Correlation of Alzheimer Disease Neuropathologic Changes With Cognitive Status: A Review of the Literature

Abstract

INTRODUCTION

Decades of research worldwide have generated a large body of clinicopathologic correlation (CPC) scholarship related to Alzheimer disease (AD). Here we review the literature from the perspective of neuropathologists and clinicians performing research in this field. The review is designed for a broad target audience. We discuss the pathognomonic features of AD (1-3), highlight some of the challenges in CPC studies, discuss the specificity of AD neuropathologic changes, and, mindful of the controversies in this field of research, describe particular combinations of concomitant lesions that are found in human brains. We then review CPC studies that have documented observations on patients along the spectrum from intact cognition to end-stage dementia linked with AD-type neuropathologic changes. Pertinent reviews of historical background are available (4-10). Terms used in this review are defined in Table 1.

OVERVIEW OF CHALLENGES TO CPC STUDIES IN AD

There are many challenges related to the study of the neuropathologic correlates of cognitive impairment in the elderly. Sources of potential bias in AD CPC studies are summarized in Table 2; some of these are virtually impossible to avoid in the design of a research study. Autopsy is required for definitive AD diagnosis, yet autopsy rates are generally low and autopsy inevitably confers a selection bias. Furthermore, CPC studies rarely are a random sample of the population and clinic- and hospital-based CPCs are subject to other potential biases. For example, because persons with behavioral problems are more likely to require help than those with isolated memory problems, the prevalence and types of Lewy body disease derived from a clinic are likely to be different from those from a broader community (11). Most prior studies have been performed in countries with the most advantageous socioeconomic conditions and have concentrated on middle- and upper-income whites. Thus, overall, ideal conditions have not yet been achieved for comprehensive, population-level epidemiologic study in which information from detailed, longitudinal neurocognitive assessments can be combined with that from state-of-the-art postmortem examinations. Another challenge of CPC studies is that the molecular, anatomic, and clinical changes of AD may not progress in a uniform fashion (7, 12-15). This problem may not be easily solved using statistical models because the progression of both the clinical and the pathologic disease are not necessarily parametrically distributed. The tendency to dichotomize the disease (i.e. “demented vs nondemented”) and overreliance on ordinal variables may also obscure important relationships. Variation in the elapsed time between final clinical evaluation and autopsy and competing mortality risks add more uncertainty. Studies on CPC are additionally complicated by the high prevalence and high morbidity of concurrent diseases (16, 17) (discussed in more detail below). Some researchers tend to focus on unusual or extreme cases with atypical presentations, whereas others describe the distribution of outcomes in larger and perhaps more representative samples. Both approaches can provide insights; CPC data must be reconciled or at least better understood with reference to multiple variables and potential confounders. In sum, both clinical and pathologic assessments are imperfect, variably applied, and constantly evolving. These complexities should be kept in mind as we assess the sometimes-controversial CPC literature.

AβPS AND NEUROFIBRILLARY TANGLES: THE HISTOPATHOLOGIC HALLMARKS LINKED TO AD

The neuropathologic hallmarks of AD are neurofibrillary tangles (NFTs; including pretangles) and AβPs (including diffuse and neuritic plaques, which may also be referred to as “senile plaques” [2, 18, 19]; Fig. 1 and Table 1). Additional changes that may also occur in the brains of AD patients include amyloid angiopathy, age-related brain atrophy, synaptic pathology, white matter rarefaction, granulovacuolar degeneration, neuron loss, TDP-43 proteinopathy, and neuroinflammation (20-26), which may contribute to cognitive impairment; however, these are not considered pathognomonic features of AD (3). Therefore, we focus this review only on the defining features of AD neuropathologic changes: AβPs and NFTs (3). Excellent reviews are available about the molecular characteristics of NFTs and AβPs (27, 28).

Open in a separate window

FIGURE 1

Photomicrograph of a section from the cerebral neocortex of an Alzheimer disease brain stained using double-label immunohistochemistry for β-amyloid (Aβ, reddish brown) and microtubule-associated protein tau (black). Aβ plaques (AβPs; blue arrows) are roughly spherical and extracellular, whereas neurofibrillary tangles (NFTs; green arrows) develop within neurons. Note that some of the dystrophic neurites in the AβPs contain aberrant tau protein pathology (black), which is biochemically identical to that seen in intracellular NFTs. These AβPs have been described to be “neuritic plaques.” Scale bar = 50 μm.

