Abstract

By integrating behavioral measures and imaging data, previous investigations have explored the relationship between biological markers of aging and cognitive functions. Evidence from functional and structural neuroimaging has revealed that hippocampal volume and activation patterns in the medial temporal lobe (MTL) may predict cognitive performance in old age. Most past demonstrations of age-related differences in brain structure–function were based on cross-sectional comparisons. Here, the relationship between 6-year intraindividual change in functional magnetic resonance imaging (fMRI) signal and change in memory performance over 2 decades was examined. Correlations between intraindividual change in fMRI signal during episodic encoding and change in memory performance measured outside of scanning were used as an estimate for relating brain–behavior changes. The results revealed a positive relationship between activation change in the hippocampus (HC) and change in memory performance, reflecting reduced hippocampal activation in participants with declining performance. Using a similar analytic approach as for the functional data, we found that individuals with declining performance had reduced HC volume compared with individuals with intact performance. These observations provide a strong link between cognitive change in older adults and MTL structure and function and thus provide insights into brain correlates of individual variability in aging trajectories.

Introduction

Decline of episodic memory and other higher order cognitive functions is commonly observed in normal aging (Park et al. 1996; Schaie 1996; Nilsson et al. 1997). The cause of age-associated cognitive decline remains to be defined. In vivo neuroimaging opens an avenue for understanding brain bases of age-related cognitive decline (Albert 1997; Buckner 2004; Cabeza et al. 2005; Raz et al. 2005).

Functional neuroimaging studies have revealed altered patterns of brain activation, particularly in the prefrontal cortex (PFC) and the medial temporal lobe (MTL), for older adults while performing cognitive tasks (for reviews, see Rajah and D'Esposito 2005; Persson and Nyberg 2006; Spreng et al. 2010). Several studies have found that older adults show relatively lower task-induced activation compared with younger adults (e.g., Grady et al. 1994; Gutchess et al. 2005; Davis et al. 2008; O'Brien et al. 2010), notably reduced encoding-related activation in the hippocampus (HC; e.g., Daselaar, Veltman, Rombouts, Lazeron, et al. 2003; Morcom et al. 2003). Such findings of age-related MTL reductions activation implicate the HC and related medial temporal regions in cognitive decline in older age. This notion is supported by findings that increasing age is correlated with reduced HC volume as measured by structural imaging (Jernigan et al. 2001; for a review, see Raz et al. 2007), and previously we found that 10-year longitudinal change in episodic memory was related to reduced HC volume (Persson, Nyberg, et al. 2006).

Thus, age-related memory decline has been linked to structural and functional changes in the HC, but several previous publications also failed to find such associations (Spreng et al. 2010; Persson et al. 2011). Between-studies differences may in part be due to differences in sample composition. There is strong evidence for substantial heterogeneity in both cognition (Habib et al. 2007) and hippocampal volume (Lupien et al. 2007) in older age. A cross-sectional finding of smaller HC volume in older than younger adults could therefore reflect early-life variability rather than age-related change (cf., Lupien et al. 2007). By contrast, longitudinal within-person changes should more directly be related to the aging process. Indeed, it has been shown that measures of age-related cognitive changes in longitudinal studies can diverge from cross-sectional measures (Schaie 1996; Rönnlund et al. 2005; Rönnlund and Nilsson 2006). Moreover, differences between cross-sectional and longitudinal estimates of age-related changes have been observed for brain structure (e.g., Raz et al. 2005, 2007) as well as functional activity (Nyberg et al. 2010). Critically, age-related structural differences in the HC are more consistently observed in longitudinal than in cross-sectional studies (Raz and Kennedy 2009).

Here, we investigate the relationship between longitudinal measures of episodic memory performance and intraindividual brain changes measured by structural magnetic resonance imaging (MRI) and functional MRI (fMRI). The participants were scanned twice, with a 6-year interval, while performing a task that previously has been associated with PFC (e.g., Logan et al. 2002; Morcom et al. 2003; Persson, Nyberg, et al. 2006) and HC (Lind, Persson, et al. 2006; Nyberg et al. 2010) activation. Participants were also scored on a composite measure of episodic memory, collected at 5 measurement points over 20 years as part of an ongoing longitudinal study (Nilsson et al. 2004), and this measure was used to assess cognitive stability or decline over time. Importantly, all participants had a stable performance up until the first fMRI data collection, and estimates of stability/decline were based on cognitive change between the first and second fMRI session. The main objective was to investigate if activation changes in regions within the MTL were related to longitudinal change in memory performance (i.e., continued stability vs. decline). A related objective was to explore the relationship between change in hippocampal volume and memory trajectories over time. Finally, we examined whether intraindividual fMRI estimates indicated frontal overrecruitment in declining individuals as has been demonstrated in cross-sectional studies (Persson, Nyberg, et al. 2006; Nyberg et al. 2010).

Materials and Methods

Participants

All subjects were recruited from “The Betula prospective cohort study: Memory, health, and aging” (Nilsson et al. 2004), an ongoing longitudinal study containing cognitive and medical data, including structural MRI and fMRI data. For the present purpose, 60 cognitively intact participants (49–79 [66.0 ± 8.1] years, 37 women) were recruited at the baseline study (2002). Of the 60 subjects who participated in the baseline study, 39 (55–84 [71.0 ± 8.1] years, 63% of the initial subject pool) participants completed the follow-up (2008) study (Nyberg et al. 2010), and 26 of these participants (55–79 [69.7 ± 8.3] years, 8 males and 18 females, 43% of the initial subject pool) were included in the current study. All 26 participants were included in the previous publications based on cross-sectional data from the baseline study (Lind, Ingvar, et al. 2006; Lind, Larsson, et al. 2006; Lind, Persson, et al. 2006; Persson, Lind, et al. 2006; Persson, Nyberg, et al. 2006; Persson et al. 2008). All participants were right-handed native Swedish speakers with no existing neurological or psychiatric illness and normal vision or vision corrected to near normal using MRI compatible glasses or contact lenses. Seventeen participants were carriers of the ϵ4-allelle of the apolipoprotein E (APOE) gene, 12 heterozygotes and 5 homozygotes. However, all participants were nondemented and scored ≥24 (mean 27.4, range 24–30) on the mini-mental state examination (Folstein et al. 1975). Also, all participants scored within 2 standard deviations of their respective age cohort on a cognitive composite measure (including measures of episodic memory, working memory, semantic memory, and cognitive speed) and within the normal range on the fMRI task. Participants were paid for their participation and signed the informed consent that was in accordance with the guidelines of the Regional Ethical Review Board. The excluded participants ranged in age from 56 to 84 years old, and the reasons for exclusion were death, health problems, suspected dementia, claustrophobia, refusals (e.g., lack of time), and missing behavioral data.

