Neuropathological basis of age-associated brain atrophy

Abstract

OBJECTIVE

To examine the association between brain atrophy during life and neuropathology in an elderly population.

DESIGN

Cohort study of community dwelling elderly

PARTICPANTS

Seventy-one healthy elderly were selected from participants of the Oregon Brain Aging Study for having an autopsy, >1 MRI scan and the last MRI scan within 36 months of death.

MAIN OUTCOME MEASURES

The associations between brain volume trajectories (ventricular, total brain and hippocampal) and time interaction terms for neurofibrillary tangles (NFTs), neuritic plaques (NPs), gross infarcts, microinfarcts, amyloid angiopathy, Lewy bodies, APOE ε4 presence, and clinical diagnosis (no cognitive impairment, mild cognitive impairment (MCI) and dementia, as time varying covariates) were examined in mixed effects models, adjusting for duration of follow up and age at death.

RESULTS

Ventricular volume trajectory was significantly associated with age, presence of infarcts, NFT and NP scores, the ε4 allele and dementia diagnosis. Total brain volume trajectory was significantly associated with age, and MCI diagnosis. Hippocampal volume trajectory was significantly associated with amyloid angiopathy.

CONCLUSION

Ventricular volume trajectory is more sensitive than total brain and hippocampal volume trajectories as a marker of accruing Alzheimer disease and vascular pathology in elderly individuals. The association between brain volume trajectories and cognitive impairment (MCI and dementia) remained after controlling for the degree of neuropathology and other covariates. This suggests that there may be other factors, not measured in this study that may be contributing to brain atrophy in those with cognitive impairment.

Introduction

There is an increasing body of evidence that the clinical syndrome of Alzheimer disease (AD) in the elderly may have a different pathologic correlate compared to younger individuals with AD^1–3. The contribution of mixed pathologies, especially AD and vascular pathology, to the clinical dementia syndrome in the elderly has been replicated in several autopsy series^4–6. These observations have led to timely editorials and reviews discussing whether: 1) the clinical syndrome of AD across the age spectrum truly represents the same pathophysiological mechanisms, and 2) whether there should be separate interventions aimed at treatment and prevention of AD in young onset (<60 years old), intermediate onset (60–85 years) and late onset (85+) cases^7–9.

A related question is that of the neuropathologic correlates of longitudinal brain atrophy in the elderly. Brain volume changes are used as surrogate markers for AD neuropathology in clinical trials and biomarker studies. Yet, the extent to which these changes are due to pathologic features of AD in the aging brain is not well established. This study aims at clarifying the neuropathologic correlates of longitudinal brain atrophy in an elderly community-based population.

To our knowledge, there are only two longitudinal studies that correlated brain atrophy during life with neuropathologic measures obtained after death ^{10, 11}. These studies had discrepant findings. One of these studies found that ventricular and total brain volume changes were associated with both NP and NFT burden in the cognitively impaired (but not intact) individuals¹¹. The other study only found a correlation between NFTs and brain volume changes, but found no correlation between a measure of total amyloid (that includes vascular and both diffuse and neuritic plaques) and brain volume changes ¹⁰. We postulate that these discrepancies are a result of the age distribution of study participants in these studies (range 49–101 versus 68.9–106.2 years old), as well as the differences in the timing of MRI scans (in relation to disease status and death). Our study specifically aims at understanding the correlation between AD neuropathology and longitudinal brain atrophy in a homogenous sample of elderly who all had their first scan before onset of dementia. Our hypothesis is that, as has been observed with neuropathologic correlates of cognitive decline in the elderly ¹², a mix of AD neuropathology and vascular disease will be associated with brain atrophy over time.

