The possibility of disease-modifying therapies for Alzheimer's disease (AD) has motivated the development of biomarkers that reflect underlying pathologic processes. The sequence of pathologic events in AD likely begins many years, perhaps decades, prior to the development of symptoms.1, 2 Amyloid-β (Aβ) deposition appears early in the disease, prior to symptoms, and then plateaus as clinical dementia emerges.3–6 In contrast, neurodegeneration, including loss of synapses, neurons, and arborization, results in brain atrophy that worsens in parallel with cognitive decline.2, 6, 7 The principal early sites of Aβ deposition are neocortical, typically in the parietal and frontal regions,4, 8, 9 whereas the sites of early atrophy include the medial temporal regions.2, 6, 10, 11 Here we relate these 2 phenomena in vivo in clinically normal (CN) older individuals and in clinically established AD patients, in order to determine the correspondence between levels of Aβ deposition and of atrophy.
It is now possible to observe the relation between Aβ deposition and atrophy in vivo with positron emission tomography (PET) imaging using Pittsburgh Compound B (PiB)9 and high-resolution volumetric magnetic resonance imaging (MRI) data.2, 6 PiB studies have confirmed what was predicted by earlier postmortem studies,13–15 that a substantial fraction (25–50%) of CN older individuals exhibit Aβ deposition.16–21 While still in an early phase, PET studies of Aβ deposition in these otherwise normal individuals suggest evidence of early brain dysfunction including disrupted default network functional connectivity,19, 21 aberrant default network activity during memory encoding,18 and even subtle cognitive impairment22, 23 that is offset by cognitive reserve.23 Here we relate the presence and pattern of Aβ-related atrophy observed in AD patients to the pattern seen in CN older individuals.
Atrophy can be quantified by automated measurement of brain MRI images, which yields estimated thickness measures of anatomically parcellated cortical regions as well as subcortical volumes.24–27 Such measurements have revealed a characteristic pattern of cortical thickness reductions and subcortical volume loss in clinically diagnosed AD patients.25, 27–29 The AD-like pattern of atrophy has also been reported in presymptomatic autosomal dominant AD,28 and in those with mild cognitive impairment who go on to develop the clinical diagnosis of AD.30 More recently, Desikan and colleagues11 identified a pattern of atrophy in the supramarginal cortex, entorhinal cortex, and hippocampus with which mild cognitive impairment (MCI) and AD could be distinguished from normal aging, and Davatsikos and colleagues31 identified a similar pattern of volume loss that related to cognitive decline among MCI as well as normal control subjects.32 However, these studies used control groups of older individuals in which amyloid was likely present, but the impact was not assessed. Studies directly relating structural data to Aβ deposition in CN subjects have yielded inconsistent results; while some have reported reduced hippocampal volume33, 34 or cortical thickness29, 34 in CN subjects with greater Aβ deposition, others have found this only among Aβ-positive CN35 or in normal individuals with subjective cognitive impairment.36 Similarly, the impact of age on amyloid and atrophy has not been consistently controlled. We sought to relate both hippocampal volume and cortical thickness reductions to a continuous measure of Aβ deposition adjusting for age in a large sample of both Aβ-positive and Aβ-negative CN subjects and in AD patients.
We first determined the pattern of cortical thickness reductions and hippocampal volume loss in mild AD patients compared to CN subjects, and then investigated whether a similar pattern of Aβ-associated volume loss was present in CN subjects. We also investigated the age-dependence of Aβ deposition and of Aβ-associated thickness reductions in both CN and AD, and quantified the extent and anatomic specificity of Aβ-related volume loss within each group. We hypothesized that Aβ deposition would be associated with local cortical thickness reductions in regions associated with the default network37 at early stages of the pathophysiological process, prior to cognitive impairment.
