Altered glucose metabolism in the hippocampal head in memory impairment

The structures of the medial temporal lobe are involved in memory processing in humans.^1,2 In particular, the hippocampus is considered important for memory retention or consolidation in amnesia or dementia.³ Several PET studies showed that normal energy consumption in the medial temporal area was relatively low^4,5 and that there was a mild reduction in hippocampal glucose metabolism in patients with AD.⁶ Many of these findings were obtained with the image reformatting technique that uses software to reslice transaxial tomographic data into secondary images parallel to the hippocampal longitudinal axis plane (HAP) or vertical to HAP (coronal reslicing) in evaluation of hippocampal metabolism. The drawback of conventional PET imaging with secondary reformatting is transformation of the tomographic data sampled using a low-spatial resolution PET scanner with a large slice interval of more than 10 mm, resulting in an increase in errors due to the partial volume effect, specifically in evaluating the metabolism of small structures. Image reformation with statistical parametric mapping is a powerful tool for predicting the locus responsible for a disease entity, but the multimorphism of the pathologic hippocampus may cause anatomic errors after normalization of all images to a standard brain template. Therefore, a high-resolution PET scanner that can focus on hippocampal formation is necessary to evaluate metabolic alterations in diseases of memory impairment.

Pure amnesia is a syndrome with selective impairment of long-term memory in the absence of other intellectual or cognitive deficits. Memory impairment is also one of the earliest manifestations of AD,^7-9 featuring intellectual deterioration. As confirmed by postmortem studies showing that the hippocampus together with its adjacent structures is essential for the formation of long-term memory,^1,2,10,11 it is important to evaluate in vivo functions of the medial temporal structures such as the hippocampus and amygdala in these patients. The head of the hippocampus consists of a greater portion of the CA1 field,¹² which is vulnerable in the Alzheimer brain¹³ or the anoxic brain causing amnesia.¹⁴ A specifically high-resolution PET scanner with a gantry tilt system permits in vivo evaluation of energy metabolism in this hippocampal head portion without reslicing tomographic data by adjusting the scanner's transaxial direction to the plane parallel to HAP, which provides a good anatomic description of the hippocampal formation from head to tail.¹⁵

In the current study, we measured the cerebral metabolic rate of glucose(CMRGlc) with PET and the volumetric shrinkage level with MRI in the hippocampus and amygdala to investigate whether metabolic reduction along with structural changes in specific parts of medial temporal structures contribute to the pathophysiology related to pure amnesia and to AD.

Methods. Subjects. Our study was approved by the Ethics Committee of the Hamamatsu Medical Center. We examined 10 patients with pure amnesia (six men and four women; mean age ± SD, 55.1± 13.9 years), eight age-matched patients with probable AD (five men and three women, 61.3 ± 5.2 years) diagnosed by NINCDS/ADRDA¹⁶ and clinical dementia rating(CDR),¹⁷ and eight age-matched normal subjects (4 men and 4 women, 58.1 ± 2.1 years) who had no neurologic problems and no abnormalities on MRI. The average education periods were 12.2 ± 2.3 years, 11.8 ± 2.2 years, and 12.6 ± 2.3 years respectively. Diagnosis of pure amnesia was established if a patient showed selective impairment of long-term memory (e.g., story recall) and normal performances on tests of short-term memory (e.g., digit span), language, and visuospatial perception. Five pure amnesics were suspected of suffering from sleep apnea, and three had severe anemia with a history of hypoxic accidents. The etiology of the other two amnesics was unknown, but these two were reported to have severe snoring. Almost all subjects performed several neuropsychological tests including the Mini-Mental State Examination, the Wechsler Adult Intelligence Scale-Revised,¹⁸ subtest batteries(figural, verbal associates I and II, visual associates I and II, and logical memories) of the Wechsler Memory Scale-Revised,¹⁹ and verbal paired associates memory recall test. Patients with AD were also assessed with the Alzheimer's Disease Assessment Scale²⁰(ADAS). The detailed subject characteristics are shown in table 1. MRI studies disclosed no abnormalities except for some levels of atrophy in the medial temporal area in patients with pure amnesia and AD.

