Trajectories of the Hippocampal Subfields Atrophy in the Alzheimer’s Disease: A Structural Imaging Study

1The pathophysiological process of Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by cognitive decline, which is thought to have begun many years before the diagnosis. With the disease progression, as the preclinical AD, subjective cognitive decline (SCD) have worse cognition than normal controls (NC), while objective examination shows that they have not yet reached the level of amnestic mild cognitive impairment (aMCI) or AD dementia. The main manifestation of SCD is the decline in memory rather than other domains of cognition. It is formally proposed and standardized by Subjective Cognitive Decline Initiative (SCD-I) in a conceptual framework for research on subjective cognitive decline. After adjustment for age, sex and education, the stage of neuropsychological examination below threshold was mild cognitive impairment (MCI) or prodromal AD. Subsequently, if there are significant interferences in the ability of work or daily activities, cognitive decline progresses onward to the stage of AD dementia. These clinical symptoms are caused by the accumulation of pathology leading to the macrostructural disorder of the brain, of which the hippocampus atrophy is the most obvious.

2The hippocampus is composed of several subfields with different histological characteristics, rather than a homogeneous structure. Hippocampal atrophy is the most significant structural biomarker of AD imaging. Differential changes in hippocampal atrophy can be relatively easily obtained from magnetic resonance imaging (MRI). The hippocampus and hippocampal subfields are found to be diversely affected in Alzheimer’s Disease (AD) and early stages of Alzheimer’s disease by neuroimaging studies. The hippocampal atrophy of AD patients was most significantly involved subiculum and CA1 subfields. Other studies have showed that there were more extensive and more evident atrophies in DG/CA3 or subiculum at the lower end of the hippocampus. Studies on prodromal AD showed that the focal atrophy of CA1-2 of MCI patients is more obvious than that of normal aging patients. The atrophy first appeared in the presubiculum and subiculum of the hippocampus at MCI. However, SCD subjects are more difficult to identify from the NC because the SCD group showed that the left total hippocampal volume was small with statistically significant difference, while the right total hippocampal volume did not change significantly. The atrophy of hippocampal surface is mainly in CA1, and the other regions have obvious overlap with AD. The atrophy of the memory-related hippocampus and hippocampal subfields is one of the earliest macroscopic features of the trajectories of Alzheimer’s disease, and has been reported in autopsies and neuroimaging studies. To our best knowledge, there is little research on the subfield of hippocampus and relationship with memory in SCD.

3We hypothesized that there may be 1) a change in the hippocampal subfields at different stages of AD in accordance to the trajectory of Alzheimer’s disease and 2) a relationship between hippocampal subfield volume and memory status. The purpose of this study was to identify which subfields of the hippocampus differ in the trajectories of Alzheimer’s disease by magnetic resonance imaging (MRI). In addition, to determine whether individual differences on memory could be explained by structural volumes of hippocampal subfields.

4We prepared 161 right-handed Chinese Han participants including 35 SCD patients, 43 aMCI patients and 41 AD patients, and 42 NC subjects from our databank (NCT: 02225964, 02353845, 02353884, and 03370744). The cognitive functions of all the subjects were assessed by experienced neurologists. Including the Clinical Dementia Rating Scale (CDR), the Chinese version of the Mini-Mental State Examination (MMSE), the Beijing version of Montreal Cognitive Assessment (MoCA), the auditory verbal learning test (AVLT), an activities of daily living (ADL) assessment, and Hamilton depression rating scale.

5The normal controls did not present cognitive decline complaints and their performance in MMSE, MoCA and AVLT were in normal range. The patients with SCD were diagnosed based on the criteria proposed by SCD-I in 2014, including (1) self-reported experience of persistent decline in memory compared to a previous state (within the last 5 years); (2) performance within the normal range on MMSE or MoCA (adjusted for age, sex, and education); (3) the Clinical Dementia Rating (CDR) score is 0. The patients were diagnosed with aMCI using the Petersen criteria, which have been described in our previous studies: (a) presence of memory complaint, confirmed by an informant; (b) presence of objective memory impairment measured by MMSE, MoCA and AVLT; (c) failure reach the standard of dementia; (d) CDR score of 0.5. The inclusion criteria for SCD were based on the recent research criteria proposed by National Institute of Aging-Alzheimer’s Association (NIA-AA) criteria for clinically probable AD: (a) meeting the criteria for dementia; (b) recessive and gradual onset for more than 6 months, not a sudden attack; (c) hippocampal atrophy confirmed by structural MRI; (d) CDR score is equal or greater than 1. Exclusion criteria were prior history of the activities of daily living disorder, stroke, mental disorders, cancer, drug abuse, epilepsy, brain tumors, Parkinson’s disease, encephalitis and hypoxic brain damage. All subjects underwent brain MRI examination. The detailed demographic and clinical characteristics of participants are shown in Table 1.

