age in the association between EM and GMV.
Finally, sex is an important factor in EM and its neural basis. Some studies have identified differences in GMV between sexes, which can serve as a foundational basis for explaining cognitive disparities [
21‐
24]. Generally, females tend to perform better than do males on verbal tasks, whereas males perform better than do females on visual-spatial tasks [
25‐
27], Several studies have provided limited neurobiological underpinnings for these differences [
28‐
30]. This difference may be due to variations in the underlying grey matter structures of the male and female brains. Specifically, compared to females, males tend to have larger volumes in the medial frontal cortex, left inferior parietal cortex (IPC), amygdala, and hypothalamus, as well as higher synaptic density in the neocortex of the temporal lobe [
31]. In contrast, females have larger volumes in the frontal and temporal lobes [
31] and the Broca’s (in the dorsolateral prefrontal cortex) and Wernicke’s (in the superior temporal cortex) areas [
32]. Male advantage in left inferior parietal cortex (IPC), an area related to spatial relationship understanding, may explain their superior spatial information processing abilities [
33], including SEM. Female advantage in the language areas (Broca’s and Wernicke’s areas) may explain their better language abilities, including VEM. Moreover, given the importance of both age and sex in EM and brain aging, it is important to consider these two factors together. Indeed, previous studies have also found that compared to females, males have a higher rate of volume loss in certain brain regions that are important for EM, including the hippocampus, amygdala [
34‐
36], and frontal lobe [
35]. Furthermore, research has shown that there is a significant interaction between age and sex in the annual percentage change of gray matter ratio (GMR) [
37].
To investigate the correlation between EM and GMV variations associated with age among older individuals. we used cross-sectional data from adults aged 55–90 years from the Beijing Aging Brain Rejuvenation Initiative (BABRI). Our study first identified brain regions whose GMV showed high correlations with EM in the whole sample (and for males and females separately) and then used the sliding window approach to examine the age-related association between GMV and EM in these brain regions. The sliding window approach’s advantage is to allow us to analyze cross-sectional data to yield a developmentally continuous picture [
38]. We specifically tested the following hypotheses. First, the two types of EM would have some shared gray matter structural bases, such as the Hippocampus MTL and lPFC, but also distinct neural bases, such as VEM being associated with Broca’s and Wernicke’s areas and SEM being associated with the IPC. Secondly, the relationship between EM and GMV is correlated with age, and supports the existence of sex differences in gray matter structure for various types of EM.
Discussion
Our study examined age-related differences in EM and GMV and their association in a Chinese sample of older community residents. The results indicated a negative correlation between EM and age, and that relevant brain regions included areas in the medial temporal lobe, frontal lobe, other neocortex, and subcortical regions. More importantly, we found that the relationship between EM and GMV depended on age, type of EM, and brain area/system. For the total sample, positive correlations were found between both types of EM and GMV in a number of brain regions located in the temporal and frontal lobes, and these associations positively correlated with age. When analyzed by sex, the association between VEM and GMVs in the insula and parietal regions became stronger with age for females but not males, whereas the association between SEM and GMVs in the parietal and occipital regions became stronger in males but not in females. When analyzed at the brain system level, we found that for males, there were positive age-related correlations in the associations between EM (either VEM or SEM) and the GMV of either the AT system or the PM system. For females, however, both VEM’s and SEM’s association with GMV became stronger with age in the AT system, but VEM’s association with GMV did not show a significant correlation with age in the PM system. The sex-specific patterns of age-related correlations in the association between two types of EM and specific regional GMV help elucidate the structural basis of EM for males and females in older individuals.
EM exhibit a negative correlation with age in older individuals is consistent with previous studies [
40‐
42]. We further found that although females showed significantly lower SEM but higher VEM than did males. However, there was no significant difference between sexes in the correlation coefficient of EM scores and age for the two types of episodic memory, thus confirming previous studies [
43]. These results suggest that sex differences in performance on different types of EM tasks are due to inherent differences between sexes, not the age [
25,
26].
