Imaging β-amyloid burden in aging and dementia

Histopathologic studies show extensive cortical β-amyloid (Aβ) deposition in Alzheimer disease (AD), but it may also be found in asymptomatic elderly persons, adults with Down syndrome, and in dementia with Lewy bodies (DLB). In Down syndrome, Aβ deposition begins over a decade prior to the dementia that develops in the majority of persons with this condition.¹ Asymptomatic cortical Aβ deposition in elderly individuals is well documented with one-fourth or more of the nondemented population age over 75 years having moderate numbers of neuritic plaques in the cerebral cortex.^2,3 The neuropathologic findings in DLB, a condition that may account for up to 15% of dementia, frequently includes Aβ plaques and neurofibrillary tangles (NFTs).^4–6 Consequently, the term “Lewy body variant of AD” is preferred by some authors.⁷ “Pure DLB” with no Aβ accumulation is seen in only 10 to 20% of clinically diagnosed cases.^4,8–12 Frontotemporal dementia (FTD) accounts for another 15 to 20% of dementia.¹³ The frontal (behavioral) and the temporal (semantic) forms of FTD rarely have Aβ plaque at post mortem.¹³ Hence, the ability to image Aβ may allow distinction of AD from FTD and give new insights into the role of Aβ in DLB.

(N-Methyl-[¹¹C])2-(4′-methylamino-phenyl)-6-hydroxy-benzothiazole ([¹¹C]6-OH-BTA-1), also known as “Pittsburgh Compound B” ([¹¹C]PIB), a carbon-11-labeled derivative of the thioflavin-T amyloid dye, binds with high affinity and high specificity to neuritic Aβ plaques.^14,15 PIB does not show significant binding to diffuse plaque, NFT, or “pure” DLB brain homogenates.¹⁶ Human PET studies have shown robust cortical binding in AD patients^17–21 and correlation with the rate of cerebral atrophy in AD subjects,²² with decreased CSF Aβ_1-42 in both demented and nondemented subjects²³ and with parietotemporal hypometabolism.²⁴

In this study we sought to compare [¹¹C]PIB PET as a biomarker for Aβ burden in AD, DLB, FTD, mild cognitive impairment (MCI), and normal elderly persons with previously established neuropathologic findings in these conditions and with clinical features.

METHODS

Twenty-seven elderly individuals with well documented normal cognitive function, 17 patients with mild to moderate AD, 10 patients with DLB, 6 patients with FTD, and 9 subjects with MCI were recruited for the study (table 1). Normal volunteers were recruited from a cohort of subjects participating in the longitudinal Healthy Aging Study at the Mental Health Research Institute of Victoria and had shown normal cognitive performance on neuropsychological tasks that included California Verbal Learning Test (CVLT II), Rey figure, Logical Memory, verbal and categorical fluency, Boston Naming Task, and digit span. All subjects in the current study were assessed with this neuropsychological test battery, the Mini-Mental State Examination (MMSE),²⁵ and the Clinical Dementia Rating²⁶ within 1 week of the PIB scan. All AD patients met National Institute of Neurological and Communication Disorders and Stroke/Alzheimer's Disease and Related Disorders Association criteria for probable AD,²⁷ whereas all subjects in the DLB group met the consensus criteria for probable DLB of cognitive fluctuation, visual hallucinations, and parkinsonism.⁴ FTD subjects had characteristic clinical presentations¹³ and frontal lobe or temporal lobe atrophy on MRI with concordant hypometabolism on fluorodeoxyglucose PET. MCI subjects met the Petersen criteria of subjective and objective cognitive difficulties, predominantly affecting memory, in the absence of dementia or significant functional loss.²⁸ All patients were recruited from the Austin Health Memory Disorders and Neurobehavioral Clinics.

Written informed consent for participation in this study was obtained prior to the scan. Approval for the study was obtained from the Austin Health Human Research Ethics Committee.

All subjects underwent a three-dimensional spoiled gradient echo T1-weighed MRI acquisition for screening and subsequent co-registration with the PET images.

APOE genotype was determined by PCR amplification of genomic DNA.

