Mapping of hippocampal cerebral blood volume (CBV) defects in human patients with probable Alzheimer disease (AD). A, Section of a T1-weighted image from a patient with AD in which hippocampal external morphologic appearance and internal architecture (outlined in white in middle panel) is best visualized. Regions of interest (right) are identified within the entorhinal cortex (red), dentate gyrus (blue), CA1 subfield (green), and subiculum (orange). B, Spatial pattern of mean relative cerebral blood volume (rCBV) (left) measured from each hippocampal subregion. Predominant dysfunction was observed in the entorhinal cortex (EC). Logistic regression analysis (right) investigating the hippocampus as a circuit showed that the dentate gyrus (DG) was the only subregion with a positive odds ratio in distinguishing the groups. Limit lines indicate standard error. C, Hippocampal CBV maps from a patient with AD vs an age-matched control subject. Warmer colors reflect more CBV.
Hippocampal cerebral blood volume (CBV) correlated with brain function. A, Results of neuropsychological testing showing that patients with Alzheimer disease (AD) performed more poorly than controls across multiple cognitive domains, including tests assessing hippocampal-dependent memory (total recall and delayed recall on the Selective Reminding Test) and tests assessing hippocampal-independent cognition, such as language (category fluency), visuospatial abilities (Benton matching), and abstract reasoning (similarities). Limit lines indicate standard error. B, Entorhinal cortex (EC) CBV and, to a lesser extent, CA1 CBV correlated with performance on hippocampal-dependent tasks (top), but not on hippocampal-independent tasks (bottom).
Mapping of hippocampal cerebral blood volume (CBV) defects in J20 transgenic mice that modeled Aβ-related neurotoxicity. A, Section of a T2-weighted image in which hippocampal external morphologic appearance and internal architecture (outlined in white in lower middle panel) is best visualized. Regions of interest (bottom right) are identified within the entorhinal cortex (EC) (red), dentate gyrus (blue), CA3 subfield (yellow), and CA1 subfield (green). B, Hippocampal CBV maps from a 21-month-old mutant mouse (left) vs a wild-type littermate (right). Warmer colors reflect more CBV. C, Spatial pattern of mean relative CBV (rCBV) (top left) measured from each hippocampal subregion in mutant J20 mice expressing a double mutation in amyloid precursor protein and wild-type littermates. Predominant dysfunction was observed in the EC. Logistic regression analysis (bottom left) assessing the hippocampus as a circuit showed that the dentate gyrus (DG) was the only subregion with a positive odds ratio in distinguishing mutants from wild-type littermates. Temporal analysis (top right) showed that EC dysfunction began at approximately 6 months of age. Temporal analysis (bottom right) showed a significant aging effect that was driven primarily by age-related decline in dentate gyrus CBV.
Effect of flurbiprofen on hippocampal cerebral blood volume (CBV) defects in transgenic mice. A, Two groups of mutant mice 6 to 8 months of age were imaged 4 times during 5 weeks. After baseline imaging, mice were imaged at 3 weeks after a course of flurbiprofen (drug group) or vehicle solution (vehicle group), and then imaged twice after the end of treatment, at weeks 4 and 5. B, Changes in entorhinal cortex CBV over time. The increase in the drug group continued after the end of treatment. Images (bottom) show 2 examples generated by subtracting week 5 CBV from baseline CBV. Warmer colors represent greater increases in CBV. The greatest drug effect was observed in the entorhinal cortex.
Moreno H, Wu WE, Lee T, Brickman A, Mayeux R, Brown TR, Small SA. Imaging the Aβ-Related Neurotoxicity of Alzheimer Disease. Arch Neurol. 2007;64(10):1467-1477. doi:10.1001/archneur.64.10.1467
Neurotoxicity related to the Aβ peptide is thought to be a primary mechanism of dysfunction in Alzheimer disease (AD). Although numerous imaging studies have observed brain dysfunction in AD, whether these imaged defects reflect Aβ-related neurotoxicity remains unknown.
To study Aβ-related neurotoxicity by means of functional imaging maps of the hippocampal formation in human patients and mouse models.
Cross-sectional study comparing humans with AD and control subjects, cross-sectional study of J20 mice, a transgenic mouse model of AD, and a longitudinal study of flurbiprofen administration to transgenic mice.
Alzheimer disease research center.
Eleven subjects with probable Alzheimer disease and 11 age-matched controls, plus J20 mice and wild-type littermates.
In the first study, human subjects and controls underwent magnetic resonance imaging. In the second study, mice underwent imaging at 3, 6, 12, 15, and 21 months of age, for a total of 57 imaging experiments. In the third study, 12 J20 mice underwent imaging repeatedly over time; 6 received flurbiprofen to ameliorate Aβ-related neurotoxicity and 6 received vehicle control.
