Effects of circadian clock genes and environmental factors on cognitive aging in old adults in a Taiwanese population

Previous animal studies have indicated associations between circadian clock genes and cognitive impairment . In this study, we assessed whether 11 circadian clockgenes are associated with cognitive aging independently and/or through complex interactions in an old Taiwanese population. We also analyzed the interactions between environmental factors and these genes in influencing cognitive aging. A total of 634 Taiwanese subjects aged over 60 years from the Taiwan Biobank were analyzed. Mini-Mental State Examinations (MMSE) were administered to all subjects, and MMSE scores were used to evaluate cognitive function. Our data showed associations between cognitive aging and single nucleotide polymorphisms (SNPs) in 4 key circadian clock genes, CLOCK rs3749473 (p = 0.0017), NPAS2 rs17655330 (p = 0.0013), RORA rs13329238 (p = 0.0009), and RORB rs10781247 (p = 7.9 × 10−5). We also found that interactions between CLOCK rs3749473, NPAS2 rs17655330, RORA rs13329238, and RORB rs10781247 affected cognitive aging (p = 0.007). Finally, we investigated the influence of interactions between CLOCK rs3749473, RORA rs13329238, and RORB rs10781247 with environmental factors such as alcohol consumption, smoking status, physical activity, and social support on cognitive aging (p = 0.002 ∼ 0.01). Our study indicates that circadian clock genes such as the CLOCK, NPAS2, RORA, and RORB genes may contribute to the risk of cognitive aging independently as well as through gene-gene and gene-environment interactions.


INTRODUCTION
Circadian rhythms are naturally recurring cycles that influence the timing of biological events such as sleepwake cycles, hormone release, and energy metabolism [1]. The intracellular molecular machinery underlying circadian rhythms indicates that circadian oscillations are stimulated and maintained by a panel of core circadian clock genes, defined as genes whose protein products are necessary components for generating circadian rhythms within individual cells [2]. Core circadian clock genes include the aryl hydrocarbon receptor nuclear translocator like (ARNTL), clock circadian regulator (CLOCK), cryptochrome circadian clock 1 (CRY1), cryptochrome circadian clock 2 (CRY2), neuronal PAS domain protein 2 (NPAS2), nuclear receptor subfamily 1 group D
The key circadian clock gene ARNTL encodes the ARNTL protein, which is an elementary helix-loop-helix protein that forms heterodimers with either CLOCK or NPAS2, two other primary helix-loop-helix proteins [2]. The ARNTL/CLOCK heterodimer initiates the transcription of numerous target genes such as PER1, PER2, PER3, CRY1, and CRY2 via E-box elements in the promoters of the target genes [2]. One after another, the resulting PER and CRY proteins inhibit further ARNTL/CLOCK transcriptional activity; a new cycle then proceeds due to the low level of ARNTL/CLOCK transcription to enhance the robustness of the oscillations based on the main loop, an additional feedback loop is formed with nuclear receptors such as NR1D1 (or Rev-erb alpha), RORA, and RORB, which initiate the circadian transcription of the ARNTL gene and thus contribute to the timing of the core clock machinery [3].
Previous works have shown that the circadian clock is involved in regulation of brain cognitive functions such as memory, mood, and adaptation to novelty [4][5][6]. Furthermore, several animal studies have indicated that impairment of memory and adaptation to novelty can result from a loss-of-function mutation in circadian clock genes, such as Arntl [7], Clock [7], Cry1 [8], Cry2 [8], Npas2 [9], and Per2 [10]. In addition, an ageassociated decline in the activity of the circadian clock can contribute to cognitive aging or declines in multiple brain functions, such as sleep, mood, and memory, indicating that disturbed circadian rhythms may affect cognitive functions [5]. Therefore, alterations of the circadian clock gene system may play a causative role in cognitive decline in neurodegenerative diseases such as Alzheimer's diseases (AD) [11,12]. Specifically, several recent association studies have indicated that single nucleotide polymorphisms (SNPs) within the ARNTL and CLOCK genes are associated with AD risk [13][14][15][16].
