Cortical thickness is differently associated with ALDH2 rs671 polymorphism according to level of amyloid deposition

Accumulating evidence indicates that amyloid-beta (Aβ) deposition and biogenic aldehyde accumulation contribute to the pathogenesis of neurodegenerative diseases. Human aldehyde dehydrogenase 2 (ALDH2) metabolizes biogenic aldehydes produced in the brain to prevent damage. However, r671G>A, a single nucleotide polymorphism of ALDH2, causes aldehyde accumulation and decreased ALDH2 activity. We aimed to investigate whether Aβ deposition and rs671 polymorphism have an interaction effect on cortical thickness (CTh). We grouped 179 participants in the Biobank Innovations for chronic Cerebrovascular disease With ALZheimer's disease Study as follows: amyloid (–) [A(–)] and amyloid (+) [A(+)] groups based on the Aβ deposition degree; A-carrier (AC) and GG (GG) groups based on the presence/absence of the rs671 A allele; and their combinations, i.e., A(–)AC, A(–)GG, A(+)AC, and A(+)GG groups. A multiple regression analysis identified nine regions of interest. Compared with the A(–)GG group, the A(–)AC group showed thinner CTh in all regions. There were no significant differences between the A(+)AC and A(+)GG groups. We observed an interaction effect of amyloid deposition and rs671 polymorphism on CTh. The CTh in the A(–) group appeared to be strongly influenced by rs671 polymorphism, which could have contributed to cortical thinning and biogenic aldehyde accumulation in the AC group. Additionally, CTh in the A(+) group appeared to be strongly influenced by amyloid deposition.

www.nature.com/scientificreports/ disease (AD) 6 . Additionally, high 4HNE levels have also been reported in hippocampal amyloid plaques and neurofibrillary tangles 7,8 . Recent studies on neurodegenerative diseases have investigated human aldehyde dehydrogenase2 (ALDH2), which is crucially involved in the metabolism and detoxification of acetaldehyde generated by ethanol consumption and biogenic aldehydes, including 4-HNE. Moreover, studies have also investigated the rs671 polymorphism, which encodes ALDH2 2, 3,9 . Specifically, the A allele (ALDH2*2) of rs671, which is mainly inherent in the East Asian population, intrinsically deregulates ALDH2 activity and induces aldehyde accumulation 10,11 . Additionally, Chinese and Japanese studies have reported that A-carriers of rs671 (AA/GA) have an increased risk of developing AD 12,13 .
Based on the aforementioned findings, we aimed to investigate whether amyloid deposition and rs671 polymorphism had interaction effects on brain structural changes. The specific brain structural change we investigated was the cerebral cortex thickness (CTh), which has been used in previous neurodegenerative studies as a diagnostic biomarker, as well as in cognitive function evaluation and prediction [14][15][16] . We made the following hypotheses: (1) there is an interaction between amyloid deposition and rs671 polymorphism; (2) under similar amyloid deposition levels, rs671 A-carriers have lower CTh values than GG carriers; and (3) within rs671 A-carriers, amyloid deposition levels are negatively correlated with CTh. The present study aimed to analyze the interaction effects of amyloid deposition and rs671 polymorphism on CTh based on the aforementioned hypotheses.

Methods
Participants. All participants were enrolled in the Biobank Innovations for chronic Cerebrovascular disease With ALZheimer's disease Study (BICWALZS) between November 2016 and December 2019. The BICWALZS was initiated in 2016 by the National Biobank of Korea and Ajou University School of Medicine as a province-wide collaborative biorepository for collecting imaging and genetic resources regarding human biological materials for research in various chronic cerebrovascular diseases. Specifically, this study was designed to identify relationships among individual genetic information, Aβ deposition, and brain structure. We included 251 participants enrolled in the BICWALZS by 2019 with genetic data obtained using DNA chips, standardized uptake value ratio (SUVr) extracted from brain Aβ positron emission tomography (PET), and cerebral CTh data obtained through FreeSurfer after brain magnetic resonance imaging (MRI). Among them, we excluded five patients with frontotemporal dementia who showed severe asymmetrical cerebral cortical atrophy since they did not fit the scope of the study. Moreover, we excluded 67 patients diagnosed with mental illness, which is known to directly affect brain CTh. Finally, we included 179 participants.