Neurofibrillary tangles are not specific for AD (29-32); indeed, they are found in almost every class of brain disease and are universal (yet topographically restricted) in normal aging subjects (33). Neurofibrillary tangles are found in the brains of individuals who experienced frontotemporal lobar degeneration with tauopathy (FTLD-MAPT), focal cortical dysplasia, myotonic dystrophy, prion diseases, metabolic/storage diseases, some brain tumors, chronic traumatic encephalopathy, viral encephalitis, and other brain diseases (34-37) (Table 3). This suggests that NFTs are, at least under some conditions, a secondary response to injury. On the other hand, tau gene mutations can produce clinical dementia with NFTs; this indicates that, under some conditions, NFTs may be directly linked to the primary or at least proximal neurodegenerative changes (29, 31, 37).

Both the density and the neuroanatomic localization of NFTs are important parameters in AD neuropathology. The findings reported by Tomlinson et al (38) in 1970 have been replicated many times: “[NFTs are] found in both controls and dements [demented subjects] in the hippocampus, but elsewhere in the cortex was only severe or widespread in the dements; severe generalized neurofibrillary change in the cortex was not seen in any control; it seems possible that its occurrence always indicates dementia.” This passage underscores the key point that may be relevant to the study of tau biomarkers, that is, a modest number of medial temporal lobe NFTs are universally present in subjects older than 70 years (33), even in persons who have intact cognition. Recently, it has been reported that NFTs are also very common in certain brainstem nuclei in subjects without dementia (39-44). As such, neurofibrillary degeneration restricted to subcortical sites is often subclinical, whereas widespread neocortical NFTs are almost always associated with severe cognitive impairment in more than 1 disease state.

In contrast to NFTs, AβPs are extracellular (45, 46) and are found in a high proportion of all elderly persons but are not universal (44, 47, 48). A particularly important subtype of AβPs are “neuritic plaques” (NPs), as they are more likely to be associated with cognitive impairment than “diffuse plaques” (49-51). Neuritic plaques are AβPs surrounded by degenerating axons and dendrites that often contain hyperphosphorylated tau aggregates. This subset of AβPs is a hallmark of the current diagnostic criteria for AD, although there is no universally accepted definition of a NP. Indeed, AβPs that lack degenerating tau-positive neurites may have dystrophic neurites that are detected using other methods, including the Bielschowsky silver method, thioflavine S, or immunohistochemistry that detects p62, ubiquitin, phosphorylated neurofilament proteins, or amyloid precursor protein (APP). Some of the variability may be due to the evolution of the NP as part of the disease process (52-54). Because non-AD tauopathies usually lack NPs, they are not an inevitable cellular response to tau pathology (Table 2).

AβPs alone do not seem to be a sufficient substrate for advanced clinical AD dementia (ADD). However, in contrast to neurofibrillary (tau) pathology, there seems to be a strong association between AD genetics and Aβ plaque formation. All high-penetrance AD genetic risk alleles (i.e. APOE ε4 allele, Down syndrome, APP mutations and duplications, PSEN1 and PSEN2 mutations) have been linked in various experimental systems with increased Aβ deposition and increased formation of putative toxic Aβ peptide species (46, 55-57). These data have strong mechanistic implications with an epidemiological scope because genetic factors confer approximately 70% of an individual’s risk for AD (58, 59).

The concordance of genetics and Aβ deposition is aligned with CPC studies that indicate that AβPs may be a temporally “upstream” feature of the neocortical disease, with the caveats that brainstem and medial temporal lobe pretangles and NFTs are seen in subjects without Aβ deposition in all age categories (33, 39, 44). Thus, AβPs and NFTs are the hallmark features of AD but do not develop in the human brain according to the same temporal or anatomic pattern (60, 61). These nuanced observations, which have been gathered and analyzed by many researchers, provide the bases to address some of the controversies that have existed in the field for decades.

