These data demonstrate that individual differences in performance on an associative episodic memory task can be predicted by individual differences in intrinsic hippocampal-posteromedial cortical connectivity during the resting state. Specifically, the intrinsic connectivity between regions that are maximally engaged during successful memory performance was examined during quiet wakefulness prior to performance of the memory task. Stronger correlations between the spontaneous fluctuations in these regions at rest were associated with better memory performance. The specificity of the finding was enhanced by demonstrating that the hippocampal-PPC intrinsic connectivity also significantly correlated with the episodic memory measure on neuropsychological tests, but not with any of the non-memory domain measures. Exploratory whole brain analyses confirmed the spatial specificity of these findings that the hippocampal connectivity with the PPC and other posterior regions of the default network was related to memory performance in these older adults. The present results provide further evidence that the coherence of spontaneous fluctuations in the intrinsic activity of components of the default network is important for episodic memory, and support the view that one of the functions of this large-scale brain network is to subserve episodic memory abilities. More generally, these findings also lend the support to the hypothesis that the spontaneous fluctuations in intrinsic brain activity may be relevant to individual difference in behavior among older individuals.
Although the literature on resting functional connectivity has been focused primarily on positive correlations between the posterior cingulate and the hippocampus (Greicius et al., 2004
; Kahn et al., 2008
), and the majority of the higher performing older subjects in our study demonstrated positive correlations between the hippocampus and the PPC, as well as other default network regions, we did observe that some low performing older subjects demonstrated negative or very low positive correlation values. There are several possible reasons for this observation. Previous work has also shown that aging is associated with reduced positive correlations between the hippocampi and several regions of the default network, and indeed that some older subjects demonstrate negative correlations, which typically results in low correlation strength among older subjects compared to young (Andrews-Hanna et al., 2007
; Hedden et al., 2009
). Our study suggests that lower positive correlations and especially negative correlations between the hippocampus and PPC are associated with poorer memory performance. Furthermore, as we defined the functional ROI based on the regions maximally engaged in successful encoding, which requires activation in the hippocampi and deactivation in the PPC, we may be especially detecting regions in which poor performing older subjects demonstrate impaired ability to rapidly modulate activity in these regions. In fact, two recent papers have suggested that the modulating from encoding deactivation to retrieval activation in the PPC may be critical to memory success (Daselaar et al., 2009
; Kim et al. in press). We postulate that our finding of decreasing positive correlation between the hippocampi and PPC may be indicative of failing integrity of memory systems in some older subjects.
It has long been noted that humans demonstrate notable inter-subject variability in cognitive task performance. Previous fMRI studies revealed that such performance variability might be related to the differences in regional brain activity elicited by a given task (Ress and Heeger, 2003
; Sapir et al., 2005
; Wagner et al., 1998
). It has been previously demonstrated that the variability in evoked brain activity can predict whether or not visual stimuli will be detected (Ress and Heeger, 2003
), and whether verbal stimuli will be remembered. However, accumulating evidence has shown that performance variability in human behavior may also be determined by ongoing fluctuations of intrinsic activity in regions of corresponding functional networks (Boly et al., 2007
; Eichele et al., 2008
; Fox et al., 2007
; Hesselmann et al., 2008a
; Hesselmann et al., 2008b
; Wig et al., 2008
). For example, spontaneous fluctuations of BOLD signal in motor cortex have been reported to account for the variability on force of button press when subjects performed a cued button-press task (Fox et al., 2007
); and prestimulus, baseline activity fluctuations in the thalamus and dorsal lateral prefrontal cortex predicted somatosensory perception (Boly et al., 2007
); while prestimulus activity in the fusiform face area biases subsequent perceptual decision on ambiguous figures (Hesselmann et al., 2008a
). Likewise, the coherent intrinsic activity in distributed functional-anatomic networks implicated in salience processing and executive control could modulate the variability in anxiety ratings and Trail Making Test performance (Seeley et al., 2007
). The present results build on previous findings relating the strength of connectivity between nodes of large-scale brain networks to individual differences in behavior by demonstrating that such relationships exist for episodic memory.
Echoing the finding that intrinsic coherent activity can be relevant to performance variability in healthy adults, disrupted intrinsic connectivity has been associated with impaired cognitive ability. For example, the disruption of intrinsic connectivity between the medial prefrontal cortex and the PCC was related to declines in cognitive performance across executive, memory and processing speed in advanced aging (Andrews-Hanna et al., 2007
). The strength in the hippocampal-PPC connectivity in these older adults was also significantly reduced when compared with results from young adults (Andrews-Hanna et al., 2007
; Buckner et al., 2008
). In the present study, we demonstrate that connectivity between two key nodes of a distributed memory network is predictive of episodic memory performance in healthy older individuals. The hippocampus has been reported to be functionally connected with the cortical regions of the default network, and is sometimes included in the definition of the default network (Buckner et al., 2008
; Greicius et al., 2004
). Our previous work with task-related fMRI in older subjects has also suggested that alterations in PPC memory-related neural activity were evident in low-performing cognitively normal older adults, requiring paradoxical increases in hippocampal activity for successful memory encoding (Miller et al., 2008
). Future work will be required to investigate whether longitudinal memory decline in aging is directly related to disruption of the hippocampal-PPC intrinsic connectivity. Furthermore, given that the amyloid-related functional disruption in the PPC is linked to the dysfunction in the medial temporal lobe memory system in cognitively normal and mildly impaired older individuals (Sperling et al., 2009
), and also given that the disruption of intrinsic connectivity between these regions has been observed in patients with MCI and mild AD (Greicius et al., 2004
; Wang et al., 2006
), the measurement of the hippocampal-posteromedial network integrity may be particularly valuable in probing alterations in episodic memory network in the earliest stage of AD.
