Regional analysis of striatal and cortical amyloid deposition in patients with Alzheimer's disease

We aimed to analyse the detailed distribution pattern of amyloid‐β (Aβ) in the striatum, and to examine whether there is any correlation between Aβ deposition levels in the striatum and cortical regions. Twenty patients with Alzheimer's disease underwent positron emission tomography using 11C‐Pittsburgh Compound B (11C‐PiB) to quantify the Aβ deposition. Volumes‐of‐interest analyses were performed on the ventral striatum (VST), pre‐commissural dorsal caudate (pre‐DCA), post‐commissural caudate (post‐CA), pre‐commissural dorsal putamen (pre‐DPU), and post‐commissural putamen (post‐PU), followed by exploratory voxel‐wise analyses. Volumes‐of‐interest analyses of 11C‐PiB binding showed: VST > pre‐DPU (P = 0.004), VST > pre‐DCA (P < 0.0001), pre‐DPU > post‐PU (P < 0.0001), and pre‐DCA > post‐CA (P < 0.0001), consistent with visual inspection of the 11C‐PiB images. Exploratory voxel‐wise analyses of 11C‐PiB binding showed a positive correlation between the VST and the medial part of the orbitofrontal area (P < 0.01 family‐wise error corrected). This study confirmed that there were ventral > dorsal, and anterior > posterior gradients of Aβ deposition in patients with Alzheimer's disease, and provided the first evidence of a robust correlation between Aβ deposition levels in the VST and the medial part of the orbitofrontal area. There are well‐known anatomical and functional links between these areas. These findings indicated that brain Aβ deposition was not randomly distributed, but had characteristic patterns related to anatomical connectivity and/or functional networks.

There are ventral > dorsal and anterior > posterior gradients of Aβ deposition in the striatum, and is a robust correlation between Aβ deposition levels in the ventral striatum and the medial part of the orbitofrontal area. There are well known anatomical and functional links between these areas. Aβ deposition may have characteristic patterns related to anatomical connectivity and/or functional networks.

11C‐Pittsburgh Compound B; Alzheimer's disease; amyloid; positron emission tomography; striatum

The positron emission tomography (PET) radioligand, 11C‐Pittsburgh Compound B (11C‐PiB) has been shown to possess high affinity and high specificity for fibrillar amyloid‐β (Aβ) (Mathis et al., [35] ; Klunk et al., [25] ; Lockhart et al., [30] ; Ye et al., [59] ; Rowe & Villemagne, [47] ). Using 11C‐PiB and PET enables us to visualise and quantify the amount of Aβ deposition, which is a useful biomarker for the diagnosis of Alzheimer's disease (AD) (McKhann et al., [37] ; Sperling et al., [52] ). The main brain regions showing 11C‐PiB binding are the frontal and posterior cingulate/precuneus cortices and the striatum (Vallabhajosula, [55] ), consistent with known patterns of Aβ deposition, as described in post‐mortem studies (Braak & Braak, [6] , [7] ). The correlations between 11C‐PiB binding across different regions, and region‐matched post‐mortem measures of fibrillar Aβ deposition levels have previously been confirmed (Bacskai et al., [3] ; Ikonomovic et al., [19] ).

The Aβ deposition begins well before the onset of cognitive decline (Sperling et al., [52] ; Villemagne et al., [56] ), and the earliest deposition appears to be in the frontal and cingulate/precuneus regions (Mintun et al., [38] ). Compared with the cortical regions, detection of Aβ deposits in the striatum occurs at a later stage in AD progression (Braak & Braak, [6] ; Thal et al., [54] ; Mintun et al., [38] ; Beach et al., [4] ). Some authors have indicated that Aβ plaques in the striatum are largely restricted to individuals with clinically‐documented dementia (Braak & Braak, [5] ; Thal et al., [54] ), although others have found plaques in clinically non‐demented individuals (Wolf et al., [58] ). Moreover, recent studies have shown that striatal Aβ plaques correlate with the presence of dementia in patients with Lewy body disease (Kalaitzakis et al., [22] , [23] ). These findings indicate that Aβ deposition and neural circuits in the striatum are important for the regulation of cognition. There are, however, neither in vivo nor post‐mortem studies that address the distribution pattern of Aβ deposition in the striatum in detail.

