Deviations in Hippocampal Subregion Associated With Cognitive and Physical Frailty in Older Adults With Cognitive Frailty

Background The hippocampus is a complex, heterogeneous structure that is composed of widely different and interacting subregions. Atrophy of these subregions has been implicated in a variety of neurodegenerative diseases. Cognitive frailty is a kind of neurodegenerative disease with unclear neuropathological changes. The aim of this study was to explore the changes in hippocampal subregions in older adults with cognitive frailty and the relationship between subregions and cognitive decline as well as physical frailty. Methods Twenty-six older adults with cognitive frailty and 26 matched healthy controls were included in this study. Cognitive function was evaluated by the Montreal Cognitive Assessment (MoCA) scale (Fuzhou version) and Wechsler's Memory Scale, while physical frailty was tested with the Chinese version of the Edmonton Frailty Scale (EFS) and grip strength. The volume of the hippocampal subregions was measured with structural brain magnetic resonance imaging. Partial correlation analysis was carried out between the volumes of hippocampal subregions and MoCA scores, Wechsler’s Memory Quotient and physical frailty indexes. subregions of older adults with cognitive frailty, and these changes are correlated with cognitive decline and physical frailty. Therefore, the atrophy of hippocampal subregions could participate in the pathological progression of cognitive frailty.


Introduction
Cognitive frailty (CF) is a major subtype of frailty. According to the International Academy on Nutrition and Aging (I.A.N.A.) and the International Association of Gerontology and Geriatrics (I.A.G.G.), in 2013, CF was rst de ned as a clinical syndrome characterized by physical frailty and cognitive impairment among older adults, excluding Alzheimer's disease and other dementias [1]. Epidemiological surveys estimated the prevalence rate of CF to be 3-9.8% in the general older adult population, whereas the gure was much higher, 10.7-40%, in the clinical setting [2][3]. CF can accelerate cognitive impairment and physical frailty in older adults, is associated with a decline in activities of daily living and quality of life, and increases the risk of dementia, falls, disability and death [4]. As human life expectancy continues to increase, the prevalence of cognitive frailty is rapidly increasing and has become one of the biggest health threats in the 21st century [5]. Therefore, it is important for early intervention to nd an effective brain biomarker that can enable early identi cation of older adults with CF.
Previous studies found that the hippocampus is the core brain area related to cognition, but it is not a uni ed brain region; rather it is composed of several subregions with speci c histological features, and distinct pathways affect its overall function [6][7]. Therefore, the function of the hippocampus depends on its own internal structures, such as hippocampal sub elds, and the connections of its surrounding structures with other parts of the brain [8]. For example, a recent work showed that the anterior hippocampus contributed to global memory, perception, imagination and recall of scenes and events [9].
The posterior hippocampus was found to support ne and perceptual detailed memory [10]. The head of the hippocampus is related to logical memory, while the body and tail of the hippocampus are related to visual memory [11]. The cornu Ammonis sub eld 1 (CA1) of the hippocampus is connected to the posterior cingulate cortex, which can regulate episodic memory [12]. Therefore, the distinct hippocampal subregions are related to different types of cognition. However, its mechanism in CF remains unclear.
Previous research has shown that physical decline in the process of aging may be at least partly due to damage to the brain or nerve function, not just disorders of skeletal muscle, and the hippocampus might be involved in the regulation of human body functions [13,14]. It is well known that the hippocampus is devoted to balance regulation and sensory motor integration [14], while the hippocampus body and anterior cingulate gyrus are involved in memory and executive function [15]. Moreover, a reduced integrity of the gray matter in these two regions was positively correlated with greater stride variability in elderly adults, which indicates the role of cognitive function in motor control [16]. A positive correlation between the left hippocampus volume, especially of the left CA1, CA2 and subiculum, among elderly adults and the balance composite score was also observed [17]. Elderly adults with a strong sense of fatigue have a smaller hippocampal volume than normal elderly adults [18]. Therefore, the hippocampus or its subregions might contribute to the process of both cognitive and physical decline. We speculated that the hippocampus or its subregions play an important role in the pathogenesis of CF. To address this hypothesis, we performed high-resolution structural MRI scans in a group of older adults with CF as well as in controls. We used volumetry analysis to assess different aspects of the hippocampus and hippocampal subregions. The above indicators as well as behavioral indicators were subjected to correlation analysis to clarify the relationship. The whole brain index evaluation included analysis of cortical thickness, the number of white matter bers in the whole brain and the proportion of brain parenchyma.

