Regional tau deposition and subregion atrophy of medial temporal structures in early Alzheimer's disease: A combined positron emission tomography/magnetic resonance imaging study.

INTRODUCTION: Molecular imaging and selective hippocampal subfield atrophy are a focus of recent Alzheimer's disease (AD) research. Here, we investigated correlations between molecular imaging and hippocampal subfields in early AD.
METHODS: We investigated 18 patients with early AD and 18 healthy control subjects using 11C-Pittsburgh compound-B (PIB) positron emission tomography (PET) and 18F-THK5351 PET and automatic segmentation of hippocampal subfields with high-resolution T2-weighted magnetic resonance imaging. The PET images were normalized and underwent voxelwise regression analysis with each subregion volumes using SPM12.
RESULTS: As for 18F-THK5351 PET, the bilateral perirhinal cortex volumes were significantly associated with the ipsilateral or bilateral temporal lobar uptakes, whereas hippocampal subfields showed no correlations. 11C-PIB PET showed relatively broad negative correlation with the right cornu ammonis 3 volumes. DISCUSSION: Regional tau deposition was correlated with extrahippocampal subregional atrophy and not with hippocampal subfields, possibly reflecting different underlying mechanisms of atrophy in early AD. Amyloid might be associated with right cornu ammonis 3 atrophy.

Introduction

Alzheimer's disease (AD) is the most common type of neurodegenerative dementia. Abnormal accumulations of extracellular amyloid β and intracellular neurofibrillary tangles (NFTs) of tau proteins are hallmarks of AD. Recently, AD research has begun taking advantage of various emerging advanced neuroimaging methods, with tau positron emission tomography (PET) considered particularly promising for in vivo estimations of AD pathology [1]. A recent study revealed that tau PET uptake patterns strongly reflect regional associations with clinical and anatomic variability, whereas amyloid PET shows a more diffuse distribution and less regional associations with other parameters [2].

Hippocampal atrophy is a key structural imaging finding in AD [3]. The selectivity of hippocampal subfield atrophy has attracted attention for its diagnostic and predictive potential [4], [5], and specific cortices such as the entorhinal cortex (ERC) and perirhinal cortex (PRC) are also considered important for memory networks [6], [7].

Because tau deposition can potentially affect clinical and morphologic parameters in AD and medial temporal subregions may also have diagnostic and/or predictive value, investigation of the relationships between subregional atrophy and abnormal accumulations such as tau deposition is required. We conducted this study to explore imaging correlations using 11C-Pittsburgh compound-B (PIB) PET, 18F-THK5351 PET [8], and automatic segmentation of hippocampal subfields (ASHS) with high-resolution T2-weighted magnetic resonance imaging (MRI), which is a reliable method for subregional volumetry in various neurologic diseases [3], [9], [10].

Methods

Patients and control subjects

We recruited 18 Japanese patients (13 women, 5 men) with early AD at our institute. AD was diagnosed based on the clinical criteria for probable AD [11] and the presence of an abnormal cortical accumulation of amyloid revealed by the visual assessment of 11C-PIB PET images. The patients were 70.3 ± 8.5 years old (mean ± standard deviation [SD]), their average Mini–Mental State Examination (MMSE) score was 22.6 ± 4.1 (mean ± SD), and their global Clinical Dementia Rating ranged from 0.5 to 1.0. Almost all patients had less than 1 year disease duration except two patients with a few years' history from the diagnoses.

For reciprocal comparisons of abnormal depositions on PET and subregional atrophy, we also recruited 18 healthy Japanese control subjects (10 women, 8 men) with normal cognition who showed visually normal 11C-PIB and 18F-THK5351 PET results. The control subjects were 66.8 ± 9.5 years old, with an average MMSE score of 29.2 ± 1.0 and a global Clinical Dementia Rating of 0. There were no significant differences in the mean age or sex proportion between the early AD patient and healthy control groups. All clinical assessments and imaging scans were performed within a 12-week period.