Given previous observations of a relationship between vascular risk and hippocampal shrinkage in aging (Raz 2005; Shing 2011) and the effects of hypertension on cerebral perfusion (Levy et al. 2008), we also obtained estimates of hypertension in the current sample. This included 2 markers of vascular risk: measures of blood pressure and use of blood pressure–lowering medication. Measures of blood pressure were taken by a nurse on a separate occasion in which 3 measures of blood pressure were obtained using a standard brachial cuff from participants 1) while standing, 2) while lying on a bed, and 3) after 3 min of resting in a supine position. For the purpose of the current study, these 3 separate measures were averaged. To classify participants as hypertensives, we used, in addition to information regarding whether a participant used any blood pressure–lowering medication, a definition of systolic blood pressure greater than 140 mmHg and diastolic pressure greater than 90 mmHg (Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure, 2004). Participants with high blood pressure and/or were using blood pressure–lowering medication were classified as hypertensives. By this definition, of the 26 participants, 15 (57%) were classified as suffering from hypertension. Of the 15 participants with hypertension, 8 were classified as showing stable cognitive performance over time and 7 were classified as showing reduced cognitive performance (see below). Importantly, there were no significant differences between participants classified as hypertensive compared with participants without hypertension in any of the brain estimates used in the current study.

Longitudinal Behavioral Measures

Five episodic memory tasks were used for the assessment of longitudinal change in cognitive performance. 1) Immediately following presentation of 16 verbal commands (e.g., point at the book, lift the cup) that were enacted by the participants (e.g., Cohen 1981), participants were requested to recall many of them orally, in any order. 2) In a second condition, the commands were studied without enactment. Two lists of commands were used (with 8 item-order variations for each list). The lists were counterbalanced across measurement occasions such that the list used for the enacted condition was studied without enactment at retest and vice versa. Number of sentences recalled (correct verb and noun) in the enacted and nonenacted conditions was entered in the present analysis. 3, 4) Following a brief retention interval, participants were asked to recall as many nouns as possible from the sentences described earlier. The 4 categories to which each noun belonged served as cues to remember the nouns. Number of nouns recalled from the enacted and nonenacted sentences served as separate measures. 5) Participants were presented auditory with a list of 12 common unrelated nouns and instructed to learn these words for an immediate free recall test. The presentation rate was 2 s per word. After the presentation, participants were asked to orally recall as many words as possible, with a time limit of 1 min. The 5 episodic memory measures were combined into a composite score, with a maximum of 76 points, which was used as an index of cognitive change in the fMRI analyses. Participants were tested 5 times with 5-year intervals, starting in 1988. Difference between the third and fifth measurement occasions (by subtracting the baseline scores from the follow-up scores) was used as an index of cognitive change (Fig. 1).

Figure 1.

Memory performance over time (assessed by a composite score including 5 episodic memory tests). (A) Mean performance for all participants for the first 3 waves of data collection. Standard bars show standard error of the mean. (B) Individual slopes for memory performance between the third and fifth wave of data collection with mean performance (in red).

Figure 1.

Memory performance over time (assessed by a composite score including 5 episodic memory tests). (A) Mean performance for all participants for the first 3 waves of data collection. Standard bars show standard error of the mean. (B) Individual slopes for memory performance between the third and fifth wave of data collection with mean performance (in red).

fMRI Data Acquisition

All images at both baseline and follow-up were collected on a same 1.5-T Philips Intra scanner (Philips Medical System, the Netherlands), equipped for echo-planar imaging (EPI), with identical pulse sequence (for previous publication of cross-sectional data using the same structural and functional protocols, see Lind, Ingvar, et al. 2006; Lind, Larsson, et al. 2006; Lind, Persson, et al. 2006; Persson and Nyberg 2006; Persson, Lind, et al. 2006; Persson, Nyberg, et al. 2006; Persson et al. 2008). To acquire blood oxygen level–dependent contrast images, a T2*-weighted single-shot gradient echo EPI sequence was used with the following parameters: time repetition (TR) 3000 ms, time echo (TE) 50 ms, flip angle 90°, field of view 22 × 22 cm, 64 × 64 matrix, and 3.9 mm slice thickness. Thirty-three contiguous transaxial slices positioned to include the whole-brain volume were acquired every 3.0 s. To avoid signals resulting from progressive saturation, 5 “dummy scans” were acquired and discarded prior to the image acquisition. In the scanner, cushions inside the head coil were used to reduce head movement and headphones were used to dampen the scanner noise. Responses were collected with a fiber-optic response box held in the right hand (Lumitouch reply system, Lightwave Medical Industries, Canada). Stimuli were projected on to a semitransparent screen at the head of the bore, viewed by the subject via a mirror mounted on the head coil. Sixty-nine functional volumes per session were collected, and 4 separate sessions for each participant were used. All images were sent to a PC and converted to Analyze format.