CONTROVERSIES IN AD CPC RESEARCH

A basic goal of AD CPC research has been to assess critically whether CPC data support the hypothesis that AD neuropathologic changes (AβPs and NFTs) correlate with clinical dementia. We note that there are publications contending that CPC data argue against a deleterious role for AD pathologic changes (i.e. that they may be an “epiphenomenon” of aging) and that the disease should not necessarily be defined by their presence (62-68). This controversy has been ongoing for many years (38). There are 4 separate assertions that seem to have gained traction in the field, thereby resulting in uncertainty:

• Assertion 1. One sometimes observes persons without dementia with advanced AD pathologic changes at autopsy.
• Assertion 2. One sometimes observes persons with cognitive impairment who clinically to have AD, but who lack AD pathologic changes at autopsy.
• Assertion 3. Clinical trials aimed at reducing AD pathologic changes have failed.
• Assertion 4. AβPs and NFTs may be neuroprotective rather than neurotoxic.

Assertions 1 and 2 are addressed directly by extensive data from AD CPC studies that support 3 points. The definition of “advanced AD pathologic changes” is critical. Some groups may consider widespread diffuse and/or NPs as advanced AD pathologic diagnosis regardless of the numbers and distributions of NFTs. It is clear, as mentioned previously, that AβPs, in the absence of any other neurodegenerative disease lesions or other pathologies, are not a sufficient substrate for severe dementia and thus do not constitute “advanced AD pathologic changes.” In contrast, dense and extensive neocortical NFTs are very consistently associated with dementia and thus, according to new diagnostic criteria, these are required to constitute high burden of AD neuropathologic change (3). As for Assertion 2, it is true that dementia, even one that may be clinically similar to probable AD, may be present without AD pathologic changes; this is simply not AD, but presumably one among many other diseases that cause dementia (e.g. hippocampal sclerosis or vascular dementia). These points will be considered in the context of the AD CPC literature.

Assertion 3 relates to what clinical trials data tell us about the impact of AD pathologic changes. There have been cases in which individuals with mid- to late-stage clinical AD were administered anti-Aβ immunotherapy that partially cleared AβPs, but this clearance did not seem to mitigate NFT formation or the inexorable disease course (69-71). There are multiple reasons why these trials may have failed, including the late timing of the therapy. Thus, conclusions drawn from these trials may be premature (72, 73). In fact, there have been hints of some beneficial effects (clinical and biochemical) linked to human anti-Aβ immunotherapies (74-79). As yet, there are no agents that are capable of reducing tangle density. In sum, clinical trials have not provided definitive answers as to the direct role of AβPs or tangles in the cognitive impairment associated with AD pathologic diagnosis (80, 81).

Assertion 4 is that AβPs and NFTs are actually neuroprotective rather than toxic. Because autopsies provide only cross-sectional data, they cannot prove mechanism. Since both tau phosphorylation and Aβ peptide generation are seen in “normal” brains, some contend that there are probably beneficial and adaptive aspects to these molecular processes. This is important to consider when developing therapies that seek to inhibit those pathways (62). Yet, there are strong arguments in favor of the hypotheses that the molecular processes that underlie tau and Aβ deposition in NFTs and AβPs are indeed pathologic. Abundant evidence from many independent studies indicates that Aβ peptides and other plaque substances (in their various molecular forms) may be neurotoxic, both directly and through secondary responses to stress, injury, and inflammation (46, 82, 83). Both gain-of-toxic-function and loss-of-normal-function deleterious effects also have been described for phospho-tau (84-89). The toxicity of oligomeric forms of tau and Aβ peptides could be connected with biochemical changes that also produce NPs and NFTs; thus, the oligomeric species may be directly related to the neuropathology, although they are not exactly correlated (90-95). It must be acknowledged that a simplistic conceptualization of “plaques and tangles” does not adequately reflect the complexity of the biochemical changes in the brains of patients with AD. Thus, peptides and higher-order aggregates derived from products of APP and MAPT genes can potentially lead to combinations of neuroprotective and toxic attributes. Here, we focus on data and analyses from particular human autopsy studies to address the specific question, “are there strong correlations between antemortem cognitive impairment and the ‘defining’ hallmarks of the disease?”

We consider it to be an important clue that most brains from elderly human subjects show abnormalities that fall somewhere along particular continua of pathologic severity (60, 96-100). These patterns of brain changes indicate specific features that can be correlated with clinical parameters, modeled to understand disease mechanisms, and targeted for therapeutics. Before addressing the “classic” AD pathologic changes and their clinical correlates, we discuss in the following section some of the less common pathologic findings in human brains, that is, the “exceptions that help test the rules.”

