A Genetic Score of Predisposition to Low-Grade Inflammation Associated with Obesity May Contribute to Discern Population at Risk for Metabolic Syndrome

Omega-3 rich diets have been shown to improve inflammatory status. However, in an ex vivo system of human blood cells, the efficacy of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) modulating lipid metabolism and cytokine response is attenuated in overweight subjects and shows high inter-individual variability. This suggests that obesity may be exerting a synergistic effect with genetic background disturbing the anti-inflammatory potential of omega-3 long-chain polyunsaturated fatty acids (PUFA). In the present work, a genetic score aiming to explore the risk associated to low grade inflammation and obesity (LGI-Ob) has been elaborated and assessed as a tool to contribute to discern population at risk for metabolic syndrome. Pro-inflammatory gene expression and cytokine production as a response to omega-3 were associated with LGI-Ob score; and lower anti-inflammatory effect of PUFA was observed in subjects with a high genetic score. Furthermore, overweight/obese individuals showed positive correlation of both plasma C-Reactive Protein and triglyceride/HDLc-index with LGI-Ob; and high LGI-Ob score was associated with greater hypertension (p = 0.047), Type 2 diabetes (p = 0.026), and metabolic risk (p = 0.021). The study shows that genetic variation can influence inflammation and omega-3 response, and that the LGI-Ob score could be a useful tool to classify subjects at inflammatory risk and more prone to suffer metabolic syndrome and associated metabolic disturbances.


Introduction
Obesity is currently considered one of the main health problems worldwide [1], and it is usually accompanied by the abnormal production of cytokines and adipokines by the immune system and adipose tissue, respectively, as well as by a disturbed profile of lipid mediators [2]. Although activation of inflammatory pathways is indispensable for the organism in specific circumstances, such as fighting against infections, undamaged inflammatory pathways are critical for tissue health and for proper homeostasis [3], including adipose tissue expansion and remodeling [4]. However, at some step that is still not well characterized, obesity results in a state of low-grade chronic metabolic inflammation involving many organs, altering energy and substrate fluxes and giving rise to the development of metabolic pathologies [5][6][7][8][9][10] (diabetes, dyslipidemia, and hypertension, among others) (see [3] for

Study Participants and Design
This genetic association study is based on three observational studies: One case-control (NUTRI-BLOOD) and two cross-sectional (OptiDiet-15 and Ob-IB) studies ( Figure 1). STREGA reporting guidelines were used.
Initial samples were genotyped from participants in the NUTRI-BLOOD study (IB 2112/13 PI), in which 18 young apparently healthy males (aged from [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36] were recruited (from October 2013 to November 2014) based on their body mass index (half of them normo-weight and the rest overweight or obese) to test the effect of DHA and EPA in isolated PBMC maintained in vitro [24]. Sample size was based on the number of subjects analyzed in similar experiments conducted in ex vivo PBMC, where the conditions were highly controlled [27,28]. Participant characteristics and data on the efficacy of the bioactive compounds have been previously published [24]. Results suggested that genetic background could influence the anti-inflammatory response observed in PBMC under omega-3 PUFA exposure, evaluated in the present study. Therefore, culture media and plasma of the participants were used for cytokine and C-reactive protein (CRP) determinations, respectively. Saliva samples were obtained for the present study and used for genotyping. Furthermore, individual records and data, including those concerning the response to PUFA treatments, were retrieved.
In addition, in the present study, in order to increase the number of subjects and to prevent a potential source of bias, participants from the running studies OptiDiet-15 and Ob-IB were incorporated to perform the genotype analysis. The respective protocols for recruitment, procedures, and sampling were approved by the ethics committee (Comitè d'Ètica de la Investigació de les Illes Balears, CEI-IB). All participants gave written informed consent. Procedures were followed in accordance with the principles of the Declaration of Helsinki and all volunteer data were codified to guarantee anonymity accuracy.
The OptiDiet-15 (IB 2569/15 PI) and Ob-IB (IB 2009/13) studies, both contributed to the present work by making available a representative sample of the general population of Mallorca (Spain). Participants were mostly Caucasian and from European origin (86% European, 11% South-American; and 3% others). Records on anthropometric, lifestyle, clinical and genetic data were collected. Individuals were recruited from September 2013 to May 2017. Only individuals from the OptiDiet-15 and Ob-IB studies with both complete clinical and genetic data sets were included in the final report (n = 376). To focus on the hypothesis, a representative subset of male individuals was further recalled according to their genetic and anthropometric characteristics from either OptiDiet-15 or Ob-IB, around 10% fulfilled the requirements (41 subjects) and were further analyzed as described below. genetic association study to build up the score of predisposition to low-grade inflammation associated with obesity (LGI-Ob). This is based on three observational studies: One case-control (NUTRI-BLOOD), and two cross-sectional (OptiDiet-15 and Ob-IB) studies. NW (Normo-weight); OW/OB (Overweight/Obese); SNP (Single Nucleotide Polymorphism); CRP (C-reactive protein); BMI (Body Mass Index); TG (Triglyceride); HDLc (Cholesterol-High Density Lipoprotein); LDLc (Cholesterol-Low Density Lipoprotein).

