Vitamin D has been traditionally known for its role in bone calcium homeostasis [1-5]. However, recent studies have highlighted its other actions in various body parts including the brain and the nervous system [4,6,7]. Notably, a number of previous studies have reported that vitamin D deficiency is associated with cognitive impairment in older adults [8,9]. It has been shown that vitamin D deficiency can increase the risk of all-cause dementia [10-12], particularly Alzheimer’s disease (AD) dementia [11,13]. In addition, several meta-analyses have reported a dose-response relationship between serum vitamin D levels and the risk of AD dementia [14,15], supporting links between lower serum vitamin D levels and an increased risk of the clinical manifestation of AD-type dementia.

Despite consistent findings on the association between vitamin D deficiency and risk of AD dementia in previous epidemiological studies, the exact mechanism underlying the link between serum vitamin D levels and AD at neuropathological level remains unclear. AD is a neurodege-nerative disease characterized by early cerebral beta-amyloid (Aβ) accumulation and subsequent regional neurodegeneration such as hippocampal atrophy [16], which occurs before clinical manifestation [17-19]. Thus, elucidating whether serum vitamin D levels are associated with in vivo AD neuropathologies such as Aβ deposition and neurodegeneration in nondemented older adults is crucial for investigating the potential therapeutic role of serum vitamin D in the treatment and prevention of AD. Some preclinical animal studies have suggested the role of vitamin D in the pathophysiology of AD by modulating the inflammatory process as well as Aβ peptides and tau phosphorylation [20-27]. Due to the advancement of AD biomarkers, it is now possible to measure in vivo AD-related brain changes using positron emission tomography (PET) and magnetic resonance imaging (MRI) [19]. How-ever, to date, only a very limited number of human studies with relatively small sample sizes have investigated the relationship between serum vitamin D level and in vivo AD pathologies, and their results have been inconsistent [28,29]. One previous study reported positive associations between serum vitamin D levels and cerebrospinal fluid (CSF) levels of Aβ, as well as brain tissue volumes in 75 older adults [28]. In contrast, another previous study found no significant relationships between serum vitamin D levels and cerebral Aβ deposition measured by [¹⁸F] florbetapir-PET among 178 dementia-free older adults [29].

In this context, we first investigated the direct association of serum vitamin D levels with in vivo AD pathologies, including cerebral Aβ deposition and neurodegeneration, in a large sample size of nondemented older adults. In addition, given that cerebral Aβ accumulation begins in the very early stage of AD and subsequently exacerbates neurodegeneration [19], the identification of moderating factors on Aβ-related neurodegeneration is important for searching potential therapeutic targets for AD. Thus, we also examined whether serum vitamin D could moderate the relationship between Aβ deposition and neurodege-neration.

Participants

This study was part of the Korean Brain Aging Study for the Early Diagnosis and Prediction of Alzheimer’s Disease (KBASE), a prospective cohort study began in 2014 [30]. It was designed to identify novel biomarkers for AD and investigate various lifetime experiences contributing to AD-related brain changes. This study protocol was approved by Institutional Review Boards of Seoul National University Hospital (C-1401-027-547) and SNU-SMG Boramae Center (26-2015-60), Seoul, South Korea. It was conducted in accordance with recommendations of the current version of the Declaration of Helsinki. All subjects provided written informed consents.

A total of 428 non-demented older adults diagnosed with cognitively normal (CN) or mild cognitive impairment (MCI) with age between 55 and 90 years (inclusive) were included for the analysis. Detailed protocols for recruitment and inclusion/exclusion criteria were described in a previous study [30]. In brief, older adults whose Clinical Dementia Rating (CDR) score was 0 and no diagnosis of MCI or dementia were included as the CN group. For MCI group, individuals with a CDR score of 0.5 who met the core clinical criteria of MCI according to the National Institute of Aging-Alzheimer’s Association diagnostic guideline were included [31]. We recruited participants with amnestic MCI whose performance score was at least 1.0 standard deviation (SD) below the age-, sex, and education-specific mean on at least one of the four episodic memory tests included in the Korean version of Consortium to Establish a Registry for Alzheimer’s Disease (CERAD-K) neuropsychological battery: Word List Memory, Word List Recall, Word List Recognition and Constructional Recall test [32,33]. Participants were excluded if there was any of the following conditions: 1) presence of a major psychiatric illness, significant neurological or medical condition (i.e., history of stroke); 2) contraindications for MRI; 3) illiteracy; 4) presence of severe communication problems due to hearing or vision impairment; and 5) in pregnancy or lactation.