All subjects gave written consent to participate in the study, which was approved by the Institutional Review Board at Japan's National Center of Neurology and Psychiatry.

Imaging acquisition

All participants underwent MRI scanning on a 3.0-T MRI system (Verio, Siemens, Erlangen, Germany). The sequence parameters were as follows. Three-dimensional sagittal T1-weighted magnetization prepared rapid acquisition with gradient echo images: repetition time/echo time, 1900 ms/2.52 ms; flip angle, 9°; in-plane resolution, 1.0 × 1.0 mm; 1.0-mm effective slice thickness with no gap; 300 slices; matrix, 256 × 256; field of view, 25 × 25 cm; acquisition time, 4 minutes 18 seconds.

High-resolution T2-weighted images were designed for hippocampal subfield segmentation and obtained as follows: repetition time/echo time, 7380 ms/76 ms; flip angle, 150°; in-plane resolution, 0.4 × 0.4 mm; 2-mm slice thickness with no gap; 30 slices; matrix of 512 × 432; 22 × 22 cm field of view; acquisition time, 6 minutes 33 seconds.

All the PET/computed tomography (CT) scans were performed on a combined PET/CT scanner (Biograph 16; Siemens). For 11C-PIB imaging, 11C-PIB at a dose of 555 MBq was injected intravenously 50 minutes before the PET/CT scan, and the emission scan duration was 20 minutes. For 18F-THK5351 imaging, 18F-THK5351 at a dose of 185 MBq was injected 40 minutes before the scan, and the scan duration was 20 minutes. PET/CT images were reconstructed using a combination of Fourier rebinning and ordered subset expectation maximization.

ASHS volumetry of hippocampal and mesiotemporal subfields

We input both the T1- and high-resolution T2-weighted images obtained from all subjects into an open-source ASHS software program (https://sites.google.com/site/hipposubfields/) [9]. The “UPenn PMC Atlas” [9] was selected as the atlas set. The software calculated the volumes of each subfield fully automatically with a combination of multiatlas label fusion and learning-based error correction. The following 10 regions of interest were delineated: cornu ammonis (CA) 1, CA2, CA3, dentate gyrus (DG), subiculum, ERC, Brodmann area (BA) 35, BA36, collateral sulcus, and miscellaneous parts. Experienced neuroradiologists visually confirmed that the parcellation quality was good or fair.

PET normalization

After partial volume correction by PETPVE12 toolbox [12], both the 11C-PIB and 18F-THK5351 PET images were normalized using the statistical parametric mapping software 12 program (SPM12; http://www.fil.ion.ucl.ac.uk/spm). The subjects' T1-weighted images were coregistered to their PET images and normalized with diffeomorphic anatomic registration using the exponentiated lie method [13]. A transformation matrix was applied to each PET image, which had been coregistered to T1-weighted image through the partial volume correction process.

After spatial normalization, all the PET images were divided by the individual's positive mean uptake value of cerebellar gray matter. Finally, each PET image was smoothed by an 8-mm full width at half maximum Gaussian kernel. This value (8 mm) was determined by doubling the full width at half maximum value of the PET scanner (4 mm).

Statistical analyses

For the correlation analyses between subregions and tau/amyloid depositions, we used the data from only the early AD patients because we designed this study to investigate the pathophysiology of AD. We determined the correlation between each subregional volume and PET images with the SPM12 voxelwise regression analyses.

We evaluated correlations using the “multiple regression” design with age, sex, and the intracranial volume calculated by ASHS as nuisance covariates and each subregional volume as the main covariate. For the reciprocal PET comparison between the AD patients and healthy control subjects, we also used SPM12 and a “two-sample t-test” design with age and sex as covariates. Correlations or differences that met the following criteria were deemed significant: a height threshold of P < .001 (uncorrected) and an extent threshold of P < .05 (familywise error).