While several encouraging findings on the relationship between circadian clock genes and cognitive aging have emerged, to our knowledge, human data is scarce in terms of SNPs. Moreover, the interplay between circadian clock genes and environmental factors such as alcohol consumption, smoking status, physical activity, and social support, has not been comprehensively evaluated in previous association studies. Given that circadian rhythms and their relevant genes may play a key role in the development of cognitive aging, we hypothesized that core circadian clock genes may contribute to the etiology of cognitive aging independently and/or through complex interactions. The gene panel investigated here comprises the 11 aforementioned core circadian clock genes (Supplementary Table 1), namely  the ARNTL, CLOCK, CRY1, CRY2, NPAS2, NR1D1,  PER1, PER2, PER3, RORA, and RORB genes. Table 1 describes the demographic and clinical characteristics of the study population, comprised of 634 subjects. The median MMSE score was 27 and the interquartile range was 25-29. First, we investigated the association between cognitive aging and the 11 circadian clock genes mentioned above. Among the 644 SNPs assessed in this study (Supplementary Table 1), there were 74 SNPs in 8 of the circadian clock genes that demonstrated evidence of association (P < 0.05) with MMSE scores (Table 2). However, only the association of the RORB rs10781247 SNP with MMSE scores nearly reached significance after applying a Bonferroni correction (P < 0.05/644 = 7.8 x 10 -5 ). As shown in Table 2, the RORB rs10781247 SNP indicated an association with MMSE scores among subjects after adjusting for covariates such as age, gender, and education in genetic models, including the additive model (P = 0.0021) and dominant model (P = 7.9 x 10 -5 ).
We then employed categorized MMSE scores as an outcome (normal: MMSE score ≥ 24; cognitive impairment: MMSE score < 24) for gene-gene and geneenvironment analysis. First, generalized multifactor dimensionality reduction (GMDR) analysis was used to assess the impacts of combinations between the four key SNPs in cognitive aging, including age, gender, and education as covariates. Table 3 summarizes the results obtained from GMDR analysis for two-way up to four-way gene-gene interaction models with covariate adjustment. There were significant two-way models involving NPAS2 rs17655330 and RORB rs10781247 (P = 0.003), RORA rs13329238 and RORB rs10781247 (P = 0.003), NPAS2 rs17655330 and RORA rs13329238 (P= Oncotarget 24090 www.impactjournals.com/oncotarget 0.004), and CLOCK rs3749473 and RORA rs13329238 (P = 0.042). These results indicate potential gene-gene interactions between NPAS2 and RORB, RORA and RORB, NPAS2 and RORA, and CLOCK and RORA, respectively, in influencing cognitive aging. Additionally, there were a three-way model (P = 0.001) and a four-way model (P = 0.007) indicating potential gene-gene interactions among CLOCK, NPAS2, RORA, and RORB in influencing cognitive aging. Table 4 shows the GMDR analysis of geneenvironment interaction models in cognitive aging using age, gender, and education as covariates. There was a significant two-way model involving CLOCK rs3749473 and environmental factors, namely smoking (P = 0.004), alcohol consumption (P = 0.01), and social support (P = 0.011), indicating potential gene-environment interactions between CLOCK and environmental factors in influencing cognitive aging. Similarly, there was a significant two-way model involving RORA rs13329238 and environmental factors such as smoking (P = 0.01), alcohol consumption (P = 0.002), physical activity (P = 0.01), and social support (P = 0.002). Finally, there was a significant two-way model involving RORB rs10781247 and the environmental factors alcohol consumption (P = 0.005) and physical activity (P = 0.035). However, there was no significant two-way model involving NPAS2 rs17655330 and environmental factors.
Finally, Table 5 demonstrates a summarized model of the associations between the MMSE scores and SNPs within the CLOCK, NPAS2, RORA, and RORB genes. Table 5 also presents a summarized model of gene-gene and gene-environment interactions among the CLOCK, NPAS2, RORA, and RORB genes.

DISCUSSION
To our knowledge, our study is the first to date to assess whether 644 SNPs in 11 circadian clock genes are significantly associated with the risk of cognitive aging independently and/or through gene-gene interactions among old Taiwanese individuals. We also examined the relationship between these genes and environmental factors to investigate whether these genes confer a risk of cognitive aging according to its effect on geneenvironment interactions. Here, we report for the first time that several SNPs of the circadian clock genes, including CLOCK rs3749473, NPAS2 rs17655330, RORA rs13329238, and RORB rs10781247, may play an important role in the modulation of cognitive aging in old adults in a Taiwanese population. Notably, the association of the RORB rs10781247 SNP with MMSE scores nearly persisted after correcting for multiple testing (P < 7.8 x 10 -5 ). Additionally, our data revealed that genegene interactions between CLOCK, NPAS2, RORA, and RORB may contribute to the etiology of cognitive aging. Our data also indicated that there were gene-environment interactions between CLOCK, RORA, and RORB with environmental factors such as alcohol consumption, smoking status, physical activity, and social support.