- Top of page
- Patients and Methods
- Potential Conflicts of Interest
- Supporting Information
The major finding of this study is that significant Aβ-associated cortical thinning occurs among CN older individuals in a pattern consistent with early AD. While this finding supports the possibility that Aβ deposition in normal individuals represents preclinical AD, direct observation with longitudinal data will be required to evaluate the strength and timing of this link. Our data suggest that Aβ-associated neurodegeneration manifests as cortical thinning in regions vulnerable to early Aβ deposition, including association cortices along the posterior medial wall and lateral parietal cortex. In particular, we observed thinning in the inferior parietal lobule and the posterior cingulate extending into the precuneus, which are regions that form nodes of a large-scale cortical system known as the default network.37, 49, 50 This system has been implicated in both memory-related function and in amyloid-related and AD-related memory dysfunction.18, 37, 50–53 Our findings are consistent with a pathophysiologic link between Aβ deposition and neurodegeneration in this network, which may anticipate memory failure and progression to clinical dementia.18, 19, 21, 37, 54
Along the posterior midline, the posterior cingulate and retrosplenial cortex are anatomically connected to medial temporal structures, and we found that while the hippocampus and medial temporal lobe (MTL) cortices demonstrated significant age-associated atrophy, the association of MTL atrophy with Aβ deposition was variable in these asymptomatic older individuals. While our results are consistent with the hypothesis that MTL atrophy coincides with the emergence of manifest cognitive impairment,2, 55, 56 we did not observe a significant difference between MTL and cortical atrophy, and thus cannot order the relative timing of effects with the present data. While macroscopically-visible cortical atrophy is associated with dementia, it is not generally observed in nondemented individuals at postmortem,57 perhaps because of an inability to differentiate it from normal age-associated atrophy. Our data suggest that the Aβ deposition commonly detected in normal older individuals is associated with subtle posterior cingulate and parietal neurodegeneration that occurs prior to, and may be a harbinger of, clinically significant impairment.29 It is possible that further investigation will reveal evidence of subtle cognitive alterations related to cortical thinning even within CN individuals, particularly when the level of cognitive reserve is considered.23
More broadly, our findings should be considered in the context of a putative sequence of events in AD pathology that can be observed with biomarkers. Using a largely biphasic model of disease sequence, Aβ deposition has been hypothesized to occur early in the sequence of AD pathology and to be followed later by neurodegeneration, which is then related to the symptomatic phases of the disease, cognitive decline and dementia.58–60 Our present findings and those of earlier studies29, 33, 34 that suggest PiB retention is correlated with cortical thinning in normal individuals raise the possibility that the hypothesized lag period between Aβ deposition and neurodegeneration may be shorter than previously thought. A precise mechanism by which Aβ deposition could be linked to neurodegeneration has not been firmly established. It is possible that toxic effects of Aβ oligomeric assemblies that surround fibrillar forms could be exerted locally early in the process and result directly in synapse and cell loss.61, 62 Such a mechanism entails a direct relationship between the presence of Aβ and neurodegeneration, which could potentially be observed with sensitive biomarkers. However, while the sensitivity of Aβ imaging may be improved in the future and permit better detection, individuals with predominantly prefibrillar or polymorphic forms of Aβ that are refractory to PiB would not be detectable with PiB or likely with other thioflavin or Congo-red derivatives.63–65
We found that Aβ and PCC thickness were more strongly correlated in AD than in CN (see Fig 1) and we evaluated these data according to a recently proposed model59 in which Aβ deposition and cortical thickness follow sigmoid-shaped curves in time. We simultaneously fit sigmoid models to PCC PiB and age-adjusted thickness data for the combined CN and AD groups, and determined the temporal lag between the dynamic phases of amyloid accumulation and cortical thinning to be approximately 35% of the total time required for amyloid to rise from its baseline to maximum. It should be emphasized that the model in its particularities as presented here is tentative, and should be considered as a schematic rather than definitive treatment of the problem. Such modeling will remain speculative as to accurate parameters of the underlying sigmoid curves until longitudinal data are available.
Whereas AD neurodegeneration is well established to occur prominently in the MTL55, 56, 66 and to be correlated with neurofibrillary pathology,67 our findings are consistent with emerging evidence that thinning in posterior association cortices is also a prominent feature of MCI and AD.11, 28, 29, 37, 48, 68 While the measured amount of age-adjusted thickness reduction per unit of PiB retention (DVR) was approximately the same in the posterior cingulate/precuneus and in MTL structures, the standard errors were larger in the MTL (see Fig 1) and the regression coefficients did not reach significance. Future work with a larger data sample will be required to order the relative emergence of effects between posterior cingulate and MTL structures. Neurofibrillary tangle pathology may partially explain this observation of greater variability in the MTL, since it is common in MTL but rarely widespread in cortex of CN subjects.57, 58, 67, 69 Our data are not consistent with previous observations that Aβ deposition is only seen after significant neurofibrillary tangle deposition and MTL atrophy,69 but instead suggest that the pathologic sequence of events in preclinical AD is one in which Aβ deposition is related to neurodegeneration in posterior cingulate and distributed regions of association neocortex that occurs along with or possibly even prior to hippocampal and entorhinal neurodegeneration.
