Alzheimer's disease (AD) is a progressive neurodegenerative disease that has as its primary symptom early in its course a profound memory loss. [1] Neuropsychological studies demonstrate that secondary memory is particularly affected, whilst aspects of primary or working memory can remain relatively normal. [2-5] Neuropathologic studies are consistent with these symptoms, with mesial temporal, temporal cortical, and diencephalic structures, which are known to be involved in normal memory functions, showing the neuritic plaques, neurofibrillary tangles, and neuronal loss characteristic of the disease. [6]
The tight coupling of local neural activity to regional cerebral blood flow (rCBF) and the ease with which multiple whole brain images of rCBF can be obtained with15 O-water using PET has allowed the development of a powerful human brain functional mapping technique. [7-9] With use of cognitive and sensory tasks, this technique has been successfully applied in normal subjects, and although there have been a variety of studies of resting rCBF in AD patients, [10-14] few studies have examined rCBF in AD patients using activation paradigms. [15-18] The results of cross-sectional studies using PET [19,20] and SPECT [21-23] show decreases in blood flow in the temporal and parietal cortices early in AD, consistent with the pattern of neurodegenerative changes in these regions, although the reliability of these findings has been recently questioned. [24] Furthermore, longitudinal PET studies have shown that these changes may parallel or precede measurable change in cognitive function. [25]
More recent PET studies of young, healthy normal volunteers have revealed distinct patterns of functional activity as the subjects engage in tasks related to auditory verbal short-term memory. [26-31] These studies are of particular relevance because the procedures have the potential to reveal specific functional defects related to specific information-processing deficits prior to any observable structural abnormalities. We report the results of a study of rCBF of AD patients during the performance of auditory verbal memory tasks that place differential demands on cognition and presumably different functional processing systems.
Methods.
Subjects.
Five women and two men meeting the NINCDS-ADRDA criteria for probable AD [1] participated in the study. They were matched in terms of age (68.7 plus minus 10.7 years), education (13.4 plus minus 1.9 years), and sex with seven normal elderly control subjects (age: 65.6 plus minus 7.7 years, education: 13.9 plus minus 1.8 years) who all completed the same PET procedures. For all patients, performance on both the Mini-Mental State Examination (MMS) (21.7 plus minus 3.4) and Mattis Dementia Rating Scale (MDRS) (120.0 plus minus 11.8) was mildly imparied relative to that of the controls (MMS: 28.6 plus minus 0.55, MDRS: 142.2 plus minus 2.6).
There was no evidence of concurrent cerebrovascular disease, except periventricular white matter intensities on MRI in some patients. The normal elderly control subjects showed no evidence of progressive cognitive decline or evidence of other psychiatric disorders. No patient or control subject was under treatment for affective disorder, and there was no history of any neuropsychiatric condition that could alter the pattern of CNS function (e.g., alcohol abuse). This research was reviewed and approved by the University of Pittsburgh Institutional Review Board and the Radioactive Drug Review Committee. All subjects signed written informed consent prior to participation in this study.
PET procedures.
Each subject was scanned eight times measuring rCBF using15 O-water with standard laboratory procedures. [26] The subjects were placed in the supine position on the Siemens 951/31 PET scanning table. An antecubital intravenous catheter was placed in the left arm for radiopharmaceutical injection. The head was positioned within the head holder and a softened thermoplastic mask fitted over the face, molded to the patient's facial contours, and fastened to the head holder. [7] The PET gantry was rotated and tilted such that the lowest imaging plane was parallel to, and approximately 1.5 to 2 cm above the canthomeatal line. Using a system of three laser lines, the face was marked with washable ink in five places to allow checks for movement of the subject during the study and to allow for positioning of the subject's head if necessary. Transmission scanning was done in all PET studies prior to radiopharmaceutical injection using three rotating rod sources of68 Ge/68 Ga.