Participants
This cross-sectional study recruited 26 older adults with CF and 26 matched healthy controls between April 2019 and September 2019 from communities in Fuzhou City, Fujian Province, China. All 52 participants included in this study participated in complete assessment scale and neuroimaging data collection. This study was approved by the ethics committee of the Second People's Hospital A liated with Fujian University of Traditional Chinese Medicine. Written informed consent was obtained from all participants before participation.
All CF participants met the following inclusion criteria: Chinese version of Edmonton Frailty Scale (EFS) score ≥ 5 points; Fuzhou version of the MoCA score ≤ 26 scores; Clinical Dementia Rating (CDR) Scale score = 0.5 (i.e., just mild cognitive impairment) and age ≥ 60 years. The inclusion criteria for the age-and education-matched controls were an EFS score < 5; a Fuzhou version of the MoCA score > 26; and a CDR Scale score = 0.
Individuals were excluded when they met one of the following conditions: history of mental illness (such as personality disorder, schizophrenia, etc.); serious depression (Becker depression scale score > 10); mild dementia and above (CDR Scale score > 0.5); history of alcohol or drug abuse; use of drugs that in uence cognitive function; serious organ failure, cerebral hemorrhage, sequelae of cerebral infarction; unsuitability for MRI scanning (such as xed metal dentures, pacemakers, etc.); and participation in another clinical trial.
Cognitive and physical frailty assessment Global cognitive ability and memory were evaluated by using the Fuzhou version of the MoCA and Wechsler's Memory Scale. MoCA scores range from 0-30, and a higher score indicates better cognitive function; scores lower than 26 points are considered to represent mild cognitive impairment [19]. Wechsler's Memory Quotient ranges from 51 to 150, with higher scores indicating better memory.
Physical frailty was assessed through the Chinese version of the Edmonton Frailty Scale (EFS) and the grip strength test.

MRI data acquisition
All participants underwent T1 and DTI imaging on a Siemens Prisma 3.0 T magnetic resonance scanner

Imaging Processing
All T1-weighted images were processed by publicly available FreeSurfer software (Version 6.0.0, http://www.freesurfer.net/) using the default settings. Before data preprocessing, image format conversion and image quality assessment were needed. Then, the command "recon-all" in FreeSurfer 6.0.0 was used for volumetric segmentation, speci cally including Talairach transformation, intensity normalization, skull stripping, volumetric registration, segmentation of gray and white matter and separation of the boundary, automatic subcortical segmentation, topology adjustment to ll and cut, and nally smoothing. Subcortical structures were segmented with a nonlinear warping atlas. Subsequently, a probabilistic atlas and a modi ed version of Van Leemput's algorithm were applied to segment the hippocampus [20,21] into 12 sub elds in each hemisphere: hippocampal tail, subiculum, CA1, hippocampal ssure, presubiculum, parasubiculum, molecular layer of the HP, granule cell layer and molecular layer of the dentate gyrus (GC-ML-DG), CA2/3, CA4, mbria and hippocampal amygdala transition area (HATA) (shown in Fig. 1). CA2 and CA3 were combined due to a lack of clear contrast, and the alveus volume was removed on account of the thin shape and unreliable segmentation. To reduce the effect of individual differences, the total intracranial volume (TIV), including the brain parenchyma and cerebrospinal uid (CSF), was estimated as a covariate.
Cortical thickness analysis was performed using the Computational Anatomy Toolbox (CAT12, http://dbm.neuro.uni-jena.de/cat/). CAT12 is based on the free and open source Statistical Parametric Mapping (SPM12, https://www. l.ion.ucl.ac.uk/spm/), which was run in MATLAB 2016 (https://www. mathworks.com/products/matlab.html). The gray matter, white matter and CSF were segmented automatically and then applied to the MNI template space with nonlinear deformation and a ne registration. CAT12 estimated cortical thickness based on the projection-based thickness (PBT) method [22], which includes partial volume correction. The gray matter and white matter were regarded as the brain parenchyma, and the ratio of brain parenchyma volume to TIV was referred to as the proportion of brain parenchyma, which was used to measure the degree of brain atrophy.
DTI images were analyzed by Fslutils (FSL, https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Fslutils), with processing including data format conversion, data quality inspection, eddy current correction, brain extraction, diffusion index estimation, diffusion tensor reconstruction, and whole brain ber track reconstruction. Then, the total number of brain bers for each subject was obtained.
Statistical analysis SPSS 21.0 (IBM Corp, Armonk, NY, USA) was used for data analysis, and a P value < 0.05 was considered signi cant. Quantitative data are expressed as the mean ± standard deviation and analyzed by independent samples t-test or the Mann-Whitney U test. Categorical variables are described as frequencies and were compared using the chi-square test. The statistical threshold was adjusted by Bonferroni correction, and a P value < 0.05/12 was taken as statistically signi cant when analyzing the volume of hippocampal subregions.
To explore the relationship between hippocampal subregion volume and cognition and physical frailty state, we conducted a partial correlation analysis between them with age, gender, years of education, Beck Depression Scale and TIV as covariates.

