Epigenetic landscape in blood leukocytes following ketosis and weight loss induced by a very low calorie ketogenic diet (VLCKD) in patients with obesity

Original article Epigenetic landscape in blood leukocytes following ketosis and weight loss induced by a very low calorie ketogenic diet (VLCKD) in patients with obesity Ana B. Crujeiras a , h , * , 1 , Andrea G. Izquierdo a , h , 1 , David Primo b , Fermin I. Milagro c , h , Ignacio Sajoux d , Amalia J acome e , Alfredo Fernandez-Quintela f , h , María P. Portillo f , h , J.Alfredo Martínez c , h , Miguel A. Martinez-Olmos a , h , Daniel de Luis b , Felipe F. Casanueva g , h a Epigenomics in Endocrinology and Nutrition Group, Epigenomics Unit, Instituto de Investigacion Sanitaria de Santiago de Compostela (IDIS), Complejo Hospitalario Universitario de Santiago de Compostela (CHUS/SERGAS), Spain b Center of Investigation of Endocrinology and Nutrition, Medicine School and Department of Endocrinology and Investigation, Hospital Clinico Universitario, University of Valladolid, Valladolid, Spain c Department of Nutrition, Food Science and Physiology, Centre for Nutrition Research, University of Navarra (UNAV) and IdiSNA, Navarra Institute for Health Research, 31009, Pamplona, Spain d Medical Department Pronokal Group, PronokalGroup, Barcelona, Spain e Department of Mathematics, MODES Group, CITIC, Universidade da Coru~ na, Faculty of Science, A Coru~ na, Spain f Nutrition and Obesity Group, Department of Nutrition and Food Science, University of the Basque Country (UPV/EHU), Lucio Lascaray Research Institute and Health Research Institute BIOARABA, Vitoria, Spain g Molecular and Cellular Endocrinology Group. Instituto de Investigacion Sanitaria de Santiago de Compostela (IDIS), Complejo Hospitalario Universitario de Santiago de Compostela (CHUS) and Santiago de Compostela University (USC), Spain h CIBER Fisiopatologia de La Obesidad y Nutricion (CIBERobn), Spain article info Article history: Received 7 January 2021 Accepted 13 May 2021 Keywords: Adiposity Methylation Nutritional intervention Circulating blood cells Biomarkers summary Background: The molecular mechanisms underlying the potential health benefits of a ketogenic diet are unknown and could be mediated by epigenetic mechanisms. Objective: To identify the changes in the obesity-related methylome that are mediated by the induced weight loss or are dependent on ketosis in subjects with obesity underwent a very-low calorie ketogenic diet (VLCKD). Methods: Twenty-one patients with obesity (n ¼12 women, 47.9 ±1.02 yr, 33.0 ±0.2 kg/m 2 ) after 6 months on a VLCKD and 12 normal weight volunteers (n ¼6 women, 50.3 ±6.2 yrs, 22.7 ±1.5 kg/m 2 ) were studied. Data from the Infinium MethylationEPIC BeadChip methylomes of blood leukocytes were obtained at time points of ketotic phases (basal, maximum ketosis, and out of ketosis) during VLCKD (n ¼10) and at baseline in volunteers (n ¼12). Results were further validated by pyrosequencing in representative cohort of patients on a VLCKD (n ¼18) and correlated with gene expression. Results: After weight reduction by VLCKD, differences were found at 988 CpG sites (786 unique genes). The VLCKD altered methylation levels in patients with obesity had high resemblance with those from normal weight volunteers and was concomitant with a downregulation of DNA methyltransferases (DNMT)1, 3a and 3b. Most of the encoded genes were involved in metabolic processes, protein metabolism, and muscle, organ, and skeletal system development. Novel genes representing the top scoring associated events were identified, including ZNF331,FGFRL1 (VLCKD-induced weight loss) and CBFA2T3, C3orf38,JSRP1, and LRFN4 (VLCKD-induced ketosis). Interestingly, ZNF331 and FGFRL1 were validated in an independent cohort and inversely correlated with gene expression. Conclusions: The beneficial effects of VLCKD therapy on obesity involve a methylome more suggestive of normal weight that could be mainly mediated by the VLCKD-induced ketosis rather than weight loss. ©2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). *Corresponding author. Instituto de Investigaci on Sanitaria, Complejo Hospitalario de Santiago (CHUS), C/ Choupana, S/N, 15706, Santiago de Compostela, Spain. E-mail addresses: [email protected],[email protected] (A.B. Crujeiras). 1 Both authors equally contributed to this work. Contents lists available at ScienceDirect Clinical Nutrition journal homepage: http://www.elsevier.com/locate/clnu https://doi.org/10.1016/j.clnu.2021.05.010 0261-5614/©2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Clinical Nutrition 40 (2021) 3959e3972 1. Introduction Ketosis has gained interest over recent years due to its induced benefits that it imparts on several health conditions [1,2]. Ketosis is associated with a delay in the onset of diseases and increased longevity [3]. Similarly, ketosis is suggested to have an extensive range of health benefits, from increased physical endurance in athletes [4,5] to delayed aging [6]. Also to improve conditions such as neurodegenerative disease [7e9] cancer [10e12], cardiovascular disease [13], and obesity [14]. Some of these studies involved high fat ketogenic diets and even though the main