Patients with MCI displayed evidence of structural compromise of the fornix compared with matched controls, based on a tract-specific measure of atrophy, tissue volume fraction. In addition, there was evidence of microstructural alteration in the residual fornix and in the other temporal association tracts, the uncinate, and PHC. The role of the fornix as the predominant correlate of recall performance—found in young29
and healthy older5
adults—was not reproduced in patients with MCI. In patients, the strongest correlate of verbal free recall was left PHC structure. Furthermore, the importance of the PHC in this group was underlined by the significant associations with recognition, which provided a more meaningful measure of residual memory performance. In MCI, both tissue fraction and microstructure of the uncinate and PHC were associated with recognition memory. An additional novel finding, consistent with current theories on the neurobiological basis of memory confidence, was an association between left PHC microstructure and confident recognition judgments.
Aside from the fornix, MTL-cortical interactions can be mediated by several pathways. The uncinate fasciculus links the anterior temporal lobe and parahippocampal region with the medial and orbitofrontal prefrontal cortices.30
The temporal (parahippocampal) portion of the cingulum (PHC)31
contains fibers that arise and project to parahippocampal cortices, and provides an indirect route to and from the prefrontal cortex in addition to direct parietal connections.32
Previous diffusion tensor imaging studies, investigating one or the other tract in isolation, have shown correlations with episodic memory.33,34
The present study extends these findings substantially both by demonstrating specificity of association—by comparison with both a control tract and control cognitive domains—and by assessing associations with multiple pathways of temporal interaction simultaneously. This is important because lesion studies in monkeys show that the significance of some connections for memory can be appreciated only when other pathways of interaction with the temporal lobe have been disconnected.35
In particular, lesions of the uncinate and temporal stem white matter in monkeys have little impact on memory when the fornix is left intact, but in combination with fornix section contribute to a severe, amnesic-like syndrome.36
The contrast in pattern of correlations between controls and MCI raises the possibility that this is also true of the breakdown of human memory in disease.
The specific association between fornix microstructure and recollection-based episodic memory was predicted by earlier MRI analyses5,29
and by the impact of lesions that selectively disrupt fornical fibers.25
One interpretation of the present results is a disproportionate recruitment of nonfornical pathways to support recognition memory. As expected,5,29
successful episodic memory in the healthy brain relied primarily on an intact fornix. In contrast, in MCI, episodic memory cannot depend primarily on fornix contributions so relies on other temporal association tracts, albeit unsuccessfully in behavioral terms.
The recruitment of nonfornical connections may be linked to a shift in memory strategies from recollection to more familiarity-based processes.37
Whereas recollection is underpinned by the extended hippocampal system, familiarity is linked to the parahippocampal region, most notably the perirhinal cortex.3
The majority of fibers in the fornix connect the hippocampus and subiculum (linked to recollection), rather than the parahippocampal region (linked to familiarity).4
The assumption is that fibers from the parahippocampal region, which predominantly contribute to the uncinate and PHC rather than the fornix, now become more critical. This interpretation is consistent with studies that have reported preserved familiarity but impaired recollection-based memory in MCI.37
The fornix was not exclusively affected in MCI: subtle microstructural alterations were observed across all reconstructed temporal association pathways. Unlike other tracts, a decrease in volume fraction was observed in the fornix. This evidence of localized atrophy may indicate a more advanced stage of structural degeneration in this tract and adds to previous diffusion tensor imaging studies that have not distinguished atrophy from true alterations in microstructure.6,7
In PHC and uncinate, AD and RD were the measures most sensitive to presumed pathology in MCI. It has been argued that these measures reflect axonal and myelin-related pathology, respectively. The pattern of results illustrates a limitation of FA as a general measure of tract “integrity.” In situations whereby both AD and RD increase, FA changes are likely to be minimal; in fact, in pathologic states that preferentially influence AD (presumed axonal degeneration), FA would be expected to increase
. The observation that FA is a relatively insensitive measure in MCI corroborates previous diffusion MRI studies.38
In the presence of axonal degeneration, the whole diffusion profile needs to be considered to capture relevant microstructural change.
Functional MRI studies have linked the posterior parietal cortex with meta-memorial processes signaling confidence levels for recognition memory.39
The PHC connects these regions to the temporal lobe. Therefore, the relationship between high recognition confidence (“sure” responses) and AD in the PHC is intriguing and suggests that parietal modulation of MTL regions could be communicated directly via this tract.39
The role of the PHC can also be understood in the context of increasingly recognized contributions of posteromedial cortex to memory.40
The contribution of the uncinate to recognition memory is less clear. In healthy older adults, there was evidence of a role for the uncinate in strategic aspects of memory.5
One possible role of the uncinate, therefore, is in mediating executive contributions to memory, which may be particularly relevant in the subset of individuals with additional executive difficulties. It is likely that there are a variety of subprocesses that support recognition in MCI, each supported by particular connections. Further studies that distinguish neuropsychologically between various aspects of recognition should begin to tease apart these differential contributions.
This study confirms that structural damage of the fornix occurs in MCI, along with a more subtle microstructural change of the uncinate and PHC. In the presence of a compromised fornix, these extrafornical tracts have a disproportionate influence on memory performance. One possible interpretation is a shift toward reliance on familiarity-based memory processes. Furthermore, we propose that the PHC has a crucial role in the modulation of confidence in memory judgments. Understanding patterns of residual memory is important in developing rehabilitative strategies. Prospective studies are now needed to examine directly how relationships between white matter tract structure and episodic memory evolve with memory decline.