|Home | About | Journals | Submit | Contact Us | Français|
Brains that have many neurofibrillary tangles (NFTs) in medial temporal lobe structures (Braak Stages III or IV) but no cortical neuritic plaques (NPs) may be a diagnostic dilemma; they also raise questions about the “amyloid cascade hypothesis” of Alzheimer disease (AD) in which NFT development is thought to occur downstream of the development of amyloid plaques. To determine the clinical, demographic, and biological factors related to NFT+/NP− cases, we analyzed 26 NFT+/NP− patient brains identified from the University of Kentucky AD Center autopsy cohort (n = 502); most of these patients were at least 85 years old and lacked profound antemortem cognitive impairment. A subset of the cases had neurofibrillary tangles in the medulla oblongata. Aberrant TAR-DNA binding protein-43 immunohistochemical staining was seen in 5 of the 26 cases with the clinical diagnoses of AD or mild cognitive impairment. We also queried cases in the National Alzheimer’s Coordinating Center Registry (n = 5,108) and found 219 NFT+/NP− cases. Those patients had a relatively high likelihood of belonging to a birth cohort with the highest incidence of influenza infection during the 1918–1919 pandemic. This observation may link the pathogenesis in NFT+/NP− cases to encephalitis during childhood. We conclude that NFT+/NP- cases comprise approximately 5% of aged individuals in multiple data sets; these cases are not necessarily within the spectrum of AD.
The diagnostic and biological relevance of histopathological markers associated with Alzheimer disease (AD) has recently been called into question (1, 2). In view of the extreme complexity of age-related human brain diseases (3), a better understanding of the neuropathology of AD and of neurodegenerative diseases that overlap pathologically with AD is needed.
The distinct pathological hallmarks of AD are neurofibrillary tangles (NFTs) and neuritic amyloid plaques (NPs). NFTs develop within neurons and are composed of filamentous tau protein polymers. As delineated by Braak and Braak, the numbers of NFTs generally increase in an anatomically predictable fashion as the disease progresses (4). The expanding distribution of NFTs in the brain is graded on a 0 to VI severity scale according to “Braak stages” (5). Braak stages 0–IV refer to NFT pathology that is mostly confined to the medial temporal lobe structures including the hippocampal formation and amygdala. The severity of neocortical NFT pathology correlates well with antemortem cognitive impairment in AD (3, 6, 7). NFTs are, however, observed in other neurodegenerative conditions (8).
In contrast to NFTs, NPs are extracellular amyloid deposits that are specific to AD. NPs contain Aβ-peptide polymers and are surrounded by degenerating neurites. The severity of NP pathology is scored according to a completely different neuropathological metric named after the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) (9). The CERAD scoring system involves a 4-tiered semiquantitative scale of neocortical NP density; the earliest changes are usually found in the neocortex rather than in the medial temporal lobes. The density of NPs in medial temporal lobe structures is typically low (10) and the correlation between NP densities in these areas with antemortem cognitive impairment is relatively weak (11, 12).
The National Institute on Aging and Reagan Institute (NIARI) consensus criteria utilizing both the Braak stages and the CERAD scales have been applied since 1997 for AD diagnosis (13). According to these guidelines cases are parsed into “low,” “intermediate,” and “high likelihood” that clinical manifestations are due to AD. “Low” likelihood corresponds to CERAD “negative,” Braak stages 0–II; “intermediate likelihood” corresponds to CERAD “possible” or “probable,” Braak stages III or IV; “high likelihood” corresponds to CERAD “probable” or “definite” Braak stages V or VI.
Most brains can be categorized readily among the NIARI diagnostic groups but some cases defy those categories and are also difficult to reconcile with current theories of AD pathogenesis. For example, up to 6% of aged persons’ brains contain relatively severe neurofibrillary pathology in the medial temporal lobes, corresponding to Braak stages III or IV, although no discernible NPs are detected elsewhere in the brains (14). We designate these cases as “NFT+/NP−” but stress that the “NFT+” refers to NFTs that are >99% confined to the medial temporal lobe structures such as the hippocampal formation and amygdala. NFT+/NP− cases have been described previously (14–19), but many questions remain about their clinical and pathological context and the relationship to newer concepts such as the aberrant expression of TAR-DNA binding protein-43 (TDP-43) (20). We used patients from the University of Kentucky Alzheimer’s Disease Center (UK ADC) cohort, the Nun Study, and pooled cases in the National Alzheimer’s Coordinating Center (NACC) Registry to determine the clinical, demographic, and biological factors related to NFT+/NP− cases.