Result
Demographic characteristics and performance variables The MoCA scores and grip strength in the cognitive frailty group were signi cantly lower than those in the healthy control group (all P value < 0.05). The score of the MQ scores in the cognitive frailty group were signi cantly lower than those in the healthy control group (all P value < 0.05). The FI score of the cognitive frailty group was signi cantly higher than that of the healthy control group (P value < 0.05). There was no signi cant difference in gender, age, years of education or Beck Depression Scale score between the two groups (P value > 0.05). Regarding certain aspects of brain macroscopic indicators, the proportion of brain parenchyma and total number of white matter bers in the cognitive frailty group were signi cantly lower than those in the healthy control group (P value < 0.05), while the total intracranial volume and cortical thickness were not signi cantly different between the two groups. See Table 1.

Hippocampal sub eld volumes
Each side of the hippocampus was divided into 12 subregions; therefore, a total of 24 bilateral subregions were analyzed (see Fig. 1). Six hippocampal subregions had signi cant differences in volume between the CF and HC groups, including 4 from the left hippocampus and 2 from the right. Speci cally, the volumes of the left presubiculum, parasubiculum, molecular layer of the HP, and HATA and the right CA1 and presubiculum in the cognitive frailty group were signi cantly lower than those in the healthy control group (P value < 0.05/12). No signi cant difference was found in the volumes of the other hippocampal subregions. See Table 2. Correlation between MoCA, MQ, FI scores and hippocampal sub eld volumes After adjusting for age, sex, years of education, Beck Depression Inventory score and TIV, the volumes of the left molecular layer of the HP and HATA and the right presubiculum were positively correlated with the MQ score (P < 0.05); the volumes of the left presubiculum, molecular layer of the HP, and HATA and the right CA1 and presubiculum were positively correlated with the MoCA scores (P < 0.05); and the volumes of the left parasubiculum, molecular layer of the HP and HATA were negatively correlated with the frailty index score (P < 0.05). See Table 3, Fig. 2.

