Brad Hyman, Massachusetts General Hospital, Charlestown. Imagining the Natural History of Alzheimer’s Disease
After well-deserved "thank you’s" to Yves Christen and Jacqueline Mervaillie of the IPSEN Foundation, who organized this conference) held on 17 March 2003, Brad Hyman’s introductory remarks reviewed his accidentally, but appropriately, titled talk, "Imagining the Natural History of Alzheimer’s Disease." Although "Imagining" was a typo for "Imaging," Hyman used this to point out that, for many years, we were forced to piece together the neuropathological findings from postmortem tissue, and had to "imagine" the natural history of the pathological changes in the entire brain. The talks in this symposium would nearly all focus on imaging this natural history. Hyman discussed the whys, whats, whens, wheres and hows of in-vivo imaging in AD.
- The reasons for why to image include definitive diagnosis, definition of end-points for clinical trials, development of clinically relevant biomarkers to follow and direct new therapies. Another motivation lies in the need for presymptomatic detection of the pathophysiologic changes of AD, which would make possible preventative therapy-noting that significant pathology precedes the onset of clinical symptoms.
- What to image? This includes "negative changes" such as atrophy reflective of neuronal loss (by MRI) and functional changes in cerebral metabolism reflective of synapse loss (by FDG PET). "Positive" changes amenable to imaging include reactive gliosis, amyloid plaques (PET amyloid imaging), and neurofibrillary tangles. Hyman stated that these positive changes may prove to be the most sensitive. Also noted were the biochemical changes in Aβ and hyperphosphorylated tau, which can be measured in CSF.
- When to look? The prodromal stages of AD would be ideal, said Hyman. It was proposed that Aβ deposits first develop in this stage. The next stage would be the first clinical (or MCI) stage, when synapse loss and NFT pathology begin to develop. Finally, in clinical AD, amyloid plaque pathology was hypothesized to plateau, making it an unlikely candidate to correlate with clinical progression. However, neuron/synapse loss (i.e., atrophy) and NFT pathology continue to worsen through the clinical course of AD, making these targets possible correlates of clinical symptomatology.
- Where to look? This was defined by the regional topology of NFT (medial temporal lobe) and amyloid plaques (neocortex as a whole). Therefore, atrophy would first be expected in limbic and association cortex. With regard to negative changes in AD, Hyman presented data showing that the entorhinal cortex first shows neuronal loss by stereological neuronal counting. This is closely followed by the superior temporal sulcus. He also reported MRI data obtained in collaboration with Marilyn Albert showing shrinkage of the entorhinal and association cortices even before clinical symptoms occur. This data could be used with over 90 percent accuracy to predict who would convert from MCI to AD. Regarding positive changes, Hyman touched on multiphoton microscopic data (Brian Bacskai) and PET data, which is discussed further below.
- How best to image AD in vivo? The answer to this includes the use of biomarkers (Papassotiropoulos), MRI and PET amyloid imaging (Wisniewski, Klunk, and Engler), structural MRI (Fox and Thompson), and FDG PET (Baron and Reiman). The next presentation by Charles Duyckaerts on the pathology of AD set the stage for the rest of the talks.
Charles Duyckaerts, Hospital de la Salpêtrière, Paris. Neuropathology: Terminus Ad Quem
Dr. Duyckaerts reviewed the pathology of AD beginning with the development in 1903 of Bielschowsky’s silver stain that was used by Alzheimer’s and his student, Perusini, in their first description of the NFT in 1906. NFT pathology first appears in the entorhinal cortex and hippocampus. These changes are not clearly unique to AD, as they commonly occur in the elderly, but they pre-date amyloid pathology by as much as 20 years. In cortex, the pathology of AD is typically first seen in the form of diffuse amyloid deposits, which Duyckaerts hypothesized to progress to focal/compact deposits stained by Congo red and silver stains. Duyckaerts discussed the cholesterol that fills plaques and, in cell membranes, is concentrated on "rafts," which also contain AβPP and BACE. With regard to temporal development of the topology of AD pathology, it was noted that NFTs progress in a regular fashion from transentorhinal cortex to neocortex; several authors have described staging systems for this. Amyloid plaque pathology appears in the entire cortex at approximately the same time, although some areas appear somewhat more affected, such as the temporal, parietal, and frontal cortices. Duyckaerts focused on area 17, the primary visual cortex, which is highly affected by amyloid pathology in advanced AD, in contrast to primary sensorimotor cortex, which is less severely affected. He described the initial stage of plaque deposition in area 17 as diffuse plaques that occur in "mid-stage" AD. These are soon followed by Congo red-positive Aβ deposits, qualifying for the term "amyloid." Coincident with Congo red-positive plaques comes evidence of inflammation. Next to appear are the "crowns" of tau-positive dystrophic axons. Duyckaerts presented very interesting data suggesting these axons originate fairly nearby from cortico-cortical projections. In area 17 isocortex, NFTs appear last (although in the brain as a whole, NFT pathology is the first to appear, namely, in transentorhinal cortex).
