Effect of HMGCR genetic variation on neuroimaging biomarkers in healthy, mild cognitive impairment and Alzheimer's disease cohorts

Alzheimer's disease (AD) has become a considerable public health issue. The mechanisms underlying AD onset and progression remain largely unclear. 3-Hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) is a strong functional AD candidate gene because it encodes part of the statin-binding domain of the enzyme, which serves as the rate-limiting step in cholesterol synthesis in all mammalian cells. Here, we evaluated the potential role of HMGCR (rs3846662) in AD-related pathology by assessing neuroimaging biomarkers. We enrolled in 812 subjects from the Alzheimer's disease Neuroimaging Initiative dataset. In general, it is possible that HMGCR (rs3846662) could be involved in preventing the atrophy of right entorhinal (P=0.03385) and left hippocampus (P=0.01839) in the follow-up research of two years. What's more, it lowered the drop rate of glucose metabolism in right temporal. We then further validated them in the AD, mild cognitive impairment (MCI), normal control (NC) sub-groups. All the results in the MCI groups confirmed the association. The results of our study indicated that HMGCR (rs3846662) plays a vital role in AD pathology mainly by influencing brain structure and glucose metabolism during AD progression.


INtrODUctION
Alzheimer's disease (AD), characterized by formation of amyloid plaque, neurofibrillary tangles and loss of neurons, is the most prevalent forms of dementia affecting the aging population, and it has become a considerable public health issue [1]. The mechanisms underlying AD onset and progression remain largely unclear. Up to now, the only genetic variant consistently shown to impact AD risk and age at onset was APOE [2]. It has been reported that variations in genes associated with cholesterol homeostasis modify the risk of AD progression [2]. 3-Hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) is a strong functional AD candidate gene because it encodes part of the statin-binding domain of the enzyme, which serves as the rate-limiting step in cholesterol synthesis in all mammalian cells [3,4]. The association of HMGCR (rs3846662) with LDL-C levels has been replicated in Japanese Populations [5]. A study identified rs3846662 as a functional SNP in intron13 [6]. Besides, it was in linkage disequilibrium with the SNPs of genome-wide significance and influenced alternative splicing of HMGCR mRNA [7]. Compounds having inhibitory effects on the enzyme HMGCR, such as statins, are thought to possess anti-inflammatory effects and may have some clinical effects in AD patient [8]. A study has clearly identified HMGCR rs3846662 as a potent genetic modifier for AD risk, age of onset and conversion in three cohorts study [4].
Neuroimaging methods increasingly are used as additional outcome measures to explore the link between HMGCR rs3846662 and AD. Therefore, this study aims to investigate the involvement of HMGCR in the development and progression of AD by studying the influence of HMGCR on neuroimaging biomarkers in the three different clinical stages (normal control (NC), mild cognitive impairment (MCI), AD) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Demographic and clinical characteristics
The dataset comprised of 48 AD patients (18 women, 75.51±9.23 years), 483 MCI samples (201 women, 72.28±7.45 years) and 281 NC (145 women, 74.51±5.56 years) at baseline in our study (Table 1). In consistent with previous studies, AD patients had the highest frequency of the ε4 allele within APOE gene and CN group had the lowest frequency. In addition, the AD patients have worst cognitive function according to the scores of the five neuropsychological scales (Clinical Dementia Rating scale sum of boxes (CDRSB), the Alzheimer's Disease Assessment Scale 11(ADAS11), the Mini-Mental State Examination (MMSE), the Alzheimer's Disease Assessment Scale 13 (ADAS13), the Rey' Auditory Verbal Learning Test (RAVLT) and Functional Activities Questionnaire (FAQ)). Furthermore, no statistical differences were observed among NC, MCI and AD patients when comparing the distribution of the tested SNP allele frequencies in our study.

