Racial Disparities in Methylation of NRF1, FTO, and LEPR Gene in Childhood Obesity

Childhood obesity has affected the health of millions of children around the world despite vigorous efforts by health experts. The obesity epidemic in the United States has disproportionately afflicted certain racial and ethnic minority groups. African American children are more likely than other children to have obesity-related risk factors such as hyperlipidemia, diabetes, cardiovascular disease, and coronavirus disease (COVID-19). For the reduction in obesity-related health inequalities to be successful, it is essential to identify the variables affecting various groups. A notable advancement in epigenetic biology has been made over the past decade. Epigenetic changes like DNA methylation impact on many genes associated with obesity. Here, we evaluated the DNA methylation levels of the genes NRF1, FTO, and LEPR from the saliva of children using real-time quantitative PCR-based multiplex MethyLight technology. ALU was used as a reference gene, and the Percent of Methylated Reference (PMR) was calculated for each sample. European American children showed a significant increase in PMR of NRF1 and FTO in overweight/obese participants compared to normal weight, but not in African American children. After adjusting for maternal education and annual family income by regression analysis, the PMR of NRF1 and FTO was significantly associated with BMI z-score only in European American children. While for the gene LEPR, African American children had higher methylation in normal weight participants as compared to overweight/obese and no methylation difference in European American children. The PMR of LEPR was significantly negative associated with the obesity measures only in African American children. These findings contribute to a race-specific link between NRF1, FTO, and LEPR gene methylation and childhood obesity.


Introduction
Type 2 diabetes, heart disease, dyslipidemia, hypertension, hepatic steatosis, and metabolic syndrome are among comorbidities that can arise because of inappropriate or excessive fat deposition in the body [1]. The frequency of childhood obesity has been rising alarmingly over the past 40 years. In the United States, childhood obesity remains an epidemic affecting the health of millions of children. According to the Centers for Disease Control and Prevention (CDC), the prevalence of obesity was 12.7% of children aged 2 to 5 years, 20.7% of children aged 6 to 11 years, and 22.2% of children aged 12 to 19 years were obese in 2017-2020 [2]. Alabama is fifth highest in the nation for the prevalence of obesity with around 22% of children and adolescents aged 10 to 17 years being overweight and obese [3]. When compared to other ethnic groups, some have a greater prevalence of obesity, as is the case with African American children, where obesity continues to be a major public health issue [4]. To successfully lower obesity-related health inequalities it is essential to identify the variables that affect various populations [5].

Study Participants
In all, 113 research participants between the ages of 6 and 10 (8.57 ± 0.13) years were recruited from Lee and Macon counties in Alabama. Following a preliminary phone interview with the parents, children with a history of diabetes or cardiovascular disease were eliminated. Parents and participants both provided their written approval. Parents brought their kids to Auburn University so that their anthropometric measurements and saliva samples were collected. The study was approved by the institutional review board of Auburn University (Protocol # 17-364 MR 1709).

Anthropometric Measurements
The World Health Organization (WHO) guidelines were followed for taking the participants' anthropometric measures. The children's body weight was measured on a Tanita digital scale to the closest 4 ounces, without shoes, and in light clothes. The height was measured with a calibrated scale that was connected to a stadiometer, and the accuracy was 0.1 cm. Children have different ratios of bone, muscle, and fat as they become older. Therefore, the BMI z-score (Body Mass Index) is a more accurate way to assess a developing child's weight than the BMI alone. Using the SPSS macro, the BMI z-score was computed using the WHO growth references adjusted for age and sex [33]. The classification for children was based on the CDC standards as underweight (<5th percentile), normal weight (≥5th percentile to ≤85th percentile), overweight (≥85th percentile to ≤95th percentile), and obese (≥95th percentile). The waist circumference was measured at the midpoint between the lower ribs and iliac crest using a non-elastic tape to the nearest 0.1 cm. R macro package was used to calculate the z-score of waist circumference (WC) and waist:height ratio (WHtR) [34].

Isolation of Salivary DNA
Saliva was collected using the Oragene Geno-Tek saliva collection kit (Catalog # OGR-500; Ottawa, ON, Canada). As per the manufacturer's instructions, saliva samples were incubated in the water bath at 50 • C for 3 h. A 500 µL aliquot was used to isolate DNA using the PrepIT.L2P DNA isolation kit (Catalog # PT-L2P-5; DNA genotek, Ottawa, ON, Canada). Each sample was labeled and stored at −20 • C until further use. Isolated DNA was quantified, and the quality was checked using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Inc., Wilmington, DE, USA), which specifically measured the double-stranded DNA.