To our knowledge, our results are the first to raise the possibility that 42 SNPs in the RORA gene and 6 SNPs in the RORB gene may contribute to susceptibility to cognitive aging, especially the RORA SNPs rs13329238, rs8040450, rs11638929, rs17237563, rs17303341, rs146660446, and the RORB SNPs rs10781247 and rs2273975. The proteins encoded by RORA (located on chromosome 15q22.2) and RORB (located on chromosome 9q2) constitute a subfamily of nuclear hormone receptors [3]. The RORA and RORB genes have been implicated in the regulation of a wide variety of physiological processes, including circadian rhythm, immunity, cellular metabolism, embryonic development, and inflammatory responses [17][18]. By using psychometric tests of cognitive function, Ersland et al. found that there was a strong association between the RORB gene and a test of verbal intelligence [19]. Consistent with this notion, Acquaah-Mensah et al. reported that the expression of the RORA gene is distinctly up-regulated in the hippocampi of AD-affected postmortem human brains, a brain region key to memory and learning, indicating a potential link  between RORA and AD [20]. In terms of brain cognitive functions such as mood, the RORA gene was reported to increase the risk of acquiring psychiatric and neurological disorders, including bipolar disorder [6], autism spectrum disorder [21], and post-traumatic stress disorder [22]. Similarly, the RORB gene was also previously linked to psychiatric and neurological disorders, such as bipolar disorder [6,[23][24] and schizophrenia [23], in independent association studies.
In addition, an intriguing finding was a positive association of cognitive aging with 19 SNPs in the NPAS2 gene, especially rs17655330 and rs62152925. The NPAS2 gene, located on chromosome 2q11.2, encodes a member of the basic helix-loop-helix family of transcription factors. The functions of the NPAS2 and CLOCK proteins are partially redundant [25]. The ARNTL/CLOCK and ARNTL/NPAS2 heterodimeric proteins bind to chromatin, resulting in the up-regulation of CRY1, CRY2, PER1, Abbreviations: A1 = minor allele, A2 = major allele, BETA = Beta coefficients, Chr = chromosome, MAF = minor allele frequency, MMSE = Mini-Mental State Examination, SE = standard error. Analysis was obtained after adjustment for covariates including age, gender, and education. www.impactjournals.com/oncotarget PER2, and PER3 gene expression [5]. Based on a series of behavioral tests, Garcia et al. found that Npas2-deficient mice may have impaired brain function and that NPAS2 may activate the neuronal gene expression required for the acquisition of long-term memory, indicating its importance for the execution of complex cognitive tasks [9]. On another note, we also observed that there was an association of cognitive aging with 3 SNPs in the CLOCK gene, particularly CLOCK rs3749473. The CLOCK gene is located on chromosome 4q12 and encodes a basic helix-loop-helix protein that constitutes the ARNTL/ CLOCK heterodimeric protein with ARNTL, another basic helix-loop-helix protein [26]. Evidence has also been reported for the association of three separate SNPs of the CLOCK gene (rs4580704, rs1554483, and 3111) with AD in Chinese populations [13][14][15]; however, these findings have not been replicated by other large AD genetics consortiums. Additionally, Kondratova et al. reported that mice with mutations in the Clock gene showed declines in cognitive performance, including increased rearing activity and impaired intersession habituation [7]. Taken together, these studies suggest the engagement of the CLOCK protein in essential processes for memory formation [7].