characteristic of ketogenic diets is the carbohydrates restriction, the specific composition in macronutrients and calories should be taken into consideration for the impact in clinical practice [15]. A very-low-calorie-ketogenic diet (VLCKD) was demonstrated to be an effective strategy in managing obesity [16], including weight loss and maintenance [17], increased preservation of muscle mass [18], and enhanced resting metabolic rate [19]. Moreover, it is able to improve metabolic parameters in patients with obesity [20,21] and type 2 diabetes [22]. Additionally, it was demonstrated that a VLCKD is able to reduce food craving and improve psychobiological parameters to help improve quality of life in patients with obesity [23]. However, the molecular mechanisms underlying these benefits of ketogenic diet remain unknown. The main molecular mechanism that links the effect of environmental factors, such as nutrition, with the regulation of the genes function is epigenetics [24]. Indeed, dietary factors or dietary patterns were identified as modulators of epigenetic mechanisms [25e28]. Recently, it was suggested that ketone bodies orchestrate gene expression via epigenomic mechanisms [29]. This molecular effect of nutritional ketosis has been reported in neurologic disorders such as epilepsy [30,31] and Kabuki syndrome and related disorders [32]. The effect of ketone bodies on epigenetic regulation was also proposed as a potential opportunity for anticancer therapies [33]. Moreover, ketogenic diets, including dietary restriction, delay aging through epigenetic effects [34]. On the other hand, obesity is associated with a specific methylation profile in several tissues [35e38] and the obesityrelated methylome is related to complications such as insulin resistance [39] and cancer [40,41]. Moreover, several methylation marks were identified as potential biomarkers for predicting the success of weight loss therapies during the active phase of treatment [42e46] or during the period of weight loss maintenance [47,48]. Therefore, DNA methylation was proposed as a target for preventing and managing obesity [48e51]. In light of the above evidence, the aim of the current study was to evaluate how a VLCKD might affect the obesity methylome. Furthermore, this study aimed to identify the changes in the obesity-related methylome that are mediated by the induced weight loss or are dependent on ketosis. 2. Materials and methods 2.1. Patient cohort The DNA and RNA for methylation and gene expression assays were isolated from blood samples of patients from a 6-month nutritional intervention study performed at the Endocrinology and Nutrition Department of the Hospital Clinico, Universitario of Valladolid; the patients were receiving treatment for obesity. In Abbreviations list ACACB Acetyl-CoA Carboxylase Beta AEBP1 AE binding protein 1 ANOVA analysis of variance BMI body mass index C3orf38 chromosome 3 open reading frame 38 CACNA1H Calcium Voltage-Gated Channel Subunit Alpha1 H CAMKK1 Calcium/Calmodulin Dependent Protein Kinase Kinase 1 CBFA2T3 CBFA2/RUNX1 Partner Transcriptional Co-Repressor 3 cDNA complementary DNA CENPF Centromere Protein F CHAT choline O-acetyltransferase CHUK Component Of Inhibitor Of Nuclear Factor Kappa B Kinase Complex COL1A1 collagen, type I, alpha 1 COL5A1 collagen type V alpha 1 chain COL9A2 collagen, type IX, alpha 2 CV coefficient of variation DMCpGs differentially methylated CpGs DNMTs DNA methyltransferases DNA deoxyribonucleic acid EWAS epigenome-wide association study FDR false discovery rate FF fresh-frozen FGFRL1 Fibroblast growth factor receptor (FGFR)-like protein 1 GAPDH Glyceraldehyde-3-Phosphate Dehydrogenase GO gene ontology HRAS HRas proto-oncogene, GTPase INSR insulin receptor JSRP1 Junctional Sarcoplasmic Reticulum Protein 1 LAMA2 Laminin Subunit Alpha 2 LRFN4 Leucine Rich Repeat And Fibronectin Type III Domain Containing 4 MKNK2 MAPK Interacting Serine/Threonine Kinase 2 PRKAG2 Protein Kinase AMP-Activated Non-Catalytic Subunit Gamma 2 PRKCZ Protein Kinase C Zeta qRT-PCR Real-time quantitative polymerase chain reaction RELA v-rel reticuloendotheliosis viral oncogene homolog A RNA Ribonucleic acid RPTOR Regulatory Associated Protein Of MTOR Complex 1 SD standard deviation SEM standard error of the mean SGCB Beta-sarcoglycan SMTN Smoothelin TRAF2 TNF Receptor Associated Factor 2 TSC2 Tuberous Sclerosis Complex 2 TSS 200 transcription start sites 200 VLCKD very-low-calorie ketogenic diet WC Waist circumference ZFHX3 zinc finger Homeobox 3 ZNF331 zinc finger protein 331 b -OHB beta-hydroxy-butyrate A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3960 addition, samples from a group of healthy volunteers were also analyzed. The inclusion criteria were: age between 18 and 65 years, body mass index (BMI) 30 kg/m 2 , stable body weight over the previous 3 months, a desire to lose weight, and a history of failed dietary efforts. The main exclusion criteria were thyroid alteration, diabetes mellitus, obesity induced by other endocrine disorders or drugs, and participation in any active weight-loss program in the previous 3 months. In addition, patients with previous bariatric surgery, reported or suspected abuse of narcotics