The included participants comprised patients with subjective memory impairment, mild cognitive impairment (MCI), Alzheimer's dementia, vascular dementia, dementia with Lewy bodies, and so on. Subjective memory impairment was diagnosed based on self and/or informant report of cognitive decline without impairment in objective cognitive tasks. MCI was diagnosed based on a clinical dementia rating (CDR) 17 score of 0.5 and the expanded Mayo Clinic criteria 18 . Patients with AD were evaluated using the criteria for probable AD following the National Institute on Aging-Alzheimer's Association core clinical criteria 19 . Vascular dementia was evaluated using the Diagnostic Statistical Manual, 5th edition (DSM-5) criteria for vascular dementia and above-moderate white matter hyperintensity (WMH). Moreover, dementia with Lewy bodies was evaluated following DSM-5 guidelines. General cognition was evaluated using the Mini Mental Status Examination (MMSE) and dementia severity was measured using the CDR sum of box score.
Single nucleotide polymorphism (SNP) genotyping and quality control. After informed consent was obtained, all participants provided blood samples for DNA collection and genotyping. The collected blood samples were stored at a biorepository, followed by genomic DNA isolation. SNP genotyping was performed through hybridization on an Affymetrix Axiom KORV1.0-96 Array (Thermo Fisher Scientific, Waltham, MA) following the manufacturer's instructions. Genotype data were produced using K-CHIPs designed by the Center for Genome Science, Korea National Institute of Health, Republic of Korea (4845-301, 3000-3031). Genotyping was performed using DNA Link (Seoul, Republic of Korea). The ALDH2 genotype was derived from rs671, which was included in the genotyping array. Given that rs671G>A significantly reduces ALDH2 activity, we divided the participants into the AC group (ALDH2*1/*2 and ALDH2*2/*2) and GG group (ALDH*1/*1) based on the presence of ALDH2*2 in rs671.
Quality control was performed using PLINK (version 1.07; Free Software Foundation, Boston, MA, USA). We excluded samples with an individual call-rate < 97%, inconsistency between the reported sex and X-chromosome SNP analysis, and extremely low or high genome-wide heterozygosity (± 2 standard deviation [SD] from the mean). Moreover, we excluded samples with SNPs that had a call-rate < 95% and a Hardy-Weinberg equilibrium test p value < 10 -6 .
MRI acquisition and measurement of CTh and WMH. MRI data were obtained using a 3.0 T MR scanner. We obtained structural MRI scans, including 3D T1, T2, and fluid-attenuated inversion recovery (FLAIR) scans. All MR images were reviewed by neuroradiologists. Detailed MRI parameters according to the site have been described in Supplementary Table S1. All T1-weighted images were preprocessed with FreeSurfer (http:// surfer. nmr. mgh. harva rd. edu/, version 6.0) using the "recon-all" pipeline 20,21 . This procedure involved automatic surface-based processing, including affine registration to the Talairach atlas, intensity normalization, bias correction, and skull stripping [22][23][24][25] . Additionally, the processing steps involved the segmentation of anatomical brain images into the gray matter (pial matter), white matter, and subcortical structures, with correction of each tissue/structure boundary. Further, T2 FLAIR images were input on the "recon-all" command line to improve between-matter contrast differentiation. Surface segmentations were visually checked, and surface misplace- www.nature.com/scientificreports/ ment errors were manually fixed by editing the "brainmask.mgz" image, which is used for skull stripping, by a well-trained researcher or physician. Subsequently, the edited images underwent further surface extraction. CTh was calculated as the mean distance between the gray and white matter surfaces. We extracted the mean CTh of 68 brain regions based on the Desikan-Killiany atlas 26 using the "mri_segstats" command 27,28 . WMH quantification was performed using a semi-automatic segmentation method in the nordicICE Basis Module 29 . Specifically, on the FLAIR images, WMH was defined as white matter pixel values > 2 SD above the mean pixel value of each slice. The total WMH area in all the slices was summed and multiplied with the slice thickness to yield the total WMH volume (mL). WMH was separately examined in periventricular and deep white matter lesions.