Main Outcome Measures
Comparison of hippocampal functional maps.
Among all hippocampal subregions, the entorhinal cortex was the dominant site of dysfunction observed in both human patients and J20 mice. Long-term administration of flurbiprofen rescued entorhinal cortex dysfunction in transgenic mice.
Our results establish that the neurotoxicity related to the Aβ peptide can be captured in vivo by functional imaging and suggest hippocampal subregions most vulnerable to its toxic effects.
During the past few decades, the field of functional imaging has identified 4 primary variables that correlate with glucose or oxygen metabolism and that can be used to map functional defects in the brain. The first is glucose uptake, a correlate of glucose metabolism, which is typically measured with positron emission tomography (PET).1 The other 3 variables—cerebral blood flow, deoxyhemoglobin content, and cerebral blood volume (CBV)—are hemodynamic correlates of oxygen metabolism2 and can be measured with PET, single-photon emission computed tomography, or magnetic resonance imaging (MRI).
All 4 functional imaging variables have been applied to Alzheimer disease (AD) and have detected disease-related defects,3 including defects in the hippocampal formation, a structure where the disease begins. Nevertheless, AD is characterized by a range of pathophysiologic abnormalities,4 and it remains unknown which underlying abnormality is driving the defects observed by functional imaging. Notably, it is generally acknowledged that AD causes neuronal dysfunction before neuronal cell death,5 and either cytopathology can affect brain metabolism and cause functional imaging defects. Furthermore, AD is characterized by a range of histologic and biochemical findings, including tau hyperphosphorylation, neurofibrillary tangles, elevated concentrations of Aβ peptide, and fully formed amyloid plaques. Among these, a convergence of studies have suggested that soluble forms of Aβ peptide represent a primary neurotoxin in the disease and that Aβ-related neurotoxicity causes neuronal dysfunction before cell death.5- 7
Accordingly, because Aβ-related neurotoxicity appears to represent a primary stage in the progression of AD pathophysiologic change, it has become important to clarify whether functional imaging can capture this stage, or whether neuronal cell death and/or neurofibrillary tangles are required for the imaged defects observed in a growing number of functional imaging studies.
With this question in mind, we became interested in transgenic mice that express disease-causing mutations in the amyloid precursor protein (APP) because they model the Aβ-related neurotoxicity of AD. These mice typically develop elevations in soluble and insoluble forms of Aβ peptide, and—as documented by behavior and by electrophysiology—develop neuronal dysfunction in the hippocampal formation.6 At the same time, however, these mice do not typically develop neuronal cell death, tau hyperphosphorylation, or neurofibrillary tangles. Thus, because of their constrained pathological features, these mice provide an opportunity to test whether Aβ-related neurotoxicity can be captured by in vivo imaging.
Among the different imaging variables, MRI-based techniques that assess regional CBV have been found well suited to image the hippocampal formation and its subregions across different species.3,8 Previous studies have already shown that MRI measures of CBV can detect AD-related dysfunction9- 11 and that, in AD, CBV tightly correlates with PET measures of glucose uptake.9 By providing proof of principle, these studies address the concern that a possible breakdown in the blood-brain barrier—an unresolved debate in AD12—might violate technical assumptions that underlie MRI measures of CBV.
Nevertheless, the hippocampal formation is made up of multiple subregions, and MRI technologies were optimized that can generate high-resolution hippocampal CBV maps in humans and in mice. In a previous study, our group used an MRI approach that relies on intravenous (IV) injections of a contrast agent containing gadolinium, which we specifically tailored so that high-resolution CBV maps of the hippocampal formation could be generated in rhesus monkeys.8 In the present study, we optimized this approach so that it is applicable to the human hippocampus.
Although similar approaches have been used to generate CBV maps of the mouse brain,13- 17 they are limited by the requirement for IV injections of contrast agents. Because of the difficulty in gaining IV access in mice, these approaches prohibit longitudinal testing over time, thereby negating one of the promised advantages of in vivo imaging. We recently introduced an MRI approach designed to overcome this limitation.18 In this approach, the gadolinium-based contrast agent is administered via intraperitoneal injections; by optimizing acquisition parameters, we have shown that this approach is capable of generating high-resolution CBV maps of the mouse brain repeatedly and safely over time.18
In this study, these technologies were applied for the first time to human patients and to transgenic mice. The cross-species studies were designed to specifically test whether functional imaging of the hippocampal formation is sensitive to Aβ-related neurotoxicity. Because high-resolution CBV maps of the hippocampal formation have not yet been reported in the disease, we began by establishing these maps in human patients and age-matched control subjects. Next, we established CBV maps of the hippocampal formation in J20 transgenic mice that express mutations in APP and model Aβ-related neurotoxicity. Finally, we show that in mice the spatial pattern of dysfunction is reversed after the administration of a drug that reduces Aβ-related neurotoxicity.