We also compared the subregional volume changes in the two groups using an analysis of covariance model with age, sex, and intracranial volume as covariates. Statistical Package for Social Science software (version 23.0; Japan, Tokyo) was used and P < .05 was considered significant. To better understand atrophic selectivity in different regions, we also calculated the z-scores of the early AD patients in each subregion using the mean and SD values of the control subjects.

Results

The calculated volume of each subregion is shown in Table 1. Patients with early AD showed significant atrophy in parts of the hippocampal subfields and in the extrahippocampal cortices. According to the z-scores, in the hippocampus, CA1 and DG showed severe atrophy, whereas CA2 and CA3 were relatively preserved.

Table 1

Volumes of the subregions in both the early AD patients and the healthy control subjects calculated by ASHS and the z-scores of the AD patients in each subregion

Left subregion	Volume (mm3)		P value	z-score of AD	Right subregion	Volume (mm3)		P value	z-score of AD
	AD	Control				AD	Control
CA1	1002.3 ± 178.5	1338.4 ± 146.7	<.001	−2.29 ± 1.22	CA1	1007.2 ± 183.7	1397.5 ± 165.8	<.001	−2.35 ± 1.11
CA2	12.5 ± 6.7	18.7 ± 7.3	.021	−0.86 ± 0.92	CA2	19.9 ± 4.9	23.9 ± 6.7	.063	−0.61 ± 0.73
CA3	71.1 ± 17.2	67.5 ± 18.0	.435	0.20 ± 0.96	CA3	73.6 ± 14.1	70.2 ± 17.6	.825	0.20 ± 0.81
DG	606.4 ± 116.3	817.3 ± 113.7	<.001	−1.85 ± 1.02	DG	608.7 ± 95.0	851.7 ± 107.0	<.001	−2.27 ± 0.89
SUB	304.5 ± 67.3	391.2 ± 63.3	.002	−1.37 ± 1.06	SUB	298.3 ± 63.3	407.7 ± 74.3	<.001	−1.47 ± 0.85
ERC	426.2 ± 97.4	504.9 ± 48.4	.024	−1.62 ± 2.01	ERC	390.7 ± 111.3	510.1 ± 62.3	.002	−1.92 ± 1.79
BA35	340.7 ± 65.4	429.5 ± 97.9	.015	−0.91 ± 0.67	BA35	311.7 ± 112.5	432.1 ± 93.0	.006	−1.29 ± 1.21
BA36	1315.8 ± 282.3	1595.5 ± 322.6	.046	−0.87 ± 0.87	BA36	1196.5 ± 402.8	1560.4 ± 329.5	.014	−1.10 ± 1.22
CS	314.8 ± 105.2	299.5 ± 88.1	.318	0.17 ± 1.19	CS	194.7 ± 86.4	216.5 ± 101.3	.969	−0.22 ± 0.85
MISC	169.4 ± 35.0	138.1 ± 51.3	.034	0.61 ± 0.68	MISC	176.0 ± 40.0	146.2 ± 49.7	.043	0.60 ± 0.80

Abbreviations: AD, Alzheimer's disease; ASHS, automatic segmentation of hippocampal subfields; BA, Brodmann area; CA, cornu ammonis; CS, collateral sulcus; DG, dentate gyrus; ERC, entorhinal cortex; MISC, miscellaneous parts; SD, standard deviation; SUB, subiculum.

NOTE. Data are expressed as the means ± SD.

The early AD patients showed significantly increased uptakes on both 11C-PIB and 18F-THK5351 PET (Fig. 1). Although amyloid accumulation was diffusely distributed, it was particularly intense in the posterior cingulate and precuneus areas, as well as the lateral temporal lobes and frontal poles. The tau depositions were also widespread, but they were intense in the inferior lateral temporal lobes.

Fig. 1

Significant increases in both (A) 11C-PIB (left) and (B) 18F-THK5351 PET (right) uptakes were identified in the early AD group. Abbreviations: AD, Alzhemier's disease; PET, positron emission tomography.