Anthropometric Measurements
Height, weight, body fat percentage (BF%), waist, and hip circumferences were measured with a non-elastic tape during face-to-face interviews by a trained investigator with the volunteer wearing minimum clothing. The waist-to-height ratio (WtHR) was calculated. The percentage of body fat was measured with a bio-impedance apparatus (OMRON BF306, Hoofddorp, The Netherlands). Body mass index (BMI) was calculated by the ratio between weight and the square of the height. Systolic (SBP) and diastolic blood pressure (DBP) were measured in triplicate with a sphygmomanometer (Omron HEM705CP, Hoofddorp, The Netherlands), measurement was made after the volunteer remained seated for ten minutes. Mean of arterial pressure (MAP) was calculated by the following formula: 2/3.DBP + 1/3.SBP [29]. The waist and hip data were only available for the OptiDiet-15 and Ob-IB studies.

Isolation, Culture Conditions and Gene Expression Determination in PBMC
Blood from NUTRI_BLOOD volunteers was collected using Vacutainer ® EDTA tubes and PBMC were isolated using Ficoll-Paque Plus density gradient media (GE Healthcare Bio Science, Madrid, Spain). Cells were incubated with RPMI-1640 medium (Sigma-Aldrich, St Louis, MO, USA) supplemented with 10% of foetal bovine serum, 1% of L-glutamine (2 mM) and a combination of 100 units/mL and 100 µg/mL of penicillin and streptomycin, respectively (Sigma-Aldrich, St Louis, United States). Incubated cells were treated with omega-3 PUFA (DHA and EPA), prepared as a 10 mM stock solution dissolved in ethanol, and administered to PBMC in a final concentration of 10 µM. Both omega-3 PUFA (>98% purity) were purchased from Sigma-Aldrich (St. Louis, MO, USA). This concentration was chosen based on our previous optimization experiments showing a maximum effect on gene expression. Control cells were incubated with ethanol.
After a 48 h-incubation period, total RNA was isolated using Direct-zol RNA Mini-Prep (Zymo Research Corp, Irvine, CA, USA). Equal amounts of total RNA (50 ng) were transcribed into cDNA using an iScrip cDNA synthesis kit (Bio-Rad Laboratories, Madrid, Spain) in an Applied Biosystems 2720 Thermal Cycler. Real-time Polymerase chain reaction (PCR) was performed for each Reverse transcription (RT) product to determine mRNA expression of genes using Power SYBR Green PCR Master Mix (Applied Biosystems, Madrid, Spain). All primers were purchased from Sigma Genosys (Sigma Aldrich Química SA, Madrid, Spain). Gene expression data were normalized against the housekeeping gene ribosomal protein, large, P0 (RPLP0).
Differences in mRNA expression of IL-6 and TNFα by isolated PBMC incubated with DHA and EPA were calculated in comparison with the respective untreated control cells. Summing up these values, an overall response in the expression of pro-inflammatory genes (∆EPG) was obtained. Therefore, ∆EPG is an integrative value that reflects the feasibility of DHA and EPA in modulating the inflammatory potential; and a more negative ∆EPG would be suggestive of better responsiveness to the anti-inflammatory potential of the tested omega-3 PUFA in the respective individuals.

Quantification of Circulating Parameters
Determination of glucose, total cholesterol (TC) and triglycerides (TG) in blood were assessed with an Accu-trend device (Roche Diagnostics, Barcelona, Spain). High-density lipoprotein cholesterol (HDLc) levels were measured in EDTA plasma by means of a colorimetric kit (Química Clínica Aplicada S.A., Tarragona, Spain) following the manufacturer's instructions. Low-density lipoprotein cholesterol (LDLc) levels were calculated from Freideweld formula [30]: LDLc = TC − (HDLc + TG/5). C-reactive protein was measured with a commercial ELISA kit of Invitrogen (Thermo-Fisher Scientific, Waltham, MA, USA).