Andreas Papassotiropoulos, University of Zurich. CSF Biomarkers for Diagnosis of Alzheimer’s Disease
Papassotiropoulos reviewed the key steps in the pathophysiology of AD, noting AβPP, Aβ, amyloid fibrils and plaques, NFT, microglial activation and cytokine release, and cell death. He pointed out that genetic modifiers exist at each step, which could serve as risk markers for AD. He hypothesized that there is a 15-30 year prodromal phase of AD, followed by a clinical phase that, at five to 10 years, is comparatively shorter. Papassotiropoulos then focused on biomarkers derived from this pathophysiology, namely tau, Aβ, NGF, interleukins, and AβPP. He focused his discussion on CSF Aβ42 and hyperphosphorylated tau. Each alone show a large overlap between AD and healthy controls, limiting their use as diagnostic markers, but the phospho-tau:Aβ ratio shows little overlap. Even so, Papassotiropoulos did not believe that the accuracy predicted for this measure added substantial accuracy to the clinical diagnosis of AD.
As a change in strategy, he focused on work using 2-D gel electrophoresis in an effort to find new biomarkers in pooled CSF from controls and AD. He described an array technology that, when coupled with mass spectrometry, produced results comparable to the tau:Aβ ratio. He suggested that combining biomarker quantification with genetic information could increase accuracy rates. In this integrated approach, he suggested that large populations be used, and that biomarkers be selected because of pathophysiological relevance. Papassotiropoulos also suggested that genes related to cholesterol metabolism might produce valuable results.
Brian Bacskai, Massachusetts General Hospital. In-Vivo Imaging of Alzheimer’s Pathology in Transgenic Mice Using Multiphoton Microscopy
Bacskai presented data related to multiphoton imaging at a one-micron (i.e., histologic) level of resolution of structural (amyloid deposits) and functional alterations (free radical/oxidative changes) in several transgenic mouse models of amyloid deposition and AD. Studies of the putative human PET amyloid imaging agent PIB showed rapid entry of PIB into the brain. Bacskai presented this data with a real-time video of the fluorescent compound actually diffusing across the blood-brain barrier of cerebral capillaries. Vascular amyloid labeled immediately, followed by plaque amyloid in an outside-to-inside fashion. PIB that was not specifically bound to amyloid cleared from the brain within 15 minutes, while the amyloid-bound compound remained detectable up to three days. After in-vivo labeling of amyloid deposits with PIB, mice were killed and amyloid deposits were dual-labeled with the red thioflavin-S-like dye, thiazine red-R, showing a 1:1 correspondence between in-vivo and postmortem amyloid labeling. Kinetic studies were done by measuring the fluorescence of individual plaques and vessels over time, thus quantifying the rapid brain entry and clearance.
Bacskai then focused on multiphoton functional imaging using the redox dyes Amplex Red and dihydro-DCF, which produce comparable results. Using these dyes, he could clearly demonstrate areas of oxidative stress, which co-localized with amyloid deposits in thioflavin-S positive (but not diffuse) plaques and vessels. The spin-trap antioxidant PBN was shown to reduce this periamyloid oxidative stress in vivo. Ex-vivo studies showed both ginkgo and PBN to reduce the oxidative stress by about 50 percent.
