In the context of the triple burden of malnutrition: A systematic review of gene-diet interactions and nutritional status

Abstract Genetic background interacts with dietary components to modulate nutritional health status. This study aimed to review the evidence for gene-diet interactions in all forms of malnutrition. A comprehensive systematic literature search was conducted through April 2021 to identify observational and intervention studies reporting the effects of gene-diet interactions in over-nutrition, under-nutrition and micronutrient status. Risk of publication bias was assessed using the Quality Criteria Checklist and a tool specifically designed for gene-diet interaction research. 167 studies from 27 populations were included. The majority of studies investigated single nucleotide polymorphisms (SNPs) in overnutrition (n = 158). Diets rich in whole grains, vegetables, fruits and low in total and saturated fats, such as Mediterranean and DASH diets, showed promising effects for reducing obesity risk among individuals who had higher genetic risk scores for obesity, particularly the risk alleles carriers of FTO rs9939609, rs1121980 and rs1421085. Other SNPs in MC4R, PPARG and APOA5 genes were also commonly studied for interaction with diet on overnutrition though findings were inconclusive. Only limited data were found related to undernutrition (n = 1) and micronutrient status (n = 9). The findings on gene-diet interactions in this review highlight the importance of personalized nutrition, and more research on undernutrition and micronutrient status is warranted.


Introduction
Malnutrition in all its forms, including micronutrient deficiencies, undernutrition and overnutrition, remains a leading cause of global mortality and morbidity (Popkin, Corvalan, and Grummer-Strawn 2020).Suboptimal diets are the critical environmental factor involved in the development of malnutrition and diet-related disease (Afshin et al. 2019).However, individual dietary requirements differ from person to person, with genetic variation influencing individual's dietary requirements and nutritional status (Stover 2006).Thus, nutritional requirements and dietary recommendations may not always be generalizable to entire populations.In addition, dietary components can modulate the expression of genes involved in metabolic pathways determining nutritional status and health outcomes (Fenech et al. 2011).This can be mediated either through nutrient-regulated transcription factors, or through epigenetic mechanisms such as DNA methylation and histone modification (Tiffon 2018).
Current government recommended nutrient intakes (e.g., recommended dietary allowances (RDAs) or safe upper limits) are designed for the general population based on different metabolic outcomes or population subgroups, such as the elderly, pregnant or lactating women (Ordovas et al. 2018).These recommendations are not optimized for genetic subgroups that may differ in response to different dietary components.The understanding that there is no one-size-fits-all diet contributed to the emergence of nutritional genomics research.This includes both nutrigenetics, which studies the role of genetic variations on nutrient metabolism and diet-related disease, and nutrigenomics, which examines the role of nutrients or dietary patterns on gene expression (transcriptomics); along with more broadly, proteomics and metabolomics (Phillips 2013).Requisite to both nutrigenomics and nutrigenetics research, is an understanding of nutrition, genetics, biochemistry and application of a range of 'omics' technologies to investigate the complex interaction between genetic and environmental factors relevant to metabolic health and disease (Matusheski et al. 2021).Precision nutrition is a more recent term, often used interchangeably with personalized nutrition, nutritional genomics, nutrigenomics or nutrigenetics (Bush et al. 2020).The objective of precision nutrition is the identification of personalized nutritional recommendations that are tailored according to an individual's biological requirements and predicted response to dietary intervention based on genetic, metabolomic and microbiome profiling (de Toro-Martin et al. 2017).Research in this area aims to provide a better understanding of nutrient-gene interactions with the ultimate goal of developing personalized nutrition strategies for optimal health and disease prevention.
A growing number of companies have offered the services of direct-to-consumer genetic testing (DTC-GT) personalized dietary advice in the past decade (Phillips 2016).Next-generation sequencing (NGS) technologies that facilitate the sequencing of whole genome at an unprecedented speed and simultaneous high-throughput testing of multiple genes (Metzker 2010), have further enabled nutrigenomics and nutrigenetics research.The advances in sequencing technology have driven down the cost of human sequencing, with a current estimated sequencing cost of $1,000 US dollars per genome (Moore 2020).These advantages have facilitated and accelerated research and development of precision nutrition strategies for the prevention and treatment for nutritional conditions.However, questions remain about efficacy, cost-benefit and accessibility, with the current scientific evidence suggesting that it is still premature to apply the use of precision nutrition in the community (Moore 2020).
The most commonly investigated type of genetic variation in genome-wide association studies (GWAS) has been single base pair differences, termed single nucleotide polymorphisms (SNPs) (Bush and Moore 2012).Other types of genetic variation include structural variants (SVs), such as insertions and deletions of short DNA fragments (INDELs), and copy number variants (CNV) where the number of copies of particular genes varies between individuals (Feuk, Carson, and Scherer 2006).To date, most of the nutritional genomics research has focused on SNPs, therefore these were the central focus for this review.Moreover, in contrast to obesity, far fewer studies have investigated genetic variants associated with under-nutrition or micronutrient status, and not all micronutrients have been studied.Therefore, this study aimed to systematically review the current evidence on the effects of gene-diet interactions on nutritional status including undernutrition (stunting, wasting and underweight), overnutrition (overweight and obesity) and micronutrient status; particularly that of iron, zinc, folate and vitamin A. This included examining the effects of genetic variants on phenotypes in response to nutrient or diet.