Volumetric Measurements

For the volumetric estimates, a T1-weighted 3D gradient echo sequence was used with the following parameters: TR 24 ms, TE 6 ms, flip angle 35°, and field of view 18 × 18 cm. One hundred and twenty-four coronal slices with a slice thickness of 1.8 mm were acquired in 160 × 160 matrices and reconstructed to 256 × 256 matrices. T1-weighted structural images were anatomically segmented with the FreeSurfer automated cortical and subcortical parcellation tools (http://surfer.nmr.mgh.harvard.edu/). The technical details of these procedures are described in prior publications (e.g., Dale et al. 1999; Fischl et al. 2002; Han et al. 2006). Briefly, this processing includes motion correction and averaging of multiple volumetric T1-weighted images, removal of non–brain tissue using a hybrid watershed/surface deformation procedure, automated Talairach transformation, segmentation of the subcortical white matter and deep gray matter volumetric structures (including HC, amygdala, caudate, putamen, ventricles), intensity normalization, tessellation of the gray matter–white matter boundary, automated topology correction, and surface deformation following intensity gradients to optimally place the gray/white and gray/cerebrospinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class. This procedure generated a set of HC masks for each participant, which was adjusted for differences in body size via the analysis of covariance according to the following formula: adjusted volume = raw volume − b × (height − mean height). This procedure removes variance associated with body size and redefines data points as the difference between an individual’s measures and others of similar size in the sample. Since many studies are using intracranial volume (ICV) to adjust for body size, we also examined if adjustment using ICV would substantially change the results. When ICV was used for correction (using a covariance method similar to the one specified for height), no substantive differences in the pattern of results were observed. Therefore, in keeping with our previous work (Lind, Larsson, et al. 2006; Persson, Nyberg, et al. 2006), we present the result from the height-corrected measures.

fMRI Behavioral Tasks

Full details of the methods for the fMRI session have been reported elsewhere (Persson, Nyberg, et al. 2006; Nyberg et al. 2010). Critically, the scanning session remained the same during both baseline (2002–2003) and follow-up (2008–2009) with the exception that a portion of the postscan recognition test was performed in the scanner during follow-up. In short, a semantic categorization task was administrated while fMRI was conducted to measure brain responses associated with incidental word encoding (e.g., Dale and Buckner 1997; Persson, Nyberg, et al. 2006). Subjects categorized each of 160 words in a word list as either abstract or concrete. Half of the words (80 words) were familiar to the subjects since they had made abstract/concrete decisions twice preceding functional scanning: the first time outside the scanner and the second time 15–20 min later, with shifted word order inside the scanner while the structural data were collected.

During the functional runs, a blocked-task paradigm was used, altering between the semantic categorization task blocks (30 s) and fixation blocks (21 s). Each session started and ended with brief fixation blocks (12 s). Four sessions were used, and they consisted of 4 categorization blocks containing 10 words each. Two of the blocks included words that had been presented twice prior to scanning and the other 2 blocks included novel words (i.e., words that had not been presented prior to the categorization task). The words were abstract and concrete nouns presented in lowercase Courier New font (font size 60). The session structure was similar for all 4 runs. After the scanning session, memory performance was tested using a self-paced old/new recognition test in which the participants indicated whether they saw a new or previously studied word. At follow-up, a portion of the words were tested in the scanner. The reason for using a different in-scanner task (from the ones used for longitudinal memory assessment) was to avoid interference with data collection procedures in the longitudinal project from which the participants were recruited.

fMRI Data Analysis

Functional images from both occasions (baseline and follow-up) were analyzed using Statistical Parametric Mapping software (SPM8, Wellcome Department of Cognitive Neurology, London, UK). All images were first corrected for slice timing in order to correct for acquisition time differences between slices. Second, the slice timing–corrected images were rigidly aligned to the first-image volume to correct for subject’s movement. Third, the realigned images were normalized into a common Montreal Neurological Institute space as defined by the SPM8 and finally smoothed using an 8.0-mm full-width at half-maximum Gaussian filter. Voxel-wised general linear models were set up for each subject to generate contrast images. Each condition was modeled as a boxcar that was convolved with the hemodynamic response function, whereas the baseline condition was implicitly modeled.

In order to assess the relationship between longitudinal change in performance and intraindividual BOLD change, we correlated the behavioral change scores (change in memory performance between T3 [1998–2000] and T5 [2008–2010]; see above for details on the behavioral measures) and BOLD signal change during semantic categorization (collapsed across familiar and novel trials) between baseline (2002–2003) and follow-up (2008–2009). Contrast images comparing change in BOLD activation between baseline and follow-up for each subject were used in a second-level regression model with change in memory between baseline and follow-up as a covariate to delineate regions with a positive and negative correlation with change in memory performance. The reason for using correlations rather than investigating brain activation differences between groups of individuals (e.g., stable and declining individuals) was that we found it difficult to reach a general agreement in how to categorize certain individuals. Therefore, in order to avoid any bias based on how participants were selected into the respective groups, we used a regression analysis with the change in BOLD and longitudinal change in behavioral scores as a covariate. Age and APOE status were included in the model as covariates of no interest because of their known effects on brain function.

Regions were considered significant at 0.001 uncorrected with an extent of 5 voxels. Percent signal change was extracted in medial temporal regions and plotted for display purposes using the MarsBar toolbox (http://marsbar.sourceforge.net/).

Results

Behavioral Results

Change in memory performance over time is shown in Figure 1. Mean performance was stable across all participants over the first decade (Fig. 1A). Variability in individual performance during the second decade was high with participants demonstrating memory decline or stability as well as increased performance (Fig. 1B). During scanning, at both test occasions, the participants performed the categorization task at a high level of accuracy and with little change in response times (RTs) across the 2 assessments (mean accuracy at baseline = 96.5% [mean RT = 1088 ms]; mean accuracy at follow-up = 98.7% [mean RT = 1074 ms]). Accuracy on the classification task was significantly higher at follow-up compared with baseline (t25 = 3.11, P < 0.01). Also, performance on the recognition task administered after the scanning session showed that participants scored generally high both during baseline (hits − false alarms = 106.1 [66.3%] of 160 total correct responses) and during follow-up (hits − false alarms = 103.3 [64.5%] of 160 total correct responses). There was no significant difference in memory performance between baseline and follow-up (t25 = 1.13, P = 0.27). In addition, the correlation between performance on the recognition task and memory performance on the composite score at follow-up was not significant (r = 0.19, P = 0.32).

fMRI Results

First, as has been reported for the larger sample (Nyberg et al. 2010) the current subset of participants recruited a highly consistent network of brain regions for both baseline and follow-up (Supplementary Table 1). These regions included left-lateralized prefrontal and parietal regions and bilateral occipital regions. Additionally, several homologous regions in the right hemisphere were activated, including right PFC (Supplementary Table 1).