Production of [¹¹C]PIB was performed in the Centre for PET, Austin Hospital, using the one-step [¹¹C]methyl triflate approach.²⁹ The average radiochemical yield was 30% after a synthesis time of 45 minutes with a radiochemical purity of >98% and a specific activity of 30 ± 7.5 GBq/μmol. Each subject received 375 ± 18 MBq [¹¹C]PIB by IV injection over 1 minute. Imaging was performed with a Phillips Allegro PET camera. A rotation transmission sinogram acquisition in three-dimensional mode with a single ¹³⁷Cs point source was performed before the injection of the radiotracer for attenuation correction. A 90-minute list-mode emission acquisition was performed in three-dimensional mode after injection of PIB. List-mode raw data were sorted off line into 4 × 30-second, 9 × 1-minute, 3 × 3-minute, 10 × 6-minute, and 2 × 10-minute frames. The sorted sinograms were reconstructed using a three-dimensional RAMLA algorithm.

Co-registration of the PET images with each individual's MRI was performed with SPM2 (Statistical Parametric Mapping, MRC Cognition and Brain Sciences Unit).³⁰ For registration purposes, the initial frames of the dynamic PET studies were summed. The early frames of the study reflect regional blood flow allowing an easy co-registration with the MRI. Regions of interest (ROIs) were then drawn on the individual MRI. Mean radioactivity values were obtained from ROIs for cortical, subcortical, and cerebellar regions as listed in table 2, and decay-corrected time–activity curves were generated. White matter ROIs were placed at the centrum semiovale, and the cerebellar regions were placed over the cerebellar cortex, taking care to avoid white matter. The orbitofrontal ROI included both the mesial and lateral aspects (including superior, middle, and inferior frontal gyri). No correction for partial volume was applied to the PET data.

Distribution volume ratios (DVRs) were determined through graphical analysis.³¹ To avoid arterial blood sampling, a simplified approach was applied using the cerebellar cortex, a region relatively unaffected by amyloid deposition, as input function.^17,20,31 Besides regional DVR values, the mean of the DVRs for frontal, cingulate, parietal, lateral temporal, and occipital cortex was calculated and termed the neocortical DVR. DVR images were created for visual inspection with a rainbow color scale.

Standardized uptake value (SUV), defined as the decay-corrected brain radioactivity concentration, normalized for injected dose and body weight, was calculated for the cerebellar cortex late in the scan when PIB binding reaches apparent steady state^20,32 to compare binding in the reference region with age, diagnosis, and cognitive status.

Statistical evaluations were performed using a Wilcoxon signed-ranks test followed by a Dunnet test to compare each group with controls and a Tukey–Kramer highly significant difference test to establish differences between group means. Pearson product–moment correlation analyses were conducted between the neocortical DVR and clinical features. Data are presented as means ± SD unless otherwise stated.

RESULTS

[¹¹C]PIB distribution.

Brain radioactivity peaked between 3 and 6 minutes post injection, and the binding appeared to be reversible with rapid washout from all areas in controls other than white matter (figure 1). [¹¹C]PIB cleared fastest from cerebellar cortex, and the rate was the same for all groups, consistent with the absence of significant Aβ in this region in all subjects (figure 1). The SUV measurements in the cerebellar cortex showed no difference between subject groups and no correlation with age or with dementia severity in the AD subjects as assessed by MMSE (r = 0.11, p = 0.21) or CDR (r = −0.08, p = 0.74).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Time–activity curves

Time–activity curves demonstrating uptake and clearance of [¹¹C]Pittsburgh Compound B (PIB) in frontal cortex, cerebellar cortex, and white matter. There is greater retention of PIB in the frontal cortex in Alzheimer disease and to a lesser extent in dementia with Lewy bodies than in frontotemporal dementia or normal elderly control subjects. In contrast, clearance is identical in the cerebellar cortex, demonstrating that this is an appropriate reference region. There were no significant differences between groups in the clearance from white matter. SUV = standardized uptake value.

On visual inspection of the DVR images, all AD subjects showed marked cortical PIB binding, greatest in the precuneus/posterior cingulate and frontal cortex and the caudate nuclei, followed by lateral temporal and parietal cortex. Occipital, sensorimotor, medial temporal, and thalamic gray matter was less affected (figure 2). Cerebellar cortex showed no uptake. Most DLB subjects also showed increased PIB binding, similar in distribution to AD (figure 2). However, the degree of binding was generally lower and varied widely between DLB subjects ranging from normal to AD levels (figure 3). One DLB subject and all six FTD subjects had no gray matter PIB binding.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 In vivo imaging of β-amyloid (Aβ) burden in aging and dementia