Methods
This systematic review was conducted by three independent reviewers following PRISMA guidelines and was prospectively registered at PROSPERO (registration no.CRD42021245115).

Eligibility criteria
The PICOS criteria for inclusion and exclusion in this review are shown in Table 1.Both observational studies and intervention trials involving human participants of all age, gender and ethnicity that investigated the effect of gene-diet interactions defined as the combined effect of the two exposures on nutritional status were included.

Search strategy
Four databases, PubMed, Scopus, Embase (Ovid) and Web of Science, were systematically searched using the combination of keywords and terms (e.g., MeSH and Emtree) through 30 th April 2021.The search was conducted using the following keywords from three main themes: i) gene exposure ("single nucleotide polymorphism*" OR SNP* OR gene* OR genetic* OR polymorphism* OR genotype* OR Table 1.PiCos criteria for inclusion and exclusion of studies.

Population
Human participants <65 years old of all gender or ethnicity.animal studies; human participants who were pregnant, using drugs or treatments in past 12 months, suffering from infectious diseases, or previously diagnosed conditions that may lead to changes in metabolic and nutritional needs including postoperative recovery, all types of cancer, renal disorders, mental health disorders, and endocrine disorders.

Exposure or Intervention
studies that investigated gene-diet interactions, which defined as the combined effects of the two exposures including genetic (e.g., single nucleotide polymorphisms, gene variation and mutation, gene expression and epigenetics such as microrna expression, dna methylation, acetylation, and histone modification) and dietary factors on nutritional status.
studies that reported individual exposure effect.

Comparison
those who did not expose to the two exposures.

Outcome
Malnutrition including undernutrition (stunting, wasting, underweight, and micronutrient deficiencies) and overnutrition (overweight and obesity).the parameters included height, weight status and adiposity such as body weight, body mass index (BMi  S1.

Study selection
Screening of the identified studies and selection of studies for inclusion in this review based on the eligibility criteria detailed above were performed independently by three reviewers using both Endnote (Endnote X7.7.1, Thomson Reuters 2016) and Rayyan (http://rayyan.qcri.org)tools.The final decision regarding the eligibility of articles was made by agreement between the three reviewers.Disagreement between reviewers was resolved by discussion and by other reviewers when necessary.

Data extraction
A standardized data extraction form was utilized to obtain the following information: author, year of publication, study design, year of study, sample size, country or population, sample characteristics (e.g., gender, age and BMI), exposure or intervention (both gene and dietary exposures), outcome measures (e.g., indicators related to malnutrition), main findings (e.g., β coefficient, odds ratio, differences in mean) and statistical analysis.In the case of missing data or unclear pieces of information, it was considered that the authors did not report such variables.