The key analyses revealed regions in which decreased BOLD signal between baseline and follow-up was “positively” correlated with change in memory performance (i.e., decreased performance over time) as well as regions showing a “negative” correlation in which increased BOLD signal was related to declining performance over time (Table 1). Two regions within the MTL that previously have been implicated in age-related changes during episodic memory tasks, the parahippocampal gyrus (PHG) and the left HC, showed significant correlations with change in memory performance. A negative correlation between brain activation and change in memory performance was found for the PHG suggesting that participants showing decline in performance over a decade had increased BOLD signal. Conversely, for the HC, participants with declining performance over time showed decreased BOLD signal from baseline to follow-up (Fig. 2a–c). No regions in lateral PFC correlated significantly with change in memory performance, suggesting limited evidence of frontal overrecruitment in participants with declining performance (cf., Persson, Nyberg, et al. 2006). To facilitate interpretation of these findings, we divided the individuals between those showing intact or declining memory performance using a median split (N = 13 in each group). Importantly, participants categorized as showing intact performance included individuals showing no longitudinal change in performance as well as participants with increased performance. As can be seen in Figure 3, these results substantiate the correlational approach by showing a differential pattern of reduced HC activation in declining individuals and no change in HC activation in stable individuals. Similarly, the negative correlation between memory performance and PHG activation was supported by the finding of increased PHG activation in declining individuals and marginal change in stable individuals (Fig. 3).

Table 1

Brain regions showing correlations with change in longitudinal memory performance and intraindividual BOLD signal

Contrast Region Side x y z BA Volume t 
Positive correlation Precentral gyrus 44 −2 26 1264 4.89 
Lingual gyrus −54 19 88 4.36 
Insula −48 −24 14  104 4.30 
Superior temporal gyrus −42 −34 12 22 112 4.27 
Precuneus 22 −48 52 168 4.26 
HC −20 −18 −10  96 3.52 
Precentral gyrus −52 18 56 3.52 
Negative correlation Superior frontal gyrus −14 66 1208 5.71 
Superior frontal gyrus 18 34 44 312 4.74 
Striatum 12  368 4.70 
Medial frontal gyrus −4 60 18 10 80 4.52 
Medial frontal gyrus 50 −6 10 72 4.36 
PHG 18 −6 −24  80 4.34 
PHG −20 −4 −26  200 4.26 
Superior occipital 32 −82 42 19 112 4.21 
Precentral gyrus 30 68 88 4.08 
Contrast Region Side x y z BA Volume t 
Positive correlation Precentral gyrus 44 −2 26 1264 4.89 
Lingual gyrus −54 19 88 4.36 
Insula −48 −24 14  104 4.30 
Superior temporal gyrus −42 −34 12 22 112 4.27 
Precuneus 22 −48 52 168 4.26 
HC −20 −18 −10  96 3.52 
Precentral gyrus −52 18 56 3.52 
Negative correlation Superior frontal gyrus −14 66 1208 5.71 
Superior frontal gyrus 18 34 44 312 4.74 
Striatum 12  368 4.70 
Medial frontal gyrus −4 60 18 10 80 4.52 
Medial frontal gyrus 50 −6 10 72 4.36 
PHG 18 −6 −24  80 4.34 
PHG −20 −4 −26  200 4.26 
Superior occipital 32 −82 42 19 112 4.21 
Precentral gyrus 30 68 88 4.08 

Note: Volume refers to the size (in mm3) of the cluster of activated voxels. BA = Brodmann area.

Figure 2.

Coronal sections showing MTL regions with positive (A) and negative (B,C) correlation between intraindividual BOLD signal change (between baseline and follow-up) and longitudinal memory performance.

Figure 2.

Coronal sections showing MTL regions with positive (A) and negative (B,C) correlation between intraindividual BOLD signal change (between baseline and follow-up) and longitudinal memory performance.

Figure 3.

Coronal sections showing activation in MTL regions for individuals with stable or declining memory performance. (A) Bar graphs represent the average activation during baseline and follow-up for each group, respectively. (B) Bar graphs represent the average change in activation between baseline and follow-up. Error bars represent standard error of the mean.

Figure 3.

Coronal sections showing activation in MTL regions for individuals with stable or declining memory performance. (A) Bar graphs represent the average activation during baseline and follow-up for each group, respectively. (B) Bar graphs represent the average change in activation between baseline and follow-up. Error bars represent standard error of the mean.

Next, we performed a series of analyses to determine if other variables might account for the current observations. First, there was no significant difference in memory performance change or MTL activation (HC or left and right PHG) between APOE ϵ4 carriers and noncarriers (change in memory performance: t25 = 1.27, P = 0.15; HC: t25 = 1.18, P = 0.25; left PHG: t25 = 1.56, P = 0.11; right PHG: t25 = 0.24, P = 0.81). In order to investigate time-on-task effects on brain activation in MTL regions, we examined the relationship between RT on the classification task and brain activation in these regions. The correlation between activation in MTL regions and the RT in the classification task was not significant, either at baseline or at follow-up (all Ps > 0.05). Also, no significant correlations were found between hippocampal volume and BOLD signal change between baseline and follow-up (all Ps > 0.05), suggesting that activation in MTL regions was not directly related to hippocampal atrophy. Also, there was no significant difference in the number of APOE ϵ4 carriers between the groups of stable and declining individuals, although there was a difference in mean age (stable participants = 66.3 and declining participants = 74.2).

Volumetric Measurements

Similar to the functional data analyses, we related change in memory performance to estimates of hippocampal volume. Given previous observations of hippocampal volume decrease in aging (e.g., Hampel et al. 2002), along with the previous findings of reduced hippocampal volume in APOE ϵ4 carriers compared with non-carriers (e.g., Soininen et al. 1995; Cohen et al. 2001), we included age and APOE status as variables in a regression model in which we investigated the relationship between hippocampal volume and change in memory performance. We found a significant positive relationship between change in memory performance and change in right hippocampal volume (r = 0.53, P < 0.01), suggesting that individuals with declining memory performance also had reduced hippocampal volume (Fig. 4A). In addition, change in memory performance was positively correlated (r = 0.41, P < 0.05), with hippocampal volume at follow-up (Fig. 4A). Again, to facilitate interpretation of these findings, we divided the individuals between those showing intact or declining memory performance using a median split (N = 13 in each group). Results from the hippocampal estimates on stable and declining individuals are shown in Figure 4B. In line with the regression analyses, we found a significant difference between stable and declining individuals for change in right hippocampal volume from baseline to follow-up (F1,25 = 3.12, P < 0.05) and right hippocampal volume at follow-up (F1,25 = 3.14, P < 0.05).