Bisulfite Conversion
The bisulfite conversion of DNA leads to the deamination of unmethylated cytosines, which are converted to uracil and subsequently to thymine in the subsequent PCR, while the cytosines that are methylated remain unchanged and this makes it possible to analyze the methylation pattern of the DNA sequence. The bisulfite conversion of non-methylated cytosines to uracil was done using the EpiTect Fast DNA Bisulfite kit (Catalog # 59720; QIAGEN, Germany). The reaction consisted of bisulfite solution, buffer, 1 ng of gDNA, and RNase-free water to make the volume up to 140 µL. Thermo Fisher Quantstudio 3 was used to run the thermal cycle under the following condition: 95 • C for 5 min, 60 • C for 10 min, 95 • C for 5 min, 60 • C for 10 min, and 20 • C for 10 min. After the bisulfite conversion, DNA was cleaned up by following the protocol provided by the manufacturer. The product was eluted using 15 µL of elution buffer and stored at −80 • C until used. Once the bisulfite conversion is done, the gDNA behaves like RNA, further quantified using nanodrop, which specifically measured the ssRNA. All samples were normalized to 5 ng/µL per nanodrop quantification before the multiplex MethyLight RT-PCR reaction.

Multiplex MethyLight Primer and Probe Design
The Multiplex MethyLight assay was carried out using two primers and a TaqMan probe for each gene. Primers and Probes were specifically designed for bisulfite-converted DNA, which can amplify methylated DNA using Beacon Designer 8.21 (Premier Biosoft International, Palo Alto, CA, USA) for three genes: NRF1, FTO, LEPR, and ALU. ALU was used as a reference gene in every well to normalize the input DNA. The primer and probe sequences for all the genes are listed in Table 1.

MethyLight RT-PCR Reaction
MethyLight is a TaqMan-based qPCR method that is extremely sensitive and depends on the hybridization and cleavage of probes that are intended to target the CpG of interest [35]. Using Quantstudio 3, fluorescence-based real-time quantitative PCR was used to amplify bisulfite converted DNA after sodium bisulfite conversion in a 96-well plate. The PCR amplification was carried out using the Epitect MethyLight PCR + ROX vial kit (Qiagen GmbH; catalog # 59496) according to the manufacturer's instructions. The final volume of the reaction was 20 µL, and it contained 400 nM of each primer, 250 nM of each probe, 10 µL of the MethyLight master mix (HotStar Taq Plus DNA Polymerase, Epitect Probe PCR Buffer, dNTP mix-dATP, dCTP, dGTP, dTTP), 0.4 µL ROX dye, RNase free water and 10 ng of bisulfite converted DNA. There were 45 cycles of the following cycling conditions: 95 • C for 5 min, 95 • C for 15 s, and 60 • C for 60 s. In addition to gDNA and the No Template Control (NTC), universally methylated and non-methylated standards were employed as a control in each plate. Additionally, commercially available human bisulfite-converted DNA (EpiTect PCR Control DNA; catalog # 59655; Qiagen GmbH) was utilized as a completely methylated control for the computation of the PMR. The PMR is a relative value of methylation in each sample as compared to the fully methylated control. PCR primers surrounding an oligonucleotide probe with a 5 fluorescent reporter dye and a 3 quencher dye are used to amplify bisulfite-converted genomic DNA. Taq DNA polymerase's 5 to 3 nuclease activity cleaves the probe and releases the reporter, whose fluorescence can be measured. The PCR amplification produces a fluorescent signal proportional to the PCR product created after crossing a fluorescence detection threshold. The cycle number at which the fluorescent signal passes a threshold in the exponential phase of the PCR reaction may be used to calculate the initial template amount.