Clinical Assessments

Board-certified psychiatrists conducted standardized clinical evaluations for participants in accordance with the KBASE clinical assessment protocol, which incorporated the CERAD-K clinical assessment [32]. A comprehensive neuropsychological assessment, CERAD-K neuropsychological battery, was also administered to all participants by trained neuropsychologists [32,33]. Vascular risk factor score (VRS) was calculated as a percentage of the number of vascular risk factors for each participant, including hypertension, diabetes, hyperlipidemia, coronary heart disease, transient ischemic attack, and stroke [34].

Measurement of Serum 25-hydroxyvitamin D and Apolipoprotein E Genotyping

Overnight fasting blood samples were collected in the morning by venipuncture. Serum levels of 25-hydroxyvitamin D (25[OH]D) were measured with a Cobas8000 e801 (Roche Diagnostics System) using an electroche-miluminescence immunoassay method. APOE genotyping was conducted as previously described [35]. APOE ɛ4 carrier positivity (APOE4 positivity) was coded if at least one ɛ4 allele was present.

Measurement of Cerebral Beta-amyloid Deposition as an Beta-amyloid Biomarker of Alzheimer’s Disease

Cerebral Aβ deposition was measured as an Aβ biomarker of AD. Participants underwent simultaneous 3-dimensional (3D) [¹¹C] Pittsburgh compound B (PiB) PET and 3D T1-weighted MRI using a 3.0T Biograph mMR scanner (Siemens). Details of PiB-PET acquisition and preprocessing were previously described [36]. Briefly, an automatic anatomic labeling algorithm [37] and a region combining method [38] were applied to determine regions of interests (ROIs) to characterize the PiB retention level in frontal, lateral parietal, posterior cingulate-precuneus, and lateral temporal regions. For intensity normalization, PiB retention in the cerebellar grey matter, a reference region extracted by a spatially unbiased atlas template of the cerebellum and brainstem, was used [39]. By dividing regional mean value by mean PiB uptake in the reference region, the mean standardized uptake value ratio (SUVR) of PiB retention in each ROI was determined. A global Aβ deposition value was the mean SUVR for all voxels in four ROIs. Aβ positivity was defined as global Aβ deposition (SUVR) greater than 1.21 [40].

Measurement of Adjusted Hippocampal Volume as a Neurodegeneration Biomarker of Alzheimer’s Disease

Adjusted hippocampal volume (HVa) was measured as a neurodegeneration biomarker of AD in all of the participants. T1-weighted MR images were also obtained from the aforementioned 3.0T PET-MR scanning and underwent automatic segmentation using FreeSurfer version 6.0 (http://surfer.nmr.mgh.harvard.edu/). After visual inspection, minor segmentation errors were manually corrected. Then, we obtained the total hippocampal volume (HV) and intracranial volume (ICV) for all participants. To control individual brain volume differences, we calculated HVa, the unstandardized residuals from a linear regression of total HV vs. estimated total ICV, using a young CN group as the reference group [41]. When using this method, if the HV of participant is smaller than the estimated HV of reference group, the HVa value could be negative. De-tailed information on MR image acquisition and preprocessing as well as calculation of HVa were described previously [36,41,42].

Statistical Analysis

We used an independent t test to examine the difference in serum vitamin D levels according to the Aβ positivity. To examine the direct association between serum 25(OH)D and two AD biomarkers, multiple linear regressions with serum 25(OH)D level as an independent variable and each AD neuroimaging biomarker including cerebral Aβ deposition and HVa as dependent variables were performed after controlling for age, sex, APOE4 positivity, and VRS.

In addition, we investigated the moderating effect of serum 25(OH)D level on the relationship between cerebral Aβ burden and HVa using a multiple linear regression analysis with an interaction variable, Aβ × 25(OH)D. This moderation analysis was performed using PROCESS macro version 4.2 for SPSS (http://processmacro.org), an observed variable ordinary least squares and logistic regression path analysis modeling tool [43]. The regression model included HVa as a dependent variable and Aβ, serum 25(OH)D, and Aβ × 25(OH)D as independent variables, with same covariates including age, sex, APOE4 positivity, and VRS. We also performed the same multiple linear regression analysis including the interaction variable Aβ × 25(OH)D, among Aβ-positive participants. All statistical analyses were conducted using IBM SPSS Statistics software version 26 (IBM Co.). A p value < 0.05 (two-sided) was considered statistically significant.