Figure 4.

(A) Correlation between change in memory performance and mean height–adjusted volume of the right HC (in mm3, top) and correlation between change in memory performance and change in volume of the right HC (in mm3, bottom). (B) Mean HC volume for left (top) and right (bottom) HC for individuals with declining and stable longitudinal memory performance during baseline (light gray) and follow-up (dark gray). (C) Individual change in left (top) and right (bottom) hippocampal volume from baseline to follow-up.

Figure 4.

(A) Correlation between change in memory performance and mean height–adjusted volume of the right HC (in mm3, top) and correlation between change in memory performance and change in volume of the right HC (in mm3, bottom). (B) Mean HC volume for left (top) and right (bottom) HC for individuals with declining and stable longitudinal memory performance during baseline (light gray) and follow-up (dark gray). (C) Individual change in left (top) and right (bottom) hippocampal volume from baseline to follow-up.

Both baseline and follow-up measures of hippocampal volume correlated negatively with age (left HC at baseline: r = −0.499 [P = 0.011]; left HC at follow-up: r = −0.457 [P = 0.022]; right HC at baseline: r = −0.476 [P = 0.016]; right HC at follow-up: r = −0.443 [P = 0.027]). Finally, we found evidence for differences in hippocampal volume between carriers and noncarriers of the APOE ϵ4 allele, where APOE ϵ4 carriers had significantly smaller volume in both right and left HC at baseline (right HC: t = 2.47, P < 0.05; left HC: t = 2.89, P < 0.01) as well as a larger longitudinal HC volume reduction in both left and right HC (right HC: t = 2.29, P < 0.05; left HC: t = 4.76, P < 0.001).

Discussion

Our results provide evidence for longitudinal functional and neuroanatomical changes associated with cognitive decline in aging. During functional scans, participants with declining performance over time show reduced activation in the left HC. No difference in hippocampal activation between baseline and follow-up was observed for participants with intact performance. In addition, individuals with declining memory performance showed overrecruitment in the right and left PHG. Longitudinal fMRI estimates were not associated with age, APOE status, or hippocampal atrophy. Also, individuals with declining performance had a more pronounced reduction of hippocampal volume between baseline and follow-up compared with individuals with preserved performance. The observed differences in structural and functional estimates suggest that multiple factors are contributing to cognitive decline.

The observed decrease in HC activation in participants with declining memory performance corroborates previous findings of age-related reduction in hippocampal activation (e.g., Grady et al. 1995; Daselaar, Veltman, Rombouts, Raaijmakers, and Jonker 2003; Gutchess et al. 2005). While several previous findings have been unable to show age differences in hippocampal activation in the absence of dementia (e.g., Sperling et al. 2003; Miller et al. 2008; Persson et al. 2011; for reviews, see Buckner 2004; Hedden and Gabrieli 2005), the present results suggest that reduced hippocampal activation underlie memory impairment in aging. Thus, variability in cognitive performance might contribute heavily to hippocampal responsivity in memory tasks and underscores the importance of relating longitudinal stability or decline in memory performance to functional brain changes. Also, HC activation at baseline was higher in declining individuals compared with stable individuals, suggesting that baseline activation predicts subsequent cognitive decline. A similar observation was found in a recent longitudinal study showing that older adults with the highest hippocampal activation at baseline and the greatest loss of activation over a 2-year period also demonstrated the greatest cognitive decline (O'Brien et al. 2010).

We also found 2 regions within the MTL in which change in activation was negatively correlated with change in cognitive performance. One key region previously associated with episodic memory, the PHG, showed such a relationship. A similar dissociation within the MTL showing reduced activation in the HC and increased activation in the PHG in older adults has been previously found during episodic memory retrieval (e.g., Cabeza et al. 2004; Daselaar et al. 2006). It has been suggested that such a pattern reflects a shift from recollection-based retrieval processes to familiarity-based processes in older adults. Since the current results were obtained during episodic encoding rather than retrieval, such an explanation is not easily applicable in the present study. However, one possibility is that increased bilateral parahippocampal activity in older adults with declining memory performance indicated attempted functional compensation for an underlying deficit in hippocampal function.

We did not find support for the previous observations of an association between longitudinal cognitive decline and increased right PFC activation (Persson, Nyberg, et al. 2006). The discrepancy between the current findings and our previous observations suggests that longitudinal fMRI data may deviate substantially from cross-sectional estimates. Indeed, recent observations with longitudinal fMRI suggest that aging is associated with reduced rather than increased frontal cortex activation (Nyberg et al. 2010).

The relationship between change in cognitive performance and hippocampal volume adds to previous cross-sectional demonstrations suggesting that the HC plays a critical role in episodic memory in aging. The findings are consistent with several cross-sectional studies showing reduced hippocampal volume and memory dysfunction in patients with mild cognitive impairment (e.g., Müller et al. 2005; Saykin et al. 2006) and Alzheimer’s disease (AD; e.g., Petersen et al. 2000). While cross-sectional studies examining brain structure–behavior relationships in healthy elderly individuals have provided more mixed results (Van Petten 2004), our findings are in line with observations showing a negative correlation between hippocampal and MTL structures and cognitive performance (e.g., Golomb et al. 1996; Rodrigue and Raz 2004; Shing et al. 2011). The current observation is further supported by results linking HC volume and decline in cognitive performance (Golomb et al. 1996; Kramer et al. 2007).