or alcohol, severe depression or any other psychiatric disease, severe hepatic insufficiency, any type of renal insufficiency or gout episodes, nephrolithiasis, neoplasia, previous instances of cardiovascular or cerebrovascular disease, uncontrolled hypertension, orthostatic hypotension, and hydroelectrolytic or electrocardiographic alterations were excluded. Females who were pregnant, breastfeeding, or intending to become pregnant and those with child-bearing potential who were not using adequate contraceptive methods were also excluded. Apart from obesity and metabolic syndrome, participants were generally healthy individuals. Under these criteria, 21 patients with obesity and 12 volunteers with normal weight were included in this study. The study protocol was in accordance with the Declaration of Helsinki and was approved by the Ethics Committee for Clinical Research of Hospital Clinico Universitario de Valladolid, Spain (C.I: 40/13, PNK-DHA2013-01). Participants provided written informed consent before any intervention related to the study. Participants received no monetary incentives. 2.2. Very-low-calorie ketogenic diet protocol Patients included in this study derived from a randomized clinical trial investigating the effect of docosahexaenoic acid (DHA) supplementation in a very low-calorie ketogenic diet. The clinical trial consisted in two arms: one arm where patients follow a VLCKD and other arms where patients followed a VLCKD þDHA [21]. Nutritional intervention was based on a commercial weight-loss program (PNK method ®), as described elsewhere [18,21]. Briefly, the intervention included an evaluation by the specialist physician conducting the study, an assessment by an expert dietician, and exercise recommendations. This method is based on highbiological-value protein preparations obtained from cow's milk, soy, avian eggs, green peas, and cereals. Each protein preparation contained 15g protein, 4g carbohydrates, 3g fat, and provided 90e100 kcal. The VLCKD þDHA arm was supplemented with 500 mg/day DHA [21]. The weight-loss program has five steps and adheres to the most recent guidelines of the EFSA (2015) on total carbohydrate intake [52]. The first three steps consist of a VLCKD (600e800 kcal/day) that is low in carbohydrates (<50g daily from vegetables) and lipids (only 10g of olive oil per day). The amount of high biological-value proteins ranged between 0.8 and 1.2g per kg of ideal body weight to ensure that patients were meeting their minimum bodily requirements and to prevent the loss of lean mass. In step 1, the patients ate high-biological-value protein preparations five times a day and vegetables with low glycemic indices. In step 2, one of the protein servings was substituted with a natural protein (e.g., meat or fish) either at lunch or at dinner. In step 3, a second serving of low-fat natural protein was substituted for the second serving of biological protein preparation. Throughout these ketogenic phases, supplements of vitamins and minerals, such as K, Na, Mg, Ca, and omega-3 fatty acids, were provided in accordance with international recommendations [53]. These three steps were maintained until the patient lost the target amount of weight, ideally 80%. Because of this, the ketogenic steps varied in time depending on the individual and the weight-loss target. The total ketosis state lasted for a maximum of 60 days. In either step 4 or 5, ketosis was ended by the physician in charge of the patient based on the amount of weight lost, and the patient began a low-calorie diet (800e 1500 kcal/day). At this point, the patients underwent a progressive incorporation of different food groups and participated in a program of alimentary reeducation to guarantee long-term maintenance of the weight loss. The maintenance diet consisted of an eating plan balanced for carbohydrates, protein, and fat. Depending on the individual, calories consumed ranged between 1500 and 2000 kcal/day, with the goal of maintaining the weight loss and promoting a healthy lifestyle. During this study, patients followed the steps of the method until they reached the target weight, or up to a maximum of 4 months of follow-up, although patients remained under medical supervision for the following months. Patients visited the research team every 15 ±2 days to evaluate adherence and potential side effects. Complete anthropometry, body composition, and biochemical assessments were performed at four of the visits, which were determined according to the evolution of each patient through the steps of ketosis and weight loss: Visit 1 (Baseline), visit 2 (Maximum Ketosis), visit 3 (Reduced Ketosis) and visit 4 (Endpoint). DNA methylationandgeneexpressionwereperformed at visits1, 2,and4. In all visits, patients received dietary instructions, individual supportive counsel, and encouragement to exercise on a regular basis using a formal exercise program. Additionally, a program of reinforcement telephone calls was instituted, and a phone number was provided to all participants to address any concerns. 