Characteristics of Participants

Demographic and clinical characteristics of participants are presented in Table 1. A total of 428 nondemented older adults were included in the current study. Mean (SD) of age was 70.54 (7.97) years and 182 (42.52%) individuals were males. Mean serum 25(OH)D level was 21.66 (9.53) ng/ml in all participants. When we compared the mean serum 25(OH)D levels between Aβ-positive (21.34 [9.42] ng/ml; N = 135) and Aβ-negative (22.42 [9.76] ng/ml; N = 292) groups, no significant difference was observed between the two groups (p = 0.276).

Direct Association of Serum 25-hydroxyvitamin D Level with Beta-amyloid and Neurodegeneration Biomarkers of Alzheimer’s Disease

We first examined the direct associations between serum 25(OH)D levels and Aβ and neurodegeneration biomarkers of AD. Regarding the Aβ biomarker of AD, serum 25(OH)D level was not associated with global Aβ deposition after controlling for the effect of age, sex, APOE4 positivity, and VRS. Additionally, in terms of neurodegeneration biomarker, serum 25(OH)D level was not associated with HVa, either, after controlling for the same covariates (Table 2).

Moderation Effect of Serum 25-hydroxyvitamin D Level on the Association between Beta-amyloid and Neurodegeneration Biomarkers of Alzheimer’s Disease

We also evaluated how the serum 25(OH)D level moderates the association between cerebral Aβ deposition and HVa through the multiple linear regression using an interaction term (i.e., global Aβ deposition × serum 25(OH)D level), after controlling for the same covariates. In this model, direct effect of global Aβ deposition (B [standardized error, SE] = −951.377 [127.497], p < 0.001), but not serum 25(OH)D level (B [SE] = 2.862 [4.253], p = 0.501), on HVa was significant. In terms of moderation effect, we found a significant moderating effect of serum 25(OH)D level (B [SE] = 34.612 [12.971], p = 0.008) on the relationship between global Aβ deposition and HVa (Table 3 and Fig. 1). A lower serum 25(OH)D level exacerbated the negative association between global Aβ deposition and HVa. Conversely, this association attenuated when serum 25(OH)D level was increased. When we classified participants into 3 groups according to the level of serum 25(OH)D and performed same multiple linear regression analyses in each tertile group, the lower (1st) tertile group (<16.09 ng/ml) exhibited a steeper decline of HVa according to global Aβ deposition compared to other tertile groups (Fig. 2 and Table 4). The moderating effect of serum 25(OH)D levels on the relationship between global Aβ deposition and HVa remains significant when we performed the same multiple linear regression model among 135 Aβ-positive participants (B [SE] = 61.711 [22.987], p = 0.008; Table 3).

In this study, we examined the association and interaction of serum vitamin D level with in vivo AD pathologies including Aβ and neurodegeneration in nondemented older adults using neuroimaging biomarkers of AD. We observed a significant moderating effect of serum vitamin D levels on the relationship between two AD biomarkers, although direct associations between serum vitamin D with each AD biomarker were not significant. Our findings indicate the lower the serum vitamin D level exacerbate Aβ-associated neurodegeneration, whereas higher serum vitamin D levels appear to attenuate Aβ-associated neurodegeneration in nondemented older adults.

Only a couple of previous clinical studies have investigated the association of blood vitamin D level with AD biomarkers, with relatively small sample sizes and conflicting results [28,29]. One previous study has reported a positive association between serum vitamin D and CSF Aβ level in 75 individuals, which partly consisted of those with AD dementia [28]. In contrast, another previous study has shown no significant association between plasma vitamin D level and cerebral Aβ deposition measured by [¹⁸F] florbetapir-PET in 178 nondemented older adults [29]. In the context of our study and a previous study that reported no significant association [29], it is of note that only nondemented older adults were included in both studies. For patients with dementia, we can postulate that their relatively lower outdoor activity and poorer nutrition, which correlates with the severity of dementia, might have contributed to a lower serum vitamin level, suggesting a potential issue of reverse causality. Thus, differences in clinical diagnosis of participants between a previous study [28] and ours might have influenced study results. Therefore, when interpreting findings from previous literature on individuals with dementia, it is essential to consider the possible influence of reverse causality. In addition, differences regarding methodologies for measuring Aβ biomarker (i.e., CSF vs. PET biomarker) and other characteristics of study samples (i.e., age, APOE4 positivity, etc.) might explain the heterogeneity between studies. A previous study reporting a positive association between serum vitamin D and CSF Aβ level [28] comprised relatively younger individuals (i.e., in their 50−60s) with higher APOE4 positivity rates (ranging from 42 to 67%) compared to another previous study [29] and the present study.