Remarkably, another intriguing finding was that we further inferred the epistatic effects between CLOCK, NPAS2, RORA, and RORB in influencing cognitive aging by using the GMDR approach. To our knowledge, no other study has been conducted to evaluate gene-gene interactions between these genes. The functional relevance of the interactive effects between CLOCK, NPAS2, RORA, and RORB on cognitive aging remains to be elucidated. At the molecular level, the circadian clock consists of an interlocking network of several positive and negative Abbreviations: GMDR = generalized multifactor dimensionality reduction. P value was based on 1,000 permutations. Analysis was obtained after adjustment for covariates including age, gender, and education. Abbreviations: GMDR = generalized multifactor dimensionality reduction. P value was based on 1,000 permutations. Analysis was obtained after adjustment for covariates including age, gender, and education. www.impactjournals.com/oncotarget feedback loops [27]. While the positive control of the circadian oscillator in the core loop involves ARNTL, CLOCK, and NPAS2 proteins, the negative control mechanism in the core loop comprises CRY1, CRY2, PER1, PER2, and PER3 proteins [27]. Additionally, an accessory pathway involving the RORs and NR1D1 nuclear receptors further stimulates the core loop [27]. Previous animal studies characterizing Clock/Clock mutant mice demonstrated that CLOCK is not indispensable for circadian rhythms, suggesting that NPAS2, as a substitute for CLOCK, can contribute to circadian rhythms in the absence of CLOCK [25,28]. By reporter gene and mutation analysis, Takeda et al. also suggested that the RORs nuclear receptors are involved in the circadian regulation of the transcription of circadian clock genes such as the Arntl, Cry1, Nr1d1, and Per2 genes [27]. Furthermore, Lai et al. found that there were potential interaction effects among the RORA, RORB, and NR1D1 genes associated with an increased risk of bipolar disorder by using the multifactor dimensionality reduction method  [6]. These studies suggest that the possible mechanisms of joint actions between these genes may synergistically incorporate other relevant circadian clock genes.
In the GMDR analysis of gene-environment interactions, we tracked down the interplay between the CLOCK gene and environmental factors (namely smoking, alcohol consumption, and social support), between the RORA gene and environmental factors (namely smoking, alcohol consumption, physical activity, and social support), and between the RORB gene and environmental factors ( namely alcohol consumption and physical activity). This interplay may manifest itself functionally through epigenetic changes. In previous animal studies, circadian deregulation was shown to affect epigenetic parameters such as DNA methylation, histone modifications, and miRNAs, suggesting that epigenetic alterations might be linked with age-related cognitive decline [29][30][31]. It has also been pointed out that core circadian clock genes may activate epigenetic modifications which are involved in circadian rhythm dysfunction with advancing age [31].
In this study, we further found evidence of potential association (P < 0.05) between cognitive aging and four other genes, the ARNTL, CRY2, PER2, and PER3 genes, which are located on chromosomes 11p15, 11p11.2, 2q37.3, and 1p36.23, respectively. The ARNTL gene encodes a basic helix-loop-helix protein that forms the ARNTL/CLOCK heterodimeric protein with CLOCK [26]. The CRY2 gene encodes a flavin adenine dinucleotide-binding protein that stimulates the circadian clock [8]. The PER2 and PER3 genes are members of the Period family, which encodes components of the circadian rhythms of locomotor activity, metabolism, and feeding behavior [10]. An association of AD with the rs2278749 SNP in the ARNTL gene in a Chinese population was previously observed [16], although this biomarker has not been identified by other large AD genetics consortiums. Contrary to our findings, Pereira et al. found no association of AD with SNPs in the PER2 and PER3 genes in a Brazilian population [32]. To our knowledge, no other study has been conducted to evaluate cognitive aging with SNPs in the CRY2 gene. It is worth mentioning that multiple potential defects underly the discordant results found among these studies, including sample size, sample deviation, environmental control, study design, covariate adjustment, phenotype definitions, population stratification, and various ethnicities [33][34][35][36].
Our analysis indicated no association of cognitive aging with three other genes, the CRY1, NR1D1, and PER1 genes, which are located on chromosomes 12q23-q24.1, 17q11.2, and 17p13.1, respectively. Like the CRY2 gene, the CRY1 gene encodes a flavin adenine dinucleotidebinding protein that stimulates the circadian clock [8]. The NR1D1 gene encodes a member of the nuclear receptor subfamily 1 that negatively activates the expression of core clock proteins [6]. The PER1 gene is a member of the same Period family as the PER2 and PER3 genes, which encodes components of the circadian rhythms of locomotor activity, metabolism, and feeding behavior [10]. To our knowledge, no other study has been conducted to evaluate the association between cognitive aging and SNPs in the CRY1, NR1D1, and PER1 genes.
The MMSE was chosen to evaluate cognitive function, because it is well-established and is the most widely used screening test of cognition. Nonetheless, MMSE reduces variability in the data because it has a floor effect in the oldest adults and a ceiling effect in healthy young adults [37]. The General Practitioner Assessment of Cognition (GPCOG) was also found to have psychometric properties similar to the MMSE, but the GPCOG needs to be further assessed for its potential cultural or language bias [38,39]. As a visual test, the Cambridge Neuropsychological Test Automated Battery (CANTAB) is a computer-based cognitive assessment system that is language independent [40]. However, aspects of CANTAB still need to be validated, and only modest correlations between CANTAB and traditional neuropsychological tests such as MMSE and GPCOG have been demonstrated [41]. Additionally, a well-validated scale in cognitive performance is the Alzheimer's Disease Assessment Scale -Cognitive section (ADAS-Cog), where a four-point change on ADAS-Cog has been established as a clinically important change in cognition [42]. Unfortunately, the length of administration time makes ADAS-Cog unsuitable for clinical practice because ADAS-Cog takes around 40 minutes to perform [37].