Brain amyloid PET acquisition. All participants underwent an 18 F-flutemetamol PET scan to quantify the cortical Aβ burden using a Discovery Ste (GE), Biograph40 (SIEMENS), or Discovery 690(GE) PET/CT scanner. A bolus of 18 F-flutemetamol was injected into an antecubital vein with a mean dose of 185 or 188.7 MBq. After 90 min, a 20-min PET scan was performed. The resulting 18 F-flutemetamol PET scans were co-registered with individual MRI scans, which were normalized to a T1-weighted MRI template. Using transformation parameters, we normalized the MRI-co-registered 18 F-flutemetamol PET images to the MRI template. To quantify 18 F-flutemetamol retention, the SUVr was obtained with the pons as a reference region. We calculated the global cortical 18 F-flutemetamol retention from the volume-weighted mean SUVr of 28 bilateral cortical volumes of interest from the frontal, posterior cingulate, lateral temporal, parietal, and occipital lobes using the Annotated Anatomical Labelling atlas 30 . We speculated that the interaction effects between Aβ deposition and rs671 polymorphism differed depending on brain Aβ deposition levels. Therefore, we performed comparisons based on the Aβ deposition level. Specifically, based on a previous report on the elderly Korean population and our observed data distribution, we divided participants into the Aβ (+) and Aβ (−) groups if their global cortical SUVr was > 0.65 and < 0.65, respectively 31 .

Cognitive function assessment. Cognitive function was evaluated using the Seoul Neuropsychological
Screening Battery, which comprises tests on language, visuospatial, memory, and frontal/executive function. General cognition and dementia severity were evaluated using the Korean-MMSE (K-MMSE) and CDR sum of box, respectively. Attention, language, and visuospatial function were evaluated using the digit span backward, Boston Naming Test (range 0-60), and Rey Complex Figure Test (range 0-36), respectively. Memory function was calculated by summing the scores of verbal (Seoul Verbal Learning Test immediate recall, delayed recall, and recognition score) and visual memory tests (Rey Complex Figure Test immediate recall, delayed recall, and recognition score; range: 0-144). Frontal/executive function was evaluated by summing the scores of the Controlled Oral Word Association Test and Stroop Color Reading Test (range 0-55).

Consideration of potential confounding variables.
In the multiple linear regression analysis, we considered several confounding variables that could affect the change of brain cortical thickness and ALDH2 rs671 polymorphism, as shown in the previous studies 32 . First, we adjusted the effects of age, sex, total intracranial volume (TIV), APOE-ε4 carrier status, and education years since the effects can basically affect cortical thickness. CDR sum of box was also considered since the severity of neurodegenerative diseases can influence cortical thickness 33,34 . Since rs671 is involved in the metabolism of aldehydes produced by alcohol consumption, the lifetime drinking status was included as a covariate 35 . The lifetime drinking status was classified into "Drinker" and "Non-drinker, " where "Drinkers" and "Non-drinkers" were defined as those who had ever and never consumed alcohol, respectively. In addition, many recent studies have reported the association of rs671 with cardiovascular disorders and cerebrovascular diseases, and these diseases are known to affect the thickness of the cerebral cortex 36 . For this reason, we tried to comprehensively correct the effects of cardiovascular disease by selecting the Framingham General Cardiovascular Risk Score (FGCRS), which has been internationally used in recent studies on neurodegenerative diseases as a covariate to adjust the cardiovascular risk as a variable 37 as well as hypertension, dyslipidemia, and diabetes mellitus. The FGCRS (%) is calculated as the sum of weighted measures of age, sex, systolic blood pressure, high-density lipoprotein levels, total cholesterol levels, smoking status, and diabetes status, which was subsequently converted to a percentage. The scores are positively correlated with cardiovascular risk 38 . Finally, we tried to correct the effect on the cerebrovascular disease by selecting the WMH score of the peripheral and deep white matter 39,40 . Group categorization and statistical analysis. The included participants were divided into the following groups: the amyloid (−)[A(−)] and amyloid (+) groups [(A(+)] based on the Aβ deposition degree; the A-carrier (AC) and GG (GG) groups based on the presence/absence of rs671 A allele; and the combination of the aforementioned groups: A(−)AC, A(−)GG, A(+)AC, and A(+)GG groups. Demographic characteristics were compared between AC and GG within each amyloid group using the t-test, chi-squared test. Also, in this study, we compared continuous demographic variables, clinical variables, and cognitive function scores among the four groups (i.e., A(−)AC, A(−)GG, A(+)AC, and A(+)GG) using ANOVA and performed post-hoc tests once each result from ANOVA showed significant difference. Post-hoc tests were adjusted with Bonferroni correction (p < 0.05).