Measurements of relative cerebral blood flow were made after an intravenous bolus injection of 50 mCi of H215 O in approximately 5 to 7 mL of saline. [32,33] Beginning approximately 5 seconds after the point when activity began to enter the brain (to allow for partial clearance of the H215 O from vascular structures), we began a 60-second sampling frame, which was used as the qualitative map of cerebral blood flow. [34]
The PET data were aligned and further spatially transformed, within group, to standard stereotaxic coordinates. Each PET image from each subject was aligned within-subject [35] and oriented to a single vertical, midsagittal plane. These images were then converted into a standard reference space based on the Talaraich & Tournoux stereotaxic atlas (STA). [36] The resulting image data were convolved with a three-dimensional gaussian filter (20 times 20 times 12 mm full-width half-maximum) to suppress noise. Differences in global activity within- and between-subject were removed by analysis of covariance (ANCOVA) on a pixel-by-pixel basis with global counts as the covariate and the regional activity across subjects for each task as treatment. For each pixel in STA space, the ANCOVA generated a condition-specific rCBF (normalized to 50 mL/100 mL/min) and an associated variance. The comparison across the selected conditions (i.e., 3WR-Rest) was done by using the t statistic, with the resulting values constituting the statistical parametric map (SPM). [37] The critical level of alpha was set at 0.001 for all comparisons. Post-hoc comparisons of rCBF across groups for selected regions were done after obtaining the scan-specific rCBF values for a given pixel location. Pixel locations correspond to peak z scores in both groups. The rCBF was recorded from images after smoothing and thus represented a weighted average similar to the size of the gaussian filter (i.e., 20 times 20 times 12 mm) centered over the voxel location.
Memory task procedures.
During the first and last scans, the subjects were instructed to lie still and fix their gaze on a cross-hair target on a video screen suspended over their head and perpendicular to their line of vision (``Rest'' condition). During the remaining six scans, the subjects were tested on single-word repetition (1WR), three-word repetition (3WR), and eight-word free recall (8WR). The scan order was counter-balanced and fixed for all subjects: Rest, 1WR, 3WR, 8WR, 8WR, 3WR, 1WR, and Rest.
After arriving at the PET facility, the subjects were given a training session for all the memory tasks. A tape recorded presentation of the instruction sets as well as practice memory lists were presented until the subjects felt comfortable with all the procedures. Of particular importance was the fact that we specifically told the subjects that the free recall task did not require serial recall, i.e., the words could be recalled in any order. For all activation conditions, the auditory-verbal memory tasks were started 20 seconds prior to the injection of 50 mCi of15 O-water so that at the beginning of the 60-second scanning frame, all subjects had completed at least one complete presentationrecall condition (esp., 8WR task).
Results.
Memory test performance.
All subjects performed the 1WR task accurately (AD patients 83% correct, controls 99% correct) (t[6] equals 1.51, p equals 0.18). There was a small difference between the patients and controls in their ability to perform the 3WR task (patients 81.5% correct, controls 98.1% correct), but this did not reach statistical significance (t[6] equals 2.05, p equals 0.086). By contrast, the performance of the patients on the 8WR task (2.06 plus minus 1.03 words) was significantly impaired relative to that of the controls (3.87 plus minus 0.40 words) (t[6] equals 5.21, p equals 0.002). The control subjects had the normal U-shaped retention function, with both primacy and recency effects, whereas the AD patients had a relatively greater loss of information in secondary memory.
Between-group comparison of rCBF.
Because this study used a yoked-control design, we were able to directly compare the rCBF between groups using a between-subject ANCOVA and subsequent paired contrasts. The results of the comparison of the Resting rCBF of the AD patients with that of the normal control subjects are shown in Figure 1. The top portion of the Figure showsthe specific pixels that have significantly more activity among the control subjects than the patients at Rest (p less than 0.01). These included the temporal-parietal-occipital border on the left and the posterior cingulate cortex and the precuneate, bilaterally (left greater than right).
Figure 1. The top of the Figure showsthe statistical parametric map (SPM) of the significant increases in rCBF in the control subjects relative to the AD patients (paired subjects comparison during the Rest condition). The images on the right-hand side of the Figure showprojections onto a representation of the medial and lateral brain surfaces. The individual significant points are referred to the closest region on the cortical surface. The ``look through'' images on the left show the pixels with significant activation projected in coronal, axial, and sagittal views. The images on the bottom compare the AD patients relative to the controls.
The lower portion of Figure 1 shows those brain regions that were relatively more active (i.e., greater rCBF) in the AD patients than the controls. This included a region of the medial frontal cortex and the lateral frontal cortex on the left. There were increases in activity relative to global blood flow also seen in the diencephalon, as well as the post-central gyrus on the right.