Thomas Wisniewski, NYU School of Medicine, New York. In-Vivo MRI of Amyloid Plaques in Alzheimer’s Disease Model Mice
Wisniewski discussed work with Gadolinium (Gd) and microcrystalline iron nanoparticle (MION)-labeled Aβ imaging of amyloid deposits in transgenic PS1/AβPP mice. Ex-vivo studies showed a good correspondence of T2 MR signal hypointensities and anti-Aβ antibody-stained deposits on postmortem sections. Gd-Aβ appeared to label both Congo red positive and negative plaques. MION-Aβ also correlated with histology, and iron could be demonstrated in postmortem plaques by special stains. Labeling with both of these agents required intracarotid mannitol in order to transiently open the blood-brain barrier (BBB). However, collaborative work with Joe Poduslo and others at the Mayo Clinic in Rochester, Minnesota, with a putrescine-modified Gd-Aβ has produced agents with higher BBB permeability. T1 imaging was not used because it required 16-hour imaging times compared to the one to two hours needed for T2 imaging. Labeling appeared more sensitive to larger lesions.
Wisniewski then presented an exciting extension of this approach to in-vivo imaging of prion pathology. The strong advantage in the application of large molecule-based imaging to prion disease is that, in the mouse model used, prion-related changes occur in the spleen long before clinical symptoms appear, and even before CNS prions appear. This circumvents the need to use large doses of MION- or Gd-Aβ (e.g., 500 mg/mouse, i.e., the equivalent of about one gram in a human on a mg/kg basis) and mannitol to open the BBB, as are required for amyloid imaging by this approach. Postmortem in-vitro studies performed using 125I-labeled recombinant PrP (rPrP) showed 76-fold brain elevations in prion-infected animals and 12-fold elevations in spleen. Studies using MION-rPrP showed enhancement of brain prion lesions and marked labeling of splenic prion deposits. Work continues to develop smaller protein fragments of rPrP that can label prion deposits and are better able to cross the BBB.
Nick Fox, National Hospital for Neurology/Neurosurgery, London. Measuring Progression in Alzheimer’s Disease Using Serial MRI: 4D MRI
Fox presented perhaps the most poignant presentation of the day, which coupled the eventual progressive brain atrophy of an early-onset familial AD subject to a video of her description of her memory concerns very early in her clinical course. Fox pointed out that adding the fourth dimension of change over time, i.e., rate of change, optimizes the usefulness of MRI-determined brain atrophy in AD. This measurement of relative difference proves much more powerful than single absolute measurements. He pointed out that the advantage of MRI measurements over clinical symptom assessment during clinical trials was the ability to detect true "disease-modifying" effects rather than symptomatic effects.
Fox presented data showing a large overlap between AD and controls using volume measurements only in hippocampus (HC). His approach extended beyond the HC to include rates of atrophy in whole brain; this was determined by digital subtraction, which avoids the tedious nature of conventional volumetry. Fox presented data using a "Boundary Shift Interval" technique, which he said has an accuracy of 0.2 percent change in whole brain. He showed no overlap between AD cases (with an average change of 2.8 percent per year) and controls (0.2 percent per year). This change correlated significantly with MMSE scores (R=0.8).
The volume loss in AD brain was noted to be nearly linear, with some hint of an acceleration of the rate of atrophy over time. Fox showed data for an increased rate of atrophy in individuals at risk for early-onset AD compared to healthy controls, and even greater atrophy in affected FAD cases. Extrapolations suggested that increased rates of atrophy may be detectable three years before the onset of clinical symptoms.
Fox then presented data using a newer "Fluid Registration" model of atrophy quantification that uses a voxel-by-voxel automated method of determining compression and expansion of brain structures over time. This technique results in pseudo-color images of brain coded for volume changes. Data was presented showing frontal atrophy in frontotemporal dementia, temporal atrophy in semantic dementia, and cerebellar atrophy in familial CJD, supporting the generality of this technology. A familial FTD case was shown to have an impressive progression of frontotemporal atrophy beginning in a single gyrus. In AD, this method showed the expected decrease in hippocampal volume over time and an increase in ventricular and sulcal volume. The atrophic changes were especially marked in medial temporal hippocampal and entorhinal cortical areas.
Fox suggested that this technology allows determination of the rate and pattern of pathology; presymptomatic detection of symptoms, and surrogate detection of the effects of therapeutic agents.
This interval change method need not necessarily take a long time, as 10 scans can be performed over a one-year period-the first six months of which would be used to define a baseline, and the second six to measure treatment effects. Importantly, Fox said the technology can be easily exported to other sites. The critical technical concern is maintaining a consistent method of MR acquisition-something that can suffer greatly by hardware or software upgrades.