DNA Isolation and Genotyping
Genomic DNA was isolated from volunteer saliva sample (400 µL) using High Pure PCR template Preparation Kit (Roche, Basel, Switzerland). The quality and the concentration of the isolated DNA were assessed using Nanodrop 1000 UV-Vis Spectrophotometer (Thermo-Fisher Scientific, MA, USA). A final concentration between 20-100 ng/µL of the isolated DNA was considered great. Then, aliquots were stored at −20 • C until the analysis of polymorphisms. Genotyping was performed by qPCR (LightCycler ® 480 FastStart DNA or Genotyping Master mix, Roche, Basel, Switzerland) using the differential melting temperature to determine the alleles of each single nucleotide polymorphism (SNP). The qPCRs were done following these conditions: 95 • C for 5 min, 45 cycles of denaturation at 92 • C for 15 s and annealing 60 • C for 1 min. The melting temperatures were recorded and analysed to determine the presence of specific alleles. The genotyping of all SNP was done using expressly designed probes (TIB Molbiol, Berlin, Germany).

Elaboration of the Genetic Score of Predisposition to Low-Grade Inflammation Associated with Obesity (LGI-Ob)
Concerning the selection of genetic variants, we focused our interest on SNP related to an altered inflammatory profile, conferring major predisposition to obesity or higher cardiovascular risk. Genome-wide association studies (GWAS), meta-analyses and genetic association studies were the primary source for selection of the genes of interest. Additionally, to be part of the genetic score we imposed that the selected SNPs fulfilled the condition of being potentially modulated by lifestyle interventions, mainly by dietary factors. Therefore, preference was focused on SNP characterized by genotype-nutrient-phenotype associations, with the potential to modulate the genetic risk by external factors, and particularly in those nutrigenetic relationships involving polyunsaturated long chain fatty acids (Table 1).
LGI-Ob genetic score is based on five SNP, a number that can be easily affordable in routinely analytical tests, each one associated with a representative gene of obesity-related metabolic disturbances and all of them known for their interaction with plasma inflammatory biomarkers.
The main characteristics of the selected genes and SNP are summarized below. Two polymorphisms are located in the promoter and influence both gene expression and protein levels. Rs1800629 entails the substitution of a G for an A in the promoter of tumor necrosis factor alpha (TNFα) gene and increases both the transcription and production rates of the cytokine. A nutrigenetic interaction with omega 3 PUFA supplementation in the modulation of plasma CRP has been observed [31]; and associated with insulin resistance, higher levels of inflammatory metabolites in plasma [32], and other age-related alterations [33]. Rs5082, located in the promoter region of apolipoprotein A2 (APOA2) gene, has been shown to interact with dietary fatty acid intake to modulate inflammation [34,35], and is associated with both lower transcription rate and lower concentration of Apolipoprotein A-II in plasma [36], conferring higher risk of obesity and diabetes mellitus [37,38]. Furthermore, non-T subjects with high omega 3 PUFA intake show greater capacity for oxidative stress neutralization, whereas T carriers have been associated with resistance to increase superoxide dismutase 2 (SOD2) activity with omega 3 PUFA intake [39] and also with an increase in plasma oxidative biomarkers when they present a rich omega 6 PUFA diet [39].
The polymorphism rs4880 on SOD2 is located in the coding region and define a missense variant, it involves the change of Alanine to Valine at position 16 of the mitochondrial SOD2, which reduces free radical scavenging efficiency in the cell [40,41] and has been associated with altered plasmatic cytokine profile [42], with major risk to cardiovascular disease [43] type 2 diabetes [44], DNA damage [45] and micro-and macro-vascular complications in diabetes [46]. Interestingly, reduced cytokine production induced by the bioactive compound resveratrol in cultured PBMCs has been shown to be associated with the presence of this polymorphism, being less evident in Valine carriers [42]. Besides, Val risk allele carriers have been associated with lower benefits on inflammatory and lipid profile after lipid-lowering drug administration [47], but this genotype has also been associated with a lower risk of breast cancer in women who usually consume fish oil [48]. The last two SNP performing LGI-Ob genetic score have been associated with metabolic disturbances in different GWAs and other association studies. First, rs1260326 located in glucokinase regulator (GCKR) gene involves the change of the amino acid Proline at position 466 for Leucine and has been strongly associated with the risk of elevated CRP [49], dyslipidemia [49][50][51][52] and type 2 diabetes [53] in different GWAs. Nevertheless, two additional studies pointed out a potential modulation of omega 3 PUFAs on lipid profile and inflammatory plasma biomarkers [54,55]. Finally, we included the FTO gene, which codifies an alpha-ketoglutarate dependent dioxygenase and is one of the genes most strongly related with metabolic syndrome and obesity traits [56,57] and risk for type 2 diabetes (T2D) [58]. The intronic variant (rs9939609) has been recently analyzed by its influence on inflammatory biomarkers [56,58]. Despite the greater predisposition to obesity of AA subjects, the risk could be increased in the case of a PUFA:SFA ratio intake lower than 0.43 [59] or depending on the quality of dietary fat [59][60][61].  [67] CRP [56,58] Lower PUFA:SFA ratio are associated with higher obesity risk than in TT subjects [59] Lower effect of diet/lifestyle interventions [68]