Within-group comparisons of rCBF.
Subspan task-activation.
Subtracting Rest from 3WR reveals those brain regions more active during auditory processing, working memory, and speech output than at rest. Figure 2 shows the SPM of the 3WR-Rest contrast for the seven elderly control subjects, and Table 1 shows the coordinates of the areas of significant activation (p less than 0.001). Previous studies from this and other laboratories have identified those brain regions activated during relatively lower-level, automatic processes associated with working memory in healthy young subjects. [26,27,29] The healthy elderly subjects showed a similar pattern of functional anatomic relationships, with significant increases in rCBF in brain regions responsible for audition, phonologic storage, and articulatory rehearsal and articulated speech.
Figure 2. The SPMs for the 3WR-Rest comparison for the normal control subjects (top) and the probable AD patients (bottom).
Table 1. Activation foci for normal elderly controls and AD patients contrasting three-word recall with rest
All seven AD subjects were able to complete the 3WR task accurately and, like the normal control subjects, showed significant activations in brain regions associated with speech and language function, and with phonologic storage (Figure 2, bottom; Table 1). However, unlike the control subjects, the AD patients did not significantly activate the hippocampal formation, a brain region implicated in secondary, but not primary, memory function [38,39] Table 1. Of particular note, however, is that in the regions that did show significant activation there was an apparent increase in the spatial extent of these regions. In all regions examined, the field of activity observed in the AD patients was greater than the activation in the same region in the normal controls. However, for the 3WR-Rest comparison, there were no brain regions activated in the AD patients that were not also activated by the control subjects.
Subspan task-deactivation.
Figure 3 shows those brain regions significantly more active when the subjects were in the Rest conditions than when they were performing the 3WR (i.e., Rest-3WR). The control subjects (top portion) had significant decreases in blood flow in the posterior cingulate cortex, the precuneate, and in an isolated deep region of the anterior cingulate. There was also significantly more activity in the cortex at the occipital-parietal border (left greater than right) when the control subjects were at rest, than when they were performing the subspan task. For the AD patients (Figure 3, bottom), the pattern of changes was very similar. There was greater activity during rest in the posterior cingulate and the cuneate, as well as an isolated region of the anterior cingulate cortex. However, unlike the control subjects, there was no significant difference in the occipital-parietal flow between Rest and 3WR.
Figure 3. The comparison of Rest-3WR shows those brain regions with greater blood flow while at rest than while performing the recall task in the normal elderly controls (top) and the AD patients (bottom).
Supraspan task-activation.
By comparing the pattern of rCBF seen in the 8WR condition with that seen in the 3WR condition, it is possible to isolate those brain mechanisms responsible for the encoding and retrieval of information from secondary episodic memory. That is, while the input and output demands of the two tasks were similar, and the demands on working memory were similar, the 8WR task required encoding of information into and retrieval of information from secondary memory. The control subjects showed significant activation primarily in Brodmann's area 10 (BA10) of the middle frontal gyrus bilaterally Figure 4, top; Table 2.
Figure 4. The SPMs for the 8WR-3WR comparison for the normal elderly subjects (top) and the probable AD patients (bottom).
Table 2. Activation foci for normal elderly controls and AD patients contrasting eight-word recall with three-word repetition
By contrast, the SPM for the 8WR-3WR comparison in the AD patients showed significant activity in the dorsolateral prefrontal cortex across the convexity bilaterally. BA8, BA9, and BA10 all showed significant activation, extending down to a level approximately 12 cm above the intercommissural (AC-PC) line. Posteriorly, the AD patients had significant regional activation in the left supramarginal (BA40) and angular gyri (BA39). Along the midline posteriorly, significant activity was observed in the precuneus (BA31). No significant activity was seen in any brain regions below 12 cm above the AC-PC line.
Supraspan task-deactivation.
Figure 5 shows those brain regions that had significantly greater blood flow during the 3WR condition relative to 8WR. For both control subjects (top portion) and the AD patients (lower portion) the only regions showing significantly different activity was the superior temporal cortex (BA42) on the right.
Figure 5. The comparison of 3WR-8WR shows those brain regions with greater blood flow while performing the subspan task than while performing the eight-word free recall task in the normal elderly controls (top) and the AD patients (bottom).