2.3. Anthropometric assessment All anthropometric measurements were performed after an overnight fast (8e10 h) under resting conditions in duplicate and performed by well-trained health workers. At each visit, patients were weighed on the same calibrated scale (Seca 200 scale, Medical Resources, EPI Inc OH, USA). BMI was calculated as body weight in kg, divided by the square of body height in meters (BMI ¼weight (kg)/height 2 (m)). Waist circumference (WC) was measured using a standard flexible non-elastic metric tape placed over the midpoint between the last rib and the iliac crest, with the patient standing and exhaling. 2.4. Determining levels of ketone bodies Ketosis was determined by measuring ketone bodies, specifically b -hydroxy-butyrate ( b -OHB), in capillary blood using a portable meter (GlucoMen LX Sensor, A. Menarini Diagnostics, Neuss, Germany; sensitivity <0.2 mmol/l). As with anthropometric assessments, all determinations of capillary ketonemia were made after an overnight fast of 8e10 h. These measurements were performed daily by each patient during the entire VLCKD, and the corresponding values were reviewed using machine memory by the research team to control adherence. Additionally, b -OHB levels were determined at each visit by the physician in charge of the patient. 2.5. DNA methylation analysis 2.5.1. DNA preparation and bisulphite conversion DNA from fresh-frozen (FF) blood samples was isolated using a standard phenol-chloroform/proteinase-k protocol according to the manufacturer's instructions, with slight modifications. Genomic DNA was isolated from leukocytes using the MasturPure™DNA purification kit (Epicentre Biotechnologies, A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3961 Madison, WI, USA). The isolated DNA was treated with RNase A for 1 h at 45  C. All DNA samples were quantified using the fluorometric method (Quan-iT PicoGreen DsDNA Assay, Life Technologies) and were assessed for purity using a NanoDrop (Thermo Scientific) to determine 260/280 and 260/230 ratio measurements. The integrity of the FF DNA was verified by electrophoresis in 1.3% agarose gel. DNA (500 ng) was bisulfite converted using the EZ DNA methylation kit Methylation-Gold (Zymo Research, CA, USA) according to the manufacturer's instructions, which converts nonmethylated cytosine into uracil. 2.5.2. Infinium MethylationEPIC BeadChip High-quality DNA samples (500 ng) obtained from blood leukocytes of patients included in the VLCKD þDHA arm of the clinical trial (discovery cohort; n ¼10 patients, 3 paired samples/patient) were selected for bisulfite conversion (Zymo Research; EZ-96 DNA Methylation™Kit) and hybridization to Infinium MethylationEPIC BeadChip (Illumina) following the Illumina Infinium HD methylation protocol. DNA quality checks, bisulfite modification, hybridization, data normalization, statistical filtering, and value calculations were performed as previously described [54,55]. Whole-genome amplification and hybridization were then performed on a BeadChip followed by single-base extension and analysis on a HiScan SQ module (Illumina) to assess cytosine methylation states. The annotation of CG islands (CGIs) used the following categorization: 1) shore, for each of the 2-kb sequences flanking a CGI; 2) shelf, for each of the 2-kb sequences next to a shore; and 3) open sea, for DNA not included in any of the previous sequences or in CGIs [54,55]. The transcription start site 200 and the transcription start site 1500 indicate regions either 200 or 1500 bp from the transcription start site, respectively. 2.5.3. Pyrosequencing analysis Pyrosequencing was used to assess selected markers in 18 patients (validation cohort: (n ¼7 derived from the discovery cohort and n ¼11 from an independent cohort of patients; 3 paired samples/patient)). DNA samples analyzed in the validation cohort were derived from patients included in the two arms of the clinical trial and merged for the statistical analysis (VLCKD: n ¼10; VLCKD þDHA: n ¼8). The primer sequences used in this analysis were designed using Qiagen's PyroMark Assay Design 2.0 software to hybridize to CpG-free sites to ensure methylation-independent amplification (details and primer sequences are available in Supplementary Table S1). Genomic DNA was isolated from FF blood leukocytes using the MasturPure™DNA purification kit (Epicentre Biotechnologies, Madison, WI, USA), according to the manufacturer's instructions. DNA methylation analyses were performed using bisulfite-treated DNA (Zymo Research; EZ-96 DNA Methylation™Kit) followed by a highly quantitative analysis based on PCR-based pyrosequencing using the PyroMark Q24 System version 2.0.7 (Qiagen). Methylation level was expressed as the percentage of methylated cytosine over the sum of methylated and unmethylated cytosines. Non-CpG cytosine residues were used as built-in controls to verify bisulfite conversion. The values are expressed as the mean for all sites. We also included human non-methylated and methylated DNA set as controls in each run (Zymo Research). The inter-assay precision (%CV) was <2.5%, intra-assay (%CV) was <1.0%. 