Effects of Dietary Fat to Carbohydrate Ratio on Obesity Risk Depending on Genotypes of Circadian Genes

Although the impacts of macronutrients and the circadian clock on obesity have been reported, the interactions between macronutrient distribution and circadian genes are unclear. The aim of this study was to explore macronutrient intake patterns in the Korean population and associations between the patterns and circadian gene variants and obesity. After applying the criteria, 5343 subjects (51.6% male, mean age 49.4 ± 7.3 years) from the Korean Genome and Epidemiology Study data and nine variants in seven circadian genes were analyzed. We defined macronutrient intake patterns by tertiles of the fat to carbohydrate ratio (FC). The very low FC (VLFC) was associated with a higher risk of obesity than the optimal FC (OFC). After stratification by the genotypes of nine variants, the obesity risk according to the patterns differed by the variants. In the female VLFC, the major homozygous allele of CLOCK rs11932595 and CRY1 rs3741892 had a higher abdominal obesity risk than those in the OFC. The GG genotype of PER2 rs2304672 in the VLFC showed greater risks for obesity and abdominal obesity. In conclusion, these findings suggest that macronutrient intake patterns were associated with obesity susceptibility, and the associations were different depending on the circadian clock genotypes of the CLOCK, PER2, and CRY1 loci.


Introduction
The circadian clock governs 24 h rhythms and regulates the sleep-wake cycle. In mammals, circadian rhythms influence metabolism and physiological processes [1]. Furthermore, the circadian clock regulates glucose and fat metabolism and energy metabolism by coordinating the expression of clock-controlled genes [1,2]. The circadian core genes, including the circadian locomotor output cycle kaput (CLOCK), aryl hydrocarbon receptor nuclear translocator-like (ARNTL, also known as BMAL1), period homolog (PER1, PER2), and cryptochrome (CRY1, CRY2) regulate the circadian rhythm mechanism [1,3]. The ARNTL-CLOCK complex drives the transcription of PER and CRY genes by binding to enhancer elements. Increased proteins of PER and CRY inhibit ARNTL-CLOCK-mediated transcription. This transcription-translation negative feedback loop leads the circadian rhythm, which takes 24 h [3,4].
The master clock located in the hypothalamic suprachiasmatic nucleus can be regulated by the light-dark cycle [1,26,27], whereas peripheral clocks in peripheral tissues, due to missing data ( Figure 1). Exclusion criteria (cancer, dementia, stroke, steroid drugs, insulin therapy, oral diabetes medication, thyroid drugs, and hormone replacement therapy) were applied for the elimination of effects derived from diseases and drugs on food intake. Finally, we investigated 5343 subjects aged 40~64 years, of which 2756 were male (mean age 48.9 ± 7.0 years), and 2587 were female (mean age 49.9 ± 7.6 years). The study was approved by the Institutional Review Board of Ewha Womans University, Seoul, Korea (IRB approval number: ewha-202105-0003-01).

Figure 1.
A flow chart of the study population.

Selection and Analysis of SNPs
Genomic DNA derived from blood samples was genotyped with the Affymetrix Genome-Wide Human SNP Array 5.0 kit (Affymetrix, Inc., Santa Clara, CA, USA) [48], and 1000 genome sequences were used for imputation [49]. After applying the Bayesian Robust Linear Modeling with Mahalanobis Distance (BRLMM) algorithm and standard quality control procedures, samples with a missing call rate >4%, heterozygosity >30%, gender incompatibility, or obtained from subjects who had cancer were excluded [50]. Among 352,228 SNPs, we selected 235 SNPs that were located in the loci of the circadian core genes CLOCK, ARNTL, PER1, PER2, PER3, CRY1, and CRY2 ( Figure 2). SNPs with a high missing genotype call rate (>5%), low minor allele frequency (MAF <0.05), and low Hardy-Weinberg equilibrium (p value < 1 × 10 −6 ) were excluded. We conducted linkage disequilibrium (LD)-based pruning (r 2 > 0.2); one SNP which had the highest MAF was selected from each LD block using PLINK software version 1.09 [51] and Haploview software version 4.1 (Broad Institute of MIT and Harvard, Cambridge, MA, USA) [52]. Utilizing the multitissue expression quantitative loci (eQTL) analysis from the Genotype Tissue Expression (GTEx) projects (release version 8) [53,54], we selected 9 SNPs related to circadian gene regulation (Tables 1 and A1, Figure 2). A recessive model was used for further investigation due to the small number of subjects of homozygous for the minor allele.