Subjects were recruited from a study investigating a random sample of elderly individuals residing in northern Manhattan in New York, New York. As previously described,19 all subjects received detailed medical, neurologic, and neuropsychological evaluations. In a weekly consensus conference composed of neurologists and neuropsychologists, each subject was examined and a clinical diagnosis was made according to the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition)20 and National Institute of Neurological and Communicative Diseases and Stroke–Alzheimer's Disease and Related Disorders Association criteria.21 Eleven subjects diagnosed as having probable AD (all with a clinical disease rating scale score of 1) were recruited to participate in this study. Then, 11 age-matched controls were selected from a group of individuals who were free of brain disease, including dementia, mild cognitive impairment, stroke, Parkinson disease, or depression. The demographic characteristics of these 2 groups of subjects are summarized in Table 1.
As part of the detailed neuropsychological battery,19 hippocampal-dependent memory was evaluated by means of the total and delayed recall components of the Selective Reminding Test. Hippocampal-independent tasks included the category fluency component of the Boston Diagnostic Aphasia Examination, which assesses language function22; the Wechsler Adult Intelligence Scale–Revised similarities test, which assesses abstract reasoning23; and the Benton Visual Retention Test,24 which assesses visuospatial ability.
The J20 transgenic strain of mice (PDGF-APPSwInd 20Lms/1J) expresses disease-causing double mutations in APP. Expression of the APP transgenic insert is directed by the human platelet-derived growth factor β-polypeptide (PDGFB) on a C57BL/6 strain background. Founder animals were bred in couples consisting of a wild-type female and a transgenic male. Offspring were identified by genotyping, with the use of tail tissue genomic DNA, following the protocol recommended by The Jackson Laboratory (http://jaxmice.jax.org/index.html). A total of 57 imaging studies were performed covering the mouse lifespan: 3-month-old mice (n = 12; 6 mutants and 6 wild types), 6-month-old mice (n = 8; 4 mutants and 4 wild types), 12-month-old mice (n = 10; 5 mutants and 5 wild types), 15-month-old mice (n = 10; 5 mutants and 5 wild types), and 21-month-old mice (n = 14; 7 mutants and 7 wild types). The mutants were equally split across sex, while the composition of the wild-type group was 35% female and 65% male (19 vs 35 mice). Although not statistically significant (χ2 = 0.06), sex was included as a covariate in all of the analyses. The weight of the animals ranged from 26 to 43 g for transgenic mice and 24 to 39 g for controls.
Subjects underwent imaging with a 1.5-T scanner (Intera; Royal Philips Electronics, Eindhoven, the Netherlands). For each subject, 2 sets of oblique coronal 3-dimensional T1-weighted images (repetition time, 20 milliseconds; echo time, 6 milliseconds; flip angle, 25°; in-plane resolution, 0.86 × 0.86 mm; section thickness, 4 mm) were acquired perpendicular to hippocampal long axis. The first series of images was acquired before and the second was acquired 4 minutes after IV administration of a standard dose of gadodiamide (Omniscan; 0.1 mmol/kg).
Images were then transferred to a workstation containing an analysis software package (MEDx Sensor Systems, Sterling, Virginia). An investigator blinded to subject grouping (W.E.W.) performed all image processing. The relatively short acquisition time minimized head motion; nevertheless, the automated image registration program was used to coregister the images.25 A Gnu plot was generated to assess the quality of the coregistration, and an individual study was rejected if a shift greater than 1-pixel dimension was detected. Three studies (2 in subjects with AD and 1 in a control subject) were rejected for poor motion correction.
To generate CBV maps, the precontrast image was subtracted from the postcontrast image, and the difference in the sagittal sinus, which serves as an estimate of the image intensity change of 100% blood, was recorded. The subtracted image was then divided by the difference in the top 4 pixels measured from the sagittal sinus and multiplied by 100, yielding relative CBV maps.26- 28 Among the series of oblique coronal images, we have consistently found that a section anterior to the lateral geniculate nucleus and posterior to the uncus provides optimal visualization of hippocampal morphologic features and internal architecture (Figure 1A). The external structure of the hippocampus was manually traced, as was the internal architecture that follows the hippocampal sulcus and the internal white matter tracts (Figure 1A). With the aid of standard atlases,29,30 regions of interest (ROIs) of 4 subregions of the hippocampal formation were identified according to the following anatomic criteria:
Entorhinal cortex: The inferolateral boundary follows the collateral sulcus; the medial boundary is the medial aspect of the temporal lobe; the superior boundary is the hippocampal sulcus and gray-white distinction between the subiculum and the entorhinal cortex.