Supraspan task-correlations with performance.
The mean rCBF change in the 8WR-3WR condition was determined for each individual subject for three brain regions: supramarginal gyrus, middle frontal gyrus, and superior frontal gyrus. There was a significantly greater increase in rCBF for the AD patients relative to the controls at the angular gyrus (t[12] equals 2.14, p equals 0.05, r equals 0.53), whereas the control subjects had significantly greater activation in the more inferior regions of the frontal cortex (t[12] equals 2.09, p equals 0.06, r equals 0.51). There was also a substantially greater increase in rCBF for AD patients over the superior frontal cortex (r equals 0.45), although the difference did not reach statistical significance, two-tailed (t[12] equals 1.77, p equals 0.10).
The capacity of primary memory was estimated for each subject for each 8WR scan using the method of Tulving and Colotla. [40] The mean primary memory capacity of the normal control subjects (2.29 plus minus 0.84 words) was significantly greater than that of the AD patients (1.25 plus minus 0.60; t[12] equals 2.64, p equals 0.02), and this was then correlated with the increases in rCBF in the frontal cortex. There was a significant positive correlation between superior frontal activity and primary memory capacity in the AD patients (r equals 0.71, p equals 0.04), and a negative correlation between primary memory capacity and middle frontal activity (BA10) (r equals minus 0.73, p equals 0.03). There was no significant association between activity in either superior (r equals minus 0.12) or middle frontal cortex (r equals minus 0.04) and primary memory capacity among the control subjects.
Discussion.
The results of this study are the first that permit comparison between AD patients and controls on primary and secondary auditory-verbal memory tasks. Although previous studies have described reduced cerebral blood flow and glucose metabolism while AD patients were at rest, the present data describe the response of the CNS to the processing demands related to the cardinal feature of the disorder, i.e., the memory loss. A study by Grady et al., [16] examining functional activity in AD patients during a perceptual face-matching task, found normal cortical activity in AD patients associated with sensory and perceptual processing, and also found additional activity (i.e., not present in normals) in the frontal eye fields, which they attributed to increased attentional load in the AD patients.
Second, we report for the first time in vivo in humans, an alteration in the size of a functional unit responsible for cognitive processing in response to brain injury. That is, when AD patients performed the 3WR task (compared with rest), the spatial extent of activation in the regions associated with phonologic storage and processing of information to-be-remembered expanded relative to the same regions in the controls. We did not see such changes in field size in either of the deactivation comparisons (i.e., Rest-3WR or 3WR-Rest), suggesting that this phenomenon may only be associated with increasing functional load.
Third, the brain regions activated by the AD patients during the performance of the memory tasks is broadly consistent with the predictions based on previous neuropsychological research. That is, those brain regions associated with the verbal subsystems of working memory, and with the lower-level processing of information to-be-remembered, were activated in the patients as they were in the controls. [4,5,16] Those brain regions involved in secondary episodic memory processing (e.g., hippocampal formation) were not activated by the AD patients.
Finally, in apparent response to dysfunction localized to the lateral frontal cortex, the dorsolateral prefrontal cortex in the AD patients became active in the 8WR task (relative to 3WR). The brain region surrounding the angular gyrus (BA39/40), which is neither necessary nor sufficient to support secondary memory processes, was also significantly active (relative to that when it was ``normally'' active, i.e., 3WR) in the 8WR task, further demonstrating functional plasticity. The extent to which these kinds of changes may reflect an abnormal alteration in processing strategy or a normal response to (what for the AD patients is now) a difficult problem is unclear and needs further study. Nevertheless, our data demonstrate that the AD patient's brain, at least in the early stages of the disease, continues to be capable of dynamic responses to information-processing challenges.
The finding that the spatial extent of activation seen in the AD patients is greater than that seen in the control subjects suggests that there is a broadening of the cortical field in response to the altered cortical connections caused by the neuropathologic changes. Although the focus, or specific locus of activity remains the same (e.g., supramarginal gyrus), the area around these foci has changed. This view is consistent with the observation that areas of cortical representation of information can be altered by changes in the input to those brain regions. [41] Further, one aspect of the dysfunction in semantic memory in AD may be a ``loss of constraints ... needed to push the network towards a stable pattern of activation,'' [42] which might be represented here as a broadening of the spatial extent of activity.