Martinez et al. ([34] ) anatomically divided the striatum into five subregions: ventral striatum (VST), dorsal caudate rostral to the anterior commissure (AC) [pre‐commissural dorsal caudate (pre‐DCA)], dorsal putamen rostral to the AC [pre‐commissural dorsal putamen (pre‐DPU)], caudate caudal to the AC [post‐commissural caudate (post‐CA)], and putamen caudal to the AC [post‐commissural putamen (post‐PU)] (Mawlawi et al., [36] ; Martinez et al., [34] ). The VST includes the nucleus accumbens, ventral caudate rostral to the AC, and ventral putamen rostral to the AC. The post‐CA and post‐PU include the caudate and putamen from the plane of the AC to the plane through the most caudal part of these regions.

The first aim of this study was to analyse the distribution patterns of Aβ deposition among the striatal subregions in patients with AD, using 11C‐PiB and PET. There is a concept that the molecular pathology leading to Aβ aggregation progresses through specific anatomical connections and/or functional networks (Seeley et al., [49] ; Raj et al., [46] ). Therefore, the second aim was to examine whether there are correlations between 11C‐PiB binding in cortical regions and any of the striatal subregions, and to discuss the relationship between any two related regions in the light of anatomical and functional connectivity. The associations between cognitive measures and 11C‐PiB binding were also assessed.

The present study was performed retrospectively. The subjects comprised 20 patients [8 men and 12 women; age 50–83 years (mean age 67.6 years, SD 9.4)] and 14 age‐matched control subjects [nine men and five women; age 52–84 years (mean age 64.6 years, SD 9.8)]. All patients had no family history of early‐onset dementia, and were diagnosed with AD at or after the time of 11C‐PiB PET scanning on the basis of clinical criteria (American Psychiatric Association, [1] ). Apolipoprotein E genotype was measured in 13 of 20 patients. Five, six and two patients presented the 3/3, 3/4 and 4/4 apolipoprotein E genotype, respectively. For each patient, the diagnosis of AD was supported by the findings from 18F‐fluorodeoxyglucose PET, which is used to estimate the cerebral metabolic rates of glucose utilisation. Briefly, all patients showed decreased 18F‐fluorodeoxyglucose uptake in the posterior cingulate, precuneus, and temporoparietal cortices, as a typical pattern of AD (Sperling et al., [52] ). All patients also underwent magnetic resonance imaging (MRI), and did not show any significant findings except for cortical atrophy, presumably due to AD. All control subjects were defined as healthy on the basis of their medical history, the results of their physical and neurological examinations, and the MRI findings.

All participants underwent 11C‐PiB PET scanning at Tokyo Metropolitan Institute of Gerontology for research purposes, during the period from December 2006 to January 2013. At the time of 11C‐PiB PET scanning, the Mini Mental State Examination (MMSE) scores of the patients were 14–30 (mean 22, SD 5). The diagnosis of AD in each patient with a relatively high MMSE score was confirmed by the clinical course after 11C‐PiB PET scanning. The Ethics Committee of the Tokyo Metropolitan Institute of Gerontology approved this study protocol, and written informed consent was obtained from all of the patients. The study conformed to the World Medical Association Declaration of Helsinki published on the website of the Journal of American Medical Association in 2013.

The radioligand, 11C‐PiB, was synthesised as described previously (Wilson et al., [57] ) with slight modifications. PET scanning was performed on a scanner (SET‐2400W; Shimadzu, Kyoto, Japan) in three‐dimensional mode. Images with 50 slices were obtained with a 2.054 × 2.054 × 3.125‐mm3 voxel size and a 128 × 128 matrix size. The transmission data were acquired using a rotating 68 Ga/68Ge rod source for measured attenuation correction. Static emission data were acquired for 40–70 min after intravenous bolus injections of 11C‐PiB. The injection doses were 462 ± 102 MBq, and the specific activities were 56.7 ± 46.4 MBq/nmol at the time of injection (mean ± SD). Data were reconstructed after correction for decay, attenuation, and scatter.

The MRI scanning was performed using a 1.5‐Tesla Signa EXCITE HD scanner (GE, Milwaukee, WI, USA) in three‐dimensional mode (3DSPGR; repetition time, 9.2 ms; echo time, 2.0 ms; matrix size, 256 × 256 × 124; voxel size, 0.94 × 0.94 × 1.3 mm), and was processed using the FMRIB Software Library (FSL; Oxford University, Oxford, UK).