Subiculum: The medial boundary is the medial extent of the hippocampal sulcus and/or the horizontal inflection of the hippocampus; the inferior boundary is the white matter of the underlying parahippocampal gyrus; the superior boundary is the hippocampal sulcus; the lateral boundary is a few pixels medial to the vertical inflection of the hippocampus.
CA1 subregion: The medial boundary is 2 to 3 pixels lateral to the end of the subiculum ROI, approximately at the beginning of the vertical inflection of the hippocampus, and the extension of the hippocampal sulcus/white matter tracts; the inferior boundary is the white matter of the underlying parahippocampal gyrus; the superior boundary is the top of the hippocampal formation.
Dentate gyrus: The medial boundary is the medial extent of the temporal lobe; the inferolateral boundary is the hippocampal sulcus–white matter tracts; the superior boundary is the top of the hippocampal formation, where the alveus is typically identified.
Of note, the border zones between any 2 subregions cannot be identified without histologic landmarks and were therefore excluded from the ROIs. Mean relative CBV from the ROI of each hippocampal subregion was measured from each subject and used for group data analysis.
Mice were imaged with a spectrometer (AVANCE 400WB; Bruker Instruments Inc, Billerica, Massachusetts) outfitted with an 89-mm bore, a 9.4-T vertical Brucker magnet (Oxford Instruments Ltd, Abingdon, England), a 30-mm–inner diameter birdcage radiofrequency probe, and a shielded gradient system (100 G/cm).
Head motion is a variable that must be minimized in high-resolution imaging of the mouse brain. Anesthesia is typically used in functional MRI studies of the rodent31,32 because it reduces head motion and minimizes the fear and anxiety induced by the loud environment of the scanner. The intended goal of any anesthetic agent is to affect pain perception and/or level of consciousness, and thus, by definition, an anesthetic agent affects neuronal function. With this concern in mind, we explored 2 alternatives to anesthesia: the use of neuromuscular blocking agents coupled with endotracheal intubation and the use of physical constraints connected to skull-implanted screws. Neither of these alternatives reduces fear and anxiety, which will also affect neuronal function. Furthermore, endotracheal intubation is an invasive procedure in mice, thus prohibiting multiple imaging over time.
As with other functional MRI studies in rodents,31,32 we used isoflurane for anesthesia (induction phase, 3 vol%; maintenance, 1.1 to 1.5 vol% at 1 L/min air flow via a nose cone). Previously, we have found that isoflurane does not significantly affect heart rate, respiratory rate, and oxygen saturation in C57BL/6 mice.18 Although isoflurane causes a decrease in global glucose uptake, it does not significantly uncouple the hemodynamic response at 1 minimum alveolar anesthetic concentration.33,34
At high fields, T2-weighted scans generate images with superior contrast compared with T1-weighted scans. As in previous studies,16 therefore, we relied on T2-weighted scans for CBV mapping in mice. Three practice scans were first acquired to position the subsequent T2-weighted images along the standard anatomic orientations in a reproducible manner. Optimal axial images acquired through the long axis of the hippocampal formation were determined empirically by repositioning the animal and manual shimming. T2-weighted images were obtained with a fast spin-echo sequence with repetition time/effective echo time of 2000/80 milliseconds, rapid acquisition with relaxation enhancement35 factor of 16, field of view of 20 mm, acquisition matrix of 256 × 256, 10 sections, section thickness of 0.6 mm, section gap of 0.1 mm, number of excitations of 28, voxel size of 0.0078 cm × 0.0078 cm × 0.7 mm (0.0042725 mm3), and an in-plane resolution of 100 μm. Each set of images required 15 minutes, and 5 sets of images were acquired sequentially. The first 2 sets were acquired before administration of contrast material, and 3 more sets of images were acquired after contrast was given. Gadodiamide (Omniscan; 10 mmol/kg) was injected via a 0.6-mm catheter placed intraperitoneally (IP). At the end of imaging, 2 mL of isotonic sodium chloride solution was injected IP to improve the clearance of the contrast agent.18 Signal to noise ratio and contrast to noise ratio were measured as previously described.18 Briefly, signal to noise ratio was assessed by dividing the mean signal intensity measured from the whole brain by the standard deviation of air surrounding the brain. Contrast-to-noise ratio was assessed by measuring the contrast-induced percentage change in signal measured observed in the thalamus (comparing the last postcontrast image to the precontrast image). If contrast- to-noise ratio was less than 20%, an extra dose of gadodiamide, 3 mmol/kg, was administered and a new postcontrast series was obtained 15 minutes later. If a 20% signal change was not achieved (seen only in 2 cases), the images were rejected for analysis.