Patients who came to autopsy from the UK ADC cohort (total n = 502) were the initial study group. Details of UK ADC IRB protocols, patient recruitment, and longitudinal follow-up have been described previously (12, 21). Briefly, non-demented subjects were contacted at 6 month intervals, had detailed annual mental status testing and had neurological and physical examinations on at least a biannual basis. Mental status testing has been described previously (12, 21). Cases with frontotemporal dementia (n = 18) were excluded. All clinical diagnoses were documented by the clinicians at a consensus conference that included neurologists, neuropsychologists, and social workers. The Nun study of the School Sisters of Notre Dame in the United States is a longitudinal study of 493 participants ranging in ages from 75 to 102 years (mean = 83 years) and has been described elsewhere (22, 23). Neuropathologic evaluations were conducted by a single blinded neuropathologist (W.R.M.) who performed all gross and microscopic examinations (9).
Detailed neuropathological studies were performed on both cohorts at the UK ADC as previously described (11, 12). Briefly, at least 24 samples were taken from each brain, including middle frontal gyrus (area 9), superior and middle temporal gyri (areas 21 and 22), inferior parietal lobule (areas 39 and 40) and occipital lobe including primary visual area (areas 17 and 18). Amyloid plaques were separated into diffuse plaques (plaques without neurites) and NPs in each region, as previously described (11, 12). An arithmetic mean was calculated from counts of diffuse plaques (number / 2.35 mm2), NPs (number / 2.35 mm2), and NFTs (number / 0.586 mm2) for each region in the 5 fields that were subjectively determined to have the greatest involvement. TDP-43 immunolabeling was performed as previously described (11).
NFT+/NP− cases were selected from the UK ADC cohort who had Braak stage III or IV with CERAD score of “negative” (n = 26). As shown in Table 3, there were 2 control groups: Braak stage 0 with CERAD “negative” (n = 21) (Control group 1) and Braak stages III or IV but CERAD “definite” (n = 15) (Control group 2).
The NACC Registry contains detailed data from more than 30 different ADCs across the United States (24). NACC data were obtained by request according to the standard application methods. Cases with no Braak stage data, no CERAD data or any subtype of frontotemporal dementia diagnosed by neuropathologic criteria were excluded. Because the NIARI consensus guidelines were published in late 1997, only cases from after 1997 were used. Based on these criteria, 6,508 cases were excluded and 5,108 cases were included.
We first evaluated data from the UK ADC (12) and the Nun Study (23, 25) to compare counts of NFTs and NPs from individual cases (Fig. 1). These data show that a subset of cases from both cohorts lack cortical NPs but have many NFTs in the medial limbic structures, i.e. the hippocampus (CA1 field and subiculum), entorhinal cortex, and amygdala. In each cohort, however, there was not an appreciable subset of cases that lacked NPs and had many NFTs in the neocortex. One case had neocortical NFTs and lacked NPs but had bilateral chronic cerebral contusions (orange arrow, Fig 1A).
Cases from the UK ADC database that were Braak stages III or IV but CERAD “negative” were defined as NFT+/NP− cases. Clinical and pathological features of these cases are presented in Table 1 and Table 2. Comparisons to the 2 control groups are shown in Table 3. The incidence of ApoE2 alleles (7 alleles in the 26 NFT+/NP− cases) was relatively high but was similar to the other CERAD “negative” cases (Control group 1). The NFT+/NP− cases tended to be relatively old (average age >88 years); few had the diagnosis of a neurodegenerative disease or mild cognitive impairment (MCI) prior to death. The average final Mini Mental State Examination (MMSE) score was 26.5 in this group. Patients in Control group 2 (Braak stages III or IV and high NPs) had an average final MMSE score of 22.1; this was not statistically different at p < 0.05 after Bonferroni correction.