We also observed significant negative correlations between some of the subregional volumes and the 18F-THK5351 PET results (Table 2, Fig. 2). The right BA35 and BA36 volumes were significantly associated with the ipsilateral temporal lobar 18F-THK5351 uptakes, whereas the left BA36 showed significant negative correlations with 18F-THK5351 in the bilateral temporal lobes. There were no significances in the hippocampal subfields. On the other hand, the 11C-PIB PET showed relatively broad correlations with the right CA3 volumes (Table 2, Fig. 2).

Table 2

Significant negative correlations between each subregional volume and 18F-THK5351/11C-PIB PET

Subregion	Cluster size (familywise error P value)	T value	x	y	z	Correlative region		BA
18F-THK5351 PET
Right BA35	4886 (.010)	5.19	40	−30	−15	Rt	FG, PHG	20, 36
		5.16	39	−43	−10		FG	37
		4.69	51	−36	−2		MTG	22
Right BA36	8616 (<.001)	6.55	51	−29	−20	Rt	FG, ITG	20
		6.55	38	−34	−20		PHG, FG	20, 36
		5.34	48	−36	−11		FG	37
Left BA36	6726 (.002)	7.19	−46	−46	−22	Lt	FG, ITG	20, 36, 37
		5.77	−38	−32	−14		PHG, FG	20, 36
		5.33	−41	−56	−12		FG	37
	3358 (.031)	6.74	56	−54	−18	Rt	FG	37
		5.00	40	−53	−9		FG	37
		4.89	38	−55	−10		FG	37
11C-PIB PET
Right CA3	11,657 (.007)	5.28	45	−54	17	Rt	STG	39
		5.22	45	−30	0		MTG, STG	21, 22
		4.91	49	−79	0		MOG, ITG	18
	8890 (.016)	4.68	27	18	−17	Rt	IFG	47
		4.61	14	27	−24		OFC, IFG, RG	11, 47
		4.59	12	5	−11		Subcallosal	34
	7999 (.022)	5.38	−27	−8	−34	Lt	Uncus	20, 36
		4.99	−50	−44	−21		FG	20, 36
		4.66	−60	−19	−17		ITG	20

Abbreviations: BA, Brodmann area; CA, cornu ammonis; FG, fusiform gyrus; IFG, inferior frontal gyrus; ITG, inferior temporal gyrus; Lt, left; MOG, middle occipital gyrus; MTG, middle temporal gyrus; OFC, orbitofrontal cortex; PET, positron emission tomography; PHG, parahippocampal gyrus; RG, rectal gyrus; Rt, right; STG, superior temporal gyrus.

NOTE. The coordinates are shown on the Talairach atlas. All the unlisted correlations were insignificant.

Fig. 2

There were several significant negative correlations between each subregional volume and 18F-THK5351/11C-PIB PET in the early AD group. Abbreviations: AD, Alzhemier's disease; CA, cornu ammonis; BA, Brodmann area; PET, positron emission tomography.

Discussion

Our present findings determined regional tau deposition patterns associated with hippocampal and parahippocampal subregional atrophy and identified significant correlations with tau deposition in extrahippocampal subregional atrophy rather than hippocampal subfields. Judging from the z-scores, CA1 or DG as hippocampal subfields showed more severe atrophy extrahippocampal subregions. Given that extrahippocampal atrophy nevertheless had a greater association with tau deposition, our findings may reflect different mechanisms that are distinct from merely atrophy progression.

According to the hallmark AD pathologic theory [14], [15], NFTs, which consist of intraneuronal aggregates of tau, first target the ERC and PRC, damaging the perforant pathway to the hippocampus. Subsequently, the NFTs affect the hippocampus, with the CA1 and subiculum targeted before the CA2 or CA3. In ASHS, BA35 and BA36 account for the PRC [9], which is a part of the memory system [7]. The atrophy of bilateral PRC was less severe than that of the CA1, DG, and ERC (Table 1), but was significantly correlated with broad tau deposition, which is consistent with the pathologic theory, suggesting that these areas are directly related to NFTs. However, hippocampal subfields except the right CA1 may have different and complicated underlying mechanisms of atrophy in AD.