2.6. Expression assay by qRT-PCR RNA from blood leukocytes (n ¼18 patients) was extracted using Trizol (Invitrogen) according to the manufacturer's recommendations. The RNA concentrations were measured with a Nanodrop 2000 spectrophotometer (Thermo Scientific). From total extracted RNA, 2 m g were DNase treated using a DNA-free kit as a template (Ambion) to generate first-strand cDNA synthesis using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time quantitative polymerase chain reaction (qRTPCR) was performed using TaqMan Universal PCR Master Mix, TaqMan Probes (Applied Biosystems) (details and primer sequences are available in Supplementary Table S1), and the Step OnePlus Real-Time PCR System (Applied Biosystems). All experiments were performed in duplicate, and gene expression levels were normalized to the levels of housekeeping gene GAPDH. The fold change in gene expression was calculated using the 2  DD Ct relative quantitation method according to the manufacturer's guidelines (Applied Biosystems), and data are reported as the geometric mean (SEM). qRT-PCR experiments were performed in compliance with the MIQE (Minimum Information for Publication of Quantitative RealTime PCR Experiments) guidelines (http://www.rdml.org/miqe). 2.7. Statistical analysis The sample size of the current study was calculated to detect differences for methylation levels taking into account published values of epigenome-wide analysis in the field of obesity [35,40,41]. The interventional differences were examined in two independent cohorts. Microarray-based DNA methylation analysis was performed in the discovery cohort (n ¼10 patients; 3 paired samples/ patient), and then the identified genes were validated in an independent cohort of patients (validation cohort; n ¼18 patients; 3 paired samples/patient). Finally, the association between DNA methylation level of the identified CpG sites and the anthropometric or biochemical parameters was assessed in the global cohort of patients included in this study (n ¼28). The methylation level of each cytosine was expressed as a b value, which was calculated as the fluorescence intensity ratio of the methylated to the unmethylated version of the probe. b values ranged between 0 (unmethylated) and 1 (completely methylated) according to the combination of the Cy3 and Cy5 fluorescence intensities. Color balance adjustment and normalization were performed to normalize the samples between the two-color channels using Genome Studio Illumina software (V2010.3). Genome Studio normalizes data using different internal controls that are present on the Infinium MethylationEPIC BeadChip. This software also normalized data depending on internal background probes [55]. b values with detected p-values >0.01 were considered to fall below the minimum intensity and threshold, and these CpGs were consequently removed from further analysis. Additionally, probes that contained single nucleotide polymorphisms (SNPs) at the 10 bp 3 0 end of the interrogating probe were filtered out. To identify consistent patterns of DMCpGs due to the nutritional intervention, a linear model was fitted using a B-spline approximation [56]. Three linear models were fitted: Model 1 was fitted by including the three points of the nutritional intervention to evaluate the general effect of VLCKD; Model 2 including baseline and maximum ketosis to evaluate the effect of ketosis and weight loss; Model 3 including methylation levels at maximum ketosis and endpoint to evaluate the effect of only weight loss, without ketosis. P values were adjusted for multiple comparisons using the false discovery rate (FDR) procedure of Benjamini and Hochberg, and results were considered statistically significant when FDR <0.10. Additionally, we applied a threshold for the significant sites based on the mean difference between visits with a minimum b value change of ±0.02. Euclidean cluster analysis of significant CpGs was performed using a heatmap function. The global methylation level was compared between the nutritional intervention visits by univariant ANOVA and a Bonferroni post-hoc analysis. All of the aforementioned statistical analyses were performed using R software (version 3.2.0). A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3962 To estimate enrichment in biological processes, a hypergeometric test was performed using the GOstats package on the biological processes defined by gene ontology (GO) [57]. This analysis detected significant over-representation of GO terms in one set (i.e., list of identified genes) with respect to the entire genome. GO terms with an adjusted p-value <0.05 were considered significant. With SPSS version 21.0 software (SPSS Inc., Chicago, IL) for Windows XP (Microsoft, Redmond, WA), the genomic distribution of the differentially methylated CpGs was compared with the distribution of the CpGs from all analyzed sites on the Infinium MethylationEPIC BeadChip. P values were computed using the chi-square test to determine overor under-representation of the CpGs. The potential association between anthropometric or biochemical parameters and DNA methylation levels ( b -values) was evaluated using the Spearman coefficient test. Differences in DNA methylation levels and expression of the identified genes during the time-course of the intervention and between the nutritional intervention visits were assessed by the nonparametric tests, Kruskal Wallis and ManneWhitney U, respectively. P0.05 was considered statistically significant. 