Heart rate, respiratory rate, and oxygen saturation were continuously monitored by means of pulse oximetry (model V33304; SurgiVet, Waukesha, Wisconsin). The probe was attached to the lower abdomen. Rectal temperature was continuously monitored with a thermistor (Precision Thermometer 4000A; YSI Temperature, Dayton, Ohio). The electrocardiogram was monitored with a vital sign monitor (Physiogard SM 785; Brucker Instruments Inc) by using subcutaneous silver electrodes from the front limbs and the reference electrode placed at the right posterior limb. Normality of physiologic measures was used as the criterion to continue with the imaging session.
Investigators blinded to subject grouping (W.E.W. and T.L.) performed all image processing. Images were motion corrected, and the same criterion for accepting or rejecting an individual study as described for human studies was applied. Relative CBV was mapped as change in transverse relaxation rate (ΔR2) induced by the contrast agent. Note that R2 = 1/T2. When the contrast agent reaches uniform distribution, then CBV maps can be measured from steady-state T2-weighted images as follows:
CBV α ΔR2 = ln(Spre/Spost)/TE,
where TE is the effective echo time, Spre is the T2-weighted signal before contrast administration, and Spost is the T2-weighted signal after the contrast agent reaches steady state. We have previously studied the kinetics of IP gadodiamide–generated CBV maps in C57BL/6 mice, in which the 37.5-minute point was identified as the optimum postcontrast time interval.18 The derived ΔR2 maps were then divided by 4 pixels with the highest ΔR2 measured from the posterior cerebral vein, yielding relative CBV maps.11,28
As in the human subjects, a section acquired through the body of the hippocampal formation consistently provides optimal visualization of the hippocampus and internal architecture. The external morphologic structure of the hippocampus was manually traced, as was the internal architecture that follows the hippocampal sulcus and the internal white matter tracts. With the aid of a standard atlas,36 ROIs of 4 hippocampal subregions were then identified by relying on the following anatomic criteria:
Entorhinal cortex: The lateral boundary follows the gray-white junction of the parahippocampal gyrus; the medial boundary is the medial aspect of the temporal lobe; the superior boundary is the beginning of the collateral sulcus; the inferior boundary is the lateral tip of the brain.
CA1 subregion: The lateral boundary is the lateral aspect of the hippocampal proper; the medial boundary is the hippocampal fissure; the superior boundary is the superior curvature of the hippocampus proper; the inferior boundary is the inferior curvature of the hippocampus proper.
CA3 subfield: The lateral boundary is the beginning of the superior curvature of the hippocampus proper; the medial boundary is the end of this curvature; the superior boundary is the superior aspect of the hippocampus proper; the inferior boundary is the superior tip of the hippocampal fissure.
Dentate gyrus: The lateral boundary is the lateral aspect of the hippocampus proper; the medial boundary is the hippocampal fissure; the superior aspect is the superior curvature of the hippocampus proper; the inferior boundary is the inferior aspect of the hippocampus proper.
Of note, the border zones between any 2 subregions cannot be identified without histologic landmarks and were therefore excluded from the ROIs. The minimum size of the ROIs was 14 pixels. Mean relative CBV from the ROI of each hippocampal subregion was measured from each subject and was used for group data analysis.
To further confirm that the observed CBV defect in mice was related to Aβ neurotoxicity, we turned to flurbiprofen, a drug shown effective in transgenic mouse models of the disease.37- 39 Guided by these studies, we designed a chronic, rather than an acute study, investigating transgenic mice when they first manifested dysfunction, between 6 and 8 months of age. Six mice (3 of each sex) assigned to the drug group received daily subcutaneous (SC) injections of generic flurbiprofen, 5 mg/kg per day (Sigma-Aldrich Corp, St Louis, Missouri), while 6 mice (3 of each sex) assigned to the vehicle group received daily SC injections of 19:1 castor oil in a dimethyl sulfoxide vehicle. Both groups received injections for 3 weeks, and imaging was repeated according to the following timeline: baseline, 3 weeks (the end of treatment), 4 weeks, and 5 weeks.
Hippocampal maps of CBV were generated at each time point and longitudinal data were analyzed by means of generalized estimated equations (GEE).40 The GEE method examines changes in CBV in each hippocampal subregion over time. The dependent variables were the subregional normalized CBV values, and the independent variables were J20 mouse group (flurbiprofen vs vehicle), time (included as a continuous variable), and interaction of group × time. The GEE analysis yields β values, which represent associations between a factor score (CBV) and variables included in the model. A significant group effect would indicate a difference between 2 groups at the baseline. A positive β value indicates that the group with a specific variable performed better than the group without that variable. A significant time effect would indicate a change in a factor score over the total duration of the experiments. A significant interaction effect (group × time) would indicate a difference in the rate of change in a factor score (CBV) between the 2 members of the group.