Selection and Analysis of SNPs
Genomic DNA derived from blood samples was genotyped with the Affymetrix Genome-Wide Human SNP Array 5.0 kit (Affymetrix, Inc., Santa Clara, CA, USA) [48], and 1000 genome sequences were used for imputation [49]. After applying the Bayesian Robust Linear Modeling with Mahalanobis Distance (BRLMM) algorithm and standard quality control procedures, samples with a missing call rate >4%, heterozygosity >30%, gender incompatibility, or obtained from subjects who had cancer were excluded [50]. Among 352,228 SNPs, we selected 235 SNPs that were located in the loci of the circadian core genes CLOCK, ARNTL, PER1, PER2, PER3, CRY1, and CRY2 ( Figure 2). SNPs with a high missing genotype call rate (>5%), low minor allele frequency (MAF < 0.05), and low Hardy-Weinberg equilibrium (p value < 1 × 10 −6 ) were excluded. We conducted linkage disequilibrium (LD)-based pruning (r 2 > 0.2); one SNP which had the highest MAF was selected from each LD block using PLINK software version 1.09 [51] and Haploview software version 4.1 (Broad Institute of MIT and Harvard, Cambridge, MA, USA) [52]. Utilizing the multitissue expression quantitative loci (eQTL) analysis from the Genotype Tissue Expression (GTEx) projects (release version 8) [53,54], we selected 9 SNPs related to circadian gene regulation (Tables 1 and A1, Figure 2). A recessive model was used for further investigation due to the small number of subjects of homozygous for the minor allele.

Macronutrient Patterns
A validated semi-quantitative food frequency questionnaire with 103 food items was used for assessing dietary data [55]. The consumption frequency and portion size of items during the previous year were investigated. The sum of the nutrient intake from each food item was calculated to evaluate the average daily energy intake and nutrient intake of each individual. Macronutrient (carbohydrate, fat, and protein) intake was presented as the percentage of total energy intake. Given the protein intake was positively correlated

Macronutrient Patterns
A validated semi-quantitative food frequency questionnaire with 103 food items was used for assessing dietary data [55]. The consumption frequency and portion size of items during the previous year were investigated. The sum of the nutrient intake from each food item was calculated to evaluate the average daily energy intake and nutrient intake of each individual. Macronutrient (carbohydrate, fat, and protein) intake was presented as the percentage of total energy intake. Given the protein intake was positively correlated with fat intake in this cohort population (data not shown), we defined fat to carbohydrate ratio (FC ratio) by dividing '% energy from fat' by '% energy from carbohydrate'. Subsequently, subjects were categorized by tertiles of the FC ratio: Very low FC (VLFC; the first tertile), Low FC (LFC; the second tertile), and Optimal FC (OFC; the third tertile).

Definitions of the Obesity and Abdominal Obesity
Anthropometric measurements were obtained (i.e., height, weight, waist circumference) by trained staff in cohort study [47]. In the present study, obesity was defined as a BMI ≥ 25 kg/m 2 according to Asia-Pacific BMI cut-off from the World Health Organization Report [56]. The abdominal obesity was defined as a waist circumference ≥90 cm for males and ≥85 cm for females according to the diagnostic criteria for Korea [57].