One of the exciting talks at the meeting was presented by Bill Klunk, who has been working for years to generate ligands that cross the blood-brain barrier and specifically bind to amyloid plaques. Dr. Klunk outlined the history of this research and set forth specific targets for compound lipophilicity, ability to bind amyloid, ability to cross the blood-brain barrier, and ability to be cleared from the CNS. He went on to describe the generation of a series of compounds, based on Congo red and the Thioflavin molecules, that were chemically modified to enhance blood-brain barrier penetration. These compounds maintain their fluorescent properties, and thus can be used as a histological stain and in multiphoton applications. Dr. Klunk presented both, demonstrating the specificity for amyloid lesions. He wrapped up with initial observations from the Center at Pittsburgh using 11C-labeled versions of the compound PIB, and showed some of the first images from living controls and Alzheimer’s patients. Dramatic difference in binding of PIB was seen in the Alzheimer’s patient. Taken together with the presentations from Uppsala, these results demonstrate, for the first time, the ability to image amyloid-β in patients, and promise exciting data to come.-Brad Hyman, Massachusetts General Hospital, Charlestown.
Henry Engler, Uppsala University, Sweden. First PET Study with a Benzothiazole Amyloid-Imaging Agent (11C-PIB) in AD Patients and Healthy Volunteers
Engler presented the first application of the benzothiazole amyloid-imaging agents described above to human studies in AD. He first presented preclinical work done in Uppsala, which confirmed the finding of the Pittsburgh group, showing increased PIB binding to homogenates of AD frontal cortex compared to controls, while binding in the cerebellum was nearly identical between AD and control brain homogenates. The Uppsala group showed similar results in an autoradiographic binding study. The Uppsala human PET study included nine AD patients and five healthy controls, two of whom were elderly and three of whom were 21. The latter subjects were included because of the certainty that these young controls would be truly amyloid-free. The AD subjects had mild disease, with an average MMSE score of 24. The control subjects retained PIB at low levels in white matter areas, at a level that was identical to that observed in AD white matter. Controls showed a virtual absence of labeling in all cortical areas, including cerebellum. There was no significant difference in PIB retention between young and old controls. In sharp contrast, AD subjects showed extensive PIB retention, more than twofold that of controls in cortical association areas expected to have substantial amyloid deposits, such as frontal, temporal, and parietal cortex. Equally important, brain areas such as cerebellum and white matter were indistinguishable between AD and control. This is consistent with the in-vitro binding studies and the absence of fibrillar plaques in these brain regions. The increased PIB retention seen in AD patients was somewhat correlated with cerebral hypometabolism determined with FDG. Engler concluded by describing plans to study subjects with FTD, DLB, CJD, and systemic amyloidosis. Discussion after the presentation focused on the potential strengths and weaknesses of this emerging technology. The question of whether imaging with PIB would be sensitive enough to detect changes in amyloid load caused by antiamyloid therapies led to a discussion of the recently released first autopsy study from the Elan vaccine trials (see ARF related news story). It was pointed out that the amyloid load remaining in this vaccine-treated case was considerably lower than that expected for an AD case, and the effect size of antiamyloid therapies could be considerable.
Jean-Claude Baron, University of Cambridge, England. Predicting Conversion to Clinically Probable Alzheimer’s Disease in Mild Cognitive Impairment (MCI) with Fluorodeoxyglucose PET
Baron suggested that FDG PET may be able to detect synaptic dysfunction prior to the ability to detect tissue atrophy. He employed a voxel-based approach able to assess the entire brain. This approach is amenable to statistical parametric mapping. In probable AD, his group finds decreased metabolism in the superior temporal region, dorsolateral prefrontal cortex, and the posterior cingulate. He reviewed previous studies and pointed out that comparing normals to MCI does not address the real clinical question of predicting conversion. To do this, he proposed comparing MCI-to-AD converters (scans performed before conversion) to MCI non-converters as well as to controls. In his 18-month study, seven of the 17 MCI patients converted to AD. When compared to the non-converters, converters had right posterior cingulate hypometabolism prior to conversion. Lesser, but otherwise similar changes were seen on the left. Non-converters showed no differences when compared to controls. Comparing converters to controls, he found differences in the posterior cingulate and superior temporal gyrus (area 42). The latter region provided 100 percent accuracy in distinguishing the seven converters from 10 non-converters and remained significant after adjusting for MMSE scores. Baron finished by presenting data from an MRI voxel-based morphometry study, which failed to predict conversion from MCI to probable AD.