MRI procedure. MRI was performed using a static magnet (0.3 T MRP7000AD, Hitachi, Tokyo, Japan) with the following acquisition parameters: three-dimensional (3D) mode sampling, TR/TE (200 msec/23 msec), 75-deg flip angle, 2-mm slice thickness with no gap, and 256 × 256 matrices to detect the longitudinal axis of the hippocampal formation. The axis was defined by the line tangent to the ventral border of the subiculum²¹ (figure 1A). The spatial relation between the locus of the center of the magnetic field and that of PET images had been calibrated in advance. This calibration permitted us to perform quantitative PET scans set parallel to an arbitrarily sectioned MRI plane by tilting PET detector rings.²² The same rigid head fixation was performed using a thermoplastic face mask designed for surgical operation during both MRI and PET measurements.

PET measurement. PET was performed using a high-resolution scanner (SHR2400, Hamamatsu Photonics K.K., Hamamatsu, Japan), with five detector rings yielding nine-slice images simultaneously, and with a tilt gantry system that could be moved from -20 deg to +90 deg.²³ The scanner's spatial resolution is 2.7 mm horizontally and 5.5 mm axially. After the gantry was set parallel to the hippocampal longitudinal axis determined by MRI, a dose of 185 MBq of[¹⁸F]-FDG was injected and arterial blood sampling was commenced simultaneously for quantitative measurement. Determination of CMRG1c was based on an autoradiographic method.²⁴

Data analysis. Boundaries of hippocampal subdivisions and amygdala were determined on MR images according to the human hippocampal atlas¹² and the results of hippocampal MRI.¹⁵ Briefly, the amygdala was outlined around the amygdaloid nuclei (ovoid mass of gray matter) on the MR images to avoid overestimation of the amygdala volume of the surrounding area.⁹ The head of the hippocampus was demarcated around the digitation part, and the border between the hippocampal head and the amygdala was determined as the uncal recess separating them.¹⁵ The body of the hippocampus was regarded as the straight part of the floor in the lateral ventricle (temporal horn), and its tail was the posterior and arched part of the hippocampus.¹²

Irregular regions of interest (ROIs) larger than two full widths at half maximum (FWHM), with a FWHM that can avoid an influence of the partial volume effect,²⁵ were drawn on the amygdala,²⁶ the hippocampal formation (head, body, tail; see figure 1, B and C),^12,26 and the surrounding cortices on the reference MR images using an image processing system (Dr View, Asahi Kasei Company, Tokyo, Japan)²⁷ on a SUN workstation (Hypersparc ss-20, SUN Microsystems, Los Angeles, CA). Reconstructed PET and MR images were obtained parallel to the hippocampal longitudinal axis so that they theoretically required no reorientation procedure. However, in some cases of head motion during scanning, we had to rearrange PET images on their sagittal views by visual inspection using Dr View imaging software. The MRI voxel size was adjusted to that of PET three-dimensionally. These MR images, with identical 3D scales and coordinates to the PET images, were used as anatomic references for the following PET ROI analysis. The ROI setting was performed primarily by two technologists who were not informed of the patient's diagnosis. If the ROI setting was completed on the reformatted MR images and the PET images were displayed alongside the MR images, then the determined ROIs were placed on the same area on both the MR and the corresponding PET images. Because the hippocampal formation is about 4.5 cm in length, 1.5 to 2 cm in width and depth at its head level, and 1 cm at its body level, one HAP PET image that condensed data of a 6.5-mm thickness contained the metabolic information of the hippocampus in the z direction. Thus, we regarded the ROI value on a single image in HAP as a regional (volume) value of CMRGlc. Regional CMRGlc in each concerned region was determined by averaging the right and left values.