Statistical Analysis
Data were presented as the mean ± standard deviation, number, and percentage. ANOVA analysis with Tukey post hoc comparison test was used to identify group differences, and Welch's ANOVA with Games-Howell test was used to adjust for unequal variances. The Chi-square test was used to analyze categorical variables. Multiple logistic regression analysis was used for exploring the associations between genotypes and disease after adjustment for covariates, such as age, body mass index (BMI), sleep duration, alcohol intake, tobacco consumption, physical activity, energy intake, and number of regular meals. Statistical analyses were performed using SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA) and RStudio ver.1.2.1335 (RStudio Inc., Boston, MA, USA). A p-value of <0.05 was considered to be statistically significant. Bonferroni correction was applied to correct for multiple testing (Bonferroni corrected p < 0.011).

General Characteristics and Nutritional Intake
The main characteristics of all the included participants are shown in Appendix B. After dividing subjects into tertiles of the FC ratio, the general characteristics according to groups were analyzed ( Table 2). Subjects in the VLFC group (T1) were older than the LFC group (T2) and the OFC group (T3) (male VLFC: 50.8 ± 7.3 years, LFC: 48.6 ± 6.7 years, and OFC: 47.3 ± 6.4 years; female VLFC: 53.2 ± 7.5 years, LFC: 49.7 ± 7.5 years, and OFC: 46.9 ± 6.3 years). The VLFC showed had a lower BMI than other groups in males (24.7 ± 2.9 kg/m 2 ), whereas female VLFC had a higher BMI (25.3 ± 3.4 kg/m 2 ). In the VLFC group, the portion of rural subjects was greater than other groups (male VLFC: 43.2% and female VLFC: 58.2%). The proportion of urban subjects was highest in the OFC group (male OFC: 84.8% and female OFC 78.0%). The VLFC had a lower lean body mass and body fat than other groups in males (52.7 ± 5.8 kg and 15.1 ± 4.7 kg, respectively). In contrast, female VLFC had a lower lean body mass (39.8 ± 4.6 kg) and higher body fat (15.7 ± 4.9 kg). Furthermore, the female VLFC showed a higher waist to hip ratio (0.91 ± 0.05) compared to other groups.
The nutritional intake including total energy, carbohydrate, protein, and fat was highest in the OFC group and lowest in the VLFC group. However, carbohydrate intake did not differ by FC group in females. The VLFC group had a significantly higher % of energy from carbohydrate intake (78.2 ± 3.0% in females) and consequently a lower % of energy from protein and fat (11.7 ± 1.4% and 8.6 ± 2.0% in females, respectively) than in other groups (Appendix C). Considering that the Korean Acceptable Macronutrient Distribution Range (AMDR) for carbohydrate is 55~65%, for protein is 7~20%, and for fat is 15~30% of the energy intake for adults [58], the OFC group's proportion fitted the Korean AMDR. In contrast, the VLFC and LFC had an inadequate composition of macronutrients, which fell outside the AMDR with a higher carbohydrate and lower fat intake. Because the OFC had a macronutrionally balanced diet with optimal proportions, we designated the OFC as the reference group in our further analysis. The FC ratio was 0.14 ± 0.03, 0.23 ± 0.02, and 0.34 ± 0.08 for male VLFC, LFC, and OFC respectively; and 0.11 ± 0.03, 0.19 ± 0.02, and 0.31 ± 0.09 for female VLFC, LFC, and OFC respectively.

Risk of Obesity by Macronutrient Intake Patterns
The prevalence of disease according to the tertiles of the FC ratio is shown in Table 3. In males, the LFC group had an increased risk of obesity (odds ratio (OR): 1.29, 95% confidence interval (CI): 1.07-1.57) compared with the OFC group. There was no effect of patterns on the incidence of abdominal obesity in males. Interestingly, in females, the VLFC group showed greater odds of obesity and abdominal obesity than in the OFC group (OR: 1.50, 95% CI:1.20-1.86; OR: 1.84, 95% CI 1.36-2.48, respectively). All odds ratios (OR) and 95% confidence intervals (CI) were calculated by performing multiple logistic regression.
(1) BMI ≥25 kg/m 2 , odds ratio adjusted for age, sleep duration, energy intake, number of regular meals, alcohol intake, tobacco consumption, and moderate physical activity. (2) Waist circumference ≥90 cm for males and ≥85 cm for females, odds ratio adjusted for age, BMI, sleep duration, energy intake, number of regular meals, alcohol intake, tobacco consumption, and moderate physical activity.