Statistical Analysis
The PMR of genes NRF1, FTO, and LEPR for each sample was calculated using the formula PMR = 100 × 2 −∆∆Ct , where ∆∆Ct = [∆Ct of sample-∆Ct of universal methylated DNA]. The calculations for PMR values were done using Microsoft Excel. IBM SPSS Statistics 25.0 was used to do independent samples t-test to see the difference between the PMR of NW and OW/OB participants. Normality tests were performed to confirm the normal distribution of the data. To analyze the effect of maternal education and annual family income on the PMR of NRF1, FTO, and LEPR hierarchical regression was performed. The value p < 0.05 was considered to indicate a statistically significant difference. Linear regression analysis was performed to determine the association between the logtransformed PMR values of each gene and BMI z-score, WC z-score, and WHtR z-score. A Receiver Operating Characteristics (ROC) curve was generated using the PMR values of NRF1, FTO, LEPR, and appropriate cut-off values for BMI categories and race were calculated for the genes using the SPSS software.

Results
The study participants included 60 NW and 53 OW/OB children aged 6 to 10 years. Table 2 shows the general characteristics of the study population. The mean age and height were not statistically different amongst the groups. As expected, when compared to NW children, OW/OB children had significantly higher anthropometric characteristics including BMI z-score, WC z-score, and WtHR z-score. Table 3 shows the general characteristics of the family income and maternal education of the study population, where a higher percentage (55.6%) of EA had a family income greater than USD 75,000 while only 20% of AA had that income. Moreover, the percentage of AA having income less than USD 25,000 was 66%, while that in EA was only 4.8%. For maternal education, a higher percentage (46%) of AA had education up to high school or less, EA only 9.8%. At the higher end, 42.9% of EA had a graduate degree compared to AA with 18% only.   Table 4 and Figure 1 show the race-specific PMR of NRF1, FTO, and LEPR genes. In all participants, NW children had significantly lower methylation for the genes NRF1 (p = 0.018) and FTO (p = 0.010), while significantly higher methylation in gene LEPR (p = 0.025) compared to OW/OB children. Further, by separating the participants by their race, in EA children our results showed a similar trend for the genes NRF1 (p = 0.002) and FTO (p = 0.001) where NW children had less methylation but there was no significant Genes 2022, 13, 2030 6 of 15 difference observed in the methylation of the gene LEPR. Interestingly, for AA no significant difference in the methylation of the genes NRF1 and FTO was found since AA NW children had higher methylation than EA NW children, while for the gene LEPR, NW children had higher methylation compared to OW/OB (p = 0.046). A multinominal linear regression between PMR of NRF1, FTO, and LEPR with the obesity measures (BMI z-score, WC z-score, and WHtR z-score) after adjusting with the A multinominal linear regression between PMR of NRF1, FTO, and LEPR with the obesity measures (BMI z-score, WC z-score, and WHtR z-score) after adjusting with the covariates (family income and maternal education) demonstrated a significant positive correlation between PMR of NRF1 in EA participants with BMI z-score in the adjusted and unadjusted model. Meanwhile, AA participants showed a negative correlation between PMR of NRF1 and obesity measures with no statistical significance (Table 5). Similar calculations were performed for PMR of FTO and obesity measures. For EA participants, a significant positive correlation was observed with the BMI z-score, WC z-score, and WHtR z-score for both adjusted and unadjusted models. While AA participants, there is a negative correlation with no statistical significance of FTO with obesity measures ( Table 6). For the gene LEPR, total participants and EA showed no significant association while in AA, BMI zscore, and WtHR z-score had a significant negative association with LEPR for both adjusted and unadjusted models. Interestingly, we could see a significant negative association between the WC z-score and PMR of LEPR only in the adjusted model (Table 7). Maternal education and family income did not affect the PMR of NRF1 and FTO. Nonetheless, some correlation was seen with PMR of LEPR and WC z-score in AA children, which could be affected by family income and maternal education. Graphical representations from linear regression analysis for all three genes are represented in Figures 2-4. The results were adjusted for annual family income and maternal education. p < 0.05 is considered statistically significant between PMR of NRF1 and obesity markers, and it is shown in bold. The results were adjusted for annual family income and mother's education. p < 0.05 is considered statistically significant between PMR of FTO and obesity markers and it is shown in bold. The results were adjusted for annual family income and the mother's education. p < 0.05 is considered statistically significant between PMR of LEPR and obesity markers and it is shown in bold. The results were adjusted for annual family income and the mother's education. p < 0 considered statistically significant between PMR of LEPR and obesity markers and it is sho bold.       The receiver operating characteristics (ROC) analysis was used to determine the cutoff value for methylation of genes NRF1, FTO, and LEPR. As shown in Table 8, the area under curve (AUC) of PMR of NRF1, FTO, and LEPR with the BMI categories is 0.565 (p = 0.236), 0.629 (p = 0.018), and 0.571 (p = 0.192), respectively. Table 9 shows the ROC analysis of methylation of NRF1 and FTO based on the race, the AUC of NRF1, FTO and LEPR was 0.574 (p = 0.179), 0.603 (p = 0.059), and 0.850 (p = 0.001), respectively. The graphical representation of AUC is shown in Figure 5. As the AUC of LEPR is 0.850, it is considered a biomarker with very good diagnostic accuracy. However, the value for FTO is between 0.6 and 0.7; it is considered a biomarker with sufficient diagnostic accuracy.  The receiver operating characteristics (ROC) analysis was used to determine the cutoff value for methylation of genes NRF1, FTO, and LEPR. As shown in Table 8, the area under curve (AUC) of PMR of NRF1, FTO, and LEPR with the BMI categories is 0.565 (p = 0.236), 0.629 (p = 0.018), and 0.571 (p = 0.192), respectively. Table 9 shows the ROC analysis of methylation of NRF1 and FTO based on the race, the AUC of NRF1, FTO and LEPR was 0.574 (p = 0.179), 0.603 (p = 0.059), and 0.850 (p = 0.001), respectively. The graphical representation of AUC is shown in Figure 5. As the AUC of LEPR is 0.850, it is considered a biomarker with very good diagnostic accuracy. However, the value for FTO is between 0.6 and 0.7; it is considered a biomarker with sufficient diagnostic accuracy.