Assessment of the Metabolic Profile and Health Status
Concerning participants from the studies OptiDiet-15 and Ob-IB, information about suffering hypertension (HT), dyslipidemia (DL), T2D, or visceral obesity (VO) was collected to assess the prevalence of metabolic syndrome. Participants were considered at risk of HT when they met any of the following WHO criteria [69]: Systolic blood pressure (SBP) ≥130 mmHg, diastolic blood pressure (DBP) ≥80 mmHg, taking medication for HT, or with a medical diagnosis of HT. The individuals who reported diagnosis for DL and T2D were considered at risk for those conditions. VO was inferred from the WtHR; those individuals with WtHR ≥0.51 (women) or WtHR ≥0.53 (men) were considered at risk. Finally, participants with two or more alterations of those named above were considered as individuals at high metabolic risk (MR) ( Table 2).

Statistical Analysis
Data analysis was performed with SPSS v21 (SPSS, Chicago, IL, United States). Normal distribution of variables was tested by the Shapiro-Wilk test when the number of cases was under 50 and by Kolmogorov-Smirnov test when the number of values was greater than 50. Homoscedasticity was tested using the Levene test; when variables did not show homoscedasticity criteria, they were either log transformed, or a non-parametric test was used.
Population was stratified by BMI group (normal-weight, NW: BMI < 25 kg/m 2 vs. overweight, OW/OB: BMI ≥ 25 kg/m 2 ) or by genetic risk score (low score: <5 vs. high score: ≥5; based on the median value of the frequency distribution in general population (see point 3.3)). Subject characteristics, mainly to assess differences between NW and OW/OB groups, were assessed using Student's t test. A Student's paired t-test was used to compare levels of cytokine production between EPA or DHA treated cells and control (no-treated) cells within high or low score groups.
Assessment of the relationship between LGI-Ob genetic score (as a continuous variable) and biomarkers (plasma parameters and gene expression data) was assessed by Pearson's or Spearman's regression tests. After stratification by BMI groups, correlation analyses adjusted by age were computed between plasma parameters and the genetic score. The coefficient of regression between LGI-Ob genetic score and the number of metabolic alterations was obtained using linear regression adjusting by confounding variables.
The contribution of the LGI-Ob genetic score (as a dichotomous variable) to find out the odds ratios (OR) for metabolic disorders or for metabolic risk were obtained by logistic regression, using low score subjects (<5 risk alleles) as reference (OR:1) and adjusting by the confounding variables age, gender, overweight (except for the variable VO), and ethnicity. Other potential confounding factors were non-reported or unknown.
Receiver operating characteristics (ROC) curve analysis was applied to test the diagnostic ability of LGI-Ob genetic score to classify potential individuals for obesity associated metabolic disturbances. No comparisons involving multiple simultaneous statistical tests were performed. The threshold of significance was defined at p < 0.05 for all analyses.