Macronutrient Intake Patterns, Genetic Variants, and Risk of Obesity
To investigate the association of macronutrient composition and genetic variations of circadian clock genes, we stratified subjects by the genotypes of nine SNPs and analyzed the risk of obesity (Tables 4 and 5). The homozygous major allele of each SNP in the OFC was used as the reference group in the regression analysis, and the Bonferroni adjustment was used for multiple testing correction.
The risk of disease was increased in the VLFC group, particularly in females (Table 5). In the male VLFC group, the minor allele carriers of CLOCK rs9312661, CRY2 rs7951225, and the GG genotype of CRY1 rs11113192 showed increased risks of obesity; however, significances were diminished after the Bonferroni correction (Table 4). An interaction between CRY1 rs11113192 and the FC on obesity was observed (p-interaction = 0.009); however, the significance disappeared after multiple corrections. No statistically significant differences were found for abdominal obesity.  All odds ratios and 95% confidence intervals were calculated by performing multiple logistic regression. p-interaction: interaction between SNP and FC tertiles. Data in bold indicate statistically significant value after Bonferroni correction for multiple comparisons (corrected p-value: 0.05/45 = 0.001). (1) BMI ≥ 25 kg/m 2 , odds ratio adjusted for age, sleep duration, energy intake, number of regular meals, alcohol intake, tobacco consumption, and moderate physical activity. (2) Waist circumference ≥85 cm for females, odds ratio adjusted for BMI and the same covariates as obesity.

Potential Links between Genetic Variants and Gene Regulation
To explore the potential role of genetic variants on circadian gene regulation, we conducted an eQTL analysis at the SNP selection step. The four SNPs (rs11932595, rs9633835, rs2304672, and rs3741892), which had association with macronutrient intake patterns and obesity risk, contributed to gene expression in various tissues involved in metabolism (Appendix A). For instance, the genotypes of rs11932595 and rs9312661 influence CLOCK gene expression in the skeletal muscle, small intestine, colon, pancreas, and subcutaneous adipose tissue (Figure 3). Moreover, thyroidal PER2 expression is impacted by rs2304672 genotypes, and the CRY1 expression of the skeletal muscle is affected by rs3741892. Interestingly, the GG genotype of PER2 rs2304672, which had a significantly increased risk of obesity in our results (Table 4), showed lower expression levels than C carriers (CC genotype: not found in the eQTL violin plot analysis, but a small portion of subjects were present in our data; n = 8 males and n = 12 females). These findings indicate that genetic variants might influence circadian gene expression levels in important metabolic tissues.
Nutrients 2022, 14, x FOR PEER REVIEW 9 To explore the potential role of genetic variants on circadian gene regulation, we ducted an eQTL analysis at the SNP selection step. The four SNPs (rs11932595, rs9633 rs2304672, and rs3741892), which had association with macronutrient intake patterns obesity risk, contributed to gene expression in various tissues involved in metabo (Appendix A). For instance, the genotypes of rs11932595 and rs9312661 influence CL gene expression in the skeletal muscle, small intestine, colon, pancreas, and subcutan adipose tissue (Figure 3). Moreover, thyroidal PER2 expression is impacted by rs230 genotypes, and the CRY1 expression of the skeletal muscle is affected by rs3741892. I estingly, the GG genotype of PER2 rs2304672, which had a significantly increased ri obesity in our results (Table 4), showed lower expression levels than C carriers (CC g type: not found in the eQTL violin plot analysis, but a small portion of subjects were sent in our data; n = 8 males and n = 12 females). These findings indicate that genetic iants might influence circadian gene expression levels in important metabolic tissues