A few caveats must be taken into consideration. As with any longitudinal study with repeated testing, retest or practice effects must be considered when interpreting the data. These effects could potentially affect both longitudinal behavioral estimates and the fMRI results. Practice can have a considerable effect on cognitive performance (Rönnlund et al. 2005) as well as brain activation (Wagner et al. 2000). Indeed, several participants showed modest increases in memory performance, possibly indicating the presence of retest effects in the behavioral data. Similarly, practice has been associated with changes in brain activity, and such changes could thus be confounded with true changes in fMRI signal related to change in memory performance. Such effects are unlikely in the present study for several reasons. First, and most importantly, we did not observe any practice effects on brain activation in either the HC or the PHG, while holding age constant (Supplementary Fig. 1). Second, the critical comparison was between individuals with intact and declining performance, and practice effects would most likely affect both groups equally. Also, although clinical screening is performed to rule out mild-to-moderate AD, decline in memory performance, low scores on global cognition, and reduced hippocampal volume are suggestive of presymptomatic dementia. It is unclear whether our findings of reduced hippocampal volume in individuals with declining performance stem from the inclusion of undiagnosed dementia cases in our sample.

In conclusion, these findings highlight the importance of examining longitudinal performance variability in combination with intraindividual brain correlates in older adults. By doing this, we could show that age-related deficits in hippocampal functioning seem to account for age-related cognitive decline.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/

Funding

Göran Gustafsson Award in Medicine (to L.N.), a grant from the Swedish Science Council (421-2006-1661 to L.N.), a Wallenberg Scholar Grant from the Knut and Alice Wallenberg Foundation (KAW2009.0070 to L.N.); a grant from the Swedish Science Council (421-2007-1895 to J.P.). The BETULA Project is supported by a grant from the Swedish Science Council (315-2004-6977 to L.-G.N. and L.N.).

We thank the staff of the BETULA Project and the staff at the Umeå Center for Functional Brain Imaging. Conflict of Interest : None declared.