LESS-PREVALENT PATTERNS OF CLINICAL AND PATHOLOGIC FEATURES FOUND IN HUMAN BRAINS

“Plaque-Only” Dementia

In 1987, Terry et al (101) described what has become known as “plaque-only dementia.” There were 58 subjects with dementia and AD changes, of which 40 had 2 or more neocortical tangles per high-magnification field in frontal, temporal, and parietal lobes; 18 had no neocortical NFTs. The cases did not differ in hippocampal tangle densities and had no significant differences in brain weights, neocortical thickness, or neocortical neuron counts. Despite this, both groups were demented and did not differ significantly on the Blessed Dementia Scale. The existence of this “plaque-only” dementia category has been controversial because some subjects with pathologic findings similar to the plaque-only group described by Terry et al have dementia and some do not. The current revision of the National Institute on Aging–Alzheimer’s Association guidelines for the neuropathologic assessment of AD stipulate that subjects with moderately dense neuritic AβPs may be regarded as having an intermediate level of AD neuropathologic change and that other potential contributors to cognitive impairment and dementia should be sought (3). Notably, prior studies necessarily included brains from persons with diseases (e.g. neocortical α-synucleinopathy or TDP-43 pathology) that were not at that time well characterized, but which can coexist with variable neocortical amyloid loads. Indeed, Hansen et al subsequently published an article titled “Plaque-only Alzheimer disease is usually the Lewy body variant, and vice versa” (102). When modern pathologic methods are used, severe cognitive impairment associated only with diffuse AβPs in the brain at autopsy is rare and has an unknown relationship to AD. Subtle cognitive impairment may be associated with diffuse AβPs, indicating that these lesions may play a direct role in incipient AD (54), although a direct role of diffuse Aβ plaques without tau pathology has not been proven. Furthermore, diffuse AβPs may be associated with cerebral amyloid angiopathy, which could produce impairments separately (103, 104), although see Nelson et al (105). As the field moves toward identifying subtle (i.e. early) cognitive impairment, more work will be required to characterize any deleterious influence of diffuse AβPs on cognition and clinical manifestations in cases of presumed incipient AD. On the other hand, the density of tau-positive NPs does correlate directly and statistically independently with some degree of cognitive impairment (105-107) (Fig. 2).

Open in a separate window

FIGURE 2

(A, B) Correlations between antemortem cognitive status (final Mini-Mental State Examination [MMSE] scores), counted neocortical neurofibrillary tangles (NFTs; A), and neuritic β-amyloid plaques (NPs; B) for 178 patients lacking concomitant neuropathologic findings (189). Each circle represents data from a single individual. Neurofibrillary tangles and NPs were counted and summed from 4 different portions of cerebral neocortex: Brodmann areas 21/22, 18/19, 9, and 35, as described (189). Data are reprinted with permission from the Journal of Neuropathology and Experimental Neurology (2007;66:1136–46). Copyright 2007, American Association of Neuropathologists. The correlation between final MMSE scores and neocortical NFT counts is stronger than that between MMSE scores and NP counts.

CLINICOPATHOLOGIC CORRELATION IN AD

Correlation Between AβPs and Cognitive Status

More than 40 CPC studies have assessed the correlation between AβPs and severity of antemortem cognitive impairment (38, 98, 103, 105-108, 158, 162, 171, 185, 189, 191, 223-260). These studies vary with respect to many factors including the research cohort characteristics, anatomic areas examined, pathologic methods, AβP subcategories and counting techniques, metrics for cognition, the range and domains of cognitive decline tested, and the rigor with which concomitant pathologic findings were evaluated and factored into the study.

Although the study designs were diverse in CPC studies involving AβPs, most studies confirm a significant correlation between antemortem cognitive impairment and neocortical AβPs, as demonstrated by Blessed et al (254) (Fig. 3). These workers used the von Braunmühl silver stain, which could not differentiate between diffuse and neuritic AβPs. This, and some other earlier studies, incorporated a bias related to the inclusion of many individuals at opposite ends of the AD continuum thereby masking the poorer correlation between plaque load and dementia severity among the impaired subjects. Subsequent studies have established that the densities of NPs correlate with antemortem cognitive impairment better than the densities of diffuse AβPs (45, 98, 106, 154, 189, 261).