Discussion
In the present study, we explored macronutrient intake patterns in a Korean mi population and observed associations between patterns and circadian clock gene var and obesity. A categorization of the three patterns by the FC ratio revealed the high bohydrate and relatively low-fat intake of subjects. The prevalence of obesity and dominal obesity increased in the VLFC compared to the OFC in females. After strati

Discussion
In the present study, we explored macronutrient intake patterns in a Korean midlife population and observed associations between patterns and circadian clock gene variants and obesity. A categorization of the three patterns by the FC ratio revealed the high carbohydrate and relatively low-fat intake of subjects. The prevalence of obesity and abdominal obesity increased in the VLFC compared to the OFC in females. After stratification by the genotypes of nine SNPs, the obesity risk according to the patterns was different according to the genetic variants of CLOCK, PER2, and CRY1. In the VLFC pattern, the major allele homozygous genotype of rs11932595, rs3741892, and rs2304672 had greater risks of obesity and abdominal obesity than the reference group, whereas minor allele carriers had no difference in risk. These findings indicate that macronutrient intake patterns were associated with obesity susceptibility, and the associations were dependent on circadian clock genetic variants, particularly in females. To the best of our knowledge, this is the first study to investigate the roles of dietary macronutrient distribution and circadian clock genes in disease risk in the Korean population.
Dietary macronutrients induced alterations of circadian clock gene expression and phase shift in tissues [30,[33][34][35]. The substitution of dietary components induced phase shifts of the hepatic circadian clock [35]. A high-fat diet altered the expression of circadian clock genes in the liver and adipose and, consequently, induced changes in the periods of circadian rhythms with advanced phase [30,32,33]. Mice fed a high-fat diet for 10 weeks revealed the reprogramming of the liver clock through the alternative oscillation of transcripts and metabolites in the liver [34]. The molecular mechanisms of reprogramming induced by high fat are the impairment of CLOCK:BMAL1 chromatin recruitment and a newly oscillating pattern of the peroxisome proliferator-activated receptor gamma (PPARγ), a nuclear receptor involved in glucose and lipid metabolism. The ketogenic diet, which consists of high fats and low carbohydrates, promotes BMAL1 chromatin recruitment in the liver and induces the tissue-specific oscillation of the peroxisome proliferator-activated receptor alpha (PPARα) and its target genes [36]. In a human study, the regulation of dietary fat and carbohydrate content altered the oscillations of peripheral clock genes and inflammatory genes [59]. A high-protein diet affected the expression of circadian genes and key gluconeogenic genes phosphoenolpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase) in liver and kidney [37]. Therefore, interactions between dietary macronutrient distribution and circadian clock genes might influence downstream clock-controlled genes, leading to changes in metabolic outcomes. In this study, we identified macronutrient intake patterns in a Korean population and observed that the VLFC pattern was associated with increased risks of obesity and abdominal obesity. Moreover, this association was dependent on circadian genetic variants of CLOCK, PER2, and CRY1. Thus, these results suggest that the identification of patterns of dietary macronutrient distribution and understanding the effects of interactions between patterns and circadian genes are essential for the prevention of obesity.
To investigate the potential contribution of genetic variants to gene regulation, we selected nine SNPs by eQTL analysis. The eQTL from the GTEx portal uncovered genetic variants, including SNPs, that influenced differential levels of gene expression [53]. In the GTEx portal, tissue-specific gene expression and SNPs associations were investigated across all 49 human tissues. A combination of eQTL and SNP is useful for the comprehensive exploration of genetic effects on phenotypic variation and disease [60]. One study, which investigated disease-associated SNPs by applying an eQTL analysis, showed that several SNPs regulated gene expression levels in a tissue-specific manner, for example, the IRS1 gene in adipose tissue and influenced the risk of obesity and type 2 diabetes [61]. Rs1801260, a CLOCK polymorphism, has a role in the development of obesity, diabetes, and metabolic syndrome [12,[18][19][20]23]. In a Korean population study, which used the same cohort data as our