Impacts of HMGCR (rs3846662) on MrI structure
Here we selected several cortical areas as the ROIs of the MRI measures analysis, including the volume of middle temporal gyrus, posterior cingulate, precuneus, parahippocampal gyrus, and hippocampus (Supplementary Table 1). In the cortical structure analysis, baseline volume of right entorhinal was identified to have strong associations with SNP rs3846662 in the hybrid population  Figure 1A). Besides, patients with CC and CT genotypes had greater volume than those with TT genotypes in right entorhinal (CC: 1808±446.2 mm 3 , CT: 1764±408.2 mm 3 , TT: 1700±410.8 mm 3 , P = 0.01002) ( Table 3) ( Figure 1A). So it was detected that right entorhinal of C-allele carriers in the subjects showed significant prevention on the percentage of atrophy. Then we compared the volume between TT, CT, and CC groups in the three different clinical stages (NC, MCI, and AD). We observed that rs3846662 only prevent the atrophy of right entorhinal in the MCI group (P = 0.02755) ( Figure  1D). In the subcortical structure analysis, we didn't find the baseline volume of any region have relationship with SNP rs3846662 in the hybrid population.
In the follow-up research of two years, we found that rs3846662 decrease the percentage of atrophy in the volume of right entorhinal (P = 0.03385) ( Figure 1B). The C-allele in remarkably protected right entorhinal from shrinking at the two-year follow-up (CC: 0.99±0.1738, CT: 0.979±0.136, TT: 0.95±0.1323) (Table 3). Similarly, we can get the results in the MCI group (P = 0.01073) ( Figure 1E). With regard to the subcortical structures, rs3846662 decreased the percentage of atrophy in the volume of left hippocampus in the hybrid population (P = 0.01839) ( Figure 1C). We can get same results in both dominant and recessive pattern (Supplementary Table 2). The C-allele prevented left hippocampus atrophy (CC: 0.9828±0.1051, CT: 0.9564±0.06219, TT: 0.9515±0.07469, P = 0.01839) ( Table 3). Furthermore, in the MCI cohort, the volume of left hippocampus was found to be related to rs3846662 (P = 0.003746) in the follow-up research of two years ( Figure 1F). In addition, the C-allele prevented left hippocampus atrophy in MCI subjects. More importantly, HMGCR (rs3846662) was not found to be related to hippocampus or hippocampus substructure (CA1) (P > 0.5) (Supplementary Table 1), the most associated substructure with the AD specific amnestic syndrome in hippocampus [9].

Impacts of HMGCR (rs3846662) on glucose metabolism
We test the relationships between HMGCR genotypes and the cerebral metabolism rate of glucose (CMRgl) on FDG-PET imaging (Supplementary Table 2). Right temporal was found to be related to rs3846662 at baseline in the hybrid population (P = 0.03468) ( Figure  2A). In the analysis of the three groups, same result was found in MCI groups at baseline (P = 0.001808) ( Figure  2C). The C-allele had significantly higher CMRgl in the right temporal (CC: 1.237±0.1166, CT: 1.212±0.1146, TT: 1.189±0.1413, P = 0.001808) (Table 2) ( Figure 2C).
In the follow-up research of two years, rs3846662 Data are given as mean ± standard deviation unless otherwise indicated www.impactjournals.com/oncotarget may lower the drop rate of CMRgl in right temporal in the hybrid population (P = 0.03078) ( Figure 2B). In the analysis of the three groups, we observed that rs3846662 may lower the drop rate of CMRgl in right temporal in MCI groups at the two-year follow-up (P= 0.01865) ( Figure 2D). Besides, subjects carrying the C-allele had significantly higher CMRgl in the right temporal (CC: 0.9965±0.06403, CT: 0.9791±0.05922, TT: 0.9731±0.07254, P = 0.03078) ( Table 3) ( Figure 2D). However, we did not observe significant difference on other regional CMRgl in the other two groups (NC or AD).

Impacts of HMGCR (rs3846662) on AV45-PEt
In the analysis of AV45-PET, HMGCR (rs3846662) did not show an effect on florbetapir retention as well as the cortical SUVR. Likewise, we did not detect any significant associations on the changes over two years of these regions (Supplementary Table 3).