Representative distribution volume ratio (DVR) PET transaxial images (top) and sagittal images (bottom) of a 73-year-old healthy control (HC) subject (Mini-Mental State Examination [MMSE] = 30), a 78-year-old patient with dementia with Lewy bodies (DLB) (MMSE = 19), an 82-year-old patient with Alzheimer disease (AD; MMSE = 22), and an 80-year-old patient with frontotemporal dementia (FTD; MMSE = 25). DVR PET images show clear differences when comparing HC or FTD subjects with DLB or AD patients, with nonspecific Pittsburgh Compound B (PIB) binding in white matter in the HC and FTD subjects compared with PIB binding in the frontal, temporal, and posterior cingulate/precuneus cortex of the AD and DLB patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Box-and-whiskers plot

Box-and-whiskers plot displaying median and 1st and 99th percentiles of neocortical β-amyloid (Aβ) burden as quantified by [¹¹C]Pittsburgh Compound B distribution volume ratio for Alzheimer disease (AD; □), dementia with Lewy bodies (DLB; ▴), frontotemporal dementia (FTD; ♦), mild cognitive impairment (MCI; ▪), and healthy control (HC; ○). †Significant results for MCI, DLB, and FTD vs AD (p < 0.05). ‡Significant results vs controls (p < 0.05).

PIB binding in MCI subjects either clustered in an “AD-like” (n = 5) pattern that was indistinguishable from AD subjects or had normal binding (n = 4) (figure 3).

Twenty-one HCs showed no binding in cortical or subcortical gray matter, and their scans were clearly distinguishable from those of subjects with AD (figure 2). However, six (22% of HCs) showed a range of increased [¹¹C]PIB binding in gray matter (figure 3). Four of these six had binding in orbitofrontal cortex with variable involvement of cingulate, precuneus, and temporal cortex, a pattern similar to that seen in AD and that resembles the stages of Aβ deposition (figure 4).³³ One subject showed isolated binding in the right occipital lobe, and one displayed focal areas of binding in several cortical regions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4 Representative sagittal PET images

Representative sagittal PET images showing the regional uptake of [¹¹C]Pittsburgh Compound B (PIB), reflecting β-amyloid (Aβ) burden in the brain, in three asymptomatic healthy age-matched control subjects (HC 1 to 3) and one patient with Alzheimer disease (AD; top), and schematics showing the stages of Aβ deposition in the human brain as proposed by Braak and Braak (bottom).³³ Twenty-two percent of HCs had PIB binding ranging from stages A to C. All AD subjects matched stage C.

Regional DVR values confirmed higher binding in neocortical areas in AD and DLB patients when compared with HC subjects and no cortical binding in FTD (table 2). DVR values clustered in MCI to either the normal or the AD range (figure 3). In contrast to the 70 to 80% greater frontal and precuneus DVR in AD over controls, binding in the medial temporal region was only increased by 14% in AD.

Correlation with clinical features.

Diagnosis.

The PIB scans of all AD subjects could be readily distinguished from normal studies both visually and by neocortical DVR values. A cut-off value for the neocortical DVR defined as 2 SD greater than the HC mean detected all AD while including one HC. Three HCs could not be distinguished from AD by visual inspection and their DVR values were very close to those of the AD subjects. Three other HCs showed cortical PIB binding of clearly lesser degree and extent than seen in AD. PIB scans were able to distinguish all FTD subjects from AD subjects. PIB scans were not able to reliably distinguish DLB or MCI from AD other than in those cases with no or low cortical binding.

Age.

The six HCs with increased PIB binding on visual inspection were older than the other HCs (78.5 vs 72.2 years; p = 0.03). On correlational analysis, there was a nonsignificant trend toward greater neocortical DVR with increasing age in the HC (r = 0.3); however, there was no correlation when all subjects were combined or examined by disease groups.

Rate of disease onset.

In the DLB subjects, high neocortical DVR correlated with shorter time between the onset of cognitive impairment and development of the diagnostic clinical features (r = −0.75, p = 0.01) (figure 5). Of all cortical areas, this correlation was strongest in the posterior cingulate cortex (r = −0.8, p = 0.005). There was no correlation in the AD group between neocortical DVR and time from onset of cognitive decline to diagnosis (r = −0.06, p = 0.9).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5 Relationship between neocortical Pittsburgh Compound B distributional volume ratio and time elapsed

Relationship between neocortical Pittsburgh Compound B distribution volume ratio and time elapsed (expressed in years) between first sign of cognitive impairment recalled by caregiver and development of the diagnostic clinical features of dementia with Lewy bodies (r = −0.75, p = 0.013). Dotted lines represent 95% confidence limits for the linear regression.

Cognitive status.