Our findings of the neural basis of EM are consistent with previous studies. Relevant brain areas included the medial temporal lobe and frontal lobe, as well as some regions of the parietal and occipital lobes, and subcortical areas [
5,
7,
9,
11‐
14]. In terms of the two types of EM, VEM involves language pathways and auditory processing in the temporal and frontal lobes [
44], whereas SEM involves visual-spatial processing in the parietal-frontal and parietal-medial temporal pathways [
45]. We also found sex differences in the neural basis of EM. Specifically, compared to females, males show greater activity in brain regions associated with both VEM and SEM, consistent with a study by Sang, Chen [
46].
In terms of the association between EM and GMV, our study yielded several important findings. First, for the total sample, the EM-GMV association typically becomes stronger with age. Specifically, the association between GMV and VEM positively correlated with age significantly in a wide range of regions in the frontal and temporal lobes, and a small number of regions in the parietal lobe, occipital lobe, and subcortical lobes. These regions are responsible for normal cognitive performance and have been linked to VEM in previous studies [
4‐
8,
11]. Similarly, the association between GMV and SEM positively correlated with age significantly in a similar regions as those for VEM, but involved more areas in the parietal lobe, occipital lobe, and subcortical regions than did VEM, perhaps because SEM involves more visual and spatial information than does VEM [
45]. Furthermore, our findings demonstrate a greater number of brain regions positively correlated with SEM comparison to VEM, specifically within the parietal and occipital lobes. These areas are associated with the processing of visual spatial information. Gorbach, Pudas [
6] speculated that after the age of 65, both brain biomarkers and cognitive measurements show more pronounced changes, leading to stronger associations between these changes. Furthermore, some studies have also found that age can induce variability in gray matter structure [
47,
48]. In our supplementary analysis, we computed the inter-subject variability of GMV within each window. There is a growing trend in variability with age, although it did not reach statistical significance (Fig. S11). Our findings provide further supplementation to this perspective, indicating a positive correlation between age and association between GMV and EM, which manifests with specificity in various types of EM.
Second, there are notable differences between males and females in the association between EM and GMV in older individuals. Specifically, for VEM, males show significantly lower age-related correlation between GMV-VEM in many regions in the frontal and parietal lobes, as well as subcortical areas (e.g., right superior frontal gyrus, bilateral inferior frontal gyrus, triangular part, left precuneus, bilateral insula, etc.). However, in the temporal lobe, males exhibit more regions with significantly higher correlations with age (e.g., right amygdala, right fusiform gyrus, right middle temporal gyrus, and left middle temporal gyrus). In the regions where females show significant differences in correlations, many of them are associated with language, such as bilateral inferior frontal gyrus, triangular part, left superior parietal gyrus and left inferior parietal gyrus [
49,
50]. For SEM, males exhibit significant positive correlations between GMV-VEM and age in most regions of the parietal and occipital lobes (e.g., bilateral superior parietal gyrus, left superior parietal gyrus, left angular gyrus, right precuneus, right lateral occipital cortex, right calcarine, left cuneus, bilateral lingual gyrus, and left superior occipital gyrus). In these areas, there is a strong association with visual and spatial information processing [
51‐
53]. On the other hand, females do not show significant GMV-VEM correlations in these regions, making it impossible to compare correlation coefficients. Instead, females demonstrate more regions with significantly lower correlations (e.g., right hippocampus, right superior temporal gyrus, right middle temporal gyrus). To further probe sex differences, we also examined age-related variations in GMV in the two sex groups (see Supplementary Material Table S9). Overall, for both sexes, the GMV in the majority of brain regions were negatively correlated with age, except for the right putamen and bilateral pallidum. Direct comparisons between the two sexes showed that, in most regions (14 regions), a greater extent of negative correlation with age compared to females (as indicated by larger absolute values of males’ t values in the regression analysis). This sex difference in age-related variations in GMV may account for the greater GMV-EM correlations for males as compared to females, which in turn may explain sex differences in VEM and SEM. The differences in these correlation patterns may stem from varying differences in GMV to support different cognitive advantages in older individuals of different sexes, resulting in distinct GMV-EM association patterns. These can be validated in longitudinal studies. The potential causes of sex differences may stem from various factors, including evolutionary demands, hormonal and endocrine influences, and the heterogeneity of older individuals. Firstly, from an evolutionary standpoint, these differences likely have an evolutionary basis related to sexual selection pressures [