Risk of bias assessment
Risk of bias in the individual studies included was assessed by independent reviewers using the Academy of Nutrition and Dietetics, Quality Criteria Checklist (2016 Evidence Analysis Manual, Academy of Nutrition and Dietetics).The 10 questions of the checklist focus on: (1) how clear the research question was, (2) selection of participants, (3) randomization/group comparability, (4) description of withdrawals, (5) how the blinding was conducted, (6) whether study procedures were described clearly, (7) whether the outcomes were clearly defined and the validity of the measurements, (8) were appropriate statistical analyses applied, (9) did the results support author's conclusions, and (10) was there funding or sponsorship bias.To be rated low risk of bias, each of criteria 2, 3, 6 and 7 must be met and the majority of 10 criteria overall.Any of criteria 2, 3, 6 and 7 not being met resulted in a neutral rating.If most criteria are not met, the article was rated high risk of bias.
In addition, a specific assessment to evaluate the methodological quality of gene-diet interaction research was performed following the criteria important for genetic association studies (Campbell and Rudan 2002;Dietrich et al. 2019).The score was based on eight items (Table S2): (1) interaction as primary study goal, (2) statistical test for interaction, (3) correction for multiple testing, (4) correction for ethnicity or population structure, (5) Hardy-Weinberg equilibrium testing, (6) test for group similarity at baseline or the comparability of case and control, (7) sample size or power analysis, and (8) sufficient details of study procedure.Based on a scoring of positive (+1), neutral (0) or negative (-1) for each item, the total points for each paper could range from −8 to 8; and were assessed as: high (6 to 8 points), neutral quality (2 to 5 points), and low (−8 to 1 points) quality.

Data synthesis
Tables were constructed to synthesize the evidence of gene-diet interactions on nutritional status based on the study design and measure outcomes, and were ordered based on the types of dietary components or interventions.GraphPad Prism 9 was used to generate a Forest plot (without calculating a summary measure given high heterogeneity of the studies) to summarize the interactions between genetic risk score (GRS), dietary components and BMI.GRS is defined the weighted sum of an individual's trait-associated alleles which derived from genome-wide associated study (GWAS) data.

Results
As shown in Figure 1, the literature search generated 11,881 records from four databases, with an additional article found through the citation lists.After removing duplicates, 7,740 articles were screened by title and abstract, and 278 full-text articles were assessed for eligibility.Of those, 167 articles were included in this review, which were comprised of: 157 articles (n = 101 observational studies and n = 56 intervention trials) that had investigated the interactions between genetic variants and dietary components on obesity; one article that reported on height; and 9 articles that reported on micronutrient status (n = 5 observational studies and n = 4 intervention trials).A few intervention trials (n = 12) investigated the interaction between weight loss outcomes and gene expression, but as beyond the scope of this review will not be discussed.
Study participants were recruited from 27 different populations, mainly from North and South American (e.g., the United States (US), including Alaska and Puerto Rican, Canada, Mexico, and Colombia) and Europe (the United Kingdom (UK), Spain, Italy, Denmark, Brazil, Sweden, the Netherland, France, Poland, Greece, Belgium and Finland).Whereas, a smaller number of studies were reported from the Middle East (Iran, Israel and Lebanon), East Asian (Korea, Japan, China), Southeast Asian (Singapore, Malaysia and Indonesia), and the Southwestern Pacific Ocean (New Zealand) populations.The age of participants ranged from 0 to 92 years, and the sample sizes ranged from 110 to 119,132 for observational studies and from 32 to 1,852 for intervention trials.

Risk of bias
Based on the Quality Criteria Checklist, 101 observational and 26 intervention studies were assessed as having low risk of bias, whereas 6 observational and 34 intervention studies received a neutral rating (Table S3).The reasons for lower quality ratings for intervention trials were typically: lack of blinding procedure, non-randomization, inappropriate statistical analysis (e.g.without intention-totreat analysis or no adjustment for baseline values or confounding factors), and no detailed description for withdrawal.In addition, the evaluation of methodological quality specifically designed for gene-diet interaction research found 97, 69 and 1 studies as having low, medium and high risk of bias, respectively (Table S3).The reasons for score reductions were mainly due to studies' lack of: power analysis or insufficient sample size, application of Hardy-Weinberg Equilibrium and correction for multiple testing.

