**3. Results**

### *3.1. 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/m2), whereas female VLFC had a higher BMI (25.3 ± 3.4 kg/m2). 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.



VLFC, Very low FC; LFC, Low FC; OFC, Optimal FC. Data are presented as mean ± standard deviation and number (percentage). ANOVA analysis with Tukey post hoc test and Welch's ANOVA with Games–Howell test for adjusting unequal variances. (1) Data were collected from subjects who completed body composition analysis; male: *n* = 725, *n* = 819, *n* = 844; female: *n* = 641, *n* = 739, and *n* = 761. (2) ≥30 min per day.

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.

### *3.2. 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).


**Table 3.** The association between tertiles of FC ratio and prevalence of disease.

All odds ratios (OR) and 95% confidence intervals (CI) were calculated by performing multiple logistic regression.(1) BMI ≥25 kg/m2, 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.

#### *3.3. 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.



*Nutrients* **2022**,*14*,

 478

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, odds ratio adjusted for BMI and the same covariates as obesity.



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/m2, 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.

### *Nutrients* **2022**, *14*, 478

In females, both genotypes of *CLOCK* rs9312661 in the VLFC showed an increased incidence of abdominal obesity compared with the reference group (AA genotype, OR: 2.26, 95% CI: 1.43–3.56, *p* = 0.0005; GA/GG genotype, OR: 2.11, 95% CI: 1.38–3.23, *p* = 0.0005). In addition, under the VLFC condition, *CRY1* rs3741892 had a significantly greater obesity risk than the reference regardless of genotype (GG genotype, OR: 1.60, 95% CI: 1.22–2.10, *p* = 0.0007; GA/GG genotype, OR: 1.76, 95% CI: 1.30–2.38, *p* = 0.0002). Intriguingly, the associations between macronutrient intake patterns and obesity risks were different depending on the genotypes of *CLOCK* rs11932595, *PER2* 2304672, and *CRY1* rs3741892. The major allele homozygous, AA genotype, of rs11932595 in the VLFC had a higher risk of abdominal obesity than the reference group (OR: 1.84, 95% CI: 1.32–2.56, *p* = 0.0003), but not in subjects carrying the minor G allele. Regarding *CRY1* rs3741892, which showed a higher obesity risk in both genotypes, the GG genotype, but not the CG/CC genotype, had a greater incidence of abdominal obesity (OR: 1.90, 95% CI: 1.30–2.76, *p* = 0.0008). Moreover, females with the GG genotype of *PER2* rs2304672 in the VLFC had significantly higher risks of obesity and abdominal obesity compared with the references (OR: 1.49, 95% CI:1.18–1.87, *p* = 0.0007; OR: 1.85, 95% CI 1.35–2.54, *p* = 0.0001 respectively), whereas no differences were detected in minor C allele carriers.

#### *3.4. 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.

**Figure 3.** Relationship between genetic variants and circadian gene regulation. Effect of genetic variants on gene expression levels are shown by expression quantitative trait (eQTL) violin plot. The plot indicates the density distribution of samples in each genotype and number of subjects shown under each genotype. The white line in the box plot (black) shows the median value of the expression at each genotype. Association between rs11932595 and rs93126661 with *CLOCK* expression (**a**), Association between rs2304672 with *PER2* expression (**b**), and rs3741892 with *CRY1* expression (**c**). Data analysis was performed using GTEx Portal and included tissue-specific information provided by the website [54].

### **4. 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 [30,33–35]. 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 (PPAR α) 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 sugges<sup>t</sup> 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 [12,18–20,23]. 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 [62–64]. 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 [69–72], 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 [73–77]. 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 [79–81]. 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 sugges<sup>t</sup> 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.

In conclusion, we investigated Korean macronutrient intake patterns and found associations between the patterns and circadian clock gene variants, and obesity risk. The VLFC pattern was related to higher incidences of obesity and abdominal obesity in females. After the genotype stratification of nine SNPs of circadian genes, the association between the FC ratio and obesity risk differed by the genetic variants of *CLOCK*, *PER2*, and *CRY1*. These findings sugges<sup>t</sup> that the low dietary FC ratio influences obesity susceptibility and the association depends on circadian clock genetic variations. Our findings highlight an important role of the association of macronutrient distribution and circadian clock on obesity.

**Author Contributions:** Conceptualization, Y.J.P. and J.S.; investigation, J.S. and Y.H.; data curation, J.S.; writing—original draft preparation, J.S.; writing—review and editing, Y.J.P.; funding acquisition, Y.J.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by Basic Science Research Programs through the National Research Foundation (NRF) funded by the Korean governmen<sup>t</sup> (2021R1A2C2012578).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Institutional Review Board of Ewha Womans University, Seoul, Korea (IRB approval number: ewha-202105-0003-01).

**Informed Consent Statement:** Written informed consent was waived by the Institutional Review Board due to all personal identifying information being removed from the dataset prior to analysis.

**Data Availability Statement:** The KoGES data are available on request from the National Research Institute of Health [47].

**Acknowledgments:** This study was conducted with bioresources from the National Biobank of Korea, the Korea Disease Control and Prevention Agency, and Korea (KBN-2021-035). J.S. was supported by NRF funded by the Ministry of Education (2020R1A6A3A13075729) and Hyundai Motor Chung Mong-Koo Foundation. Y.H. was supported by Brain Korea Four Project (Education Research Center for 4IR-Based Health Care).

**Conflicts of Interest:** The authors declare no conflict of interest.