6The study approved by the medical research ethics committee and the institutional review board of Xuanwu Hospital, Capital Medical University, Beijing, China. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

7The 3T magnetic resonance imaging system (MAGNETOM Trio Tim; Siemens, Erlangen, Germany) was used for image acquisition at the Department of Radiology, XuanWu Hospital, Capital Medical University. T1-weighted MRI scans were acquired at the sagittal plane by using a magnetization prepared rapid acquisition gradient echo sequence with the following parameters: TR = 1900 ms, TE = 2.2 ms, FA = 9°, inversion time (TI) = 900 ms, matrix = 256 × 256, slices = 176, thickness = 1.0 mm and Voxel size = 1 × 1 × 1 mm3.

8Structural MR Images were processed by the FreeSurfer image analysis suite, which can be downloaded free of charge from the website (version 6.0.0, http://freesurfer.net/).

9First, the entire hippocampal formation was segmented using the routine volumetric FreeSurfer pipeline. Briefly, T1-weighted MR images were corrected for within-subject head motion; then, non-brain tissues were removed using a hybrid watershed/surface deformation algorithm. The resulting images were further affine registered to the Talairach space. Subsequently, segmentation of the subcortical and cortical structures (including the hippocampus) was conducted using a probabilistic brain atlas. The estimated total intracranial volume (eTIV) of each subject was also calculated using the standard FreeSurfer processing pipeline by exploiting the relationship between the intracranial volume and the linear transformation to the atlas template. The eTIV was used to correct for individual differences in head size in the subsequent statistical analysis. Automated segmentation of hippocampal subfields was performed using a built-in module of FreeSurfer, in which a Bayesian statistical model with Markov random field priors was used to estimate the label of each subfield. This method has been successfully applied to detect hippocampal abnormalities in specific subfields in many neuropsychiatric diseases. A bounding box containing the hippocampus that was upsampled to a 0.5 mm isotropic resolution was applied to this module. This approach relied on a tetrahedral mesh-based probabilistic atlas of the hippocampal formation, which was constructed from the manual delineation of the right hippocampus based on ultra-high-resolution T1-weighted scans (0.38 × 0.38 × 0.8 mm3) of 10 normal subjects. By maximizing the posterior probability of a segmentation, the left and right hippocampi were automatically segmented into twelve subfields: hippocampal tail, parasubiculum, presubiculum, subiculum, CA1, CA3, CA4, hippocampus–amygdala transition area (HATA), granule cell layer of dentate gyrus (GC-DG), molecular layer, fimbria, and hippocampal fissure. In this manuscript, the method for automated segmentation is standard. Additionally, the method for segmentation is validated to be accurate by. The hippocampal subfield segmentation results are illustrated in Figure 1. The entire hippocampal volume was defined as the sum of the volume of all hippocampal subfields.

10Statistical analysis was carried out using Statistical Package for Social Sciences software (SPSS, version 21.0). All the statistical tests were two-tailed. Categorization of demographic variables was assessed using Chi-square test. Continuous demographic variables were evaluated through ANOVA. In this study, the estimated total intracranial volume (eTIV) was used as a covariate to control head size. Statistically significant differences based on ANOVA (P < 0.05) were further explored using Bonferroni post hoc analysis. In the post hoc analysis, the differences between the individual experimental group and the control group were assessed. The left and right hemisphere measurements were analyzed, respectively. In addition, covariance analysis was used to analyze the volume differences in individual hippocampal subfield with age, sex, years of education and eTIV as covariates. Furthermore, we investigated relationships between hippocampal subfield volumes and memory test variables (AVLT-immediate recall, AVLT-delayed recall, AVLT-recognition) through the regression model analyses controlled for age, gender, education and eTIV.

11The demographic characteristics of the normal control, the patients of SCD, the patients of aMCI and the patients of AD are shown in Table 1. Four groups of age, sex, and educational level were well-matched (P > 0.05 for each group comparison). Comparing SCD and NC groups, there were no significant differences in MoCA, MMSE, immediate recall part of AVLT, the recognition part of AVLT, while significant difference (P = 0.012) in the delayed recall part of AVLT. The patients with AD and aMCI had significant lower scores in MoCA, MMSE, and AVLT compared with the healthy control participants (P < 0.005).

12We tested differences in whole hippocampal volume and all subfields among four groups using ANCOVA with age, years of education, and eTIV as covariates. Table 2 shows the statistical results of hippocampal subfields and hippocampal volumes. The volume of the left whole hippocampus was significantly different between NC, SCD, aMCI and AD in Figure 2. However, there was no statistically significant difference in the right whole hippocampus between NC and SCD. Compared with NC, aMCI group and AD group showed significant decreases in right whole hippocampal volume in Figure 3. In addition, the significant decreases were found for SCD and NC in the volume of hippocampal tail, subiculum, presubiculum, molecular layer HP, GC-ML-DG and CA4 of left hippocampal subfields, right presubiculum and fimbria of right hippocampal subfields. Most of the hippocampal subfields showed significant volumetric difference except hippocampal fissure and left parasubiculum between aMCI and NC groups. The significant differences in the hippocampal volume were detected between the AD and NC except right hippocampal-fissure. Furthermore, in our study, CA1, subiculum, presubiculum, molecular layer and fimbria showed the trend toward significant volume reduction among four groups with the trajectories of Alzheimer’s disease.