27,
54]. Secondly, sex hormones modulate brain structure and function through different receptors or pathways, affecting mechanisms such as neurogenesis, dendritic spine density, and synaptic plasticity [
27]. Thirdly, brain structural variability between individuals tends to increase with age [
47,
48]. In our supplementary analysis, we conducted correlation analyses between the variability of GMV and age among participants of different sexes. The results indicated a significant positively correlation in variability with age among males, whereas females exhibited the opposite trend. This suggests that heterogeneous differences may not be the sole age-related factor influencing the relationship between GMV and EM. We further explored the impact of different overlaps on the GMV-EM relationship with age based on sliding windows. Specific results can be found in Fig S12 (70% overlap) and Fig. 13 (60% overlap). We observed that smaller overlaps resulted in a reduced number of delineated windows along the age axis. Despite the challenge posed by the decreased window count, the positive correlation between GMV-EM and age remained stable.
Finally, we investigated sex differences in age-related variations in the association between EM and GMV at the level of two brain systems (AT and PM). At the entire sample, significant positive correlations with age were observed for both VEM-GMV and SEM-GMV associations with the AT and PM systems. However, VEM-GMV association with the PM system did not exhibit a significant correlation with age. Sex-specific analyses showed that in males, both types of EM demonstrated a significant positive correlation with age in both systems. In females, both types of EM displayed a significant positive correlation with age in the AT system, while VEM’s association with the PM system did not significantly correlate with age. Notably, the SEM-GMV association with the PM system did not yield significant results across age groups. Differences in correlation coefficients revealed that females had significantly higher correlations with age for both types of EM in the AT system compared to males. Conversely, in the PM system, the GMV-VEM relationship showed significantly lower correlations with age in females compared to males. These findings suggest sex differences in the age-related patterns of the relationships between EM types and GMV in the AT and PM systems, with males receiving stronger support from both systems, and females showing a stronger inclination toward the GMV of the AT system. The AT system, associated with processing verbal information, and the PM system, inclined towards handling visual-spatial information [
16]. Our analysis focused exclusively on regions positively correlated with EM. Building upon the observed sex-specific patterns mentioned earlier, it may unveil that the consistent predominance of stronger SEM in males is linked to a more robust correlation of the GMV of the PM system with age. Conversely, the stable manifestation of stronger VEM in females is associated with a stronger correlation of the GMV of the AT system with age. We offer a new perspective on previous functional studies of the AT and PM systems and EM from the standpoint of GMV [
55,
56]. In summary, our findings suggest a higher correlation between GMV-EM in regions associated with the EM in the elderly population, indicating that GMV is more supportive of cognitive performance in older individuals.
In summary, we observed stronger correlation between EM and specific region GMV in the elderly population. Previous research indicates that one of the hallmarks of brain aging is dedifferentiation [
57], suggesting a convergence of function across a wide range of brain areas. Our findings reveal that with older age GMV exhibits a broader and stronger correlation with EM, consistent with the phenomenon of brain functional dedifferentiation. Further exploration of this phenomenon necessitates utilizing multimodal data in subsequent studies to investigate its underlying mechanisms.
Several limitations of this study need to be mentioned. First, our age-related conclusions were based on cross-sectional data. They should be validated with large-scale longitudinal data in future studies. Second, the male sample in our data had more years of education than did the female sample. We used statistical control to handle this issue, but we did not explore how sampling methods and education policies affected the characteristics of our sample and the results. Thirdly, our study did not encompass additional variables such as lifestyle habits, diseases, biological information, etc., which can influence the causes of cognitive aging. These valuable variables will be taken into account in future research to provide a more comprehensive characterization of the relationship between cognition and brain structure.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.