Eleven patients with probable AD (mean age, 83.9 years) and 11 age-matched controls (mean age, 84.5 years) (Table 1) underwent imaging before and after IV administration of gadodiamide, and, as previously described,8 contrast-enhancement values were used to derive CBV maps of the brain. Images were acquired perpendicular to the hippocampal long axis and, for each case, an optimum section was identified that contained all hippocampal subregions and in which the external morphologic features and the internal architecture were best visualized. According to these landmarks, anatomic criteria were used to parcel the hippocampal formation into the following hippocampal subregions: the entorhinal cortex, the dentate gyrus, the CA1 subfield, and the subiculum (Figure 1A). The CA3 subfield could not be reliably identified. The ROIs were manually identified and mean CBV values were measured from each hippocampal subregion. An interclass correlation showed a high interrater reliability when using 3 blinded investigators and 8 subjects (4 with AD and 4 controls) (entorhinal cortex, r = 0.88; subiculum, r = 0.85; CA1, r = 0.84; and dentate gyrus, r = 0.81).
An analysis of variance (ANOVA), including sex, education, and ethnicity as covariates, showed that patients with AD had predominant reduction in entorhinal cortex relative CBV compared with controls (F = 6.7, P = .02) (Figure 1B). Logistic regression analysis, including group as the dependent variable and CBV from the different hippocampal subregions and demographics as covariates, showed that, compared with the entorhinal cortex alone, group classification increased from 81% to 100% when CBV values from all hippocampal subregions were included in the model. Interestingly, in this circuit-based analysis, the dentate gyrus was distinguished by having a positive odds ratio, suggesting that a relative increase in dentate gyrus CBV among the patients enhances group classification (Figure 1B).
On cognitive assessment with a neuropsychological battery, as expected, an ANOVA showed that patients with AD performed more poorly across multiple cognitive domains (Figure 2A), including hippocampal-dependent memory tasks (total recall on the Selective Reminding Test, F = 11.7, P = .003; and delayed recall on the Selective Reminding Test, F = 11.6, P = .003), and hippocampal-independent tasks such as language (category fluency, F = 5.4, P = .03), visuospatial abilities (Benton matching, F = 7.8, P = .01), and abstract reasoning (similarities, F = 4.7, P = .04). Linear regression analysis was used to test whether hippocampal CBV reflects cognitive function, including CBV values from all subregions as dependent variables and demographics as covariates. Results showed that hippocampal-dependent memory tasks were positively correlated with CBV values measured from the entorhinal cortex and, to a lesser extent, with values measured from the CA1 subfield (total recall: entorhinal cortex, β = 0.51, P = .007; CA1, β = 0.53, P = .01; delayed recall: entorhinal cortex, β = 0.51, P = .02; CA1, β = 0.41, P = .08) (Figure 2B). Correlation analyses were performed between CBV values from each of the hippocampal subregions and the different neuropsychological measures. No other statistically significant correlations were observed.
Beyond mapping a spatial profile, the ability to determine the precise temporal pattern with which dysfunction emerges is a potential advantage of imaging mouse models. Mapping a temporal course requires imaging a large number of mice, assessing both mutants and their wild-type littermates at multiple time points across their lifespan. Needing to focus our efforts on one transgenic strain, we chose J20 (PDGF-APPSwInd 20Lms/J) mice, which express disease-causing double mutations in APP. Previous studies have documented that this strain develops dysfunction in a delayed fashion, providing the opportunity to capture a predysfunction baseline. More importantly, because these mice have been studied extensively, their temporal profile of cognitive dysfunction, electrophysiologic defects, Aβ production, and plaque formation has been established.6,41
Mutant J20 mice and wild-type littermates were imaged at 3, 6, 12, 15, and 21 months of age, for a total of 57 MRI experiments. All mice underwent imaging before and after an IP injection of gadodiamide, and, as previously described,18 contrast-enhancement values were used to derive CBV maps of the hippocampal formation.