NFT+/NP− cases tended to have an apparent dropout of neurons in the cornu ammonis of the hippocampus but there was not enough cell loss or astrocytosis for the designation of hippocampal sclerosis (Fig. 2). Most cases that were immunostained for TDP-43 did not show aberrant staining but most of the cases with antemortem cognitive impairment (cases 9, 14, 18, 20, 26, Table 1 and Table 2) did have aberrant staining. This proportion of aberrant TDP-43 staining (over half of cases with any degree of cognitive impairment including all with hippocampal sclerosis) is comparable with our prior studies of other cases (11); aspects of TDP-43 immunohistochemistry will be described in greater detail in a subsequent study. There was no consistent evidence of synucleinopathy, hippocampal sclerosis or argyrophilic grains in the NFT+/NP− cases (Table 2).
Eight out of 23 NFT+/NP− cases that were assessed had immunohistochemical staining for phospho-tau antibody PHF-1 (26) in the ventral midline of the medulla oblongata (Table 2; Fig. 3); the appropriate tissue blocks were not available in 3 NFT+/NP− cases. The stained cells had the histopathological appearances of glial tangles. This staining was also observed in 2/19 controls (p value vs. NFT+/NP- cases <0.07 by Chi-square); another case with this finding had cerebellar and inferior olivary degeneration.
We used cases from the NACC Registry to assess demographic indices of NFT+/NP− cases in a larger sample. Table 4 shows the relative representation of different cases stratified by Braak staging and CERAD scores. As in the UK ADC data, approximately 5% of cases are Braak stages III or IV although CERAD “negative”. There were also cases with higher Braak stages (V and VI) that also were CERAD “negative” in the NACC data; these were not evaluated since they were not represented in the UK ADC or Nun Study data sets and we were unsure what biological processes might underlie this pathology. We did not pursue detailed follow-up studies in the NACC data because of differences in neuropathological diagnostic approaches across ADCs; it has been shown that Braak staging can be different between individuals even in the best circumstances (27).
When the years of birth of persons from the NACC Registry were compared across different groups, the NFT+/NP− cases had a statistically different distribution (p < 0.001 by the Kolmogorov-Smirnov statistic; Fig. 4). One possible explanation for this difference appeared to be the tendency for NFT+/NP− cases to have been born during the years 1911 and 1912. Results according to individual year of birth “bins” are shown in Table 5.
We used multiple data sets to characterize the demographic and neuropathologic features of NFT+/NP− cases. Patients with extensive involvement of the cerebral neocortex by NFTs without concurrent NPs are extremely rare; there was only 1 case in our cohort that had bilateral chronic cortical contusions. By contrast, cases with NFTs mostly in medial temporal lobes represented approximately 5% of our cohort. Most of these patients lacked profound antemortem cognitive decline and most of their brains showed no diffuse plaques, hippocampal sclerosis, synucleinopathy or argyrophilic grains. A few cases with antemortem cognitive impairment also had aberrant TDP-43 staining, the significance of which is unclear; TDP-43 immunostaining is seen in a variety of conditions (20, 28–31). There were also some differences in ApoE alleles between NFT+/NP− cases and controls but this also is of unknown significance. There was also a tendency for some NFT+/NP− cases to exhibit neurofibrillary pathology in the medulla oblongata. These demographic and pathological features allow us to formulate a novel hypothesis about NFT+/NP− pathogenesis.
Previous studies have described brains from aged individuals with medial temporal lobe NFTs without any amyloid plaques (14, 15); the incidence in various series has ranged from 0 to approximately 6% (14–19, 32, 33). Notably, the neuroanatomical pattern of NFTs and the tau isoforms in this condition are similar to those in AD but are distinct from those in progressive supranuclear palsy, parkinsonism-dementia of Guam, corticobasal degeneration and other tauopathies (14, 16, 32).
It is unclear how NFT+/NP− cases relate to AD. Most prior studies have assumed that NFT+/NP− cases represent a neurodegenerative disease separate from AD (14, 33), but it is possible that NFT+/NP− cases belong within the AD spectrum. From this perspective the existence of these cases would argue against the amyloid cascade hypothesis (34) by indicating that NFTs do not develop downstream of amyloid plaques in AD. A number of authors have described NFT+/NP− cases in the context of “atypical AD” (35, 36) or a “variant”/“subgroup” of AD (37, 38), which can be considered a middle ground between the hypotheses that NFT+/NP− cases are separate from, or directly related to, all other AD cases.