Methods

Participants

Participants were healthy elderly followed as part of the Oregon Brain Aging Study (OBAS) ¹³. OBAS began in 1989 at the National Institute on Aging’s Alzheimer’s Disease Center at Oregon Health & Science University (OHSU) ¹⁴. Community dwelling elderly ≥ 55 years old, who had no functional impairment, with no comorbid illnesses; a baseline Mini-Mental State Examination (MMSE) score >23 ¹⁵, a Clinical Dementia Rating (CDR) Scale score of 0 ¹⁶ and exhibited no depression by screening with the Geriatric Depression Scale were included ¹⁷. In 2004, entry criteria were modified to include subjects with well-controlled, chronic medical conditions common with advanced age, such as hypertension and coronary artery disease, in order to include participants more representative of the general population. Of the 376 subjects evaluated between 1989 and 2005, 305 met inclusion criteria and were enrolled. Of those 305 who were enrolled, 293 were ≥ 65 years old. Attrition rates caused byloss to follow-up other than death were less than 1% per year. The Institutional Review Boards of OHSU and the Portland Veterans Affairs Medical Center approved the studies. All volunteers signed written informed consent. Volunteers were examined every six months with the MMSE ¹⁵, CDR scale¹⁶, Cognistat¹⁸ and neurological examination. Annually, a detailed battery of neuropsychological tests and brain MRI scans were performed ^13,19. Performances on the MMSE and Cognistat were included in the CDR rating. APOE was genotyped using standard methods. Of the 193 OBAS participants who died during follow up, 125 had an autopsy at death. Individuals with autopsy were older at entry to the study, and at death, but showed no differences in sociodemographic parameters, APOE4 allele distribution, duration of follow up and diagnoses compared to those that did not have an autopsy. In this study, participants were categorized into three cognitive groups based on their CDR rating: no cognitive impairment (NCI) (CDR=0), mild cognitive impairment (MCI)(CDR = 0.5 and not meeting diagnostic criteria for dementia), and dementia (CDR ≥ 0.5 and meeting diagnostic criteria for dementia). Diagnoses of possible or probable AD, vascular dementia and mixed dementia were made according to established clinical criteria ^{20, 21}. Each patient had a diagnosis assigned at each MRI time point.

Inclusion & Exclusion Criteria

OBAS volunteers were included if they: 1) had ≥ 2 MRI scans and, the last scan was performed within 36 months of death, 2) had cognitive evaluation done within 24 months of death, 3) had brain autopsy at death. Of the 125 subjects with autopsy, 74 met the above criteria. Three outliers with brain volumes > 3 standard deviations from the mean at any time point were excluded. The remaining 71 volunteers were included in this study.

MRI Methods

MR images were acquired with a1.5-T magnet. The protocol consisted of continuous-slice, multiecho, multiplanar image acquisition, with 4-mm-thick coronal slices and a 24-cm² field of view using a 256 x 256 acquisition matrix with two excitations (repetition time, 2800 msec; echo time, 30 and 80 msec). To orient the coronal plane, TI-weighted sagittal images (TR = 600 msec, TE = 20 msec); centered in the midsagittal plane were used. REGION, a semi-automated image assessment program, was used for quantitative measurement of brain volumes^{19, 22}. Image analysts were blind to the volunteers’ clinical data and previous imaging results. Image analysis methods have been described previously²². Intra-class correlation coefficient, assessing inter-rater reliability for volume measurements for all regions, was ≥ 0.9. Hippocampal, total brain and ventricular volumes were measured as a proportion of intracranial volume.

