Hippocampal adaptation to face repetition in healthy elderly and mild cognitive impairment

Abstract

We examined the dynamic process of encoding novel repeating faces using functional MRI (fMRI) in non-demented elderly volunteers with and without diagnosed memory problems. We hypothesized that adaptation (repetition dependent reduction in activity) would occur in the mesial temporal lobe (MTL), and that this would be associated with cognitive status. Twenty-three right-handed volunteers were studied with an experimental encoding paradigm during fMRI scanning. Twelve participants had the diagnosis of mild cognitive impairment-amnestic type (MCI). The remaining 11 were cognitively healthy. All were diagnosed with a comprehensive neuropsychological battery and neurological evaluation prior to the study; they also received a brief cognitive battery the day of the scan. During the event-related fMRI task, participants viewed unfamiliar faces that repeated on average every 25 s with seven repetitions. The reduction in activation response as a function of repetition of unfamiliar faces was modeled in SPM99. Statistical parametric maps of adaptation slopes reflecting the decrease in activation with stimulus repetition were calculated for each participant, followed by a random-effects group analysis in which slope images were tested for significant group differences. Significant differences in adaptation slopes, with more negative slopes in the controls, were found in the medial temporal search region in the hippocampus and parahippocampal gyrus bilaterally, right more than left. Gray matter density analyses suggest the adaptation difference is not due to atrophy. Results suggest that the medial temporal response over repeated presentation is related to clinical status. Probes of incremental encoding processes over trials may be useful markers of medial temporal lobe integrity.

Introduction

It is well established that memory impairment is one of the earliest symptoms of Alzheimer’s Disease (AD) and is associated with deficits in both encoding and retention of new information (Albert, Moss, Tanzi, & Jones, 2001; Petersen, Smith, Ivnik, Kokmen, & Tangalos, 1994; Petersen et al., 1999; Swearer, O’Donnell, Drachman, & Woodward, 1992; Welsh, Butters, Hughes, Mohs, & Heyman, 1992). Mesial temporal lobe (MTL) structures, such as the hippocampus and entorhinal cortex, are critical for encoding new memories (Squire, 1992) and show atrophic changes at the earliest stages of AD (Jack et al., 1999). The volume of these structures has been shown to correlate with memory and learning in several studies of aging and AD (Jack et al., 1999, Jack et al., 2000, Laakso et al., 1995, Petersen et al., 2000). Functional imaging studies of encoding have found hippocampal activation in young adults (e.g. Constable et al., 2000; Grady, McIntosh, Rajah, & Craik, 1998; LePage, Habib, & Tulving, 1998; Saykin et al., 1999) as well as elderly (Bookheimer et al., 2000; Grady, McIntosh, Rajah, Beig, & Craik, 1999; Small, Perera, DeLaPaz, Mayeux, & Stern, 1999; Sperling et al., 2003).

Mild cognitive impairment-amnestic type (MCI) (Petersen et al., 2001a, Petersen et al., 2001b) may represent the earliest clinical stage of AD. In this stage, observable dysfunction should exist on tests of learning and recall of new information, in addition to memory complaints by the individual. A recent study by Albert et al. (2001) found that among a battery of neuropsychological tests administered to persons with memory difficulty, indicators of learning over repeated trials were the best predictors of conversion to AD. It may be that measuring the neural response over repeated trials in this population may provide important information related to the neural substrates of memory decline, which may in turn be helpful in early identification of a neurodegenerative process such as AD.

The majority of the cognitive brain activation paradigms in the literature using either positron emission tomography (PET) or functional MRI (fMRI) have relied on the steady state assumption—that the brain makes a consistent neural response within a particular cognitive condition over repeated occurrences of the condition. This assumption may not always be true, and does not lend itself to temporally dynamic phenomena such as adaptation, habituation, or other learning-related cerebral functions that might occur with time, repetition, or exposure. Different cognitive activation paradigms are needed to examine cognitive processes that do not maintain a steady state.

Repetition-related dynamic changes in the mesial temporal region have been detected on the neuronal level using single-cell recording. Event-related potentials in the mesial and inferior temporal lobes of non-human primates show a reduction in amplitude as the stimulus is repeated (Ringo, 1996). This stimulus specific adaptation has generally been interpreted as a learning effect (Baylis & Rolls, 1987; Brown, Wilson, & Riches, 1987; Desimone, 1996, Henson & Rugg, 2003; Riches, Wilson, & Brown, 1991; Wiggs & Martin, 1998). Furthermore, using depth electrodes in the hippocampi of patients with temporal lobe epilepsy, it has been found that anterior mesial temporal N-400 responses to novel items decrease with repetition in the intact hippocampus but not in the epileptic hippocampus (Grunwald, Elger, Lehnertz, Van Roost, & Heinze, 1995; Grunwald, Lehnertz, Heinze, Helmstaedter, & Elger, 1998). Other electrophysiology studies (Elger et al., 1997, Fernandez et al., 1999) demonstrated that the magnitude of the N-400 response in the anterior left hippocampus during the presentation of novel words was predictive of subsequent recall performance.

Recent fMRI studies (Dolan & Fletcher, 1999; Strange, Fletcher, Henson, Friston, & Dolan, 1999) demonstrate a technique that models dynamic change in activity in the hippocampus associated with repetition. These studies modeled changes in activity associated with repeated exposure to the same stimuli using fMRI and demonstrated incremental decrease in mesial temporal lobe activation as a function of repetition within a time frame of minutes. Those results illustrate that statistical processing of functional images can model changes in voxel behavior within a cognitive condition, in addition to steady state voxel behavior (Buchel, Coull, & Friston, 1999; Henson & Rugg, 2003).

Assessing hippocampal functioning using incremental models of encoding may be helpful in early identification of AD. The current study used such a model in elderly adult volunteers with and without a diagnosis of mild cognitive impairment-amnestic type. We hypothesized that elderly individuals with better learning ability would exhibit a greater hippocampal adaptation response as a function of repetition.