Volumes of interest (VOIs) included the striatal subregions as target regions, with the cerebellum as a reference region. A whole‐striatum VOI was created by combining the caudate, putamen, and nucleus accumbens VOIs, anatomically defined in native space using FSL FIRST. The whole striatum was divided into five anatomical VOIs (Fig. [NaN] A, C, and D): the VST, pre‐DCA, post‐CA, pre‐DPU, and post‐PU, as described in the Introduction (Mawlawi et al., [36] ; Martinez et al., [34] ). A cerebellum VOI (Fig. [NaN] B and D) was created by transforming a bilateral VOI, drawn manually on the cerebellar cortex in Montreal Neurological Institute (MNI) 152 space, to native space using FSL FNIRT.

The static PET images were co‐registered to the corresponding structural MRI using FSL FLIRT. The VOIs placed on the MRI were moved onto the PET images, and VOI‐based analyses were performed. Activity data within the anatomically defined VOIs (see above) were extracted from co‐registered PET images. 11C‐PiB binding for each VOI was quantified, with the cerebellum used as a reference region (Lopresti et al., [32] ; Ikonomovic et al., [19] ). The ratio (activity in the target region:activity in the cerebellum) was described as the standardised uptake value ratio (SUVR).

The VOI‐based analyses were performed to test the difference in SUVR values between the two striatal subregions using the paired t‐test with Bonferroni adjustment as a post‐hoc test after one‐way repeated measures anova in SPSS (IBM Corp., Armonk, NY, USA). The differences in SUVR values between patient and control groups were tested using the independent t‐test. Relationships between SUVR values and the MMSE scores in each striatal subregion were tested using Pearson Correlation analysis. Statistical significance was set at P < 0.01 (two‐tailed).

Exploratory voxel‐wise analyses were then performed to determine the cortical region where 11C‐PiB binding was correlated with SUVR values in each of the striatal subregions. The SUVR images of 11C‐PiB were transformed into MNI152 space from native space using MRI‐guided spatial normalisation (FSL FNIRT). After smoothing with a Gaussian kernel of sigma 4 mm to improve the signal‐to‐noise ratio, cortical areas on the smoothed images were masked with a Harvard–Oxford cortical atlas (included in FSL). All masked voxels were then subjected to linear regression analyses with SUVR values in each of the five striatal subregions, which were obtained by VOI‐based analyses, using Statistical Parametric Mapping 8 (Wellcome Trust Center for Neuroscience, London, UK) implemented in MATLAB 7.0 (MathWorks Inc., Natick, MA, USA). Statistical t maps of positive contrast were calculated using a height threshold of P < 0.01 family‐wise error (FWE) rate corrected (T > 6.38), excluding clusters smaller than 200 voxels. The Statistical Parametric Mapping t maps were transformed to FWE corrected P maps.

The VOI‐based analyses showed that the SUVR values of 11C‐PiB in patients with AD were significantly greater than those in healthy controls in each of the five subregions using the independent t‐test (Table [NaN] ). For patients with AD, there was a significant difference in the SUVR values between the five subregions using one‐way repeated‐measures anova (P < 0.0001, Fig. [NaN] ). The SUVR values of 11C‐PiB decreased in the following order: VST > pre‐DPU > pre‐DCA > post‐PU > post‐CA. A post‐hoc paired t‐test with Bonferroni adjustment produced: VST > pre‐DPU (P = 0.004), VST > pre‐DCA (P < 0.0001), pre‐DPU > post‐PU (P < 0.0001), and pre‐DCA > post‐CA (P < 0.0001). There were no significant correlations between MMSE scores and the SUVR values in each of the five subregions: VST (r = −0.13, P = 0.58), pre‐DCA (r = −0.05, P = 0.85), pre‐DPU (r = −0.24, P = 0.32), post‐CA (r = 0.19, P = 0.43), and post‐PU (r =−0.23, P = 0.32).

Comparison of 11 C‐PiB binding between the patients with AD and healthy controls

1 *T‐values and **P‐values from independent t‐test between patient and control groups in each subregion.

The 20 transformed SUVR images of 11C‐PiB were averaged in MNI152 space and are displayed in Fig. [NaN] A, visually confirming that 11C‐PiB binding was greatest in the ventral part of the striatum. Exploratory voxel‐wise analyses revealed positive correlations between the SUVR values in the VST, which were obtained by VOI‐based analyses, and 11C‐PiB binding in the medial part of the orbitofrontal area at P < 0.01 FWE corrected (T > 6.38), and MNI coordinates of the peak‐level voxel were x = −2 mm, y = 40 mm and z = −18 mm (Fig. [NaN] B and C). There were no other cortical regions where 11C‐PiB binding was significantly correlated with the SUVR values, for each of the other four subregions at P < 0.01 FWE corrected. There were no significant correlations between the MMSE scores and 11C‐PiB binding in cortical regions at P < 0.01 FWE corrected.