Eric Reiman, University of Arizona, Phoenix. Positron Emission Tomography Studies of Cognitively Normal Persons at Genetic Risk for Alzheimer’s Disease
Reiman presented FDG PET data obtained in subjects at risk for AD based on their ApoE genotype. Previously, statistical parametric mapping analysis of cerebral metabolic rate for glucose (CMR) data determined by FDG PET in Reiman’s lab has shown that there is hypometabolism and an abnormally high rate of decline in CMR in the parietal, temporal, posterior cingulate and pre-frontal cortices in AD patients compared to controls. Reiman mentioned that these changes decreased by five-11 percent after one year of follow-up. In a group of 50- to 65-year-old people stratified by ApoE genotype, he found 4/4 subjects to have an increased rate of decline in CMR in the posterior cingulate and pre-frontal cortex compared to 3/3 subjects. Subjects with a 3/4 genotype showed an abnormal decline in CMR in the posterior cingulate compared to 3/3 subjects. A similar finding was seen in 3/4 subjects who were as young as 20-39, raising questions about how decreases in CMR of this magnitude can be sustained over the period of time prior to the onset of clinical AD and through the clinical course.
Reiman presented new data from a study of 36 4/4 subjects, 44 3/4 subjects and 78 controls. ApoE4 gene dosage was correlated to hypometabolism in posterior cingulate, parietal, temporal and prefrontal cortices. Estimates were made of the reductions in numbers of subjects (from the typical 1,000 or more required for clinical trials) needed to obtain statistically significant findings in a clinical trial of a therapeutic agent if ApoE genotype was taking into account in enrollment. Depending on the effect size of the therapeutic agent, the numbers of subjects needed to demonstrate a significant effect can drop anywhere from a factor of five to nearly a factor of 100 (with a 50 percent effect size) when combinations of changes in CMR in the brain areas mentioned above are employed. Reiman ended with a discussion of preliminary studies of CMR in transgenic mice. Early studies were complicated by structural changes in corpus callosum in the strain originally employed, but that he hopes will not be present in other transgenic strains.
Paul Thompson, University of California, Los Angeles, School of Medicine. Dynamics of Gray Matter Loss in Alzheimer’s Disease, Mapped with a Population-Based Brain Atlas
In what was literally the most colorful presentation, Thompson presented data on dynamic atrophic changes in AD, schizophrenia, and adolescence based on MR morphometry. The technique was difficult for many members of the audience to appreciate, owing to the sophisticated mathematical models and intensive computation involved in the "high-dimensional elastic deformation mapping" technique this group employs. This technique employs a complex alignment, segmentation, and "warping" process to map an individual MRI onto a population average MRI. This technique and its application to AD was published last month (see Thompson et al.). As quoted recently in the press, Thompson described the progressive atrophy measured in AD brain as a "lava flow" of atrophy from the limbic areas out and up into the cortex. Fascinating videos were included that compressed the atrophy process from years into seconds.
In contrast to other structural and functional imaging techniques presented at this meeting, the emphasis of this presentation was less on quantifying changes in AD brain for use in early diagnosis or evaluation of disease progression and disease-modifying therapies; rather, it provided qualitative insight into the topographic and temporal pattern of atrophy in AD brain. Streaming videos are available on this group’s website. When viewing these videos with the swirling pseudo-color changes, one must remember that the source of the data is basic structural MRI data acquired in a longitudinal fashion. The computational and graphic display is what’s unique to this approach, and it is a powerful process for the purposes described. Videos from schizophrenia patients acquired and displayed in the same manner show an equally fascinating, but almost inverse pattern of atrophic change beginning in the superior aspects of cortex and progressing down through the brain. Thompson finished his presentation by noting that this technology is "10 years behind" more standard techniques with respect to its application to the study of AD. It may well also be ahead of its time in regard to the global nature of the changes portrayed, and easily appreciated when viewing the videos-even if you don’t understand the math of the warping process.-William Klunk, University of Pittsburgh Medical Center, Pennsylvania.
- Thompson PM, Hayashi KM, de Zubicaray G, Janke AL, Rose SE, Semple J, Herman D, Hong MS, Dittmer SS, Doddrell DM, Toga AW. Dynamics of gray matter loss in Alzheimer's disease. J Neurosci. 2003 Feb 1;23(3):994-1005. PubMed.
No Available Further Reading