3. Results 3.1. Patient characteristics Samples from a total of 21 patients who followed the nutritional intervention based on a VLCKD were compared with samples from 12 healthy volunteers with normal weight and evaluated in this study (Supplemental Fig.1). We first evaluated the discovery cohort (n ¼10 participants with obesity who followed a VLCKD þDHA (n ¼5 women) and n ¼12 (6 women) volunteers with normal weight). An extended validation cohort composed of 11 patients (7 women) with obesity that followed a VLCKD þDHA or a VLCKDDHA was also analyzed. Statistically significant differences were not observed between either cohort in age, gender, height, body weight, BMI, waist circumference, ketosis, or the response to weight loss treatment (Table 1). Differences statistically significant were only detected in body weight, BMI and waist circumference between patients with obesity and subjects with normal weight (Table 1). All patients lost weight after nutritional intervention (21.8 ±4.9%), together with reductions in BMI (21.9 ±5.1%) and waist circumference (19.3 ±4.4%). 3.2. DNA methylation changes during the global VLCKD intervention DNA methylation profiles of blood leukocytes involving approximately 850 thousand CpGs were analyzed after VLCKD intervention. This analysis revealed statistically significant differences (cut-off point D 0.02; FDR 0.10) at 988 CpG sites, from a total of 739,222 valid CpGs (see detailed list in Supplementary Table S2). The differentially methylated CpGs were mostly characterized as changes towards CpG hypomethylation occurring after nutritional intervention, in both total DMCpGs (Fig. 1A) and in the DMCpGs located in promoters, and islands or shores (Fig. 1B). This result of global hypomethylation was correlated with the downregulation in the expression of DNA methyltransferase (DNMT) 1 and the de novo methyltransferases DNMT3A and DNMT3B (Fig. 1C and Supplemental Fig. 2). The identified CpG sites mapped to 786 unique genes and we were able to separate the samples according to nutritional intervention visits using a hierarchical cluster approach (Fig. 1D). It should be noted that the methylation levels of the 988 CpG sites differentially methylated after nutritional intervention were altered to resemble methylation levels observed in samples from subjects with normal weight (Fig. 1C). The 20 DMCpGs with the highest difference with respect to baseline among genes that are represented by more than 2 CpGs are represented in Table 2. Regarding the functional distribution (Fig. 2), the differences in DNA methylation were mainly observed in open sea regions (Fig. 2A), with 43.2% of the DMCpG sites located in promoter regions and the majority of DMCpGs in the body (Fig. 2B). Moreover, the DMCpGs were mainly found in chromosomes 2, 9,16,17,19, and 22 when compared with all CpGs analyzed (Fig. 2C). Among the 988 DMCpGs, we found 886 (89.7%) with decreased and 102 (10.32%) with increased levels of DNA methylation after nutritional intervention. Moreover, the DMCpGs that lost methylation with respect to baseline were located mainly in the open sea (Fig. 2D) and body (Fig. 2E). In contrast, the DMCpGs that gained methylation after the nutritional intervention were mostly located in promoters (TSS 200 and 1st exon) (Fig. 2D) and in CpG islands (Fig. 2E). Regarding chromosome distribution, the DMCpGs with decreased methylation levels after nutritional treatment were found on chromosomes1,4,67,8,9,14,and16,whereasDMCpGswithincreased Table 1 Clinical characteristics of patients at baseline and during the intervention with a VLCKD. Discovery cohort Validation cohort P value Normal weight Obesity Obesity Baseline (day 0) Maximum ketosis (day 30) Endpoint (day 180) Baseline (day 0) Maximum ketosis (day 30) Endpoint (day 180) Adiposity Time Cohort Time x Cohort N1210101018*18 18 eeee Age (years) 50.3 ±6.2 48.8 ±9.20 ee 47.1 ±9.8 ee 0.654 e0.678 e Gender (men/women) 6/6 5/5 ee 7/11 ee eeee Height (m) 1.67 ±0.08 1.68 ±0.08 ee 1.65 ±0.10 ee 0.773 e0.446 e Body weight (Kg) 63.8 ±8.7 93.4 ±9.9 a 84.3 ±8.4 73.9 ±6.7 b 90.8 ±11.6 81.9 ±10.4 70.4 ±10.4 b <0.0001 <0.001 0.488 0.649 BMI (Kg/m 2 ) 22.7 ±1.49 32.9 ±1.4 a 29.8 ±1.7 26.1 ±1.8 b 33.2 ±1.7 30.06 ±1.7 25.7 ±1.7 b <0.0001 <0.001 0.986 0.334 Waist circumference (cm) 77.4 ±7.2 111.1 ±6.6 a 102.3 ±7.4 90.2 ±3.1 b 108.3 ±8.5 100.1 ±7.3 86.9 ±7.4 b <0.0001 <0.001 0.325 0.836 Ketonemia (mM) —0.15 ±0.07 2.11 ±1.11 b 0.15 ±0.07 0.18 ±0.08 1.71 ±1.08 b 0.18 ±0.08 e<0.0001 0.445 0.374 Weight loss (%) ee 9.69 ±2.51 20.58 ±5.17 c e9.80 ±1.99 22.52 ±4.75 c e<0.0001 0.336 0.455 Data shown are mean ±SD (standard deviation).*The sample size of the validation cohort was completed with n ¼7 patients from the discovery cohort and 11 patients from an independent cohort. Abbreviations: VLCKD, very-low-calorie ketogenic diet. a Statistically significant differences compared with control Normal weight in discovery cohort. b Statistically significant differences compared with Baseline in both cohorts. c Statistically significant differences compared with Maximum Ketosis in both cohorts. A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3963 methylation levels were mainly found in chromosomes 3, 5, 10, 11, 12, 17, 21, and 22 (Fig. 2F). 