Pathological Methods

Brains were examined for NFT and NP pathology and staged by Braak and Braak, and Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) systems ^23–25. Neuropathological evaluation has been described previously ²⁶. Brains were examined grossly as well as microscopically after being fixed in neutral-buffered formaldehyde solution for at least 2 weeks. For microscopic evaluation, tissue samples were taken from all cortical lobes bilaterally, frontal lobe white matter, anterior cingulate gyrus, hippocampus, amygdala, bilateral striatum and thalamus, midbrain, pons, medulla, and cerebellum. Plaques and tangles were assessed unilaterally in most cases; cerebrovascular disease was assessed bilaterally. Six-micrometer sections from all these brain regions were stained with hematoxylin and eosin with Luxol fast blue stain for myelin (H&E/LFB). An occipital cortical section was stained with Congo red to evaluate the degree of amyloid angiopathy. A modified Bielschowsky silver impregnation method was used to identify diffuse and neuritic plaques in the frontal and parietal cortex. Selected sections of the hippocampus and neocortical regions were immunostained with antibody to tau (tau2, Sigma, St. Louis, MO). The presence of lacunar infarcts and large vessel strokes were evaluated both by visual inspection and microscopic examination. Lesions were classified as microinfarcts if there was histologic evidence of tissue necrosis and gliosis that was not identified at gross examination of the brain. Subjects were coded to have gross infarcts if lacunes or large vessel infarcts were present. The presence of Lewy bodies was determined histologically in the midbrain and nucleus locus coeruleus by evaluation of H&E/LFB-stained sections and by anti-alpha-synuclein immunohistochemistry performed on sections of amygdala and frontal cortex. Hippocampal sclerosis was coded as present or not present. Braak stages were coded into three groups, low (no NFTs or stage I or II), intermediate (stage III or IV), and high (stage V or VI). NP burden was coded as none or sparse versus moderate or frequent. Because infarcts that occurred acutely or sub-acutely around the time of death were more likely to have occurred after a subject’s last MRI scan, only more remote infarcts were considered for this study. Similarly, cerebellar or brain stem infarcts were not considered for this study, since they would not substantially affect the brain volumes of interest for this study.

Statistical Analyses

Pearson chi-squares, Fisher’s exact test or student’s t-tests were used as appropriate to compare subject characteristics between clinical diagnostic categories (cognitive impairment versus NCI). Mixed-effects models, which allow random intercept and age slope, with each brain volume (ventricular, total brain and hippocampal) as the outcome were run with the following independent variables: time (age at each MRI assessment, centered at age 85), the interaction terms of time with the following covariates: Braak stage (3 categories), plaque burden (2 categories), microinfarcts (3 categories), presence of gross infarcts, Lewy bodies, and amyloid angiopathy, diagnosis at each MRI assessment (time-varying covariate), and APOE genotype (one or more ε4 versus none). We also controlled for age at death and duration of follow-up (time from first MRI to death) as potential confounders. Covariates were selected based on their previously established associations with longitudinal brain volume changes. JMP version 5.0.1a, SAS version 9.2 (SAS Institute, Cary, NC) and R version 2.11 were used for statistical analyses. Significance was set at p<0.017 (0.05/3), the p-value adjusted for three comparisons (i.e., three volumetric measures) using the Bonferroni method.

Results

Participant Characteristics

The mean age at death was 94.7 (SD 5.46) (range 80.1–108.1) years old (Table 1). Volunteers were followed for a mean of 7.68 years (SD 3.59, range 1.5–16.1 years). Of the 71 participants who enrolled cognitively intact, 44 developed cognitive impairment. Of these, 20 had probable or possible AD, 5 had vascular dementia, 16 had MCI, and three had a mixed dementia as their last clinical diagnosis. The mean interval between last MRI and neuropathological evaluation was 15 months.

Table 1.

Subject characteristics

	Total N=71	Cognitive impairment (CDR 0.5) N=44	No cognitive impairment (CDR=0) N=27	p-values a
Women	57.75%	59.09%	55.56%	0.77
Duration of follow up, yrs	7.68 (3.59)	8.42 (3.38)	6.45 (3.63)	0.02
Education	14.50 (2.91)	14.30 (3.15)	14.82 (2.49)	0.48
Age at death	94.72 (5.46)	96.36 (5.05)	92.05 (5.11)	<0.001
Age at onset of cognitive impairment	n/a	92.47 (5.29)	n/a	n/a
Age at last cognitive evaluation	94.35 (5.49)	96.01 (5.06)	91.68 (5.17)	<0.001
Last cognitive evaluation to death, months	4.44(3.24)	4.32(3.48)	4.68 (2.88)	0.63
Age at first MRI	88.36 (5.77)	89.21 (5.32)	86.98 (6.29)	0.18
Age at last MRI	93.38 (5.27)	94.79 (4.95)	91.09 (5.05)	0.003
Time from first MRI to death, yrs	6.85 (3.86)	7.77 (3.56)	5.34 (3.92)	0.009
Time between first and last MRI, yrs	5.51 (3.61)	6.25 (3.46)	4.32 (3.59)	0.03
Last MRI to death, months	15.36 (9.96)	17.76 (9.96)	11.52 (8.88)	0.009
APOE ε4 one or more present	21.13%	18.18%	25.93%	0.44