Anti-inflammatory Potential of EPA and DHA is Influenced by LGI-Ob Genetic Score
Analysis of gene expression focusing on the feasibility of omega-3 PUFA to modulate inflammatory response (∆EPG) and taking into account genotype characteristics of the individuals indicated that cellular response to DHA and EPA was dependent on the number of risk alleles carried. Thus, modulation of the expression of pro-inflammatory genes (∆EPG) by treatment with the omega-3 fatty acids showed a different profile related to the LGI-Ob genetic score of the individuals (Figure 2A). ∆EPG correlated positively with the LGI-Ob score in subjects with a high score (≥5) and reached statistical significance in the case of treatment with EPA (r = 0.669, p = 0.049; Figure 2A). However, a negative, although not significant, correlation was found in subjects with a low score (<5), which was in agreement with decreased pro-inflammatory cytokines in the medium of the cells (PBMC) treated with the fatty acids ( Figure 2B). In particular, both DHA and EPA were effective in decreasing the production of cytokines in culture media of PBMC from subjects with a low genetic score. Decreased IL6 was found after incubation with both DHA (∆ = −1.057, p = 0.035) and EPA (∆ = −1.450, p = 0.036) and decreased TNFα concentration was observed after DHA treatment (∆ = −6.175, p = 0.043) in samples from individuals with low score. However, in subjects with a high score only EPA decreased IL6 concentration in the media (∆ = −2.041, p = 0.042) ( Figure 2B). Thus, these data show that a previous heterogeneous inter-individual response to omega-3 PUFA observed in an in vitro system of PBMC of the characterized subjects [24] could be related to the presence of genetic variants linked with a status of low-grade inflammation associated with obesity. The selected genes and role of the variants included in the LGI-Ob score shaped the anti-inflammatory response to PUFA, showing that individuals with elevated score were to some extent resistant to the anti-inflammatory potential of EPA and DHA, an effect observed at the level of both transcriptional and translational processes. Pearson correlations were used to assess linear association between LGI-Ob genetic score and ∆EPG; (B) IL-6 and TNFα production in PBMC culture media by LGI-Ob genetic score: Low (<5) or high (≥5). Two-way repeated measures ANOVA was performed to compare differential cytokine production from control cells vs EPA or DHA treated cells in both LGI-Ob genetic score groups (* p < 0.05). Boxes represent the dispersion (maximum and minimum) and the horizontal line represents the arithmetic mean in each treatment. Data correspond to subjects from NUTRI-BLOOD study (n = 18).

Plasma Biomarkers of Inflammation are Influenced by LGI-Ob Genetic Score
To further assess the extent of genetic impact on inflammation, CRP levels were measured in plasma from the above individuals (volunteers of the NUTRI-BLOOD study, n = 18). A positive correlation (Rho = 0.580 p = 0.015) was observed between CRP concentration and the LGI-Ob score. Interestingly, stratification by BMI gave a stronger correlation when only OW/OB (BMI ≥ 25) individuals were included (Rho = 0.737, p = 0.023, n = 9), which was lost for NW subjects (BMI < 25) (Rho = 0.275, p = 0.509, n = 8). To rule out these data as a result of the specific population studied or associated with the low number of individuals analysed, data from the initial core group of participants (NUTRI-BLOOD study, n = 18) were pooled with a subset of individuals from Optidiet-15 study (n = 41), all of them adult men and representative of the general population (Table 3). Accordingly, subjects covered a wider range of age (19-74 years old) and overweight and obesity were more prevalent in individuals 10 years older than in normo-weight subjects (p = 0.002). After grouping by BMI, higher MAP (p = 0.037), plasma triglycerides (p = 0.016), and CRP (p < 0.0001) levels were observed in OW/OB subjects, but no differences were found in glucose, or cholesterol plasma levels (Table 3). Positive correlations were found between CRP and obesity related indicators, such as BMI (Rho = 0.664, p < 0.0001), body weight (Rho = 0.554, p < 0.0001), and %BF (Rho = 0.612, p < 0.0001; n = 52). Correlation analyses confirmed the positive relationship between plasma CRP and LGI-Ob genetic score (Rho = 0.309, p = 0.026; n = 52) ( Figure 3A), as well as the positive correlation found in OW/OB subjects (Rho = 0.526, p = 0.003; n = 29), and the lack of it in NW individuals (Rho = 0.027, p = 0.903; n = 23). Analyses performed after adjusting by age, a potential confounder co-variable, remained significant (r = 0.503, p = 0.005) in the overweight/obese group, but not in NW group (r = 0.056, p = 0.806) ( Figure 3B). In addition, we tested the association between CRP in the non-segregated population using the simple linear regression model adjusted by age and by BMI (p = 0.059) and adjusted by age and presence of overweight (p = 0.018). Therefore, the risk of higher basal inflammation entailed by adiposity seems to affect individuals unequally depending on their LGI-Ob genetic score. Furthermore, NW subjects with a high genetic score would be more susceptible to low-grade chronic inflammation in case of increasing body fat.
Elevated levels of TG and low levels of HDLc have been described as key contributors to the development of inflammation and atherosclerosis [70], being associated with insulin resistance and T2D [71][72][73][74]. Although HDLc was not altered in the obese population studied, TG levels were 36% higher than in NW subjects (Table 3). Therefore, a positive correlation between the TG/HDLc ratio and LGI-Ob score was also found in overweight/obesity group (r = 0.418, p = 0.030; n = 28), adjusting by age, but not among NW subjects ( Figure 3C). Using the same model (adjusted by age and overweight), the association between TG/HDL ratio and LGI-Ob score remained significant (p = 0.038); and if adjusted by BMI, the significance was not reached either (p = 0.061). Then, to determine the potential interaction between BMI and the genetic risk score, in relation to the levels of CRP and TG/HDL, both variables were dichotomized, and a two-way ANOVA test was performed, adjusting by age. Under these conditions, simple effects were statistically significant, but interaction effect was not seen.
Recently, the TG/HDLc ratio has been proposed as a reliable biochemical index predictor for insulin resistance, particularly in overweight diabetic patients [75]. Our results allow for a further level of precision, suggesting that the presence of the risk alleles considered in LGI-Ob genetic score is able to influence the TG/HDLc ratio. Therefore, individuals with a high genetic score are more susceptible to showing a disturbed ratio, and would thereby be at higher risk to develop insulin resistance and associated co-morbidities [71].