Discussion
CF is a simultaneous state of physical weakness and cognitive dysfunction. The pathogenesis of CF is related to changes in brain structure. Our previous study found that the volumes of certain subcortical nuclei in CF was smaller than those in healthy controls, indicating that the brain structure in CF has indeed changed [23]. The current study showed that there was an obvious decrease in the volume of 6 hippocampal subregions, including the left presubiculum, parasubiculum, molecular layer of the HP, and HATA and the right CA1 and presubiculum. Furthermore, the volumes of the left molecular layer of the HP and HATA and the right presubiculum were positively correlated with MQ score, the volumes of the left presubiculum, molecular layer of the HP, and HATA and the right CA1 and presubiculum were positively correlated with the MoCA score, and the volumes of the left parasubiculum, molecular layer of the HP and HATA were negatively correlated with the frailty index. We also found that the CF group showed a decreased proportion of brain parenchyma and total number of white matter bers. These ndings suggest that atrophy in some hippocampal subregions may be a potential mechanism underlying cognitive frailty.
To our knowledge, this is the rst study to compare changes in hippocampal sub eld volumes between CF and HC. Anatomically, atrophied structures, including the presubiculum and parasubiculum, were situated at the medial portions of the hippocampus, while the molecular layer of the HP, CA1 and HATA were situated at the lateral portions. In fact, the volume of the presubiculum or parasubiculum, which is related to cognitive level, has been found to be decreased in many diseases, such as Parkinson's disease, diabetes and Alzheimer's disease [24][25][26]. The presubiculum and parasubiculum play an important role in cognitive processing and visual spatial function [24]. The volume of the presubiculum is considered to be a promising marker of imminent memory in Alzheimer's disease [27]. Additionally, the left presubiculum volume is positively correlated with MoCA score in MCI patients, which is in line with our study [28]. The connection between the presubiculum and the retrosplenial cortex is the primary site of lesions in most forms of amnesia, and stimulation in this region is reported to enhance memory [29]. Information transfer from the parasubiculum and presubiculum to the medial entorhinal cortex is key to controlling spatial navigation, an important cognitive function [30]. Due to atrophy of the presubiculum and parasubiculum damaging these pathways, cognitive function might be weakened. Although we were unable to nd any research on the relationship between the decline in physical function and the subiculum, some studies claimed that exercise could improve the functional connectivity or structural brain health of the parahippocampal gyrus and dentate gyrus of the hippocampus with areas related to motor, sensory integration and mood regulation [31,32]. The parasubiculum might affect physical function through the above brain areas in CF.
The hippocampal CA1 region is an important part of the medial temporal lobe memory circuit. It is selectively vulnerable to attack in the process of cognitive decline, which can also predict episodic memory impairment [33,34]. Anatomical and physiological studies con rm that CA1 regulates hippocampal circuitry function and cognitive behavior by interacting with the entorhinal cortex, CA3, subiculum and dentate gyrus [35]. The results of animal experiments also showed that improvement in neuronal inactivation and apoptosis in the hippocampal CA1 area was signi cantly positively correlated with an improvement in cognitive function [36]. The CA1 region has been highlighted in most studies as a focal atrophy sub eld in the early stages of AD [37]. Our study shows that compared with that of HCs, the CA1 region of CF patients was smaller and related to cognitive dysfunction, which was consistent with previous studies [38]. As mentioned above, the CA1 region is an important node of information input and output. The results of this study showed that atrophy of the CA1 region in the CF group was closely related to lower MoCA scores, which may be the secondary result of a decline in information processing ability stemming from this atrophy.
The HATA is located in the medial region of the hippocampus and is the transitional area between the hippocampus and amygdala. A study suggested that in Parkinson's disease subjects with cognitive impairment, the volume of the left mbria, right CA1, and right HATA were decreased compared with those in normal cognition subjects, the volumes of the left parasubiculum and HATA were predictive of the conversion from normal cognition to mild cognitive impairment, and the CA1 area was associated with baseline attention [6]. The atrophy of the parasubiculum and HATA might destroy the integrity of the hippocampal-amygdala network, which is in charge of information processing [39]. In research on memory recall across the adult lifespan, it has been proposed that HATA is closely related to memory function, which is consistent with our study [39]. In addition, the HATA plays an important role in fear regulation, the underlying mechanism of situational learning and emotional memory [40]. Atrophy of the HATA might be related to a decline in the adaptability of elderly CF adults to new environments, but further research is needed to con rm this hypothesis.
The molecular layer of the HP is located above the subiculum and underneath the ssure, which includes part of the subiculum and CA elds. A study on the development of hippocampal subregion volumes across adolescence found that CA1 and molecular layer of the HP were nonlinear developmental trajectories in early volume increases, and global cognitive ability was positively associated with molecular layer of the HP development [41], while the numbers of synapses in the molecular layer of the HP showed a signi cant correlation with cognitive ability in subjects with early Alzheimer's disease, mild cognitive impairment, or no cognitive impairment [42]. Additionally, the loss of synapses might be an early event in the disease process, and this structural loss was correlated with cognitive function [42]. The current study found for the rst time that the volume of the molecular layer of the HP was not only related to cognitive function but also negatively correlated with the frailty index. The reason for the negative correlation between the molecular layer of the HP and physical frailty is not clear and needs further study.
In terms of global brain parameters, we found that compared with that in the HC group, brain parenchyma atrophy was more severe and the total number of white matter bers was smaller in the CF group. These ndings indicate that the degree of aging and complexity of the brain in elderly CF patients are decreased from the overall level. In our study, the left side of the hippocampus lost more volume than the right side, which was different from previous work on Alzheimer's disease [43]. This might mean that physical frailty may be more associated with changes in the left side of the hippocampus.

Limitations
In the present study, we focused on investigating alterations in hippocampal sub elds, and the interpretation was largely limited to hippocampal function. The limited sample size of our research may have a certain impact on the accuracy of the results. A future study with a larger sample size is needed to replicate our ndings. Second, due to the cross-sectional design of the study, the dynamic process of CF could not be observed in real-time; therefore, the causality of the relationship was not clear. Since crosssectional studies provide a limited ability to investigate relationships, a longitudinal design should be attempted in the future to gain deeper insights. Even so, the current study provides new information about the pathological progression of cognitive frailty.

Conclusion
Our study shows that hippocampal sub eld atrophy is more associated with the degree of cognitive decline and physical frailty in the brains of older adults with cognitive frailty than in those of healthy controls. These ndings indicate that speci c hippocampal sub eld volume changes might be involved in the pathological progression of cognitive frailty.

Declarations
Ethical Approval and Consent to participate

Consent for publication
Not applicable.

Data Availability Statement
Data sets used and / or analyzed in the current study are available from the appropriate authors on request. Figure 1 Diagram of hippocampal segmentation T1 images of hippocampal subregions from view angles of sagittal, coronal, and axial planes and their magni cations.