Both the CDR and the MMSE correlated with neocortical DVR when all subjects were pooled (CDR: r = 0.49, p < 0.001; MMSE: r = −0.42, p < 0.005). However, there was no correlation in any group when analyzed separately. For example, in the AD group, there was no correlation between PIB DVR and MMSE score (r = 0.06, p = 0.81) or with specific domains of cognitive function such as episodic memory (CVLT II short delay recall r = −0.32, p = 0.30).

The cognitive performance of the HCs with cortical PIB binding compared with those without was not significantly different. The mean MMSE was 29 for both groups, and all subjects performed within the expected range for age and education on neuropsychological tasks. There was a minor but nonsignificant reduction in performance on several cognitive tests in those with PIB binding such as for the CVLT II long delay (9.8 ± 3.1 vs 11.8 ± 2.8). Likewise, no difference in cognitive deficits was observed between MCI subjects with and without cortical PIB binding.

APOE status.

An APOE ε4 allele was present in 56% of AD, 78% of DLB, 44% of MCI, 33% of FTD, and 30% of HC subjects (table 1). With all subjects combined, the presence of an APOE ε4 allele was associated with higher PIB binding (neocortical DVR 1.61 ± 0.41 vs 1.38 ± 0.39, p = 0.02). Within groups there was no significant difference.

Medication.

Ten of the 17 AD subjects were taking acetylcholine esterase inhibitor medication at the time of the PIB scan. Those on this medication showed no difference in PIB DVR vs the other AD subjects (1.97 ± 0.21 vs 2.05 ± 0.2; p = 0.45).

DISCUSSION

In vivo amyloid imaging with PET is allowing new insights into Aβ deposition in the brain, facilitating research into the causes, diagnosis, and future treatment of dementias where Aβ may play a role. In this study we have demonstrated that extensive Aβ deposition is present in the neocortex and caudate nuclei of all patients with AD, even when the severity of the disease is classified as mild. This is consistent with postmortem studies that suggest the deposition of Aβ is well advanced prior to the onset of dementia. Support for this hypothesis is provided by our observation of cortical PIB binding in 22% of the normal elderly subjects in this study. This accords well with the reported prevalence of AD at age 85 years of 15 to 25% and postmortem reports that 30% of nondemented older persons over age 75 have moderate numbers of neuritic plaques in the cerebral cortex.^2,34 The mean age of our normal cohort was 73 years and the mean age of those normal subjects with cortical PIB binding was 78 years. In a recent study of nondemented subjects, PIB binding was found in 3 of the 20 subjects over age 65.³⁵ These subjects showed no difference in cognitive performance vs those with no PIB uptake. Our HCs with PIB binding also showed no significant reduction in cognitive performance, though there was a trend on several measures that may have reached significance with larger subject numbers. A recent report of neuropathologic findings in 134 elderly persons without cognitive impairment found a slight but significant reduction in episodic memory test scores in those with moderate numbers of neuritic plaques.³ It is possible that PIB PET is detecting preclinical AD. Longitudinal observation of a large cohort of cognitively normal elderly subjects studied with PIB PET is required to confirm or refute this hypothesis.

We did not find a correlation between the Aβ burden and the severity of dementia in the AD cases. Neuropathologic reports are not consistent on this question, with studies reporting greater plaque density with worsening dementia,^36–38 no correlation,² or a correlation only in those under age 75.³⁹ Some have found a correlation with levels of soluble but not insoluble Aβ.⁴⁰ Our data suggest that Aβ deposition is an early event and likely to occur prior to demonstrable cognitive impairment with little subsequent change in Aβ burden. A recent report on PIB PET repeated 2 years apart in 16 AD subjects supports this interpretation, finding no change between scans.⁴¹ The authors did find a correlation between the degree of PIB binding on the first scan and the likelihood of progression over this 2-year period. Caution is required in interpreting the apparent lack of correlation between PIB binding and dementia severity observed in our study. Studies to date, including ours, have not corrected the PET data for atrophy. Partial volume effects,⁴² where increasing cortical atrophy results in an apparent reduction in radioactivity, could mask an increase in binding over time or with increasing dementia severity. Lack of sufficient cases with moderate or severe dementia and relatively small subject numbers may also result in failure to detect a correlation. Increased PIB binding in the cerebellar reference region would mask an increase in cortical binding. There are reports of Aβ in the cerebellum in advanced AD predominantly in diffuse plaques to which fibrillar dyes and PIB related compounds bind poorly.^16,43,44 Our SUV measurement of cerebellar PIB binding and our time–activity curves for cerebellar cortex found no evidence of increased uptake in AD with age or with increasing severity of dementia.