research but utilized a different genotype array chip, CLOCK rs1801260 affected the incidence of metabolic syndrome, and the association was more apparent after the stratification of monounsaturated fatty acid intake [22]. Moreover, the haplotype of three SNPs (rs1801260-rs11932595-rs4580704) influenced the risks of overweight and hyperglycemia. Considering the eQTL information of rs1801260 and rs11932595 was related to the differential expression of CLOCK in various tissues, these results imply that circadian genetic variants might regulate circadian genes as well as clock-controlled genes, resulting in different metabolic phenotypes. Having investigated the effects of genetic variants and macronutrient patterns on obesity risk, we found four significant SNPs. According to the eQTL analysis, the four SNPs influenced gene expression in various tissues (Appendix A). Genetic variants of CLOCK, PER2, and CRY1 are associated with gene expression in muscle, adipose, and thyroid, which are known to regulate metabolism. In particular, the rs2304672 genotypes showed differential PER2 expression levels, which were lower in the GG genotype compared with the GC genotype. PER2 rs2304672 genetic variants were previously associated with psychiatric disorders including bipolar disorder, depression, and diurnal preference [62][63][64]. Two studies reported that the G allele of rs2304672 had morning preference [64,65], but no significance was found in a young Korean population [66]. In overweight/obese participants on a weight-reduction program, the G allele carriers of rs2304672 showed a lower waist to hip ratio values but had a greater probability of dropping out from the program with constant snacking and skipping breakfast than the CC genotype [21]. Moreover, the interactions between rs2304672 and plasma fatty acids on the modulation of lipoprotein-related biomarkers were reported [67]. Among metabolic syndrome patients with high plasma saturated fatty acid levels, the G allele carriers had higher plasma triglycerides, apolipoprotein C, and apolipoprotein B-48 concentrations than the CC genotype. Given that PER2 also interacts with nuclear receptors including PPARα and can regulate the expression of nuclear receptor target genes involved in lipid metabolism, PER2 polymorphisms could contribute to metabolic disorder vulnerability [68]. In addition, rs2304672, which is located in the 5 untranslated region of the PER2 gene, was suggested to alter the secondary structure of the transcript or change the folding of PER2 mRNA, resulting in differential translation levels or functionality of proteins between the genotypes [64,67]. Although the mechanisms underlying disease susceptibility is not fully understood, these results support an important role of PER2 genetic variants on obesity by regulating circadian gene expressions and functions. Further analysis is required to investigate the gene regulatory mechanisms of these SNPs.
We displayed distributions of Korean macronutrient intake patterns by the FC ratio stratification (Appendix C). The notable features in our study were a high proportion of carbohydrate intake and a positive correlation between protein and fat intake. The VLFC group, which had a low fat to carbohydrate ratio, had the highest carbohydrate intake and relatively low intake level of fat and protein. In contrast, the OFC group had a lower carbohydrate intake and increased fat and protein intake than the VLFC. Moreover, the OFC group had a balanced distribution with appropriate proportions of macronutrients that met the Korean AMDR.
The dietary intake proportion differed across populations. Western diets are characterized as having a high dietary level of saturated fats and refined carbohydrates and low levels of fiber. Previous studies have reported the effects of conventional dietary approach which applied a low-carbohydrate or low-fat diet to weight loss and improvement of obesity [38,69]. The types of intervention diets usually suggested for controlling weight can be categorized into three types: low-carbohydrate, low-fat, and moderate macronutrients [38]. Low-carbohydrate diets including Atkins and Zone diets contain 15~40% energy from carbohydrates, 30% energy from proteins, and 30~55% energy from fats. The low-fat diet is composed of 60~70% of energy from carbohydrates, 10~15% from proteins, and 10~20% from fats. In addition, a high-protein, low-fat diet had positive effects on body weight loss and metabolic benefits [69][70][71][72], providing 44%, 31%, and 25% of energy from carbohydrates, proteins, and fats, respectively. These results imply that previously utilized intervention diets are designed for western-style macronutrient distribution. For instance, there is a large difference in distribution between 'low-carbohydrate diets' or 'high-protein and low-fat' diets and Asian populations who have a much higher carbohydrate intake.