LGI-Ob Genetic Score May Contribute to Identify Subjects at Higher Risk to Develop Metabolic Syndrome
All the above data point to the fact that the LGI-Ob genetic score could be relevant in mediating a higher rate of adverse events in the relationship with obesity and cardio-metabolic health. The next step was to assess the association of genetic score with known risk factors of metabolic syndrome in a wider, representative sample of the general adult population, incorporating the female gender. Information about health status and genotypes of interest was obtained from 376 subjects (63% women) ( Table 4). The frequency of LGI-Ob score followed a normal distribution with a median of 4 (IQR = 2) and a mean of 4.36 (SD = 1.7) alleles of risk ( Figure 4A). Therefore, the median value was used to stratify the subjects in two subgroups of genetic risk: Low (<5) and high score (≥5) throughout the study.
Additionally, information concerning the presence of metabolic alterations (i.e., hypertension, dyslipidaemia, diabetes, and visceral obesity) in these subjects was recorded and analysed. Interestingly, the number of altered metabolic alterations showed a positive linear regression with the LGI-Ob genetic score (β = 0.145, p = 0.001). In fact, higher prevalence of two or more metabolic alterations was observed in individuals with a high genetic score, whereas individuals with a low score were more represented in the categories with no or only one metabolic alteration ( Figure 4B). In order to characterise the prevalence of metabolic risk, a logistic regression analysis was undertaken. When compared to low score subjects (OR = 1), the high scoring group had higher odds ratio for all metabolic alterations registered, and reached statistical significance for hypertension (OR = 1.58, p = 0.047), T2D (OR = 3.26, p = 0.026), and for the presence of two or more risk factors (relative metabolic risk, MR) (OR = 1.87, p = 0.021) ( Figure 4C). To check the accuracy of predicting the incidence of metabolic disturbances associated with obesity, MR variable was selected as a binary outcome and ROC analysis was performed. The ROC analysis showed a significant discriminant accuracy of the LGI-Ob in identifying subjects having metabolic disturbances related with obesity (Area Under Curve, AUC= 0.613, p = 0.001, confidence interval, CI = 0.551-0.674) ( Figure 4D). Genetic variants in FTO gene are highly relevant in obesity and to assess whether LGI-Ob score would perform better than FTO on its own, we tested the removal of FTO polymorphism from the model. This had a minor impact on AUC value (AUC = 0.582, p = 0.036), suggesting that the remaining SNP in LGI-Ob model had major influence with the inflammatory profile associated with obesity than FTO per se. Therefore, the LGI-Ob score as initially defined (i.e., including the influence of FTO polymorphism) would support a higher degree of confidence in the discriminant accuracy of LGI-Ob score identifying subjects having metabolic disturbances related with obesity.