Observational studies: Gene-diet interactions and weight status
The evidence for the effects of gene-diet interactions on weight status from observational studies in the context of overnutrition (n = 101) are summarized in Table 2.We identified only one study reporting the effect of gene-diet interaction on undernutrition, assessed by body height.The authors reported that children in Greece carrying the risk allele (A allele) of insulin-like growth factor II (IGF) rs680, who consumed high intake of dairy product were taller compared to those with low intake of dairy product (Dedoussis et al. 2010).
With respect to dietary patterns, studies showed those who either had higher GRS or were carrying the risk alleles of the genetic variants examined had unfavorable outcomes on body weight, BMI or adiposity when following unhealthy dietary patterns.This included meal skipping, snacking and diets with high inflammatory index, as well as meat-based diets and Western diets (Hosseini-Esfahani et al. 2019;Jaaskelainen et al. 2013;Masip et al. 2020;Wang, Tang, et al. 2014;Yarizadeh et al. 2021;Zhang et al. 2015;Zhu, Xue, Guo, Deng, et al. 2020).On the other hand, increased adherence to healthy dietary patterns rich in vegetables, fruits and whole grains was more likely to be associated with lower BMI or body fat (Barchitta et al. 2014

GRS and dietary components on BMI from observational studies
A total of 29 studies reported the interactions between GRS comprised of multiple SNPs and dietary components on obesity related phenotypes.These mainly investigated interactions with dietary fats (n = 11) (Alsulami, Nyakotey, et  The number of SNPs included in the GRSs utilized in these studies ranged from 2 to 1,148,565 (Masip et al. 2020).A forest plot illustrates the GRS-diet interaction on BMI using β coefficient and 95% CI (Figure 2).Given the tremendous heterogeneity in the investigated dietary components, as well as the genetic variants included, and approaches Figure 2. Forest plot of the interactions between Grs and dietary components on BMi (kg/m 2 ) from observational studies.Ci; confidence interval, Grs; genetic risk scores, PuFa; polyunsaturated fats, sFa; saturated fats, snPs; single nucleotide polymorphisms taken to estimate the GRS the results are presented without calculating a summary estimate.In Europe, studies from Finnish populations, including FinnTwin16 (FT16) and Northern Finland Birth Cohort of 1986 (NFBC1986), reported that unhealthy dietary practices such as snacking (β = 0.27, 95% CI = 0.21, 0.34) (Masip et al. 2020) and meal skipping (β = 0.06, 95% CI = 0.02, 0.09) (Jaaskelainen et al. 2013) were associated with increased BMI among those with higher GRS based on 996,919 and 8 SNPs, respectively.Separately, findings from the UK Biobank reported that intakes of total energy (β = 0.56, 95% CI = 0.50, 0.63), total fat (β = 0.58, 95% CI = 0.51, 0.64), and SFA (β = 0.65, 95% CI = 0.59, 0.72) were associated with increased BMI among those with higher GRS based on 93 SNPs (Celis-Morales et al. 2017).
Findings from an East Asian population who were genetically predisposed to obesity, the Takahata cohort study, showed that fiber intake was positively associated with BMI (β = 0.15, 95% CI = 0.04, 0.26); however, surprisingly per gram increases in vegetable fat and animal protein intake resulted in lower BMI (Nakamura et al. 2016).The authors opined that dietary fiber alone may not be sufficient to control weight among the participants with higher GRS, which may be due to the associated macronutrient intake that collectively affected the body weight.Furthermore, the authors note there are different types of fiber e.g., soluble or insoluble fiber, which were not differentiated in their analysis.
However, other studies reported low-carbohydrate-high fat diets (30-45% of energy from carbohydrate and 40-45% energy from fat) were more effective in reducing body    risk of bias (general) 1 was assessed by Quality Criteria Checklist and risk of bias (genetic) 2 was assessed by a methodology quality evaluation method specifically tailored to gene-diet interaction research ↓ indicates significant greater reduction in weight loss outcomes among the risk alleles carriers; ↑indicates significant smaller reduction in weight loss outcomes; ↔ indicates no significant post-intervention changes in weight loss outcomes between the risk alleles and non-risk alleles carriers.
weight and fat mass among those carrying the minor alleles of HNF1A rs7957197, Neuropeptide Y (NPY) rs16147, TFAP2B rs987237 in US and European populations (Huang et al. 2018;Lin et al. 2015;Stocks et al. 2012), but not among the risk alleles carriers of FTO rs9939609 (A), Angiotensin II Receptor Type 2 (AGTR2) rs5950584 (T) and VEGFA rs1358980 (T) (de Luis, Aller, Izaola, de la Fuente, et al. 2012;Seip et al. 2008;Svendstrup et al. 2018).This may be dependent on fat quality, as other studies reported that MUFA-and PUFA-enriched diets were found to be effective in reducing body weight and fat mass among the Spanish who carrying the risk alleles of Fatty Acid Amide Hydrolase (FAAH) rs32440 and Fatty Acid Binding Protein 2 (FABP2) Ala54Thr, respectively (de Luis, Aller, Izaola, Sagrado, et al. 2012;de Luis, Aller, Izaola, Conde, et al. 2013).
Findings from the US POUND LOST trial (n = 692; 61% females) showed that overweight and obese individuals with a lower GRS computed from 7 SNPs had significantly greater reduction in body weight (p = 0.003) and WC (p = 0.014) after 6 months of consuming low-fat diets compared to those with higher GRS (Li et al. 2020).On the other hand, the US MOVE! Programme (n = 51 overweight and obese adults; 25% females) reported no significant difference in weight loss outcomes after 24 weeks between those following nutrigenetic guided diets and standard balanced diet (Frankwich et al. 2015).In a European population, the NUGENOB trial reported that women with the highest decile of a GRS had significantly greater reduction in weight compared to the lowest decile (7.3 ± 3.0 kg versus 4.9 ± 3.0 kg) after 10 weeks of low-fat or low-carbohydrate hypocaloric diets (reduction of 600 kcal/day) (Svendstrup et al. 2018).Plus greater weight loss was observed in Dutch adults with higher GRS (β ± SE= −0.52 ± 0.18, p = 0.004) after consuming protein-enriched VLCD for 5 months (Verhoef et al. 2014).However, individuals with low GRS and high adherence to MD had a greater reduction in BMI and WC compared to low MD (San-Cristobal et al. 2017).However, in an Asian population, no significant effect of GRS was found on weight loss outcomes in response to both the HIPCREF (Individualized high-protein, energy-restricted, high-vitamin E and high-fiber) diet and a standard diet (based on Malaysian Dietary Guidelines 2010) among the overweight and obese Malaysian adults (Tan and Mitra 2020), but those with higher GRS had a significantly greater reduction in C-reactive protein levels after HIPCREF diet compared to the standard diet.