Discussion
This study explored the differences and the association of methylation of NRF1, FTO, and LEPR genes between normal weight and overweight/obese children taking into consideration of their race, maternal education, and family income. For the genes NRF1, FTO, and LEPR, our study demonstrates a race-specific difference in the DNA methylation of these obesity-related genes. Previously, increased NRF1 methylation in OW/OB children was also reported by Rushing et al., who found that NRF1 methylation was significantly associated with a higher probability of childhood obesity [21]. Additionally, DNA methylation and miRNA expression in skeletal muscle samples from sedentary obese and diabetic people, before and after 16 weeks of aerobic or resistance training, showed that NRF1 methylation was reduced after aerobic exercise [36]. It is important to note here that we used saliva to measure methylation, therefore even though direct comparison cannot be made with studies using blood and skeletal muscle, the fact that we found consistent results opens a window for saliva to be the epigenetic marker, especially for children where collecting tissue and blood is very difficult and painful. Supporting this, Oelsner et al. examined the methylation of obesity-related genes in saliva samples from preschool-age Hispanic children and showed that methylation of one of the CpG sites of NRF1 was significantly associated with increased BMI [37]. NRF1 plays a major role as a transcription factor in metabolic regulation and stimulates the expression of PPARG, a gene that is highly expressed in adipose tissue controlling numerous genes involved in metabolic homeostasis, lipid, glucose, and energy metabolism, adipogenesis, and inflammation which looks like a promising target and important for future research [38].
Czogala et al. demonstrated higher methylation of gene FTO along with increased expression of FTO gene in obese children compared to the control group with a significant correlation between FTO methylation and gene expression [39]. The gene FTO plays a major role in regulating hepatic gluconeogenesis and lipid metabolism, where an increase in FTO expression leads to reduced fatty acid oxidation and lipolysis along with increasing triglycerides [40]. In our study, we found that AA normal weight children had almost similar methylation in the FTO gene as overweight and obese, signifying a higher gene expression of FTO leading them towards increased risk of obesity while FTO gene methylation for EA normal weight children was extremely lower. Interestingly, the ROC curve for the gene FTO showed statistical significance. However, due to its low prediction capability of <63% of BMI categories correctly, we would suggest further research on FTO gene methylation as an early predictor of childhood obesity. Additionally, the results also indicate that in the AA population not only the obesity status, but also other factors such as different health disparities could be playing a role in causing increased DNA methylation, as we did not see a significant association between obesity markers and DNA methylation of NRF1 and FTO gene in AA children. While on the other end, in EA children we saw an increase in DNA methylation of both NRF1 and FTO genes as their BMI z-score increased, implying that in the EA population, the methylation status depends more on their obesity status.
To our knowledge, this is the first study demonstrating the differences in LEPR methylation between NW and OW/OB children among races. Even though we saw a significant difference in the methylation of the gene, there was no significant association found between methylation and obesity markers. However, in AA children, we could see a significant association between WC z-score and LEPR methylation after adjusting for maternal education and family income, indicating the important role of health disparities amongst races causing epigenetic modifications. Moreover, AA children generally have higher LEPR methylation compared to EA children, signifying that the methylation of LEPR in AA does depend more on their BMI status. The leptin receptor protein, which is important in controlling body weight, is made mostly by following the instructions provided by LEPR. Through the activation of several signaling pathways, including Janus Kinase 2/Signal transducer and activator of transcription 3 (JAK2/STAT3) and Mitogen Activated Protein Kinase (MAPK cascade), it serves as a receptor for the hormone leptin and mediates LEP central and peripheral effects [41]. It is also known that higher body mass is associated with increased leptin hormone in the body. LEPR expression is most likely decreased as a result of methylation [42], and therefore higher methylation in normal weight AA could be due to decreased leptin levels as a result of lower body fat. This could be a new window of opportunity to explore whether the obesity associated CpG methylation is caused by obesity, and not causing obesity [43].
Further, obesity prevalence rates differ significantly by race and ethnicity, with African Americans being 50% more likely to be obese than non-Hispanic whites [44], and African American children are at higher risk of being obese [45][46][47]. These results are in consistence with another study, affirming that AA is at a higher risk of obesity [48] and that race, along with the BMI status, is linked with DNA methylation of genes. Epigenetic regulation "levels" differ between racial and ethnic groups, according to research, even after accounting for access to care and genetics, implying that epigenetic regulation may be a contributing reason to health disparities [49]. The age group of 6-10 years can also provide us with an opportunity for early detection of obesity, especially in highly susceptible populations with a potential of methylation of LEPR as a biomarker, as the analysis of the ROC curve showed LEPR methylation has very good diagnostic accuracy.
Although our study provided a race-specific difference in DNA methylation and a potential link to health disparities, the study's relatively small sample size may have made it harder to find significant correlations with obesity measures. The results would be strengthened and validated if the current research were to be expanded to cover bigger sample numbers. This study does not consider the genetic effects of these genes on DNA methylation, which can further give more insight on the biological processes. Although there was a precise technique for collecting saliva, there is always a possibility of contamination and human collection error when collecting salivary DNA and doing further clinical procedures. Even though prior research suggests that DNA methylation in saliva and blood samples is similar, the current study only looked at methylation patterns in saliva and cannot be used to directly compare DNA methylation in blood and other tissues. Additionally, this sample provides information on children aged 6-10 years, yet DNA methylation patterns in children of other ages and races/ethnicities should be researched. Further, while collecting the saliva samples, we also recorded the dietary intake of the children, and the association between dietary intake and methylation patterns is underway.