Many figures in previous studies demonstrate that 11C‐PiB binding is relatively high in the anterior and ventral parts of the striatum in patients with AD (Klunk et al., [26] , [27] ; Nordberg, [40] ; Mintun et al., [38] ). To our knowledge, however, no authors have yet addressed the distribution pattern of 11C‐PiB binding in detail. This is an initial study investigating the regional distribution of striatal Aβ deposition levels, and we have confirmed that there are ventral–dorsal and anterior–posterior gradients of 11C‐PiB binding in patients with AD, and that the binding of 11C‐PiB is greater in the ventral and anterior parts than in the dorsal and posterior parts. We have also shown that 11C‐PiB binding in patients with AD was significantly greater in each of the five striatal subregions compared with control subjects, indicating that there can be Aβ deposition in every area of the striatum. The striatum is actually known to contain extensive fibrillar Aβ plaques in virtually all patients with AD (Braak & Braak, [5] ; Brilliant et al., [8] ).

The Aβ deposition may present as either diffuse or dense plaques (Duyckaerts et al., [15] ; Rowe & Villemagne, [47] ). Diffuse plaques are assumed to be an early phase of plaque formation. Dense plaques may be described as focal, spherical, or cored, and are called neuritic when they are associated with local neuronal damage and inflammation. Dense plaques are visible in sections stained with hematoxylin and eosin, compared with diffuse plaques, which are unstained and their apparent number depends on the quality of the immunohistochemistry or other staining techniques (Duyckaerts et al., [15] ). Dense plaques are characteristic of AD, whereas diffuse plaques have also been found in many normal subjects, contributing to the idea that they may not be directly toxic (Delaere et al., [11] ; Dickson et al., [13] ). Histopathological studies have revealed that there is a wide range of plaque morphology in the striatum, including diffuse, dense, and neutritic types (Rudelli et al., [48] ; Bugiani et al., [10] ; Suenaga et al., [53] ; Selden et al., [50] ; Brilliant et al., [8] ). Dense plaques have been reported more frequently in the ventral part of the striatum than in the dorsal part (Suenaga et al., [53] ). However, a PET amyloid tracer, 11C‐PiB, can detect both diffuse and dense plaques, although its affinity for dense plaques is much higher than that for diffuse plaques (Lockhart et al., [31] ; Ikonomovic et al., [19] ). The intensity of in vivo11C‐PiB binding presumably reflects the number of Aβ‐immunoreactive fibrils in plaques, from lightly labeled diffuse plaques to intensely fluorescent dense plaques. Our findings may imply that the number of dense plaques is greater in the ventral and anterior parts than in the dorsal and posterior parts. Future histopathological studies are required to determine the number and types of plaques that are present in each of the five striatal subregions.

The present study showed that 11C‐PiB binding in the VST is highest within the striatum, and strongly correlates with that in the medial part of the orbitofrontal area (Fig. [NaN] ). The anatomical location of the peak‐level voxel (MNI coordinate: x = −2 mm, y = 40 mm and z = −18 mm) is consistent with the location of the frontal medial cortex, according to the Harvard–Oxford cortical atlas (included in FSL). The frontal medial cortex corresponds to Brodmann areas 11 and 12 located in the medial part of the orbitofrontal area (Rademacher et al., [45] ). The VST consists of the nucleus accumbens and ventral parts of the caudate nucleus and putamen (Mawlawi et al., [36] ; Martinez et al., [34] ), and its associations with cognition, as well as behavior, are well known (Diekhof et al., [14] ; de Jong et al., [20] ; Li et al., [29] ). Anatomically, the VST receives both focal and diffuse projections from different prefrontal subregions, predominantly from the medial part of the prefrontal area (Ferry et al., [17] ; Ongur & Price, [41] ), and it is also functionally linked to the medial part of the orbitofrontal area (Di Martino et al., [12] ; Jung et al., [21] ). Meanwhile, there is a concept that brain Aβ deposition is not randomly distributed, but has characteristic patterns through specific anatomical connections and/or functional networks (Arnold et al., [2] ; Braak & Braak, [6] ; Seeley et al., [49] ; Raj et al., [46] ; Sepulcre et al., [51] ). One example is the region related to the default mode network that is preferentially vulnerable to Aβ deposition (Klunk et al., [26] ; Buckner et al., [9] ), and disruption of activity and metabolism (Lustig et al., [33] ; Greicius et al., [18] ). The robust correlation between the VST and medial part of the orbitofrontal area for Aβ deposition levels in this study supports the idea described above. Future histopathological studies are, however, required to confirm the phenomenon observed in the study.