To determine the interaction effects of Aβ deposition and rs671 polymorphism on CTh, we performed multiple regression analysis with false discovery rate (FDR) correction. Statistical significance for multiple comparisons was set at an FDR threshold of < 0.05. FDR thresholding controls the expected false-positive proportion to only brain regions showing significance. After identifying a specific brain region through multiple regression analysis, we performed between-group comparisons to determine significant differences in the CTh of each brain region,  Fig. 1, visualization of brain maps was performed using R-package ggseg (https:// github. com/ ggseg/ ggseg) 41 and bar plots were drawn using the R-package ggplot2 42 .

Ethical approval. The BICWALZS was registered at the Clinical Research Information Service (CRiS) of
Republic of Korea (identifier: KCT0003391) and approved by the Institutional Review Board of Ajou University Hospital (AJIRB-BMR-SUR-16-362). It complied with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all participants and caregivers.

Results
Sample characterization. Table 1  In both amyloid groups, there were no significant differences in K-MMSE scores, CDR sum of boxes scores, hypertension, diabetes mellitus, and dyslipidemia between AC and GG. However, the FGCRS(%) was significantly higher in the A(−)AC group (16.560%) than in the A(−)GG group (14.194). Regarding lifetime alcohol drinking, there was a significantly higher ratio of non-drinkers in the AC group than in the GG group for both amyloid groups, with this difference being significant in the A(+) group [A(+)AC, 87.1%; A(+)GG, 59.0%]. This is consistent with previous reports of a low incidence of alcohol use disorder and alcohol consumption in patients with rs671G>A 35 .
Regarding age, the A(−)GG group was significantly younger than the A(+)GG group (70.774 years vs. 74.508 years). The A(−)AC group had significantly lower SUVr values than the A(+)AC group (0.576 vs. 0.865); Figure 1. The results of the regression analyses of the interaction effect of amyloid deposition and rs671 polymorphism on brain cortical thickness using the Desikan-Killiany atlas and violin plot. After correction at FDR < 0.05, nine regions were observed, with six and three in the left and right hemispheres, respectively (Lt. cuneus, Lt. pars opercularis, Lt. pars triangularis, Lt. rostral middle frontal, Lt. superior parietal, Lt. insula, Rt. Inferior parietal, Rt. pars triangularis, Rt. superior parietal). The nine regions in the brain maps were colorcoded using scales of -log10 (uncorrected p) from the multiple regression analysis. The violin plot was made with the estimated values obtained from regression analysis, and Lsmeans with standard errors were marked with error bars. Brain maps and bar plots was visualized using R-packages ggseg (https:// github. com/ ggseg/ ggseg) 41 Table 2. In the comparison among 4 groups, there were differences in visuospatial function, memory function and frontal/executive function. In addition, as a result of the post-hoc test, the memory function was statistically significantly decreased in the A(+)AC and A(+)GG groups compared to the A(−)AC and A(−)GG groups, respectively.   www.nature.com/scientificreports/ The interaction effect between Aβ deposition and rs671 polymorphism on brain cortical thickness. To determine the interaction effects between Aβ deposition and rs671 polymorphism on brain CTh, we performed multiple regression analyses with FDR correction. As shown in Table 3, we identified nine areas with FDR corrected p values < 0.05. The analysis results for a total of 68 areas divided into Desikan-Killiany atlas are displayed in Supplementary Table S4. Subsequently, we performed among-group comparisons of the CTh in the nine regions. Table 4 presents the Lsmeans and their respective p-values, which are visualized in Fig. 1. A similar pattern was observed in all nine Table 2. Cognitive function of study participants grouped according to the SUVr of amyloid PET and the presence of rs671 A allele (N = 179). AC, A-carrier; PET, positron emission tomography SD, standard deviation; SUVr, standardized uptake value ratio. All values are z-score results. a ANOVA test was conducted for continuous variables.    Table 4. Lsmeans of cortical thickness in the four groups and among-group comparisons in each brain region. Lsmeans, Least-Squares means; SE, Standard errors. We marked * for p value below 0.05 after Bonferroni correction.