The differences in patterns of activation in the 8WR condition relative to 3WR may be related to specific neuropathologic changes in frontal cortex. Among the normal elderly control subjects, the lateral orbital frontal cortex, bilaterally, was specifically activated during the 8WR task relative to 3WR Figure 4. The finding that this brain region was not activated by the AD patients, and that the adjacent dorsolateral cortex was activated, would be consistent with the hypothesis that the orbital region was dysfunctional secondary to neuropathologic changes, and that related areas of the frontal cortex were attempting to compensate for this functional loss. Using the nomenclature of Braak and Braak, [43] an abnormality in this region of frontal cortex (with relative sparing of dorsolateral convexity) would occur in stage B of amyloid deposition and stages V-VI of neurofibrillary changes. Although there is little other evidence consistent with the notion that extracellular amyloid can affect neuronal function, the early stage of the dementia in these patients leads one away from focusing exclusively on the neurofibrillary changes. For example, synapses are a major user of energy metabolism, and a loss of frontal cortical synapses [44] could partially explain these findings. However, this cortical reallocation may also be secondary to a functional or cognitive reallocation, that is, the AD patients are actually applying a different cognitive strategy to solve the 8WR problem, and this results in a different area of frontal cortex being activated.
Why there was consistent and significant activation in the region of the angular gyrus among the AD patients is less clear. This brain region is most often associated with phonologic storage [45-47] and was activated by both patients and controls when the 3WR was contrasted with Rest. Our observation of significant activation in the 8WR condition, over and above that seen in the 3WR, would also be consistent with the hypothesis that the AD patients altered their cognitive strategy to deal with the demands of the 8WR task.
It is important to note that these data address the magnitude of changes in cerebral blood flow in particular brain regions, but do not address the relationships between functionally active regions. [48] This may be particularly important in the case of the control subjects. In our previous report, [26] young normal subjects showed a bifrontal activation similar to that seen in the AD patients in the present study when we compared supraspan with subspan conditions. It may be the case that while a 12-word recall task is relatively difficult for young subjects (requiring prefrontal cortical involvement), the eight-word task used here does not significantly challenge the older subjects, and thus frontal activation is not seen. However, the performance of the control subjects in this study was far from perfect (i.e., approximate 48% correct recall), and the normal subjects in neither study showed the biparietal activation. Therefore, the issue may be more complicated than it initially appears.
In summary, patients early in the course of AD show normal patterns of cortical activity when performing low-level, automatic cognitive operations. When task demands increase, and additional cognitive resources are necessary, the AD patients show an abnormal activation of dorsolateral prefrontal cortex and of the parietal-temporal border. These data demonstrate that early in the course of the dementia, AD patients' brains retain a significant degree of functional plasticity, and this ability to reallocate cognitive or cortical resources may be of benefit in planning therapeutic interventions. More detailed specification of the precise nature of these preserved neural systems, and their neurochemical correlates, may lead to a better understanding of the pathophysiology of the disorder.
Footnotes
-
.AB.-Conscious recall of past events that have specific temporal and spatial contexts, termed episodic memory, is mediated by a system of interrelated brain regions. In Alzheimer's disease (AD) this system breaks down, resulting in an inability to recall events from the immediate past. Using subtraction techniques with PET-acquired images of regional cerebral blood flow, we demonstrate that AD patients show a greater activation of regions of cerebral cortex normally involved in auditory-verbal memory, as well as activation of cortical areas not activated by normal elderly subjects. These results provide clear evidence of functional plasticity in the AD patient's brain even if those changes do not result in normal memory function, and provide insights into the mechanisms by which the AD brain attempts to compensate for neurodegeneration.| NEUROLOGY 1996;46:692-700
- Copyright 1996 by the Advanstar Communication Inc.
REFERENCES
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Alert Me
Recommended articles
Medial temporal atrophy on MRI in normal aging and very mild Alzheimer's disease
Diagnostic utility of abbreviated fluency measures in Alzheimer disease and vascular dementia
Prosody impairment and associated affective and behavioral disturbances in Alzheimer's disease
Imaging β-amyloid burden in aging and dementia