References

Albert
MS
The ageing brain: normal and abnormal memory
Philos Trans R Soc Lond B Biol Sci
 , 
1997
, vol. 
352
 (pg. 
1703
-
1709
)
Buckner
RL
Memory and executive function in aging and AD: multiple factors that cause decline and reserve factors that compensate
Neuron
 , 
2004
, vol. 
44
 (pg. 
195
-
208
)
Cabeza
R
Daselaar
SM
Dolcos
F
Prince
SE
Budde
M
Nyberg
L
Task-independent and task-specific age effects on brain activity during working memory, visual attention and episodic retrieval
Cereb Cortex
 , 
2004
, vol. 
14
 (pg. 
364
-
375
)
Cabeza
R
Nyberg
L
Park
DC
Cognitive neuroscience of aging
 , 
2005
New York
Oxford University Press
Cohen
RL
On the generality of some memory laws
Scand J Psychol
 , 
1981
, vol. 
22
 (pg. 
267
-
281
)
Cohen
RM
Small
C
Lalonde
F
Friz
J
Sunderland
T
Effect of apolipoprotein E genotype on hippocampal volume loss in aging healthy women
Neurology
 , 
2001
, vol. 
57
 (pg. 
2223
-
2228
)
Dale
AM
Buckner
RL
Selective averaging of rapidly presented individual trials using fMRI
Hum Brain Mapp
 , 
1997
, vol. 
5
 (pg. 
329
-
340
)
Dale
AM
Fischl
B
Sereno
MI
Cortical surface-based analysis. I. Segmentation and surface reconstruction
Neuroimage
 , 
1999
, vol. 
9
 (pg. 
179
-
194
)
Daselaar
SM
Fleck
MS
Dobbins
IG
Madden
DJ
Cabeza
R
Effects of healthy aging on hippocampal and rhinal memory functions: an event-related fMRI study
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
1771
-
1782
)
Daselaar
SM
Veltman
DJ
Rombouts
SA
Lazeron
RH
Raaijmakers
JG
Jonker
C
Neuroanatomical correlates of episodic encoding and retrieval in young and elderly subjects
Brain
 , 
2003
, vol. 
126
 (pg. 
43
-
56
)
Daselaar
SM
Veltman
DJ
Rombouts
SA
Raaijmakers
JG
Jonker
C
Deep processing activates the medial temporal lobe in young but not elderly adults
Neurobiol Aging
 , 
2003
, vol. 
24
 (pg. 
1005
-
1011
)
Davis
SW
Dennis
NA
Daselaar
SM
Fleck
MS
Cabeza
R
Que PASA? The posterior-anterior shift in aging
Cereb Cortex
 , 
2008
, vol. 
18
 (pg. 
1201
-
1209
)
Fischl
B
Salat
DH
Busa
E
Albert
M
Dieterich
M
Haselgrove
C
van der Kouwe
A
Killiany
R
Kennedy
D
Klaveness
S
, et al.  . 
Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain
Neuron
 , 
2002
, vol. 
33
 (pg. 
341
-
355
)
Folstein
MF
Folstein
SE
McHugh
PR
“Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician
Journal of Psychiatric Research
 , 
1975
, vol. 
12
 (pg. 
189
-
198
)
Golomb
J
Kluger
A
deLeon
MJ
Ferris
SH
Mittelman
M
Cohen
J
George
AE
Hippocampal formation size predicts declining memory performance in normal aging
Neurology
 , 
1996
, vol. 
47
 (pg. 
810
-
813
)
Grady
CL
Maisog
JM
Horwitz
B
Ungerleider
LG
Mentis
MJ
Salerno
JA
Pietrini
P
Wagner
E
Haxby
JV
Age-related changes in cortical blood flow activation during visual processing of faces and location
J Neurosci
 , 
1994
, vol. 
14
 (pg. 
1450
-
1462
)
Grady
CL
McIntosh
AR
Horwitz
B
Maisog
JM
Ungerleider
LG
Mentis
MJ
Pietrini
P
Schapiro
MB
Haxby
JV
Age-related reductions in human recognition memory due to impaired encoding
Science
 , 
1995
, vol. 
269
 (pg. 
218
-
221
)
Gutchess
AH
Welsh
RC
Hedden
T
Bangert
A
Minear
M
Liu
LL
Park
DC
Aging and the neural correlates of successful picture encoding: frontal activations compensate for decreased medial-temporal activity
J Cogn Neurosci
 , 
2005
, vol. 
17
 (pg. 
84
-
96
)
Habib
R
Nyberg
L
Nilsson
LG
Cognitive and non-cognitive factors contributing to the longitudinal identification of successful older adults in the Betula study
Aging Neuropsychol Cogn
 , 
2007
, vol. 
14
 (pg. 
257
-
273
)
Hampel
H
Teipel
SJ
Bayer
W
Alexander
GE
Schwarz
R
Schapiro
MB
Rapoport
SI
Möller
HJ
Age transformation of combined hippocampus and amygdala volume improves diagnostic accuracy in Alzheimer's disease
J Neurol Sci
 , 
2002
, vol. 
194
 (pg. 
15
-
19
)
Han
X
Jovicich
J
Salat
D
van der Kouwe
A
Quinn
B
Czanner
S
Busa
E
Pacheco
J
Albert
M
Killiany
R
, et al.  . 
Reliability of MRI-derived measurements of human cerebral cortical thickness: the effects of field strength, scanner upgrade and manufacturer
Neuroimage
 , 
2006
, vol. 
32
 (pg. 
180
-
194
)
Hedden
T
Gabrieli
JD
Healthy and pathological processes in adult development: new evidence from neuroimaging of the aging brain
Curr Opin Neurol
 , 
2005
, vol. 
18
 (pg. 
740
-
747
)
Jernigan
TL
Archibald
SL
Fennema-Notestine
C
Gamst
AC
Stout
JC
Bonner
J
Hesselink
JR
Effects of age on tissues and regions of the cerebrum and cerebellum
Neurobiol Aging
 , 
2001
, vol. 
22
 (pg. 
581
-
594
)
Kramer
JH
Mungas
D
Reed
BR
Wetzel
ME
Burnett
MM
Miller
BL
Weiner
MW
Chui
HC
Longitudinal MRI and cognitive change in healthy elderly
Neuropsychology
 , 
2007
, vol. 
21
 (pg. 
412
-
418
)
Levy
BI
Schiffrin
EL
Mourad
JJ
Agostini
D
Vicaut
E
Safar
ME
Struijker-Boudier
HA
Impaired tissue perfusion: a pathology common to hypertension, obesity, and diabetes mellitus
Circulation
 , 
2008
, vol. 
118
 (pg. 
968
-
976
)
Lind
J
Ingvar
M
Persson
J
Sleegers
K
Van Broeckhoven
C
Adolfsson
R
Nilsson
LG
Nyberg
L
Parietal cortex activation predicts memory decline in apolipoprotein E-4 carriers
Neuroreport
 , 
2006
, vol. 
17
 (pg. 
1683
-
1686
)
Lind
J
Larsson
A
Persson
J
Ingvar
M
Nilsson
LG
Backman
L
Adolfsson
R
Cruts
M
Sleegers
K
Van Broeckhoven
C
, et al.  . 
Reduced hippocampal volume in non-demented carriers of the apolipoprotein E4: relation to chronological age and recognition memory
Neurosci Lett
 , 
2006
, vol. 
396
 (pg. 
23
-
27
)
Lind
J
Persson
J
Ingvar
CM
Larsson
A
Cruts
M
Van Broeckhoven
C
Adolfsson
R
Backman
L
Nilsson
L-G
Nyberg
L
Cognitively intact carriers of the Apolipoprotein E4 show reduced functional brain activity response
Brain
 , 
2006
, vol. 
129
 (pg. 
1240
-
1248
)
Logan
JM
Sanders
AL
Snyder
AZ
Morris
JC
Buckner
RL
Under-recruitment and nonselective recruitment: dissociable neural mechanisms associated with aging
Neuron
 , 
2002
, vol. 
33
 (pg. 
827
-
840
)
Lupien
SJ
Evans
A
Lord
C
Miles
J
Pruessner
M
Pike
B
Pruessner
JC
Hippocampal volume is as variable in young as in older adults: implications for the notion of hippocampal atrophy in humans
Neuroimage
 , 
2007
, vol. 
34
 (pg. 
479
-
485
)
Miller
SL
Celone
K
DePeau
K
Diamond
E
Dickerson
BC
Rentz
D
Pihlajamäki
M
Sperling
RA
Age-related memory impairment associated with loss of parietal deactivation but preserved hippocampal activation
Proc Natl Acad Sci U S A
 , 
2008
, vol. 
105
 (pg. 
2181
-
2186
)
Morcom
AM
Good
CD
Frackowiak
RSJ
Rugg
MD
Age effects on the neural correlates of successful memory encoding
Brain
 , 
2003
, vol. 
126
 (pg. 
213
-
229
)
Müller
MJ
Greverus
D
Dellani
PR
Weibrich
C
Wille
PR
Scheurich
A
Stoeter
P
Fellgiebel
A
Functional implications of hippocampal volume and diffusivity in mild cognitive impairment
Neuroimage
 , 
2005
, vol. 
26
 (pg. 
1033
-
1042
)
Nilsson
L-G
Adolfsson
R
Bäckman
L
de Frias
C
Molander
B
Nyberg
L
Betula: a prospective cohort study on memory, health, and aging
Aging Neuropsych Cogn
 , 
2004
, vol. 
11
 (pg. 
134
-
148
)
Nilsson
L-G
Bäckman
L
Erngrund
K
Nyberg
L
Adolfsson
R
Bucht
G
Karlsson
S
Widing
M
Winblad
G
The Betula prospective cohort study: memory, health and aging
Aging Neuropsych Cogn
 , 
1997
, vol. 
4
 (pg. 
1
-
32
)
Nyberg
L
Salami
A
Andersson
M
Eriksson
J
Kalpouzos
G
Kauppi
K
Lind
J
Pudas
S
Persson
J
Nilsson
LG
Longitudinal evidence for diminished frontal cortex function in aging
Proc Natl Acad Sci U S A
 , 
2010
, vol. 
107
 (pg. 
22682
-
22686
)
O'Brien
JL
O'Keefe
KM
LaViolette
PS
DeLuca
AN
Blacker
D
Dickerson
BC
Sperling
RA
Longitudinal fMRI in elderly reveals loss of hippocampal activation with clinical decline
Neurology
 , 
2010
, vol. 
74
 (pg. 
1969
-
1976
)
Park
DC
Smith
AD
Lautenschlager
G
Earles
JL
Frieske
D
Zwahr
M
Gaines
CL
Mediators of long-term memory performance across the life span
Psychol Aging
 , 
1996
, vol. 
11
 (pg. 
621
-
637
)
Persson
J
Kalpouzos
G
Nilsson
LG
Ryberg
M
Nyberg
L
Preserved hippocampus activation in normal aging as revealed by fMRI
Hippocampus
 , 
2011
, vol. 
21
 (pg. 
753
-
766
)
Persson
J
Lind
J
Larsson
A
Ingvar
CM
Cruts
M
Van Broeckhoven
C
Adolfsson
R
Nilsson
LG
Nyberg
L
Altered brain white matter integrity in healthy carriers of the APOE4 allele: a risk for AD?
Neurology
 , 
2006
, vol. 
66
 (pg. 
1029
-
1033
)
Persson
J
Lind
J
Larsson
A
Ingvar
M
Sleegers
K
Van Broeckhoven
C
Adolfsson
R
Nilsson
LG
Nyberg
L
Altered deactivation in individuals with genetic risk for Alzheimer's disease
Neuropsychologia
 , 
2008
, vol. 
46
 (pg. 
1679
-
1687
)
Persson
J
Nyberg
L
Altered brain activity in healthy seniors: what does it mean?
Prog Brain Res
 , 
2006
, vol. 
157
 (pg. 
45
-
56
)
Persson
J
Nyberg
L
Lind
J
Larsson
A
Nilsson
LG
Ingvar
M
Buckner
RL
Structure—function correlates of cognitive decline in aging
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
907
-
915
)
Petersen
R
Jack
CRJ
Xu
YC
Waring
SC
O'Brien
PC
Smith
GE
Ivnik
RJ
Tangalos
EG
Boeve
BF
Kokmen
E
Memory and MRI-based hippocampal volumes in aging and AD
Neurology
 , 
2000
, vol. 
54
 (pg. 
581
-
587
)
Rajah
MN
D'Esposito
M
Region-specific changes in prefrontal function with age: a review of PET and fMRI studies on working and episodic memory
Brain
 , 
2005
, vol. 
128
 (pg. 
1964
-
1983
)
Raz
N
Kennedy
KM
Jagust
W
D’Esposito
M
A systems approach to age-related change: neuroanatomical changes, their modifiers, and cognitive correlates
Imaging the aging brain
 , 
2009
New York
Oxford University Press
(pg. 
43
-
70
)
Raz
N
Lindenberger
U
Rodrigue
KM
Kennedy
KM
Head
D
Williamson
A
Dahle
C
Gerstorf
D
Acker
JD
Regional brain changes in aging healthy adults: general trends, individual differences and modifiers
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
1676
-
1689
)
Raz
N
Rodrigue
KM
Haacke
EM
Brain aging and its modifiers: insights from in vivo neuromorphometry and susceptibility weighted imaging
Ann N Y Acad Sci
 , 
2007
, vol. 
1097
 (pg. 
84
-
93
)
Rodrigue
KM
Raz
N
Shrinkage of the entorhinal cortex over five years predicts memory performace in healthy adults
J Neurosci
 , 
2004
, vol. 
24
 (pg. 
956
-
963
)
Rönnlund
M
Nilsson
LG
Adult life-span patterns in WAIS-R Block Design performance: cross-sectional versus longitudinal age gradients and relations to demographic factors
Intelligence
 , 
2006
, vol. 
34
 (pg. 
63
-
78
)
Rönnlund
M
Nyberg
L
Bäckman
L
Nilsson
LG
Stability, growth, and decline in adult life span development of declarative memory: cross-sectional and longitudinal data from a population-based study
Psychol Aging
 , 
2005
, vol. 
20
 (pg. 
3
-
18
)
Saykin
AJ
Wishart
HA
Rabin
LA
Santulli
RB
Flashman
LA
West
JD
McHugh
TL
Mamourian
AC
Older adults with cognitive complaints show brain atrophy similar to that of amnestic MCI
Neurology
 , 
2006
, vol. 
67
 (pg. 
834
-
842
)
Schaie
KW
Intellectual development in adulthood: the Seattle longitudinal study
 , 
1996
New York
Cambridge University Press
Shing
YL
Rodrigue
KM
Kennedy
KM
Fandakova
Y
Bodammer
N
Werkle-Bergner
M
Lindenberger
U
Raz
N
Hippocampal subfield volumes: age, vascular risk, and correlation with associative memory
Front Aging Neurosci
 , 
2011
, vol. 
3
 pg. 
2
 