During volumetric analysis, in reference to the hippocampal atlas¹² and the results of MRI,¹⁵ we drew ROIs on the hippocampus and the amygdala in five or six contiguous MRI slices covering these areas. This procedure was reevaluated occasionally by referring to the secondarily reformatted coronal images vertical to HAP.²⁶ The volume was measured by multiplying the ROI by the slice thickness. Preliminary studies on reproducibility of this volume assessment using small rectangular objects scanned on separate days showed that the coefficient of variation expressed in percent SD of the mean was 8.3%. This value was consistent with that determined in a previous MRI volumetric study.⁹

Statistics. Statistical analysis for CMRGlc was first performed using one-way ANOVA with post hoc Scheffe's F-test with respect to one intersubject factor (group) with three levels (AD, pure amnesia, control) and one intrasubject factor (cerebral region consisting of five cortical areas after the values from all structures in the medial temporal area were averaged as a value of the medial temporal lobe) to evaluate the level of metabolic reduction in the medial side of the temporal lobe. In addition, because no close reaction was observed in two-way ANOVA between the locations(i.e., medial or lateral side of the temporal lobe and types of diagnosis), one-way ANOVA was performed to compare the metabolic effects in either side separately. Statistical analysis for volume was performed using one-way ANOVA with post hoc Scheffe's F-test. Because post hoc multiple comparisons were performed in both analyses, statistical significance was taken as p < 0.05.

Results. Clinical characteristics. There was no difference in the intelligence scores between patients with pure amnesia and normal controls, whereas marked intellectual deterioration was observed in patients with AD (see table 1). Short-term memory (digit span) was preserved in patients with pure amnesia. Performance on verbal and visual memory tests was equally deteriorated in patients with pure amnesia and AD. ADAS and CDR disclosed that Alzheimer's patients participating in this study were within the moderately severe range.

MRI volumetric study. One-way ANOVA showed that the mean volumes of the hippocampus in pure amnesics did not differ from normal control subjects, whereas the left hippocampal volume in AD patients was significantly lower than that in other groups (table 2). The right hippocampal volume tended to be lower in AD patients. The average amygdaloid volumes in pure amnesics tended to be lower than those in control subjects, but this difference was not significant. In AD patients, however, bilateral amygdaloid volumes were reduced significantly compared with control values. Regression analysis showed no significant correlations between these shrinkage levels of either hippocampus or amygdala and CMRGlc values (not shown).

PET metabolic study. The cerebral metabolic rate of glucose in the medial temporal area was significantly lower than that in the lateral temporal area in normal control subjects (table 3^,figure 2). Post hoc testing showed that there was a significant reduction in CMRGlc in the lateral temporal and parietal cortex in AD patients compared with normal control subjects and pure amnesics.

Within the medial temporal area, the value of CMRGlc in the hippocampal head was a little higher than that in normal control subjects, although this difference was not significant. One-way ANOVA indicated significant CMRGlc reductions in the hippocampal head in pure amnesics, and in the amygdala and the hippocampal head in AD patients.

Discussion. The current study focused on in vivo quantitative measurement of glucose metabolism in the whole hippocampal formation in patients with memory impairment. Our findings are consistent with the results from other PET studies^4,28,29 that show that glucose consumption in the human mesiotemporal region is lower than in the lateral cortices. Our observation of a slight increase in metabolic rate in the hippocampal head in normal subjects suggests that the head region requires more energy than other portions of the hippocampal formation. Comparisons among constituents of the amygdalohippocampal system indicated that only hippocampal head glucose metabolism was significantly lower in patients with pure amnesia, whereas extensive metabolic reductions were observed over the mesial temporal area including the hippocampal head and the amygdala, and the temporoparietal association cortex in patients with AD. These findings suggest that metabolic alterations in the hippocampal head may be important for the pathophysiology of patients with memory impairment and that altered functions in the amygdala and the parietotemporal association area may lead to clinical manifestations of AD. This hypometabolic pattern in the posterior part of the brain in AD patients was reported previously.^30-33 Glucose metabolism in the entorhinal cortex might have been affected because of its proximity to the amygdala and the hippocampal head, but this parahippocampal metabolism was not examined in our study because of the difficulty in identifying the field on MR images.