Although accumulating evidence supports the contribution of dietary macronutrient distribution to the development and prevention of metabolic diseases, the relationship between macronutrients and metabolic benefit is still controversial. Several research groups demonstrated that a low-carbohydrate diet is more effective at reducing weight, fat mass, and serum triglycerides and improving metabolic syndrome than a low-fat diet [73][74][75][76][77]. In contrast, other results showed both diets led to similar effects on weight control or clinical markers including glucose level, lipid profile, and blood pressure [40,74]. A metaanalysis study comparing 14 popular dietary programs found that most diets reduced weight and improved blood pressure at 6 months; however, the effects disappeared at 12 months [38]. One issue to consider is that previously conducted intervention diets modifying macronutrient distribution were usually based on energy restriction and have a short-term design. However, there were mouse studies with diets varying in protein to carbohydrate ratio, which examined the interactive effects of dietary macronutrient distribution and metabolic outcomes under ad libitum conditions [42,78]. Short-term 'high-protein and low-carbohydrate' diets decreased insulin sensitivity, impaired glucose tolerance, and increased triglycerides, resulting in metabolic dysregulation [78]. In contrast, 'low-protein and high-carbohydrate' diets prevented adiposity gain and improved metabolic health including insulin, glucose, and lipid levels, despite increased energy intake. As a result of chronic feeding over a lifetime in mice, 'high-protein and low-carbohydrate' diets reduced food intake and adiposity; however, they caused negative outcomes in metabolic health and shortened longevity [42]. Long-term 'low-protein and high-carbohydrate' diets increased food intake, body weight, and adiposity, but there were positive impacts on health and a longer lifespan, possibly through the regulation of mammalian target of rapamycin (mTORC1) activation [42].
Low-carbohydrate diets replaces carbohydrates with proteins or fats, a typical example is a ketogenic diet. The metabolic benefits of the low carbohydrate diets are inconsistent. Low-carbohydrate diets with increased fat or protein have been reported to be effective for weight loss and improving the lipid profile [39,75,76]. A meta-analysis comparing 'low-carbohydrate, high-fat' and 'high-carbohydrate, low-fat' diets found that the lowcarbohydrate diet had a greater effect on weight loss than the high-carbohydrate diet, but no differences were observed for fat mass, glucose, and triglyceride levels, and blood pressure [41]. Results from prospective cohort studies, which investigated the effect of longterm dietary macronutrient distribution without calorie restriction, reported an association between low-carbohydrate intake and increased mortality [79][80][81]. Conversely, multinational and Asian studies have suggested that a high-carbohydrate intake contributed to increased mortality [82,83]. Interestingly, in a large prospective cohort study with a 25-year follow-up, midlife participants who had low (<40%) or high (>70%) energy from carbohydrate consumption were associated with increased mortality [84]. Moreover, those with a 50~55% carbohydrate intake showed the greatest lifespan, a level that might be considered moderate in the West but low in Asia. These conflicting results suggest the fact that the effects of macronutrient challenge in the short term, or energy restriction conditions might be different to those under long-term dietary intake and free-living individual conditions. Although our study analyzed multiple variants of circadian core clock genes in Korean population cohort data, there were some limitations. The SNPs from the genomic data of the cohort did not cover the full list of variants, resulting in missing SNPs reported in previous studies. Therefore, the analysis of comprehensive genetic variant data including crucial variants will provide additional important SNPs. Secondly, our study analyzed local community-based cohort data because of the availability of genomic data. To confirm these findings, futures studies based on a national representative cohort study with a larger sample size are required. Third, even though we included the covariates (i.e., age, BMI, and energy intake) for adjustment in a statistical analysis process, the possibility of effects induced by potential confounding factors, such as residential area, socioeconomic position, and health-related behaviors, should be considered.