DIscUssION
Most brain cells generate cholesterol by two separate yet interrelated processes: synthesis by HMGCR and internalization of lipoproteins by means of the APOE/ low density lipoprotein receptor cascade which is damaged in AD [4]. Cholesterol synthesis may be involved in AD neurodegeneration because of polymorphism in HMGCR was found to be related with AD [10]. Some variants of HMGCR have been identified as AD susceptibility loci, whereas little is known about the interaction between the HMGCR and risk of AD [2, 10]. Rs3846662 SNP in intron 13 of the HMGCR gene was found to significantly associate with sporadic AD [4]. This present study investigated the effects of HMGCR rs3846662 (TT, CT, It is recognized that regional brain atrophy could be an important underlying pathology in functional imaging studies of neurodegenerative diseases [11,12]. A study has demonstrated that structural brain changes occur years before the clinical onset in AD and were localized to regions influenced by AD neuropathology [13]. Therefore, in our study we detected the influence of rs3846662 on different brain volumes. Our study showed that variants at rs3846662 would change the volume of right entorhinal in the two-year follow-up study and the C-allele remarkably prevented the atrophy of right entorhinal. Then the finding was confirmed by the stratified analyses of MCI group. In addition, rs3846662 had influence on the volume of left hippocampus in the two-year follow-up study and C-allele showed protective effect on the volume of left hippocampus. This was in accordance with previous MRIbased studies that hippocampal atrophy was a recognized biological marker of AD. The finding was confirmed by the stratified analyses of MCI group. But we did't found any relation in AD group due to the limited sample size. Apart from hippocampal, middle temporal has been implicated as a key brain region involved in the pathogenesis of AD [14,15]. But we didn't confirm it in our study.
Cerebral glucose metabolic rate may be related with AD. We investigated the mechanism underlying the interaction between rs3846662 and AD in this study, and found that right temporal reveal higher CMRgl both at baseline and the two-year follow-up. In addition, in the stratified analyses of AD, MCI and NC group, we also discovered C-allele carriers increased CMRgl in right temporal in the follow-up study of two years.
Some previous studies have paid their attentions to the effects of cholesterol on Aβ metabolism. They have shown that brain cholesterol metabolism may play a certain role in AD development, with increased cellular cholesterol levels leading to high amyloid beta (Aβ) production [16][17][18]. Another study believe Aβ directly affects the activity of the HMGCR enzyme and the metabolism of cholesterol in neuronal cells, being γ-secretase dependent [8]. In our study, AV45-PET imaging was utilized to mark the presence and deposition of Aβ. Nevertheless, no significant effect was detected for rs3846662 SNP on florbetapir retention. In addition, in the stratified analyses of AD, MCI and NC group, we also didn't discover rs3846662 have connection with Aβ deposition. More evidence will be needed to illustrate the interactions between rs3846662 and Aβ deposition.
There are several potential limitations in our study. First, our results are obviously limited to the ADNI dataset and its small sample size. Second, a follow up of two years may be too short to detect the significant influence of HMGCR on the AD process, and further study need an increase of follow-up. Moreover, our results cannot represent other ethnicities, because our sample was restricted to Caucasians. Finally, unbalanced number of subgroups (AD, MCI and NC) also makes the conclusions of total cohort tends to the results arisen from MCI, which has the most case number. This might lead to untrue finding for the entire cohort. Hence, more studies are needed to perform to fill the gaps.
In general, it is possible that HMGCR (rs3846662) could be involved in the structural and functional modification of right entorhinal and left hippocampus throughout the AD physiopathological process. What's more, it influedced glucose metabolism of right temporal. HMGCR (rs3846662) plays an important role in AD-related neurodegenerative processes. Therefore, identification of specific factors that regulate HMGCR alternative splicing and elucidating the underlying mechanism may lead to a better understanding of its impact on regulating cellular cholesterol homeostasis and neurodegenerative processes. It is necessary to explore more clear understanding on the mechanisms in a larger sample, with longer follow-up and a wider range of people.