Lsmeans of cortical thickness (SE) p (Bonferroni) A (−) A (+)
AC www.nature.com/scientificreports/ regions; moreover, the A(−)AC group showed significantly lower CTh values in all regions than the A(−)GG groups at a Bonferroni's corrected p value < 0.05. Contrastingly, there were no significant differences between the A(+)AC and A(+)GG groups. Compared with the A(−)GG and A(+)GG groups, the A(+)GG group showed lower Lsmeans of CTh in eight of the nine areas; among them, the differences were significant in the Lt. superior parietal, Rt. inferior parietal, and Rt. superior parietal. In contrast to our expected finding that the A(+)AC group had the lowest Lsmeans of CTh, this was observed in the A(−)AC group. Moreover, compared with the A(−)AC group, the A(+)AC group showed significantly higher CTh in the Lt. cuneus, Lt. pars opercularis, Lt. pars triangularis, and Lt. insula.

Discussion
Our findings suggest that amyloid deposition and rs671 polymorphism have interaction effects on CTh in older patients. These findings are important since they were obtained after adjustment for various clinical factors, including the vascular load, which affects CTh. Specifically, we observed a genotype-based (presence/absence of rs671G>A polymorphism) influence on the amyloid deposition effect on CTh among individuals in East Asia, including the Republic of Korea. Therefore, future studies on brain volume reduction resulting from amyloid deposition in older patients should consider their genetic characteristics.
First of all, we observed interesting results compared between the A(−) and the A(+) groups within the same rs671 genotype. Participants in the A(−)GG and A(+)GG groups who did not show a functional decline in aldehyde processing, revealed cortical thinning in the bilateral superior parietal and Rt. inferior parietal. This is consistent with previous reports regarding the areas where initial cortical thinning occurs due to amyloid deposition 43 . However, we observed notable findings after comparing amyloid deposition effects within the AC group. Specifically, there were higher CTh values in the A(+)AC group than in the A(−)AC group in all nine regions, with four regions showing significant differences. However, the A(+)AC group showed lower scores for all cognitive domains, which was significant for memory function. This is indicative of differences between the effects on CTh and cognitive function scores. Even if the CTh values contradicted the hypothesis, our findings can be explained with the biphasic change trajectory for CTh reported in recent animal 44,45 and human studies 46,47 . The biphasic trajectory in cross-sectional 46,48 and longitudinal studies 49 suggests that pathological cortical thickening occurs during the initial period of AD progression, which is more pronounced in the temporal lobe than in the frontal lobe 50,51 . In biological terms, amyloid-related CTh can be described as an inflammatory response to oligomeric Aβ, Aβ-induced neuronal hypertrophy, pathogenic synergies between Aβ and Tau, and relationships between other factors 52,53 . These findings suggest that the effect of amyloid deposition on CTh varies depending on rs671 polymorphism. Several recent studies on CTh have suggested that synergistic effects of amyloid and tau, rather than amyloid deposition alone, are crucially involved in cortical thinning; moreover, atrophy in the early disease stages is associated with these synergistic effects 46,54 . However, tau pathology primarily occurs in the late AD stage; on the other hand, neuroinflammation and astrocyte activation are major pathological occurrences in the early AD stages 55 . Specifically, numerous studies have shown that inflammation causes neuronal and glial swelling, as well as glia recruitment and activation, which presents increased CTh in the early AD stages 47,56 . Further, biogenic aldehydes are strongly associated with oxidative stress and inflammation; moreover, they are known to accelerate Aβ-mediated fibril formulation 4,5 . Taken together, these findings indicate that although cognitive decline was relatively more advanced in the A(+)AC group, there were alternative interactions other than the aforementioned interaction between amyloid deposition and rs671 polymorphism in the A(−) group. This suggests that excessive inflammation related to abnormalities in biogenic aldehyde metabolism observed in the AC group may result in increased paradoxical cortical thickening during early-stage degenerative changes caused by cell swelling, glial cell recruit/activation, and Aβ fibril formation. Moreover, these phenomena have been mainly reported in the insular and inferior frontal regions rather than the areas presenting initial cortical thinning due to amyloid deposition.