Open in a separate window

FIGURE 3

Correlations between antemortem cognitive status (“dementia scores”) and counted amyloid plaques from the 1968 article by Blessed et al (254). Dementia scores were derived from “psychological tests of orientation, remote memory, recent memory, and concentration.” Amyloid plaques were visualized using the von Braunmühl silver stain. Each circle represents data from a single individual. There is reasonable correlation between the dementia scores and the number of plaques, although this work antedated the era of neocortical synucleinopathy, TDP-43, and other factors now known to both clinicians and neuropathologists. This figure was reproduced with permission from The British Journal of Psychiatry (1968;114:797–811). Copyright 1968, The Royal College of Psychiatrists.

It is important to note, however, that many prior studies have only considered limbic and neocortical plaques. The distribution of AβP pathology may occur throughout the entire neuraxis in advanced disease. Elderly without dementia show early stages of this hierarchical process. The progression of AβP pathology seems to begin in the neocortex with a few diffuse plaques and then sequentially progress to also include 1) the hippocampus, entorhinal cortex, cingulate cortex, and amygdala; 2) basal ganglia and diencephalon; 3) midbrain and medulla oblongata; and 4) pons and cerebellum (262). A recent study with large numbers of subjects assessed with standardized antemortem cognitive testing found that the amyloid stage that has progressed to involve the striatum is highly predictive of dementia (263). Correlations between amyloid stages (Thal phases), NFT burden, and cognitive impairment are shown in Figure 4.

Open in a separate window

FIGURE 4

β-Amyloid (Aβ) phase showing the relationship between cognition, as represented by the retrospectively determined clinical dementia rating scale (CDR) score, and the phase of Aβ deposition determined in the medial temporal lobe (MTL) in 202 cases (103, 262). Non–Alzheimer disease (AD) dementia cases were excluded from this analysis. (A) Almost all demented cases exhibited end phases of the expansion of Aβ deposition without major differences; cases without dementia had early-phase Aβ pathology (partial correlation controlled for age and sex for all cases: r = 0.582, p < 0.001; only for AD cases: r = 0.086, p = 0.605). (B) Relationship between Braak neurofibrillary tangle (NFT) stage and Aβ phase in 201 AD and control cases. With increasing Braak ? NFT stage, the distribution of Aβ plaque deposition expanded to end-phase Aβ deposition reached with Braak NFT stage IV (partial correlation controlled for age and sex for all cases: r = 0.621, p < 0.001; only for AD cases: r = 0.07, p = 0.671).

A critical point supporting the importance of plaques is that widespread neocortical NFTs are virtually nonexistent without the presence of widespread AβPs, except in the minority of cases that show a clear-cut non-AD tauopathy (e.g. progressive supranuclear palsy). Owing in part to the scarcity of neocortical NFTs in the absence of AβPs, and the relatively modest apparent direct impact of AβPs on cognition, it has been suggested that the pathogenetic effect of AβP-related substances may be mediated by “seeding” neurofibrillary pathology (46, 88, 189, 264, 265). The current clinical trial literature is compatible with the idea that AβP-related substances kindle a self-propagating process related to neocortical NFTs, as has been hypothesized (266-269). Regarding AβPs in AD, an analogy can be made to the established role of high blood cholesterol in heart disease. Hypercholesterolemia promotes atherosclerosis that over time can cause myocardial infarction. Therefore, high blood cholesterol is considered a causative risk factor for heart disease, although it does not correlate with clinical heart disease in all groups (many people live with hypercholesterolemia for years without experiencing myocardial infarction). In the same way, AβPs may be a key pathologic factor in AD without equating with the clinical features of ADD.

Often neglected is the question of lesion turnover (i.e. AβPs being possibly reabsorbed during life so that they would not be “counted” at autopsy), which seems to occur in some subjects (70, 270). For this reason, the number of AβPs might plateau (or at least the rate of increase flatten out), leading to a “ceiling effect” relatively early in disease process that confounds linear correlations between plaque number and severity of cognitive change (253). Aβ vaccine trials also attest to the ability of the immune system to clear diffuse AβPs (69, 70, 271), but it is not known whether toxic APP metabolites or other plaque components remain after clearance of fibrillar Aβ-peptides (73, 272). The active “turnover” during life could reduce the correlation between the lesion numbers observed at autopsy and clinical features of the disease.