Unlike in our series where most cases were not cognitively impaired, clinical and pathological series have referred to “NFT-predominant dementia,” “limbic NFT dementia,” “senile dementia of the NFT type,” “senile dementia with tangles,” and ”tangle-predominant senile dementia” (14, 16, 32, 38, 39). In the UK ADC autopsy cohort, only 4 cases had clinical diagnosis of AD and one had the diagnosis of MCI. Similar cases have been reported in the context of clinicopathological case series with MCI (17–19); this is more expected since Braak stages III and IV are not usually substrates for severe cognitive impairment (3); there are no UK ADC cases with higher Braak stages and no NPs. We note that the NACC data contain a few dozen Braak stages V or VI cases that are CERAD “negative”; however, prior studies appear to refute the presence of these cases (14) and we are uncertain about the clinicopathological features of these cases since we have not studied them ourselves. In contrast to NFT+/NP− cases, virtually all Braak stage VI cases have many NPs and diffuse plaques. Moreover, NPs do not develop in response to tauopathies. Furthermore, the NFT+/NP− patients are quite old; thus, there are reasons to suggest that NFT+/NP− cases are unlikely to represent “pre-clinical” stages of AD.
We and others have hypothesized that NFTs and/or their precursor tau protein elements are injurious to neurons in AD and across many different conditions (3, 37, 40–46). Aside from the “tauopathies,” NFTs are seen in brains in association with a variety of developmental, neoplastic, traumatic, and prion diseases including Niemann-Pick disease type C, pantothenate kinase-associated neurodegeneration, myotonic dystrophy, chronic traumatic encephalopathy, gangliogliomas and Gerstmann Straüssler Scheinker disease (3, 8, 47–49). There are many subtypes of neurodegeneration without NFTs, but there is no condition characterized by widespread neocortical NFTs without cognitive impairment and neuron loss. In addition, there are at least 2 viral illnesses characterized by brain NFTs: subacute sclerosing panencephalitis and post-encephalitic parkinsonism (PEP) (48, 50–52).
The clinical and pathological features of PEP deserve special note. PEP occurred in individuals who survived encephalitis lethargica (Von Economo’s encephalitis, or the “sleeping sickness”), which was linked to the 1918–1919 influenza pandemic. The influenza pandemic affected ~30% of Americans and killed approximately 20 to 40 million worldwide (53, 54) and encephalitis lethargica killed as many as 500,000 Americans (55). Clinically, PEP refers to a spectrum of extrapyramidal movement dysfunction. Pathology was mostly subcortical, but cortical changes could include hippocampal NFTs (involving the same tau protein isoforms as AD) but no amyloid plaques (52, 56–58). The disease was progressive over decades and incidence of PEP peaked in the 1950s, over 30 years after the pandemic started.
Even outside of encephalitis lethargica, the influenza virus can induce profound stress on the human central nervous system (59). Further, widespread behavioral sequelae from encephalitis lethargica were more pronounced than generally thought, affecting children differently from adults (55, 60). To the best of our knowledge, however, there has not been a systematic clinical or pathological study of survivors of the 1918–1919 pandemic who did not have encephalitis lethargica or frank PEP. Notably, the age group with peak incidence of illness from the 1918–1919 influenza pandemic in the U.S. was those born between 1909 and 1914 (53, 54). This group also coincides with the lowest mortality rate from the disease (53, 54); hence, many individuals survived from this birth cohort.
We found that NFT+/NP− cases had a greater tendency than controls to be in the above-mentioned high-risk birth cohort for infection during the 1918–1919 influenza pandemic (Table 5). This provides circumstantial evidence to support a novel hypothesis that the NFT+/NP− pattern might occur in individuals infected during the pandemic. According to this hypothesis (Fig. 5) there was a relatively “short” (~30-year) lag time to autopsy for the individuals affected by encephalitis lethargica and PEP, with commensurately dramatic clinical and pathological disease. For persons affected differently by the influenza virus, however, there might be a longer subclinical phase and patients may be very long-lived and manifest far milder NFT+/NP− pathology. European studies performed over a decade ago provided a similar age range for tangle-predominant dementia cases as in the current series (15, 61). It should be noted, however, that the cognitive impairment seen in those studies appeared more severe that in our cohort, which could indicate that milder cases survived longer with cognition more intact.