This study has both strengths and limitations. The main weakness is that our observations require much further research to determine whether the present research findings are sustained in diverse ethnic groups [34,43]. To support the statistical analysis results, it is also desirable to have some additional evidence of biological functions, because the reported SNPs that influence the expression of the genes of interest have been shown to be importantly enriched for association studies [44]. In future work, prospective clinical trials with other ethnic populations would provide a comprehensive evaluation of the associations and interactions of the investigated genes with cognitive aging [35,45,46]. On the other hand, a major strength of our study is that we employed environmental data, which provided the opportunity to study the interplay between the investigated genes and environmental factors.

CONCLUSIONS
In conclusion, we conducted an extensive analysis of the association as well as gene-gene and gene-environment interactions of the circadian clock genes with cognitive aging in old Taiwanese subjects. Overall, results from the current study revealed that the CLOCK, NPAS2, RORA, and RORB genes may affect the prevalence of cognitive aging independently and/or through complex gene-gene www.impactjournals.com/oncotarget and gene-environment interactions. Future independent replication studies with larger numbers of subjects will likely lead to further insights into the roles of the circadian clock genes described in this study.

Study population
This study incorporated subjects from the Taiwan Biobank [47][48][49]. The study cohort consisted of 634 participants. Recruitment and sample collection procedures were approved by the Internal Review Board of the Taiwan Biobank before conducting the study. Each subject signed the approved informed consent form. All experiments were performed in accordance with relevant guidelines and regulations.
Education was defined based on whether or not high school was attended. Current alcohol drinker was defined as currently drinking 150 ml of alcohol per week for more than six months. Current smoker was defined as currently smoking for more than six months. Physical activity was defined by the amount of exercise activity exceeding three times and more than 30 minutes each time in a week. Social support was assessed based on marital status and whether or not the subject lived alone.

Cognitive assessment
Global cognitive assessment was performed using the 30-point Mini-Mental State Examination (MMSE), which includes questions based on the five domains of orientation, registration, attention and calculation, recall, and language. We analyzed MMSE as a continuous outcome, as well as according to categories based on previously defined MMSE thresholds [37]: MMSE score ≥ 24 (normal) and MMSE score < 24 (cognitive impairment).

Genotyping
DNA was isolated from blood samples using a QIAamp DNA blood kit following the manufacturer's instructions (Qiagen, Valencia, CA, USA). The quality of the isolated genomic DNA was evaluated using agarose gel electrophoresis, and the quantity was determined by spectrophotometry [50,51]. SNP genotyping was carried out using the custom Taiwan BioBank chips and run on the Axiom Genome-Wide Array Plate System (Affymetrix, Santa Clara, CA, USA). The SNP panel consisted of variants from the following 11 core circadian clock genes: ARNTL, CLOCK, CRY1, CRY2, NPAS2, NR1D1, PER1, PER2, PER3, RORA, and RORB.

Statistical analysis
In this study, we evaluated the association of the investigated SNP with MMSE scores by a general linear model using age, gender, and education as covariates [36,33]. The genotype frequencies were assessed for Hardy-Weinberg equilibrium using a X 2 goodness-of-fit test with 1 degree of freedom (i.e. the number of genotypes minus the number of alleles). In order to correct for multiple testing, we applied a conservative Bonferroni correction factor for the number of SNPs employed in the analysis. The criterion for significance was set at P < 0.05 for all tests. Data are presented as the mean ± standard deviation.
To investigate gene-gene and gene-environment interactions, we employed the generalized multifactor dimensionality reduction (GMDR) method [52]. We tested two-way up to four-way interactions using 10-fold crossvalidation. The GMDR software provided some output parameters, including the testing accuracy and empirical P values, to assess each selected interaction. Moreover, we provided age, gender, and education as covariates for gene-gene and gene-environment interaction models in our interaction analyses. Permutation testing obtained empirical P values of prediction accuracy as a benchmark based on 1,000 shuffles.