Participants
Data used in this article were got from the ADNI database (www.loni.ucla.edu\ADNI). The ADNI is an ongoing, longitudinal, multicenter study aimed at developing clinical, genetic, serial magnetic resonance imaging (MRI), positron emission tomography (PET), and biochemical biomarkers to measure the progression of MCI and early AD [19]. ADNI is the result of efforts of many co-investigators in the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, the Food and Drug Administration, private pharmaceutical companies and nonprofit organizations and subjects have been recruited from over 50 sites across the United States and Canada [20]. The initial goal of ADNI was to recruit 800 subjects, but it has been followed by ADNI-GO and ADNI-2. To date, three protocols have recruited over 1500 adults, ages 55 to 90, to participate in the research, consisting of cognitively normal older subjects, people with early or late MCI, and people with early AD. For more information, see http:// www.adni-info.org.
We enrolled participants according to criteria demonstrated in the ADNI study protocol (http://www. adni-info.org/scientists/adnistudyprocedures.aspx). The present analysis was restricted to participants whose genotype data of HMGCR single nucleotide polymorphisms (SNPs) were available. Furthermore, we selected only non-Hispanic (Caucasian) participants in order to avoid the population stratification effects which www.impactjournals.com/oncotarget can lead to spurious findings. Finally, 812 individuals including 281 NC, 483 MCI (including 63 who converted to AD and 420 who did not) and 48 AD were included in our study. Baseline and longitudinal data including structural MRI and PET were collected as parts of this study. Furthermore, all participants underwent a series of baseline clinical tests including CDRSB, ADAS11, MMSE, ADAS13, RAVLT, FAQ.

sNP selection and genotyping
We performed the quality control (QC) procedures using PLINK software, and the inclusion criteria were as follows: minimum call rates > 90%, minimum minor allele frequencies (MAF) > 0.01, Hardy-Weinberg equilibrium test P > 0.001. We only enrolled SNPs which have been studied, large case-control trials or other experimental studies. After quality control procedures, one tag SNP which captured the greatest amount of common variations in HMGCR (rs3846662) was remained for data analysis (Table 2).

MrI structure
UCSF FreeSurfer datasets were used to conduct association test of HMGCR genotypes with brain structure and the cerebral image segmentation and analysis were performed with the FreeSurfer version 5.1 (http://surfer.nmr.mgh.harvard.edu/) based on the 2010 Desikan-Killany atlas [21]. This process mainly contained motion correction and averaging of multiple volumetric T1 weighted images (when more than one is available), removal of non-brain tissue using a hybrid watershed/surface deformation procedure, automated Talairach transformation, segmentation of the subcortical white matter and deep gray matter volumetric structures (including hippocampus, amygdala, caudate, putamen, ventricles) [22], intensity normalization, tessellation of the gray matter white matter boundary, automated topology correction, and surface deformation following intensity gradients to optimally place the gray/white borders at the location where the greatest shift in intensity defines the transition to the other tissue class. The technical details of these procedures are described in prior publications [23]. we selected regions have been reported to link to AD closely, such as hippocampus, parahippocampus, middle temporal, posterior cingulate, precuneus, and entorhinal cortex as our regions of interest (ROI) to analyze their associations with HMGCR genotypes [24][25][26].

Glucose metabolism on imaging
FDG analysis data were from UC Berkeley and Lawrence Berkeley National Laboratory on the website (http://adni.loni.usc.edu/data-samples/access-data/) [27]. We extracted the mean counts about the five ROIs (left angular gyrus, right angular gyrus, bilateral posterior cingular, left inferior temporal gyrus, right inferior temporal gyrus) for each subject's FDG scans at baseline and 24-month. These images were spatially normalized in Statistical Parametric Mapping (SPM) to the MNI PET template. Then the intensity values were computed with SPM subroutines. Finally each ROIs mean were normalized by dividing it in pons/vermis reference region mean.

AV45-PEt
We obtained PET imaging data with amyloid tracer, florbetapir (AV-45), from UC Berkeley-AV45 analysis dataset on website (http://adni.loni.usc.edu/data-samples/ access-data/). For AV45-PET, mean florbetapir uptake within 4 cortical regions (frontal, anterior/posterior cingulate, lateral parietal, and lateral temporal) was extracted. Cortical standardized uptake values ratios (SUVR) were calculated by averaging across the 4 cortical regions and dividing this average by whole cerebellum. Each mean florbetapir uptake of the 4 main regions and cortical SUVR were used for analysis.

statistical analysis
We used one-way analysis of variance (ANOVA) to test the differences in continuous variables. Categorical data were tested using chi-square test to analyze