The entorhinal cortex, the dentate gyrus, the CA1 subfield, and the CA3 subfield were mapped (Figure 3A). Signal from the subiculum was found to be unreliable. The ROIs were manually identified and mean CBV values were measured from each hippocampal subregion. All measurements were made by investigators blinded to subject grouping (H.M. and T.L.). Intraclass correlation coefficients were used to compute interrater reliability estimates. A group of 12 (6 in the J20 group and 6 in the control group) 1-year-old mice were analyzed by 4 different raters. The correlation coefficients were high (r = 0.97, 0.91, 0.92, 0.95, and 0.86 for the entorhinal cortex, CA1, CA3, dentate gyrus, and posterior cerebral vein, respectively). To estimate measurement reliability, normalized CBV values for the same mouse were compared between 2 imaging sessions separated by a brief period, during which it was reasonable to expect minimal or no changes. Nine 6-month-old mice (5 in the J20 group and 4 controls) underwent imaging within 96 hours. The interclass correlations between values paired by mouse but randomly assigned to test or retest were r = 0.93, 0.92, 0.96, and 0.95 for the entorhinal cortex, CA1, CA3, and dentate gyrus, respectively.
An ANOVA including group and age as the 2 fixed factors showed a main effect for group, driven primarily by CBV changes in the entorhinal cortex (F = 11.3, P = .002) (Figure 3B); a main effect for age, driven primarily by CBV changes in the dentate gyrus (F = 2.0, P = .01) (Figure 3B); and a group × age interaction, observed in the entorhinal cortex and other hippocampal subregions (entorhinal cortex: F = 4.1, P = .006; dentate gyrus: F = 3.4, P = .02; CA3: F = 4.1, P = .006). The observed group effect demonstrated that, in this transgenic strain, the entorhinal cortex is the dominant site of hippocampal dysfunction (Figure 3B), while the group × age interaction suggests that over time dysfunction spreads throughout the hippocampal circuit. Indeed, post hoc analyses revealed that by 15 months of age CBV reductions were observed in all hippocampal subregions of J20 mice. The CBV changes in the entorhinal cortex were best modeled as a quadratic decline, establishing that entorhinal decline begins at approximately 6 months of age (Figure 3B). The observed aging effect agrees with previous studies suggesting that the dentate gyrus is a hippocampal subregion particularly vulnerable to the effect of aging.8,42- 44
Logistic regression analysis, including group as the dependent variable and CBV from the different hippocampal subregions as the covariate, showed a minor increase in diagnostic accuracy (from 81% to 87%) when CBV values from all hippocampal subregions were included in the model. Nevertheless, as in humans, the dentate gyrus was unique in having a positive odds ratio (Figure 3B), suggesting an equivalent circuit-based pattern of dysfunction in humans and mice.
Both flurbiprofen and vehicle groups received injections for 3 weeks and underwent imaging at baseline, 3 weeks, 4 weeks, and 5 weeks (Figure 4A). Results of GEE studies showed a significant group × time interaction only for entorhinal cortex CBV (β = 0.016, P = .03) (Figure 4B), demonstrating that, compared with vehicle control, flurbiprofen significantly and selectively improved entorhinal cortex CBV (Figure 4B). Interestingly, improvement in entorhinal cortex CBV continued 2 weeks after flurbiprofen was stopped, illustrating the additional information that can be gained by a longitudinal design. No effect was observed for CBV values measured from any other hippocampal subregion (Table 2).
In an effort to clarify the histopathologic changes that underlie functional imaging defects in AD, a series of cross-species imaging studies were completed by using a unified functional imaging variable in humans and mice. We began with human patients, replicating previous functional imaging studies (as reviewed by Wu and Small3) that have found entorhinal cortex defects in AD. In contrast to previous studies, which relied on alternative imaging variables (glucose uptake or deoxyhemoglobin content), entorhinal cortex defects were detected by means of CBV. The concordance across human studies and the fact that CBV tightly correlates with fludeoxyglucose F 18 PET9 suggests that, as previously reported, the entorhinal cortex is particularly vulnerable to AD. Moreover, showing that entorhinal cortex CBV selectively correlates with hippocampal-dependent memory validates CBV as a functionally meaningful measure of the brain.