Numbers are frequency or mean (SD)

CDR: Clinical Dementia Rating

^at-test for continuous variables and Pearson chi-square test for categorical variables, unless noted otherwise.

The NCI group had more individuals with lower Braak scores and less individuals with gross infarcts (Table 2).

Table 2.

Pathological parameters in study subjects

	Total N=71	Cognitive impairment N=44	No cognitive impairment N=27	p-valuea
Braak 1–2 or no tangles	38.03 %	25.00%	59.26%	0.02
Braak 3–4	43.66%	52.27%	29.63%
Braak 5–6	18.31%	22.73%	11.11 %
CERAD none or sparse	56.34%	54.55%	59.26 %	0.69
CERAD moderate or frequent	43.66%	45.45%	40.74%
Gross Infarcts	33.8 %	43.18 %	18.52 %	0.04 b
Microinfarcts none	53.52%	52.27%	55.56%	0.52
Microinfarcts one	21.13%	18.18%	25.93%
Microinfarcts two or more	25.35%	29.55%	18.52%
Amyloid angiopathy present	36.62%	34.09%	40.74%	0.62 b
Hippocampal sclerosis present	7.04%	9.09%	3.70%	0.64b
Lewy bodies present	19.72%	22.73%	14.81%	0.54 b

^aPearson Chi-square test unless noted otherwise.

^bFisher’s exact test.

Mixed-Effects Models

More ventricular volume enlargement over time was significantly associated with older age, diagnosis of dementia, having gross infarcts, a Braak score of 5 or 6, moderate or frequent NPs, and the presence of the ε4 allele (p<0.017). There was a trend for association with MCI diagnosis (p=0.04). That is, enlargement was accelerated during the period of MCI compared with that of normal cognition.

More total brain volume loss over time was significantly associated with older age and MCI diagnosis (p<0.017). There was a trend for association with dementia diagnosis (p=0.018), presence of gross infarcts (p=0.04), and amyloid angiopathy (p=0.03).

More hippocampal volume loss over time was significantly associated with presence of amyloid angiopathy (p=0.009) (Table 3).

Table 3.

Results of mixed effect models for the outcomes %^* ventricular, total brain and hippocampal volume change: interaction effects of age and postmortem pathologies, APOE ε4 and cognitive status.