Our finding of neurofibrillary pathology in the medulla oblongata in NFT+/NP− brains may also support the hypothesis of a connection with PEP, which was characterized by subcortical NFTs without extracellular amyloid (56, 58, 62). The appearance of this pathology in NFT+/NP− brains was not statistically different from the comparison group although a trend was noted (8/23 NFT+/NP− cases, versus 2/19 cases in the control group, p < 0.07), perhaps because some cases that also have AD pathology show some of the features of this process too. Notably, Rub et al looked carefully for medulla oblongata neurofibrillary pathology in the earliest stages of AD using sensitive tau antibodies and did not report this pattern of staining (63).
Testing the hypothesis that NFT+/NP− pathology is linked to the 1918–1919 influenza pandemic would be extremely challenging. Unfortunately, reliable personal recollections or medical records are not adequate to perform a retrospective correlation study. Moreover, prior studies that have tried to isolate evidence of viral nucleic acid in PEP brains were unsuccessful (64–66). Therefore, we cannot address the question using current molecular techniques. If this hypothesis is valid, however, it would predict that the incidence of NFT+/NP− persons coming to autopsy would decrease in coming years with the passing of the members of this birth cohort. There is also debate about the causative link between influenza and encephalitis lethargica (62, 67). This concern would not obviate the point that long-term neuropathological sequelae may be associated with the influenza pandemic of 1918–1919, or, alternatively, with whatever induced PEP.
In summary, we cannot yet prove the lack of a direct relationship between AD and NFT+/NP− cases since the tau protein isoforms and anatomical distribution of NFTs in these cases are essentially identical to those in early AD (14). Further, there is suggestion from human A-beta vaccination studies that amyloid-negative NFT-positive cases can progress to end-stage AD (68).
We note that approximately one-fifth of the cases with NFT+/NP− pathology in our autopsy cohort were born after 1919 so that the birth cohort effect is not absolute. Hence, it is quite possible that NFT+/NP− cases are not due to a viral illness or other environmental effect and are instead directly relevant to AD pathogenesis. Even if the early pathogenesis of neurofibrillary pathology in NFT+/NP− cases is notably different from what occurs in AD brains, since some pathways must be relevant to both, the study of these cases may provide important clues about how AD develops.
From a diagnostic standpoint, it is notable that the large majority of NFT+/NP− cases in the UK ADC cohort had a high cognitive status. We propose that Braak stage III or IV cases that are CERAD “negative” should still fall in the NIARI “low likelihood” category in terms of AD, as so designated in prior studies (18, 19). Even if the brain shows no other contributing pathologies and the patient was profoundly cognitively impaired, we agree with Jellinger and others (14) in recommending a diagnosis of some version of non-AD dementia.
Limitations to this study include the challenges inherent in studying an uncommon, sporadic, human-specific disease process that requires expert neuropathological evaluation for detection. Most of the conclusions that we draw are based on correlations that will require validation from other data sets. The UK ADC is limited mostly to relatively well-educated Caucasians, without Parkinson disease or neuropsychiatric disorders and it is difficult to generalize beyond these data. It should be emphasized that inter-rater reliability for diagnosing Braak stages also is imperfect, even in under the best circumstances (27). Finally, since an association with NFT+/NP− pathology was seen in both the 1903–1908 and 1909–1914 birth cohorts, who likely represent the oldest subjects autopsied in this study population, an additional possibility is that this pathology is associated with extreme age, independent of exposure to influenza. We look forward to further studies from other data sets of patients with extensive medial limbic NFTs without appreciable amyloid plaque pathology.
We are deeply grateful to all of the participants in our longitudinal aging study and to the patients in our Alzheimer’s Disease Center’s research clinic. We thank Ann Tudor, Paula Thomason, Dr. Huaichen Liu and Sonya Anderson for the technical support, Gregory Cooper, MD, PhD, Nancy Stiles, MD and Allison Caban-Holt, PhD for the clinical evaluations, and Daron Davis MD for pathological evaluations. We also thank Leslie E. Phillips, MS for help with NACC data.
This study was supported by grants R01 NS061933, K08 NS050110, and P30-AG028383 from the National Institutes of Health, Bethesda, MD.