PIB binding in our AD patients closely matched the reported histopathologic distribution of neuritic plaques.^33,45,46 In all of the AD subjects in our study, the distribution of Aβ as measured by [¹¹C]PIB binding was extensive and corresponded closely to stage C of the Braak Aβ deposition categories (figure 4).^33,45 The high binding of PIB in the caudate nucleus has been observed by other groups.^17,35 Neuropathologic studies have reported high levels of Aβ in the caudate, less in the putamen, and little in the globus pallidus.^46,47

The relatively small increase in binding in the medial temporal lobe agrees with the histopathologic observation that in mildly demented patients, the plaque density in the hippocampus is one-third that in the orbital cortex.² In our affected normal elderly subjects, the area that most frequently displayed PIB binding was the orbitofrontal cortex followed closely by the posterior cingulate/precuneus, consistent with the neuropathologic stages of neuritic plaque deposition.^33,45 White matter binding was observed in all subjects, and though it is likely to represent nonspecific binding, no explanation for it has yet emerged.

In vitro studies indicate that PIB binding is highly specific for Aβ when using a tracer dose at nanomolar concentration, and our robust difference between the binding in AD vs HC and FTD shows that this specificity holds for in vivo studies. At the nanomolar concentrations attainable in human PET studies, [¹¹C]PIB and related benzothiazole derivatives bind at 10-fold higher levels to AD frontal cortex homogenates of extracellular plaque and perivascular Aβ than the background binding observed in amyloid-free control brain frontal cortex.¹⁶ Under these same conditions, benzothiazole compounds do not produce detectable binding to brain homogenates from the frontal cortex of pure DLB brain.¹⁶

In this study, PIB binding was consistent with the histopathologic observation that the density of classic Aβ deposits is significantly lower in the cortex of DLB subjects compared with AD and absent in a small proportion of DLB.^4,8–12,48

Our observation that higher Aβ burden is associated with more rapid development of the full DLB phenotype has not been previously reported. Data suggest that Aβ may influence the development of DLB. The cortical Aβ deposits in PD with dementia, a condition with many similarities to DLB, are associated with extensive α-synuclein lesions and higher levels of insoluble α-synuclein.⁴⁹ Further evidence comes from studies in transgenic mice. Cross-breeding of mice that develop α-synuclein related lesions with mice that develop cerebral Aβ deposits results in greater aggregation of α-synuclein and exacerbation of α-synuclein-dependent neuronal injury.⁵⁰ These data suggest a possible role for Aβ in DLB. An alternative explanation is that factors favoring amyloid deposition also promote synuclein pathology. Either way, our findings suggest that strategies to reduce Aβ may also be of benefit in DLB. Forty to sixty percent of subjects with MCI progress to AD.²⁸ Longitudinal follow-up studies will determine if the two well differentiated patterns of PIB binding observed in our MCI subjects reliably predict those who will progress to AD.

None of the FTD subjects in our study showed cortical PIB binding. This suggests that PIB PET may assist the differential diagnosis of AD from FTD.

A potential limitation of our study is the reliance on clinical diagnosis as the “gold standard” rather than neuropathology. Consequently, we aimed for specificity rather than sensitivity, selecting only those subjects with characteristic clinical presentations supported by appropriate structural and functional imaging findings. For example, all the DLB subjects had all of the three consensus diagnostic features of persistent visual hallucinations, parkinsonism, and fluctuation in cognition. Neuropathologic studies have shown that the presence of two or more of these features results in a specificity of greater than 85% for DLB.⁴ Our observation that all of the AD subjects had extensive PIB uptake and none of the FTD subjects had PIB uptake indicates that our approach to subject selection produced reasonably pure cohorts. At this time postmortem confirmation of the diagnoses is available in only one subject. This patient had a clinical diagnosis of DLB and had high cortical PIB binding indistinguishable from the AD subjects. Postmortem neuropathologic examination confirmed the presence of large numbers of cortical neuritic Aβ plaques, Lewy bodies, and NFTs. Another limitation of our study is the relatively small number of subjects with DLB and MCI with great variability in PIB binding. Therefore, though PIB PET images correlate well with neuropathologic observations of the incidence and distribution of Aβ plaques in aging and neurodegenerative diseases in this study, confirmation with larger sample sizes is warranted before it can be claimed that in vivo imaging of Aβ will allow more accurate and specific diagnosis of dementia.

ACKNOWLEDGMENT

The authors thank Jessica Sagona, Kunthi Pathmaraj, Tim Saunder, Bridget Chappell, Jason Bradley, and Gareth Jones for their help with PET examinations and image processing.