The density of Aβ deposition in the cortical regions has not consistently been shown to correlate with the degree of cognitive impairment in post‐mortem studies (Parvathy et al., [42] ; Prohovnik et al., [44] ) or in vivo studies using 11C‐PiB and PET (Edison et al., [16] ; Pike et al., [43] ). This is probably because Aβ deposition in some cortical regions begins in the early phase of AD pathology, and almost plateaus by the onset of cognitive decline. Compared with the cortical regions, the striatum is involved in Aβ pathology in the later stages of AD (Braak & Braak, [6] ; Thal et al., [54] ; Kemppainen et al., [24] ; Mintun et al., [38] ; Beach et al., [4] ). The presence of striatal plaques has reportedly been associated with measures of memory impairment in patients with AD (Wolf et al., [58] ), and the occurrence of dementia in patients with Lewy body disease (Kalaitzakis et al., [22] , [23] ). In addition, the striatum is known to play an important role in cognition and behavior (Nakano et al., [39] ; Kalaitzakis et al., [22] ; Lee et al., [28] ). These findings collectively seem to indicate a relationship between cognitive decline and the density of Aβ deposition in the striatum. However, this has not previously been examined in detail. As we could not find a significant correlation between 11C‐PiB binding in each of the five striatal subregions and the MMSE scores, further detailed studies will be required to investigate the relationships between Aβ deposition levels in the striatal subregions and cognitive domains including memory.

The present study confirmed that the amount of Aβ deposition in the VST is highest within the striatum, and provides the first evidence that there is a positive correlation between the VST and the medial part of the orbitofrontal area for Aβ deposition levels. Because the VST is well known to have anatomical and functional links with the medial part of the orbitofrontal area, our findings indicate that the brain Aβ deposition is not randomly distributed, but has characteristic patterns related to anatomical and/or functional networks.

The authors thank Ms Hatsumi Endo, Ms Hiroko Tsukinari and Mr Kunpei Hayashi for their technical assistance.

Abbreviations

11 C‐PiB, 11 C‐Pittsburgh Compound B

AC, anterior commissure

AD, Alzheimer's disease

Aβ, amyloid‐β

FWE, family‐wise error

MMSE, Mini Mental State Examination

MNI, Montreal Neurological Institute

MRI, magnetic resonance imaging

PET, positron emission tomography

post‐CA, post‐commissural caudate

post‐PU, post‐commissural putamen

pre‐DCA, pre‐commissural dorsal caudate

pre‐DPU, pre‐commissural dorsal putamen

SUVR, standardised uptake value ratio

VOI, volume of interest

VST, ventral striatum

The authors declare no financial or other conflicts of interest.

Graph: An example of VOIs. VOIs placed on the VST (blue), pre‐DCA (yellow), pre‐DPU (red), post‐CA (green), post‐PU (light blue), and cerebellum (orange) are displayed in the axial (A and B), coronal (C) and sagittal (D) sections of a representative MRI image. One side of the VST, pre‐DCA, pre‐DPU, and post‐CA VOIs, and both sides of the cerebellum VOI are shown.

Graph: Comparison of 11 C‐PiB binding in the patients with AD and in healthy controls. Open circles and × symbols represent SURV values in the patients with AD and control subjects, respectively. Vertical bars represent mean ± SD.

Graph: 11 C‐PiB images in MNI space and the results of voxel‐wise analyses. (A1–3) Averaged SUVR image of 11 C‐PiB in 20 patients with AD is superimposed on an MNI standard brain and is displayed in coronal sections. (B1–3) Voxel‐wise analyses showed positive associations for 11 C‐PiB binding between the VST and medial part of the orbitofrontal area. Voxels with significant correlations ( P  <   0.01 FWE corrected) are overlaid on the MNI standard brain and are displayed with MNI coordinates of the peak‐level voxel ( x  = −2 mm, y  = 40 mm, and z  = −18 mm). (C1–3) The averaged SUVR image superimposed on an MNI standard brain is displayed for the corresponding MNI coordinate. The yellow–red and rainbow scales represent the magnitude of P ‐values (B) and SUVR values (A and C), respectively.