3.3. Biological significance of the dietary intervention-related DMCpG sites and associated genes Interestingly, most differentially methylated genes belonged to a network significantly enriched in protein interactions (p <0.001) according to STRING analysis (Fig. 3A). GO analysis was performed to test whether certain molecular functions or biological processes were significantly enriched within the 786 genes associated with the 988 DMCpGs discovered through VLCKD intervention (Fig. 3B). Among the categories of functional processes that exhibited statistical significance (FDR <0.05), we found processes related to regulation of transcription, signal transduction, cell differentiation, proliferation and apoptosis, metabolic processes, response to hypoxia, protein processing, muscle organ and skeletal system development, nervous system development and axon guidance (Fig. 3B). Among these, we highlighted relevant pathways in the field of obesity physiopathology such as insulin signaling, protein digestion and absorption, adipocytokine signaling and muscle development (Fig. 4AeD). To investigate biological relevance, the CpG sites representing promoter regions (TSS1500, TSS200, 5 0 UTR and 1st Exon) at CpG islands/shores were selected. This selection yielded 141 CpGs representing 150 unique genes. Based on this filter we identified CpG sites that could be genetic targets whose methylation is associated with VLCKD responses. Among these, the most representative gene was ZNF331, which was represented by 2 CpG sites located in the promoter and island with the highest difference with respect to baseline. Furthermore, FGFRL1 was also selected from among the DMCpG sites because of its biological relevance in metabolic pathways and obesity pathogenesis and because the methylation level of this gene in leukocytes was previously proposed as an episignature that mirrors methylation levels of dysfunctional adipose tissue in obesity [35]. 3.4. DNA methylation changes associated with the effects of dietary-induced ketosis An analysis comparing baseline (day 0) with maximum ketosis (day 30) yielded 1365 DMCpGs. With respect to all CpGs analyzed, these CpGs were found mainly in the open sea (Fig. 5A) and body (Fig. 5B) and in chromosomes 1, 10, 11, 17, 19, and 22 (Fig. 5C). The DMCpGs that gained methylation were found in the island and Fig. 1. Profile of DNA methylation following the VLCKD (n ¼10; Discovery cohort). (A). Global differences in methylation levels of 988 DMCpGs identified by Infinium MethylationEPIC BeadChip analysis. (B). Global differences in methylation levels of DMCpGs located in the promoter and island/shore. (C). Expression levels of DNA methyltransferase (DNMT) 1 and the de novo methyltransferases DNMT3A and DNMT3B (D). Supervised clustering of the 988 CpGs that were found to be differentially methylated between patients with obesity and healthy volunteers with normal weight (n ¼12) groups. The identified CpG sites mapped to 786 unique genes. Differences statistically significant detected (P <0.0001). DMCpGs, differentially methylated CpGs; VLCKD, very-low calorie ketogenic diet. A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3964 Table 2 List of DMCpGs with the highest differences for Baseline - Maximum Ketosis - Endpoint between genes that are represented by more than 2 CpGs. TargetID CHR Position Gene Name Gene region CpG context Methylation levels (mean) Differences p value FDR BMKEMK-BEB cg04254103 19 1794380 ATP8B3; ATP8B3; ATP8B3 Body; Body; Body N_Shore 0.826 0.786 0.796 0.039 0.030 0.001 0.091 cg06643002 14 105735797 BRF1; BRF1; BRF1; BRF1; Body; Body; Body; Body; Body 0.782 0.741 0.748 0.040 0.034 0.001 0.099 cg12063937 1 7731375 CAMTA1 Body 0.837 0.785 0.802 0.051 0.034 0.000 0.054 cg00035197 16 88962986 CBFA2T3; CBFA2T3 Body; Body Island 0.803 0.748 0.763 0.056 0.040 0.001 0.082 cg09962824 21 44479417 CBS Body N_Shore 0.754 0.707 0.714 0.047 0.040 0.000 0.051 cg23790296 10 73233854 CDH23; CDH23; CDH23; CDH23; CDH23 Body; Body; Body; Body; Body 0.846 0.814 0.820 0.032 0.026 0.001 0.095 cg26493726 19 36508614 CLIP3 Body S_Shelf 0.808 0.771 0.773 0.037 0.035 0.000 0.069 cg01818220 9 23821773 ELAVL2; ELAVL2; ELAVL2; ELAVL2 5 0 UTR; TSS1500 Island 0.092 0.148 0.134 0.056 0.042 0.000 0.008 cg01933710 5 132596864 FSTL4 Body 0.828 0.778 0.799 0.049 0.029 0.001 0.076 cg07390459 2 121582002 GLI2 Body 0.768 0.720 0.727 0.048 0.041 0.001 0.097 cg14777822 2 121728297 GLI2 Body 0.770 0.728 0.726 0.042 0.044 0.001 0.093 cg16120742 7 50345049 IKZF1 5 0 UTR S_Shore 0.114 0.058 0.041 0.057 0.073 0.000 0.026 cg02999309 10 134502626 INPP5A Body N_Shore 0.836 0.799 0.807 0.037 0.029 0.000 0.024 cg14010696 19 5119250 KDM4B Body Island 0.853 0.812 0.819 0.041 0.034 0.000 0.008 cg16333587 14 101369826 MEG8 Body 0.857 0.817 0.828 0.040 0.029 0.000 0.051 cg20963002 14 101360088 MEG8 TSS1500 0.862 0.821 0.828 0.041 0.034 0.001 0.098 cg18026309 2 26683651 OTOF; OTOF; OTOF; OTOF; OTOF Body; Body; Body; Body; Body 0.884 0.851 0.857 0.033 0.027 0.000 0.051 cg01764953 11 70814657 SHANK2 Body 0.914 0.887 0.887 0.027 0.026 0.000 0.067 cg19696891 19 54057705 ZNF331; ZNF331; ZNF331 5 0 UTR; 