The current findings may have implications for our understanding of the functional role of spontaneous brain activity. The spontaneous BOLD signal fluctuations were thought to originate from a range of processes from intrinsic low-level physiological processes that are independent of mental activity, to spontaneous cognition (Buckner et al., 2008
; Fox and Raichle, 2007
; Larson-Prior et al., 2009
; Vincent et al., 2007
). Recent evidence, however, suggests that this “spontaneous” activity may be dynamically associated with preceding experience, or may serve a role in the consolidation of previous experience (Hasson et al., 2009
; Lewis et al., 2009
). Because the resting run in this paradigm was scanned prior to performance of the memory task, the intrinsic connectivity measured in the present study was unlikely to be modulated by the task or reflect the subjects’ “recapitulating” the task. In this case, given that the intrinsic connectivity correlated with the two episodic memory measures acquired minutes and 3 months within resting-state scanning, which may be reflective a persistent behavioral trait, the present results may suggest that the coherence of spontaneous fluctuations in intrinsic brain activity can be, at least in part, a reflection of the stable trait of the large-scale functional-anatomic network in these older adults.
The interpretation of current results may also have implications for our understanding of the neural mechanism underlying episodic memory, and the role of the default network or core network in supporting memory function. The PPC, as well as regions in the lateral parietal and medial prefrontal cortices, has been observed to show greater activity during resting state as compared with a variety of goal-directed tasks, thought to subserve the default mode of brain function (Raichle et al., 2001
). Recent fMRI studies of episodic memory revealed that the PPC shows greater activity during successful episodic retrieval but during unsuccessful episodic encoding, suggesting that activity in key nodes of the default network is associated with retrieval success and encoding failure (Daselaar et al., 2004
; Daselaar et al., 2009
; Kim et al. in press). Our recent studies have suggested that successful memory formation requires coordinated activity between the hippocampus and the PPC in both young and older subjects (Miller et al., 2008
). Whereas the present results showed that stronger temporal coupling of default network activity between the PPC and hippocampus was associated with better performance on the episodic memory task. Further studies investigating the relationship between correlated intrinsic activity in the default network activity and encoding/retrieval invoked activity in the PPC are needed. Furthermore, the hippocampus and PPC have also been found to be the components of the core network, hypothesized that a set of regions serves as the common neural underpinnings to support a broad number of cognitive domains, including autobiographical memory, prospection, navigation, theory of mind, and default mode (Spreng et al., 2009
). Likewise, the present findings relating the intrinsic connectivity of two key nodes of the default network or core network to individual memory performance may support the view that one of the functions of these large-scale brain networks is to subserve episodic memory abilities in healthy older adults (Buckner et al., 2008
; Greicius et al., 2003
There are several limitations to the present study. First, the current study includes a relatively small number of older subjects and because these subjects were well characterized and specifically selected to be cognitively normal, the ranges in behavioral measures were limited, both of which may be the factors requiring relatively liberal threshold (although a cut-off of p < 0.001 is commonly used in studies with older subjects or patients). Future studies including subjects with mild cognitive impairment are likely to have more statistical power to observe such relationships. Nevertheless, even with this small number of subjects, we also found that hippocampal-PPC intrinsic connectivity was correlated with two measures of episodic memory but not with any of the non-memory domain measures, including Trails B, Digit Symbol, and MMSE.
Second, it is important to note that the hippocampal-PPC intrinsic connectivity is likely only one of the factors contributing to episodic memory performance. The variance in observed connectivity-behavioral relationship might suggest that successful episodic memory is the result of integrated activity of multiple brain systems. For example, those subjects with relatively weak hippocampal-PPC connectivity may compensate a compromised episodic memory system by engagement of alternative prefrontal systems to maintain a relatively high memory score. Future studies examining the integrity of the prefrontal control systems may prove valuable in understanding the interactive or compensatory relationship between the executive control and episodic memory systems and how such a relationship can affect episodic memory performance in these older subjects.
Lastly, it is noteworthy that both inter-individual variability in human behavior and alterations of intrinsic functional connectivity in large-scale brain networks may be relevant to the activity of neurotransmitter systems. For example, activity in the striatal dopamine system, measured by dopaminergic D2 receptor binding, correlated with episodic memory performance in older adults (Backman et al., 2000
). The experimental manipulation of dopaminergic neurotransmission could change the intrinsic connectivity in striatal cognitive and motor networks (Kelly et al., 2009
; Li et al., 2000
). Given that multiple neurotransmission systems, such as cholinergic, GABAergic and glutamatergic systems, are involved in the human memory network, and in particular, that the hippocampus and posterior cingulate are densely innervated with cholinergic fibres (Selden et al., 1998
), the current connectivity-behavioral relationship remains to be fully understood by exploring the relationships of neurotransmission systems with hippocampal-PPC intrinsic connectivity, as well as with episodic memory ability in these older subjects.
In summary, the current findings provide further evidence for the hypothesis that coherent fluctuations of spontaneous brain activity may be relevant to individual difference in specific cognitive behavior among older adults. The localization of the relationship observed from intrinsic connectivity-episodic memory performance promotes our understanding of the default network functions, and the vulnerability of this network in the aging brain. Given the converging evidence that intrinsic connectivity between the hippocampus and PPC is critical for preserved memory function in older individuals, and the vulnerability of these regions to Alzheimer’s disease pathology, longitudinal fMRI studies are crucial to determine whether impaired connectivity will serve as a sensitive predictor of memory decline related to early Alzheimer’s disease.