13In a first step, all potential risk factors (age, education years, sex, GM volume of hippocampal subfields, TIV) were correlated with AVLT scores and only variables correlated with AVLT score at P < 0.2 were used in subsequent stepwise linear regressions. This was performed to avoid too many independent variables. In the regression model, variables were removed when P > 0.05. Table 3 presents the results of the linear regression analyses. In our study, volume of left subiculum of all the four groups was most strongly and actively correlated with performance of AVLT three measures.

14In this study, we investigated the volumetric difference of hippocampus and hippocampal subregions among AD, aMCI, SCD, and NC subjects. There were also trends in some hippocampal subregions with the trajectories of Alzheimer’s disease in addition to the volumetric differences between the four groups. Furthermore, we studied AVLT and typical hippocampal subfields related with memory. It also shown trends with the trajectories of Alzheimer’s disease.

15In our study, we found that the differences of hippocampus and hippocampal subfields with age, years of education, and eTIV as covariates. The effect of the size of the brain in different subjects was excluded. Our study showed that the difference in volumes was in the left whole hippocampus as that of previous studies. We further divided the volume of the hippocampus, and the volumetric subfields of SCD, aMCI and AD were compared with the volumetric subfields of the NC. The hippocampal subfields volume of AD had significant differences except for right hippocampal fissure. There were also volumetric differences of aMCI in hippocampal tail, subiculum, presubiculum, molecular layer HP, GC-ML-DG, CA4, CA3, fimbria, HATA and right parasubiculum. These were consistent with previous studies. Previous studies had shown that the volume of the whole hippocampus and hippocampal subfields of SCD and NC were not consistent. But our research found that the volumes of SCD were different from those of NC in left whole hippocampus hippocampal tail, subiculum, presubiculum, molecular layer HP, GC-ML-DG and CA4 of left hippocampal subregions, right presubiculum and right fimbria. Of note, we observed the trend in the CA1, subiculum, presubiculum, molecular layer and fimbria subregions, which were in line with the previous studies, but their studies rarely involved the trajectories of Alzheimer’s disease. The obvious atrophic structures in AD are located at CA1, subiculum and the presubiculum. The atrophy of CA1 in MCI has also been reported, which is related to the increased risk of conversion from MCI to AD. In our study, we found that the hippocampus-related subfields had changed as early as SCD stages, however, not all of them showed trend changes. Trend-changing parts are rich in fibers and synapses, which also provide intrahippocampal connections and receive inputs from the hypothalamic lobe and thalamic nucleus. This is strongly correlated with memory impairment in AD patients. Our finding about the hippocampal volume reduction are consistent with neuropathological findings in the progression of AD disease. In our study, the atrophies of CA1, subiculum, presubiculum, molecular layer and fimbria subregions among SCD, aMCI and AD groups suggest that they may be a potential early biomarker for detecting AD at the SCD stage. These results similarly suggest that, compared with normal control subjects, the difference in the volumes of hippocampal subfields and the trend of these changes could show the evolution of AD in the earlier stage.

16The functions of the hippocampal subfield were different, which were related to memory, executive function, attention deficits and so on. The analysis of subfield volumes has been applied to memory neuroscience suggesting that subregion such as CA1, CA3 and dentate gyrus in memory is important. In our study, the scores of delayed recalls of AVLT were more closely related to the changes of hippocampal subfields than the score of immediate memory and recognition. As we all knew delayed recalls reflect the episodic memory which was impaired first in AD. Furthermore, the scores of delayed recalls of AVLT were better correlated with left subiculum. It implied that left subiculum might tell diseases earlier as an imaging biomarker.

17There are limitations in our study. Firstly, the main limitation is the lack of high risk group but asymptomatic control group besides the four groups (AD, aMCI, SCD and NC). In future design, we will collect the high risk but asymptomatic control group. Furthermore, this study was based on cross-sectional data, longitudinal follow-up studies of the same cohort are conducted to identify early imaging markers for disease transformation and prediction. Finally, we only studied hippocampal subregion volume by structural MRI. The combination of the multimodal imaging (i.e., structural, functional MR imaging and positron emission tomography technique) could be used in our future research.

18Our findings show that the trend changes in the hippocampus subfield and further illustrate that SCD is the preclinical stage of AD earlier than aMCI. The susceptibility of hippocampal subfield to AD pathological damage is different, so the volume of hippocampal subfield is better than the total volume of hippocampus in identifying early AD. It can better review the trajectory of AD, understand the mechanism, and identify sensitive biological indicators at different stages of AD.

19All datasets generated for this study are included in the manuscript and/or the supplementary files.

20WZ contributed to the main body of the manuscript. YH and SL conceptualized, designed, supervised the study and approved the final version of the manuscript. CY and MH contributed to study design and data collection. XW contributed to data analysis and manuscript preparation. All the authors interpreted the data, provided important feedback, and revised the manuscript.

21The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.