Correlation Between NFTs and Cognitive Status

Numerous CPC studies have assessed the association between NFTs and cognitive status (38, 98, 154, 158, 162, 185, 188, 189, 191, 223-232, 235-240, 244, 250-252, 258, 261, 264, 287-296). As with studies on AβPs, those of NFTs have used diverse study designs. Regardless of methodology, the correlation between neocortical NFTs (but not necessarily allocortical or subcortical NFTs) and antemortem cognitive status is strong in studies that span the spectrum of cognitive impairment (provided there is not a strong influence by other diseases that can cause cognitive impairment). Results from the University of Kentucky Alzheimer’s Disease Center (189) corroborate previous findings (Fig. 2). Duyckaerts et al (287) and Sabbagh et al (252) independently arrived at the conclusion that the density of NFTs in selected cerebral cortical fields significantly (p < 0.01) correlated with Blessed or Mini-Mental State Examination scores (297). Likewise, Dickson et al (251) described a strong correlation between cortical NFT counts or phospho-tau ELISA measurements and Blessed Information, Concentration, and Memory test scores. Similar results were obtained by Bennett et al (264) who found a strong correlation between global cognitive status and densities of neocortical tau-positive tangles.

On the basis of the review of an enormous sample of cross-sectional autopsy data, the pattern of NFT development across the clinical spectrum of disease has been integrated in Braak neuropathologic staging (60, 298, 299). Neurofibrillary pathology is observed in the human brain long before Aβ plaques develop (39, 44, 300). At 40 years, all individuals studied by Braak and Del Tredici (39) and Braak et al (44) exhibited at least initial neurofibrillary pathology in the brainstem (Fig. 5). The relationship between the near-universal tau pathology seen in the locus coeruleus of middle-aged individuals and AD is not known. As AD progresses, the NFTs become numerous in locus coeruleus which may suggest the same pathogenetic process is becoming more severe (301).

Open in a separate window

FIGURE 5

Development of phospho-tau (AT8)-immunoreactivity (ir) versus β-amyloid (Aβ) pathologic findings. (A) White columns indicate the relative frequency of 2,332 nonselected autopsy cases devoid of any abnormal intraneuronal tau deposits. Columns in shades of blue indicate the relative frequency of cases with all types of intraneuronal lesions (Braak NFT stages). (B) Development of extracellular Aβ deposits. Purple areas within the columns indicate subgroups of cases showing plaque-like Aβ-amyloid deposits in temporal neocortex (Phase 1, light purple), allocortex and neocortical association areas (Phases 2 and 3, middle purple and dark purple), or in virtually all cerebral cortical regions (Phase 4, black). Note the relatively late appearance of Aβ plaques in comparison to subcortical neurofibrillary tangles. This figure is reproduced with permission from the Journal of Neuropathology and Experimental Neurology (2011;70:960–99) (44). Copyright 2011, American Association of Neuropathologists.

In more advanced disease, the neuroanatomic distribution of NFTs correlates both with the location at which neurons die (302-305) and with the cognitive domains affected in patients with ADD. For example, early AD symptoms tend to relate to memory, when the anatomic substrates of memory in the medial temporal lobe are selectively affected by NFTs (Braak stages III–V). The cognitive domains affected in mid-and late-stage ADD expand to include areas of executive function, visuospatial capacities, and speech. These manifestations occur in synchrony with the development of NFTs in the neocortical areas responsible for those functions (Braak stages V–VI). Braak staging provides a useful ordinal system for describing the topographical distribution of pretangles and NFTs in human brain. Nevertheless, Braak staging also masks the fact that marked variability in pathologic lesion density across cases exists within a given Braak stage (7, 12) and that not all AD cases fall within the same spatial continuum with regard to brain NFT distributions (12, 15, 151).

A key question in this regard is whether a heavy neocortical NFT load is present in individuals with intact cognition. In a review of 11 different studies comprising 555 subjects without dementia, a total of 12 brains were assessed as Braak stage V (2.2%); 3 were assessed as Braak stage VI (0.5%) (142). However, not all Braak stage VI brains were the same; there was significant variation in NFT burden in these cases (142, 151). In the Braak stage VI cases studied where neocortical NFTs were most abundant, cognitive impairment was found on testing (151). It is important to emphasize that, among thousands of cases from dozens of CPC studies worldwide, there never has been a report of a thoroughly documented individual with truly “end-stage” neocortical NFT pathology who lacked antemortem cognitive impairment proximal to death (142, 151). Even the single case report “outlier,” a Braak neuropathologic stage VI case without full-blown dementia, had some degree of cognitive impairment within expectations for late preclinical AD (306).