Conclusions
To conclude, OW/OB children had significantly higher methylation of NRF1 and FTO and lesser methylation of LEPR. The methylation of LEPR is significantly higher in AA children than EA. Only EA OW/OB children had significantly higher levels of methylation in the NRF1 and FTO genes compared to NW, whereas the AA NW children had higher methylation of the LEPR gene than the OW/OB children. After controlling for maternal education and annual family income using regression analysis, the PMR of NRF1 and FTO was significantly associated with BMI z-score in EA children but not in AA children. However, the PMR of the gene LEPR was significantly associated with all the obesity measures after adjusting the maternal education and family income only in AA children. These findings contribute to a racial difference in the correlation of methylation of the NRF1, FTO, and LEPR genes with childhood obesity.
Author Contributions: Conceptualization, J.R.B., X.W. and T.G.; formal analysis, P.P. and V.S.; funding acquisition, J.R.B., X.W. and T.G.; methodology, P.P. and V.S.; project administration, T.G.; supervision, T.G.; validation X.W. and T.G.; visualization, J.R.B. and T.G.; writing-original draft preparation P.P.; writing-review and editing, all. All authors have read and agreed to the published version of the manuscript. Informed Consent Statement: Informed consent was obtained from all participants involved in the study.

Data Availability Statement:
The study datasets of the current manuscript are available from the corresponding author upon request.