A previous pathologic study revealed significant associations among NFTs, neuronal loss, and cognition, and NFTs in the parahippocampal area were considered the best predictor of the MMSE score [16], although controversy persists regarding whether NFTs are causative or due to a protective process [17]. Because increased tau in cerebrospinal fluid reflects not regional tau deposition, but neuronal damage and disease progression, tau PET is expected to be useful for estimations of regional tau deposition in AD [1]. Thus, the correlations between tau PET and various clinical anatomic parameters should be a fruitful area of investigation. The significant correlations of atrophy levels in several subregions and the locations of relevant tau depositions revealed in the present study contribute to the understanding of AD.

We also found a significant correlation between the right CA3 atrophy and broad amyloid depositions. The right CA3 neuronal loss is considered to have an association with amyloid β and NMDA receptors' functions [18], [19], which might be the cause of our results. However, there still remains a question about this result, given that the ASHS showed the lowest performance for CA2/3 segmentation [9] and our patients showed no significant CA3 atrophy as a group comparison (Table 1).

According to pathologic studies, amyloid plaques accumulate mainly in neocortices and are less associated with AD progression and cognitive impairment [14], [16], and a recent tau and amyloid PET imaging study confirmed such a tendency [2]. We also found no significant correlations between amyloid deposition and any subregional atrophy. The pathologic process of AD precedes the clinical diagnosis by several years [20], and amyloid deposition may also reach a plateau at diagnosis following such a preclinical stage [21]. On the other hand, our use of 11C-PIB PET for the diagnosis of AD may possibly have skewed the results about amyloid.

Hippocampal subfields are receiving widespread attention in the neuroimaging of AD [4]. The ASHS software shows reliable concordance with manual segmentation even in the parahippocampal areas [9]. Most studies suggested that CA1 in the hippocampus shows the most severe neuronal loss and strongest associations with cognition and the pathologic stage [22], [23], whereas some studies reported significant subiculum and ERC atrophy in AD [23], [24]. Our ASHS results also showed volume reductions in these areas, which would be consistent with previous studies. However, the correlations with regional tau deposition varied (Table 2, Fig. 2), suggesting different degenerative mechanisms in each subregion.

The main limitation of this study is the relatively small sample size (18 AD patients and 18 control subjects) and the use of uncorrected P values for the height threshold and the lack of correction of accumulated alpha errors because of multiple comparisons, which could make our results exploratory or preliminary. We speculate that the extent threshold with familywise error correction would overcome this shortcoming, and our findings may thus provide additional new knowledge on AD. Moreover, because of potential beta errors, we cannot conclude that there are absolutely no correlations in our insignificant analyses. The lack of genetic data (e.g., APOE ε4 gene) is another limitation, given that such parameters may affect subfield atrophy patterns [24]. In addition, a most recent report has suggested that 18F-THK5351 PET might bind monoamine oxidase B [25]. We used the cerebellar cortex as the reference, which is considered the least-affected region by this problem [25], and no participant took monoamine oxidase B inhibitors in the present study. However, this issue should be addressed and resolved by future studies.

Conclusions

Regional tau deposition was correlated with PRC atrophy rather than hippocampal subfields, suggesting different underlying mechanisms of atrophy in early AD. Amyloid deposition showed a correlation with right CA3 atrophy.

Systematic review: We searched Medline and PubMed databases for studies on in vivo tau positron emission tomography, hippocampal subfields, and relevant pathologic findings in Alzheimer's disease. Relevant studies were additionally found in the reference lists of articles or citation lists on PubMed.

Interpretation: We demonstrated that regional tau deposition patterns were associated with extrahippocampal subregional atrophy and not with hippocampal subfields, suggesting different underlying mechanisms of atrophy in early AD. Amyloid positron emission tomography showed a broad correlation only with the right CA3 subfield.

Future directions: Further studies including larger and multistage cohorts with the analysis of genetic data would be helpful to determine these correlations and mechanisms more precisely.