Nevertheless, as with all human studies, the underlying histopathologic mechanism that drives the observed functional defects remains unknown. As implied by its name, functional imaging is, in principle, sensitive to neuronal dysfunction; however, cell death is an early occurrence within the entorhinal cortex of patients with AD, and cell death might be required for the functional imaging defects. In an attempt to address this question, we studied transgenic mice, which express disease-causing mutations in APP and which model the Aβ-related neurotoxicity of the disease. Particularly important to us, these mice develop neuronal dysfunction in the hippocampal formation but are notably free of cell death or neurofibrillary tangles. By imaging a large group of mutant mice and their wild-type littermates across their lifespan, we observed that these mice develop CBV defects at approximately 6 months of age. This is the approximate age in which these mice are reported to manifest Aβ-related neurotoxicity, as evidenced by accelerated Aβ production, hippocampal-dependent memory deficits, and hippocampal cell dysfunction measured by electrophysiology and histologic markers.6 Providing further support that the imaged CBV defect was, at least in part, related to Aβ neurotoxicity, transgenic mice receiving long-term administration of flurbiprofen showed a reversal of the entorhinal cortex CBV defect. Interestingly, the analysis suggested that the dentate gyrus was the primary hippocampal subregion affected by the aging process per se, agreeing with previous studies in humans, nonhuman primates, and rats.8,42- 44
Taken together, our studies demonstrate that Aβ-related neurotoxicity in mice is sufficient to cause functional imaging defects as detected by CBV maps of the hippocampal formation. The findings suggest that CBV is a functional imaging variable that can detect Aβ-related hippocampal dysfunction and does not require cell death or neurofibrillary tangles. Furthermore, because CBV correlates with all other functional imaging variables—cerebral blood flow, deoxyhemoglobin content, and glucose uptake—our results inform other functional imaging studies that have detected AD-related brain defects.3
Remarkably, the transgenic mice were found to have the same spatial pattern of CBV defects observed in the hippocampal formation as human patients. The cross-species similarities support the likelihood that the patterns represent the same underlying Aβ-related pathophysiologic mechanism. Interestingly, previous human studies have documented that levels of Aβ40 and Aβ42 peptide are abnormally high in the entorhinal cortex of brains of patients with AD and that, more so than insoluble amyloid plaques, soluble Aβ levels correlate with histologic indicators of neuronal dysfunction.45 Nevertheless, despite the similarities observed in humans and mice, we cannot exclude the possibility that neuronal cell death and neurofibrillary tangles are contributing to the functional imaging defects observed in patients with AD. Establishing that functional imaging with CBV is, however, sensitive to Aβ-related neurotoxicity clarifies the utility of in vivo imaging. Detecting AD at its earliest stages is one of the greatest challenges facing the field. Insofar as Aβ neurotoxicity is one of the earliest stages of AD, our findings suggest that functional imaging will play an important role in early detection. More importantly, compared with neuronal cell death, neuronal dysfunction is considered the disease's cytopathological feature most amenable to pharmacologic intervention. Thus, cross-species CBV maps of the hippocampal formation can be used to develop, optimize, and test for effective drugs against this devastating, and increasingly common, illness.
Finally, mapping a spatial pattern of hippocampal dysfunction is an important step toward a greater mechanistic understanding of the disease process. Within the microanatomy of the hippocampal formation, our findings suggest that in mice, as in humans, the entorhinal cortex is the hippocampal subregion dominantly affected by Aβ neurotoxicity, an observation that agrees with results of previous mouse studies.7 This finding sets the stage for future studies to focus on cellular levels of analysis. In particular, whether dysfunction in entorhinal cortex neurons reflects dendritic or axonal pathological changes remains unknown, a cellular question that has emerged as an important and unresolved issue in the pathophysiologic abnormalities of AD. Understanding the molecular basis for the observed patterns of regional selectivity is perhaps a more important question. Future studies are required to determine whether the molecular profile of the entorhinal cortex is such that cells in this hippocampal subregion differentially produce more Aβ or clear Aβ more slowly, or are simply more vulnerable to the downstream toxic effects of Aβ. In any case, uncovering the molecular factors that account for the disease's regional selectivity will help clarify the molecular biology of the disease.
Correspondence: Scott A. Small, MD, Department of Neurology, Columbia University, College of Physicians and Surgeons, 630 W 168th St, PH #19, New York, NY 10032 (firstname.lastname@example.org).
Accepted for Publication: March 26, 2007.
Author Contributions:Study concept and design: Moreno, Mayeux, Brown, and Small. Acquisition of data: Moreno and Small. Analysis and interpretation of data: Moreno, Wu, Lee, Brickman, Mayeux, and Small. Drafting of the manuscript: Moreno, Mayeux, and Small. Critical revision of the manuscript for important intellectual content: Moreno, Wu, Lee, Brickman, Mayeux, and Brown. Statistical analysis: Moreno and Small. Obtained funding: Mayeux and Small. Administrative, technical, and material support: Wu, Lee, Mayeux, and Brown. Study supervision: Moreno, Wu, and Small.
Financial Disclosure:None reported
Funding/Support: This work was supported in part by grants AG025161 and AG027476 from the National Institutes of Health, the McKnight Neuroscience of Brain Disorders Award, and the James S. McDonnell Foundation.
Additional Contributions: Kenneth Hess and Fan Hau assisted with mouse imaging, Christiane Reitz, MD, PhD, assisted with data analysis, Ana Pereira, MD, edited the manuscript, and Lennart Mucke, MD, Gladstone Institute of Neurological Diseases, provided the PDGFB on a C57BL/6 strain background.