Soininen
H
Partanen
K
Pitkänen
A
Hallikainen
M
Hänninen
T
Helisalmi
S
Mannermaa
A
Ryynänen
M
Koivisto
K
Riekkinen
PS
Decreased hippocampal volume asymmetry on MRIs in nondemented elderly subjects carrying the apolipoprotein E epsilon 4 allele
Neurology
 , 
1995
, vol. 
45
 (pg. 
391
-
392
)
Sperling
RA
Bates
JF
Chua
EF
Cocchiarella
AJ
Rentz
DM
Rosen
BR
Schacter
DL
Albert
MS
fMRI studies of associative encoding in young and elderly controls and mild Alzheimer's disease
J Neurol Neurosurg Psychiatry
 , 
2003
, vol. 
74
 (pg. 
44
-
50
)
Spreng
RN
Wojtowicz
M
Grady
CL
Reliable differences in brain activity between young and old adults: a quantitative meta-analysis across multiple cognitive domains
Neurosci Biobehav Rev
 , 
2010
, vol. 
34
 (pg. 
1178
-
1194
)
Van Petten
C
Relationship between hippocampal volume and memory ability in healthy individuals across the lifespan: review and meta-analysis
Neuropsychologia
 , 
2004
, vol. 
42
 (pg. 
1394
-
1413
)
Wagner
AD
Koutstaal
W
Maril
A
Schacter
DL
Buckner
RL
Task-specific repetition priming in left inferior prefrontal cortex
Cereb Cortex
 , 
2000
, vol. 
10
 (pg. 
1176
-
1184
)