As with AβPs, NFTs are quantified with certainty only at autopsy. Therefore, it is necessary to know whether some NFTs are removed (or reabsorbed) from the brain before death. Some, but not all, studies have indicated that more neurons disappear in brains of demented subjects than can be explained directly by the number of NFTs and “ghost tangles” observed at autopsy (285, 304, 307-310) (but see [305]). Thus, some evidence exists for NFT “turnover,” neuronal shrinkage, and/or non-NFT cell death mechanism(s).

EVOLVING CONCEPTS AND UNANSWERED QUESTIONS

The field of AD research has made progress with respect to in vivo diagnostic methods to track the progression of AD and to identify patients who may benefit from candidate therapies. For the most part, these methods, which include neuroimaging, cerebrospinal fluid, and blood markers, are beyond the scope of the current review. Nonetheless, the strong focus on neuroimaging and biomarkers has raised important issues about the basic concept and definition of AD. There may be a future method that does not involve brain autopsy and that is optimized for diagnosing specifically the devastating illness that we refer to as AD. However, until such a method becomes available, neuropathologic examination remains the criterion standard for disease definition. Because cognitive impairment in aging may be attributed to many different conditions other than AD, current biomarkers for AD are limited to predicting which patients may become cognitively impaired with neocortical AβPs and NFTs, rather than just predicting cognitive impairment alone. For example, hippocampal atrophy is visualized on magnetic resonance imaging in both AD and hippocampal sclerosis patients and is therefore not a distinct AD biomarker. This is an important consideration because up to 20% of individuals in advanced age are affected by hippocampal sclerosis that correlates with cognitive impairment independently of AD pathologic diagnosis (105, 211, 321-323). A patient lacking AD pathologic diagnosis at autopsy, despite “biomarker positivity,” is not an AD case and suggests the biomarker has imperfect specificity. Conversely, AD pathologic diagnosis at autopsy in a biomarker-negative case implies imperfect sensitivity. Moreover, it is important to remember that AD is often mixed with other pathologies in the aging brain; thus, the finding of a positive biomarker for AD will not eliminate the need for autopsy studies. Indeed, there are still also no recognized biomarker-based criterion standards for a large number of other neurodegenerative tauopathies, synucleinopathies, and vascular brain changes.

There are limitations to CPC studies and significant questions remain unanswered at this time. Pathologic assessments cannot prove molecular mechanisms; instead, they identify and describe deviations from what is recognized as “normal.” We try to understand these changes by using an array of immunohistochemical and experimental methods. Although the density of neocortical NFTs and neuritic AβPs on autopsy correlate with the severity of cognitive dysfunction, and the observed results align with a plausible hypothesis, this does not prove that these pathologic hallmarks are proximate to the ultimate neurotoxic species. Moreover, we still seek answers to the following critical questions: What biochemical changes are upstream of the formation of NFTs and AbPs What exactly is the role of pretangles in the pathologic process? What processes occur in parallel with “classic” neuropathologic lesions? And, why are some cell populations differentially vulnerable?

SUMMARY AND CONCLUSIONS

Summary points are presented in Table 6. Neuropathology of aging-related brain disease must take into account diverse medical, technical, biochemical, and anatomic considerations. Concomitant pathologies are very common in aging brains. These diseases and many other factors collectively are formidable obstacles to aligning pathology and cognition along strictly linear scales. Although copathologies (CVD, hippocampal sclerosis, synucleinopathies, etc.) contribute to cognitive impairment, the universal observation of a strong independent correlation between plaques and tangles with dementia severity (despite the “background noise”) means that AD neuropathologic changes are likely significant. An extensive literature on CPC correlations related to AD dementia, a body of research-based on work performed around the world with many thousands of patients, indicates a sequence that may begin with specific genetic and environmental factors that increase risk of widespread AβPs. Neuritic plaques combine extracellular Aβ deposition and intracellular neurofibrillary (tau protein) pathology. Cross-sectional data indicate that pretangles and NFTs develop first in the brainstem and medial temporal lobe structures. Among patients with measurable clinical disease and lacking comorbidities, the extent of cognitive impairment parallels the severity of neocortical NFT pathology. In sum, there is a complex but predictable correlation between AD pathologic hallmarks and cognitive impairment.

Acknowledgments

References