5 0 UTR; TSS1500 Island 0.379 0.431 0.447 0.052 0.068 0.000 0.015 cg03643149 19 54041519 ZNF331; ZNF331; ZNF331; ZNF331; ZNF331; ZNF331 TSS1500; 1stExon; TSS200; TSS200; 5 0 UTR; 5 0 UTR Island 0.397 0.445 0.458 0.048 0.061 0.000 0.019 Data shown are mean. Abbreviations: B, baseline; CHR, chromosome; E, endpoint; FDR, false discovery rate; MK, maximum ketosis. Fig. 2. Characterization of the DMCpGs after VLCKD in the discovery cohort (n ¼10). (A) Genomic distribution of the DMCpGs and their respective locations regarding the broader CpG context, (B) gene region and (C) chromosome. (D) Genomic distribution of DMCpGs comparing those that exhibited an increase with those that exhibited a decrease in methylation levels following VLCKD, and their respective locations in the broader CpG context, (E) the gene region and (F) chromosome. DMCpGs, differentially methylated CpGs; VLCKD, very-low calorie ketogenic diet. A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3965 Fig. 3. Biological implications of the DMCpGs following a VLCKD. (A) Summary of the GO analysis of the biological process categories representing the differentially methylated genes associated with DMCpG sites. (B) Gene-protein interaction network-STRING analysis. Most of the genes regulated by methylation belonged to a network significantly enriched in protein interactions (p <0.001) according to STRING analysis. DMCpGs, differentially methylated CpGs; GO, gene ontology; VLCKD, very-low calorie ketogenic diet. Fig. 4. Genes with known functions that present DMCpGs following a VLCKD. (A) Novel genes epigenetically regulated following VLCKD belonged to the insulin signaling pathway, (B) adipocytokine signaling pathway, (C) protein digestion and absorption functions, and (D) muscle organ development functions. DNA methylation values are expressed as b-values from the Infinium MethylationEPIC BeadChip. *Denotes differences statistically significant (P <0.05). DMCpGs, differentially methylated CpGs; VLCKD, very-low calorie ketogenic diet. A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3966 promoter and in chromosomes 1, 2, 3, 4 and 7. The DMCpGs that lost methylation were found in the open sea and body and in chromosomes 9, 10, 11, 12, 17, 21, 22 (Fig. 5DeF). With the aim of isolating the specific effects of ketosis on methylation profile, additional analysis was performed. A Venn diagram was created by including: DMCpGs Baseline eMaximum Ketosis ¼1365, DMCpGs Baseline eEndpoint ¼405, and DMCpGs Maximum Ketosis eEndpoint ¼21. Using these conditions, we identified 280 CpGs whose methylation was affected by the induced weight loss per se (see detailed list in Supplementary Table S3) and 1239 CpGs were identified as VLCKD-induced ketosis-related DMCpGs (Fig. 6A). These VLCKD-induced ketosisrelated DMCpGs corresponded to 966 annotated genes, and 161 of these were represented by 137 VLCKD-induced ketosis-related DMCpGs located in the promoter and island or shore (see detailed list in Supplementary Table S4). Among the categories of functional processes that exhibited statistical significance (FDR <0.05), we found processes related to regulation of transcription, signal transduction, cell adhesion, cell differentiation, proliferation and apoptosis, as well as nervous system development and axon guidance. Moreover, these ketosis-related differentially methylated genes belonged to pathways involved in focal adhesion, insulin and adipocytokine signaling pathways, MAPK and P53 signaling pathways and in cancer and type II diabetes mellitus-related pathways (Fig. 6B). Overall pathways associated with obesity physiopathology. Among these, we highlighted relevant pathways in the field of obesity physiopathology such as insulin signaling, protein digestion and absorption, adipocytokine signaling and muscle development. To identify potentially novel signatures of DNA methylation associated with VLCKD-induced ketosis, those genes with more Fig. 5. Characterization of the DMCpGs during ketosis induced by a VLCKD from the discovery cohort (n ¼10). (A) Genomic distribution of the DMCpGs and their respective locations regarding the broader CpG context, (B) gene region and (C) chromosome. (D) Genomic distribution of DMCpGs comparing those that exhibited an increase with those that exhibited a decrease in methylation levels during the VLCKD-induced ketosis and their respective locations in the broader CpG context, (E) the gene region and (F) chromosome. DMCpGs, differentially methylated CpGs; VLCKD, very-low calorie ketogenic diet. Fig. 6. Biological implications of the DMCpGs related to VLCKD-induced ketosis. (A) Venn diagram of the DMCpGs detected between baseline and maximum ketosis, between baseline and endpoint, and between maximum ketosis and endpoint. From this analysis, 1239 CpGs were identified as nutritional ketosis-related DMCpGs. (B) Summary of the GO analysis of the biological process categories representing the differentially methylated genes associated with nutritional ketosis-related DMCpG sites. DMCpGs, differentially methylated CpGs; GO, gene ontology; VLCKD, very-low calorie ketogenic diet. A.B. Crujeiras, A.G. Izquierdo, D. Primo et al. Clinical Nutrition 40 (2021) 3959e3972 3967