	% Ventricular Coefficient (SE), p-value	% Total Brain Coefficient (SE), p-value	% Hippocampal Coefficient (SE), p-value
Age	0.08(0.02); <0.0001 a	−0.28 (0.07); 0.0001 a	−0.0006(0.0004); 0.14
Age* mild cognitive impairment	0.009 (0.005);	−0.07 (0.02);	0.00002 (0.0002);
Reference: no cognitive impairment	0.04	0.004a	0.89
Age* dementia	0.03 (0.007);	−0.08 (0.04);	−0.000007 (0.0002);
Reference: no cognitive impairment	<0.0001a	0.018	0.97
Age* Braak 5–6	0.06(0.02);	−0.11(0.09);	0.0005 (0.0004;)
(Reference: Braak none, 1 or 2)	0.015a	0.2	0.26
Age * Braak 3–4	−0.004(0.02);	−0.09 (0.06);	0.0003 (0.0003);
(Reference : Braak none, 1 or 2)	0.79	0.11	0.35
Age* CERAD moderate or frequent	0.05 (0.02);	−0.05 (0.06);	−0.0002 (0.0003);
(Reference: CERAD none or sparse)	0.001 a	0.39	0.48
Age* ε4 present	0.1(0.02);	−0.12 (0.08);	−0.0005 (0.0004);
(Reference: no ε4 present)	<0.0001 a	0.11	0.18
Age * gross infarct present	0.07 (0.02);	−0.12 (0.06);	−0.0004 (0.0003);
(Reference: no gross infarct present)	<0.0001 a	(0.04)	0.17
Age* Lewy Bodies present	−0.008 (0.02);	0.01 (0.07);	−0.0001 (0.0004);
(Reference: no Lewy Bodies present)	0.65	0.84	0.73
Age* Amyloid Angiopathy present	0.01 (0.01);	−0.13 (0.06);	−0.0008 (0.0003);
(Reference: no amyloid angiopathy)	0.44	0.03	0.009
Age* microinfarcts 2 or more	−0.013 (0.02);	−0.03 (0.07);	−0.0003 (0.0003);
(Reference: no microinfarcts)	0.45	0.63	0.39
Age* microinfarcts one	−0.005 (0.02);	−0.006 (0.08);	−0.0003 (0.0004);
(Reference: no microinfarcts)	0.82	0.94	0.43

In all models, duration (time from first MRI to death) and age at death were controlled.

^*proportion of intracranial volume

^aSignificant using Bonferroni multiple comparison adjusted p-value (0.017)

Discussion

In this clinicopathological correlation study examining the neurodegenerative basis of brain atrophy, we observed that of the three volumes measured, ventricular volume trajectory compared to total brain and hippocampal volume trajectories had the strongest association with neuropathologic measures of AD and vascular disease.

There are several plausible explanations why ventricular volume trajectory showed a stronger association with pathologic measures. First it may be due to the simple fact that ventricular volume measurements are more robust compared to other volumetric measurements. Because of the relatively smaller initial volume of the cerebrospinal fluid (CSF) space, the percentage of increase in ventricular volume is proportionally more rapid than percentage loss of brain tissue²⁷. Thus changes over time are easier and more robust to measure with less noise associated with repeated measurements. Alternatively, there may be a biologic explanation: ventricular enlargement is a summary marker of atrophy in both gray matter (as a result of neuropil and neuronal loss) and white matter (as a result of axonal loss) associated with aging as well as changes in CSF production and flow that occur during the aging process ^{27, 28}. It is possible that ventricular volume change correlates well with all of these aging related processes. We also found that the presence of high NP scores was more significantly associated with ventricular volume enlargement compared to NFTs. This contradicts one previous study that found NFTs, but not a measure of total brain amyloid, to be associated with total brain atrophy and ventricular volume change ¹⁰. Several differences exist between this previous study and our study that may have led to this discrepancy: the ages of the cohorts (median age at last scan was on average 10 years older and the age range of study participants was narrower in our study (80–106 versus 49–101)), timing of observations (in our study participants were monitored for a longer period of time with initial imaging obtained before onset of dementia), and differences with regard to the assessment of amyloid pathology (NP score in our study, versus a measure of total amyloid density). Two other studies found associations between measures of amyloid (NP scores in one study and 11C-PIB positron emission tomography in the other) and ventricular volume enlargement rates as we did ^{11, 29}.

Another finding worth further discussion is that the presence of amyloid angiopathy was significantly associated with increased hippocampal volume loss, and showed a trend for association with increased total brain volume loss. Two previous studies also found that presence of amyloid angiopathy correlated with smaller cross-sectional hippocampal volumes ^{30, 31}. Some studies suggest that amyloid angiopathy is associated with reduced brain perfusion ³². Additionally, hippocampus seems to be more vulnerable to brain hypoperfusion ³³. It is possible that subtle continuous chronic ischemia from amyloid angiopathy may result in atrophy, more selectively observed in the hippocampus. This observation needs further investigation in future studies. Why presence of amyloid angiopathy and not other AD neuropathologic markers correlated with hippocampal atrophy is not clear. One previous study also did not find an association between longitudinal hippocampal volume changes and NFTs and NPs ¹¹. One possible explanation is that hippocampal volume changes due to AD neuropathology occur early on, and plateau preceding changes in other volumes, where as in those with amyloid angiopathy hippocampal volume loss continues. Results also may be influenced by the fact that hippocampal volume may be more prone to variability in longitudinal measurements. This may be related to more noise due to measurement techniques as well as possible biologic fluctuations ³⁴.