Subsequently, a comparison between AC and GG within the same amyloid group observed different results depending on the amyloid group. There was no significant difference in CTh between the A(+)AC and A(+)GG groups in all nine areas. Contrastingly, the A(−)AC group showed significantly lower CTh values than the A (−)GG group in all nine regions ( Fig. 1; Table 4). The A(−) group comprised a heterogeneous population mainly diagnosed with MCI and subjective memory impairment. Notably, the global SUVr value in the A(−)AC and A(−)GG groups were 0.576 and 0.579, respectively, which is similar to the value of 0.58 reported by a previous study on elderly Korean individuals with normal cognitive function 31 . Our results suggest that in normal individuals with amyloid deposition, differences in rs671 polymorphism significantly affect the CTh. This can be interpreted by the relationship between neurodegenerative changes and biogenic aldehydes, which has recently accumulated extensive evidence 6,8,57,58 . Biogenic aldehydes generated in the brain are metabolized by ALDH2 expressed in neurons and glial cells present in the frontal lobe, temporal lobe, hippocampus, midbrain, basal ganglia, cerebellum, etc. This process facilitates the minimization of aldehyde-induced brain damage 59 . However, ALDH2 activity is significantly reduced if rs671, which is known as the SNP of ALDH2, is AG and AA; specifically, rs671G>A is associated with an increased incidence of neurodegenerative diseases 60 . Our findings suggest that compared with the A(−)GG group, the A(−)AC group showed lower CTh values, which resulted from brain damage caused by abnormalities in the processing of biogenic aldehydes. The nine identified areas were mainly frontoparietal; moreover, previous studies have reported various areas with cortical thinning associated with aging, dementia with Lewy bodies, or other mental disorders 16,61,62 . This could be attributed to the heterogeneity in the A(−) group. We corrected FGCRS and WMH with reference to a recent study that reported a relationship between vascular load and lipid profile in relation to rs671 10,39,40,63,64 . Despite performing these corrections, we assumed that several factors, including other vascular factors not addressed in this study, may have affected the findings. www.nature.com/scientificreports/ This study has several limitations. First, this was a small-scale cross-sectional study. Additionally, the study population was heterogeneous, especially in the A(−) group, and comprised elderly individuals with various clinical diagnoses and cognitive symptoms. Second, since we employed the global SUVr, we could not determine the correlation between SUVr and the CTh of each brain region. Third, other than FGCRS(%) and WMH, we did not analyze tau protein, inflammatory factors, and other vascular factors that could directly affect brain blood flow, including vessel tortuosity, lipohyalinosis, arteriolosclerosis, venous collagenosis, blood-brain barrier disruption, and vessel obliteration 65 . Finally, this study is a candidate gene study conducted by selecting rs671. A candidate gene study has been criticized for its gene selection process and nonreplication of results. In addition, it has a limitation in that it shows a high number of false positives because all genes that contribute to the cause cannot be considered 66 .
In summary, we observed interaction effects of amyloid deposition and rs671 polymorphism on CTh. The CTh in the A(−) group appeared to be heavily influenced by rs671 polymorphism, which could have contributed to cortical thinning and biogenic aldehyde accumulation in the AC group. Additionally, CTh in the A(+) group appeared to be heavily influenced by amyloid deposition. The paradoxical phenomenon of decreased and increased cognitive function and CTh, respectively, in the AC group suggests that the interaction between amyloid deposition and a certain inflammatory response level in the AC group, which presented reduced ALDH2 activity, increases the CTh during the initial stage of degenerative changes though cell swelling, as well as glial cell recruitment and activation. Future studies on brain volume reduction related to amyloid deposition in older patients should consider their genetic characteristics. In addition, research focusing on a number of factors that can affect an individual's CTh, such as inflammation and tau pathology, as well as amyloid deposition and aldehyde accumulation, needs to be conducted to clarify the relationship between these factors.