In our moderation analysis, we observed significant interaction of serum vitamin D level with in vivo cerebral Aβ deposition on hippocampal volume reduction in nondemented older adults. The degree of negative association between brain Aβ deposition and hippocampal volume was greater when the serum vitamin D level is lower. In contrast, Aβ-associated hippocampal neurodegeneration was attenuated when serum vitamin D levels were higher in nondemented older adults. This finding suggests that serum vitamin D level might be involved in aggravation of Aβ-related neurodegeneration rather than direct contribution to each AD-related brain changes. Given a positive correlation between serum and CSF total 25(OH)D level reported by a previous study [44], our finding is consistent with previous preclinical studies supporting a protective role of vitamin D against Aβ-related neurotoxicities [45,46]. A previous study had conducted an in vitro study using a neuronal cell line and found that vitamin D administration could prevent cytotoxicity of Aβ by modulating vitamin D receptor and specific calcium channel expression [45]. Another previous study had also reported a protective effect of vitamin D against Aβ-related cytotoxicity through cell signaling pathway modification and found that long-term vitamin D treatment could reduce Aβ-induced damage to hippocampi of rats [46].

Meanwhile, serum vitamin D level did not show a direct association with HVa, a neurodegeneration biomarker. Considering that the majority of participants in our study were CN older adults who might have relatively less severe hippocampal atrophy compared to cognitively impaired ones, our finding was in line with a previous study reporting no association between vitamin D level and hippocampal volume in nondemented older adults with normal neuropsychological test results [47]. Although some previous studies have reported smaller hippocampal volume in a vitamin D deficient group [48,49], these studies were conducted in those with cognitive decline including dementia. Thus, different sample characteristics among studies could be one potential factor for explaining our findings in the context of previous literature.

The present study has several strengths. To the best of our knowledge, this is the first study examining the moderating effect of vitamin D on in vivo AD pathologies using AD biomarkers in nondemented older adults. Moreover, the sample size of our study with a total of 428 participants was much larger than previous research studies [28,29]. Nevertheless, our study has some limitations. First, as the current study had a cross-sectional design, a causal relationship could not be inferred. Second, it remains unclear whether lower serum vitamin D levels also aggravate Aβ-related neurodegeneration in brain regions other than the hippocampus. Thus, future studies investigating brain regions vulnerable to the moderating effect of serum vitamin D levels will be valuable to determine whether this effect is specific to the hippocampal region. In addition, the relationship between serum vitamin D and brain tau deposition was not explored in this study. Further studies with a longitudinal design are needed to infer the etiological contributions of vitamin D to diverse AD-related neuropathologies, including tau deposition.

Our study observed moderating effect of serum vitamin D levels on in vivo AD pathologies, by exacerbating Aβ-associated neurodegeneration when serum vitamin D level was low in nondemented older adults. This provides a new perspective on the role of vitamin D in the pathophysiology of AD. Considering that vitamin D deficiency is a modifiable risk factor of AD which can be corrected by its supplementation, future studies should investigate the potential therapeutic effect of vitamin D supplementation on the progression of AD pathology.

The authors are grateful to all participants in this study. A complete list of KBASE research group members can be found at http://kbase.kr/eng/about/research.php.

No potential conflict of interest relevant to this article was reported.

Conceptualization: Junha Park, Min Soo Byun. Data acquisition, analysis, and interpretation: Junha Park, Min Soo Byun, Dahyun Yi, Hyejin Ahn, Joon Hyung Jung, Nayeong Kong, Yoon Young Chang, Gijung Jung, Jun-Young Lee, Yu Kyeong Kim, Yun-Sang Lee, Koung Mi Kang, Chul-Ho Sohn, Dong Young Lee. Statistical analysis: Junha Park, Min Soo Byun. Supervision: Min Soo Byun, Dong Young Lee. Writing—original draft: Junha Park, Min Soo Byun. Writing—review & editing: Junha Park, Min Soo Byun, Dahyun Yi, Hyejin Ahn, Joon Hyung Jung, Nayeong Kong, Yoon Young Chang, Gijung Jung, Jun-Young Lee, Yu Kyeong Kim, Yun-Sang Lee, Koung Mi Kang, Chul-Ho Sohn, Dong Young Lee.