We used a high-resolution PET scanner equipped with a gantry tilt system, which enabled us to focus on small structures in arbitrarily sectioned plane images. However, regardless of whether a high-resolution scanner is used, the partial volume effect is always a problem in radioisotope-related tomographic studies. In the current study we found that CMRGlc of the hippocampal head was decreased in patients with pure amnesia relative to normal subjects. This significant metabolic reduction could not be ascribed to the partial volume effect due to its small size or existing atrophy, because volumetric MRI analysis showed no significant differences in volume of the hippocampus between normal subjects and pure amnesics. The volumes of the hippocampus and amygdala in AD patients were, however, significantly lower compared with the other two groups. The degree of shrinkage was 25 to 30% in total, supporting the results of a previous MRI volumetric study.³⁴ Consistent with another PET study of pure amnesia,³⁵ regression analysis in the current study disclosed no correlation between the whole hippocampal metabolic depression and its structural alterations in either pure amnesics or AD patients (not shown). The limitation was that our MRI did not allow volumetric measurement of the hippocampal subdivisions. However, this lack of a correlation also suggested that the effect of partial volume was not a critical factor affecting the CMRGlc values in our PET study. Although there were significant reductions in the volumes of amygdala and hippocampus in AD patients, the metabolic reductions in these areas were essential to the pathology of AD because it was reported that hypometabolism in atrophic brain regions was consistently observed as a true biochemical abnormality after correction for the partial volume effect of PET data in patients with AD.³⁶

The head of the hippocampus consists of an anterior part of the arc of the hippocampus or hippocampal digitations.¹² The digitations cover a large part of the CA1 field,¹² which is subject to selective damage in patients with AD¹³ and postanoxic amnesia.¹⁴ The current study showed that the energy metabolism of the hippocampal head was higher (although not significantly) than that in other hippocampal substructures in normal subjects. This suggests that the energy consumption of synaptic transmission through the perforant pathway³⁷ is higher in this area to maintain appropriate cognitive functions, such as memorization of new events, and that conversely metabolic depression in this region can cause the memory dysfunction seen in amnesic patients. This speculation is supported by the results of a previous electrophysiologic study showing that the slow, rhythmic activity that is active during exploratory behavior was localized in CA1 and the dentate gyrus.³⁸ Lesion of the CA1 field also caused severe impairment of the retention of new tasks³⁹ and anterograde memory impairment.⁴⁰ The hippocampal head is adjacent to the amygdala anteriorly and surrounded superficially by the entorhinal cortex, which sends major excitatory projections into CA1 of the hippocampus. This close physical proximity along with the strong connective relationship that reflects a tight link among these structures⁴¹ may explain the necessity of the high-energy metabolism in the hippocampal head.

It is interesting that amygdaloid metabolism was preserved in pure amnesics but not in AD patients in our study. It is widely accepted that the amygdala is involved in the neural substrate of emotional behavior⁴² and recognition of emotional changes in humans.⁴³ In nonhuman primates there is also evidence that the amygdala is an important component of the neural system involved in social behavior.⁴⁴ These findings support our results, which show no reductions in amygdaloid metabolism or volume in pure amnesics who had neither emotional nor social problems except for memory disturbance. This suggests that the amygdala is not primarily involved in sustaining memory function. However, the amygdala was reported to undergo marked degeneration in patients with senile dementia⁴⁵ and shrinkage in patients with AD.⁹ Because the amygdala receives direct projections from the temporoparietal association area,⁴⁶ which connects the hippocampus indirectly via the entorhinal cortex, functional alterations in the temporal and parietal cortices can alter metabolic flow in the amygdalohippocampal system and vice versa. Therefore, the amygdaloid metabolic and volumetric reductions in dementia potentially have a wide field of influence and may contribute to the widespread behavioral disturbances characteristic of AD.^47,48

Our findings suggest that quantitative measurement of the brain glucose or oxygen metabolism using a sophisticated PET scanner designed to focus on small structures of interest, such as cerebral deep nuclei, hippocampus, and amygdala, in conjunction with volumetric MRI analysis, is important in detecting in vivo focal metabolic abnormalities that may contribute to the pathophysiology of disease.

Acknowledgment

We thank Mr. Kanno for his technical support.