Previous reports of the relation of PiB retention and hippocampal volume in CN subjects have been inconsistent. Some reported an inverse relation (ie, decreased volume with increased PiB retention) in CN subjects,33, 34 while others found such a relation only among the Aβ-positive CN group35 or only in CN subjects with subjective cognitive impairment.36 These studies have differed in their treatment of the potentially confounding effect of age on both Aβ level and atrophy. We evaluated hippocampal volume, cortical thickness, and PiB retention for evidence of age-dependence and found evidence for an age effect in all domains. The age-dependence of volume/thickness across a broad age range has been previously reported,70 and although some investigators have not found a significant impact of age within a more restricted older age range,29 others have applied an age-adjustment to thickness/volume data.11 The strong age-dependence of Aβ deposition we observed in CN subjects is consistent with neuropathological studies that inferred from cross-sectional data that Aβ gradually accumulates with age.68 Several PiB studies9, 16, 71 did not report evidence of a significant relationship with age, perhaps due to the small sample sizes with limited age ranges, although a recent study did demonstrate an age association.72 Morris and colleagues73 found that the age-dependence of PiB data was largely accounted for by the strong age association with PiB retention among carriers of the APOE4 allele,34 likely a reflection of the sample enrichment for younger CN subjects with positive family histories. We did not find an interaction of the age-Aβ relationship with APOEe4 carrier status among the subset of subjects on whom genotypes were available (data not shown).
Interestingly, the age-dependence of Aβ deposition among CN subjects was reversed in AD patients, such that greater age was associated with lower levels of Aβ deposition. The reversal of the Aβ-age coefficient in the AD group compared to the CN group could be due to a survivor effect, such that older subjects with greater amounts of Aβ were too impaired (eg, MMSE < 18) to have been included in our study. Other potential factors, including an age-related change in PiB binding sites or affinity or a change in the production or clearance of Aβ,74, 75 will require further study. Moreover, longitudinal observation over long intervals may be required to determine whether individual AD patients' levels of Aβ decline over time, which has not been observed in longitudinal PiB data that has spanned 1 to 2 years.2, 5, 76
In summary, our findings provide support for the hypothesis that Aβ is associated with local neurodegeneration in key nodes of a distributed network supporting memory processes, and that this process begins prior to clinically-evident cognitive impairment, but continues into the stage of clinical dementia. Longitudinal follow-up of these CN older individuals is ongoing to determine if the combination of Aβ burden and volumetric loss is predictive of incipient cognitive decline, and progression to AD dementia.
Potential Conflicts of Interest
- Top of page
- Patients and Methods
- Potential Conflicts of Interest
- Supporting Information
J.C., T.H., D.P., J.A.B., J.M., D.M.R., G.A.M. and R.L.B. have none to report. B.F. has received one or more grants and has a grant pending from the NIH. D.M. is involved with an ongoing phase 3 clinical trial with Elan Pharmaceuticals and Janssen Immunotherapy, and is on the speakers' bureaus of Novartis and Pfizer. S.S. serves on the scientific advisory boards of Elan Pharmaceuticals, Sanofi-Aventis, Pfizer Inc., Eisai Inc., and Bristol-Myers Squibb; serves as Associate Editor for the Journal of Neuropsychiatry and Clinical Neurosciences; receives publishing royalties for The Frontal Lobes and Neuropsychiatric Illness (American Psychiatric Press, Inc., 2001), The Neuropsychiatry of Limbic and Subcortical Disorders (American Psychiatric Press, Inc., 1997), and Vascular Dementia (Humana Press, 2004); receives honoraria from Eisai Inc., Pfizer Inc, Novartis, Forest Laboratories Inc., Elan Pharmaceuticals, and Athena Diagnostics, Inc.; holds corporate appointments with Merck Serono and Medivation, Inc.; receives research support from Elan Pharmaceuticals, Wyeth, Bristol-Myers Squibb, Janssen Immunotherapy, Pfizer Inc., Bayer, and Eisai Inc.; received research support from Myriad Genetics, Inc., GlaxoSmithKline, Neurochem-Alzhemed, Cephalon, Inc., and Forest Laboratories Inc.; receives research support from the Alzheimer's Disease Neuroimaging Initiative, Dominantly Inherited Alzheimer's Network (NIA 1U01AG032438-01); received research support from Aging Brain: DTI, Subcortical Ischemia and Behavior (NIA 1 R03 AG023916-01A1); and receives research support from The Norman and Rosalie Fain Family Foundation, the Champlin Foundation, and the John and Happy White Family Foundation. D.G. has a grant from the NIH. K.A.J. has received grants and has grants pending from the NIH/National Institute on Aging and The Alzheimer Association. R.A.S. has a grant pending from the NIH/National Institute on Aging.