Additionally, we found that ventricular volume enlargement accelerated during periods of dementia diagnosis and total brain volume atrophy increased during periods of MCI diagnosis after adjusting for the presence of the neuropathologic measures. This suggests that brain volume changes in those with MCI or dementia cannot be solely explained by the amount of AD or vascular neuropathology. There likely are other factors not measured in this study that may be mediating the degree of brain atrophy in the presence of similar neuropathologic burden. For example, functional polymorphisms in genes in the apoptotic pathway or environmental influences on neurotrophins may mediate the degree of cell loss and in turn brain atrophy in the context of AD neuropathology ^34,35. Alternatively, timing of the MRI observations in relation to cognitive evaluations may have contributed to this observation.

We also observed significant associations between the presence of gross infarcts and APOE ε4 allele and ventricular volume trajectories. While it is not surprising to find that macroscopic infarcts are correlated with brain atrophy, it was surprising to find that the presence of the APOE ε4 allele was very strongly associated with rate of change in ventricular volume. This finding was unexpected because it is thought that the effect of the APOE ε4 allele is most salient in young old individuals compared to old old individuals ³⁶. A previous study from our center found a similar APOE ε4 effect on ventricular volume in women ³⁷.

This study has several limitations. Our sample size, although among the largest to assess the clinicopathological correlation of longitudinal brain volume trajectories, was nevertheless too small to examine the effects of less frequent pathologies, such as HS, or categorizing Lewy bodies as cortical versus non-cortical. Additionally, assessments for other markers of pathogenic contributors to dementia such as TDP-43 were not included. Our sample data is derived from volunteers in the ninth through eleventh decades of life, a segment of the population at highest risk for cognitive impairment and death. Conclusions relative to younger at risk individuals remains speculative at this point. The strengths of this study are: 1) use of a well described set of participants who were all enrolled as cognitively intact elderly and followed over a mean of 7.7 years until death with multiple MRI scans and clinical assessments during life; 2) the inclusion of other confounding pathological measures such as vascular disease and Lewy bodies in the analyses; and 3) use of mixed-effects models, which allowed for including participants with different durations of follow-up and for individual differences in time slope (random effects), and adjusting for the changing diagnoses during the duration of follow up.

Our findings suggest that in individuals in the ninth through eleventh decades of life, the trajectory of ventricular volume, but not other regional or total brain volumes, shows the strongest correlation with neuropathologic measures of AD and vascular disease. Given their relative ease of measurement, ventricular volume trajectories can be widely used when providing care or assessing potential new treatments to an increasingly elderly population. Additionally, our data suggest that the brain volume changes over time in those with cognitive impairment are different than those without cognitive impairment even after accounting for the degree of the neuropathologic burden. This suggests that other mechanisms possibly related to cognitive reserve or resistance to atrophy may be modulating brain atrophy.

Figure 1.

Trajectories of ventricular volume change. Dotted lines represent the predicted trajectories based on estimates using the coefficients generated in the mixed effects models and solid lines represent the observed trajectories.

Figure 1 A and B represent trajectories for the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuritic plaque categories; CERAD none or sparse (A), and CERAD moderate or frequent (B).

Figure 2.

Trajectories of ventricular volume change. Dotted lines represent the predicted trajectories based on estimates using the coefficients generated in the mixed effects models and solid lines represent the observed trajectories.

Figure 2 A, B and C show trajectories for the three Braak groups: Braak none, 1 or 2 (A), Braak 2 or 3 (B), Braak 5 or 6 (C).