Discussion
Obesity is a multifactorial condition associated with metabolic disturbances, such as type 2 diabetes and cardiovascular diseases. Its treatment and prevention are difficult as it is highly influenced by both genetic and environmental factors, as well as by their interactions. In fact, the contribution of genetics to obesity and adiposity traits has been extensively analyzed and, to date, over 500 loci have been identified and constitute targets for active research (see [76,77] for recent reviews). Interestingly, functional characterization of the genes involved has revealed the role of immune-related cells and pathways, which pin-points the relevance of low-grade inflammation observed in obesity [3]. Thus, preventive or therapeutic strategies aiming to counteract immune-metabolic dysfunction may constitute an essential step in resolving obesity. In this respect, long chain omega 3 PUFA have been shown to induce an anti-inflammatory profile at the level of gene expression in immune cells [78][79][80]. In a previous study, we contributed to show the anti-inflammatory potential of EPA and DHA in humans, measuring the expression of inflammatory key genes and determining levels of cytokines in the medium of an in vitro system of PBMC isolated from apparently healthy adult men. However, the attenuated anti-inflammatory reaction was observed in cells from overweight/obese individuals, together with highly heterogeneous inter-individual responses [24]. Thus, in the present study we aimed to investigate to what extent inflammatory status and the beneficial effect associated with long chain omega 3 PUFA would be modulated by genetic background. A number of nutrigenetic studies have implemented the use of genetic risk scores, as they seem to have greater power than single variants and may be useful in elucidating the biological function of the SNP [77,81]. Therefore, a genetic score expected to identify the risk of Low-Grade Inflammation associated with Obesity (LGI-Ob) was elaborated. The score was based on the presence of 5 common variants known for their influence on plasma inflammatory biomarkers (TNFα and CRP), some of which have already been analyzed as regards PUFA interaction [31,34,42,53,56]. Application of LGI-Ob genetic score was initially performed in the core group of participants of the previous study, then it was tested in a medium-sized group of men, and finally in a totally heterogeneous sample of people representative of the general Caucasian population. The LGI-Ob genetic score followed a bell-shaped distribution with an observed range from 0 to 10. Taking the median value of the whole population, the score was predictive of higher risk (odds ratio) of metabolic disturbances, reaching statistical significance for HT, T2D, and MR. Furthermore, when the genetic predictor was contrasted with phenotypic features in the smaller subsets of individuals, it was positively correlated with a disturbed TG/HDL ratio, higher CRP, and also even decreased anti-inflammatory response to EPA and DHA. ROC analysis of the LGI-Ob score as a predictor of developing metabolic disturbances and associated low-grade inflammation gave an AUC ROC of 0.613, which is not a very high value. However, the current genetic models to predict obesity based on SNP associated with BMI loci deal with AUC ROC between 0.546 (when FTO is the only gene considered) up to 0.575 (when considering 12 or 32 loci) [82]. Therefore, available genetic information is relatively poor and does not allow for accurate discrimination between those at high risk of obesity. The LGI-Ob score identified in our study may represent a good genetic marker and may be useful at earlier steps aimed at evaluating an individual's predisposition to metabolic disturbances associated with obesity and low-grade inflammation. Thus, an individual's genetic information may enable the prevention and management of metabolic syndrome, following the current trend towards personalized nutritional approaches, which in addition may efficiently contribute to tailored dietary advice aiming for precision nutrition [12,83]. In summary, our results show that genetic data could be combined with other usual analytical parameters (i.e., CRP, HDLc, and TG) and/or contrasted with a functional test assessing in vitro performance of DHA/EPA concerning anti-inflammatory response. In this regard, the possibility of using the results from functional assays, for example testing in PBMC the response to omega 3, or other bioactive compounds with anti-inflammatory potential, can help determine the most appropriate intervention for each individual, as stated in previous works [24,84,85]. The knowledge provided by introducing the analysis of specific genetic variants could help to clarify some of the null or contradictory results found on Omega-3 supplementation studies [86,87].
Obesity may be exerting a synergistic effect with genetic background to disturb the anti-inflammatory potential of DHA and EPA. Hence, this study highlights that a set of relevant genetic variants associated with genes involved in energy metabolism and inflammatory processes could contribute to a better prediction of an individual's metabolic resistance to the anti-inflammatory capacity of omega 3 PUFA. Therefore, the genetic score of predisposition to low-grade inflammation associated with obesity (LGI-Ob) defined in this study may efficiently contribute to discern population at risk for metabolic syndrome.