Effects of gene-diet interactions on micronutrient status
Very few studies (5 observational studies and 4 intervention trials) investigated gene-diet interactions on micronutrient status (Table 4).From these, only Aldehyde Dehydrogenase 2 Family Member (ALDH2) rs671 was found to modulate both BMI and micronutrient status (Tao et al. 2019;Wang et al. 2016).Chinese men carrying A alleles of ALDH2 rs671 who consumed alcohol (both > 0 < 10 g/day and ≥10 g/day groups) had significantly lower serum ferritin levels compared to men carrying GG genotypes (Tao et al. 2019), but this association was not observed in the non-alcohol drinkers.While in a separate study, Chinese men carrying A alleles of rs671 were observed to have lower visceral fat accumulation with lower alcohol consumption (OR = 0.27, CI = 0.09-0.23,per copy of A allele), suggesting a genetic interaction mediating BMI in the context of alcohol consumption (Wang et al. 2016).
Higher red blood cell folate levels were observed with increased folate intake in the US population carrying the TT genotype of Folate Hydrolase 1 (FOLH1) T484C, compared to C allele carriers (Cummings et al. 2017).However, Japanese women carrying the TT genotype of Methylene tetrahydrofolate reductase (MTHFR) 677 C > T had significantly lower serum folate with increased intake of folic acid compared to non-carriers (Hiraoka 2004).With respect to the intervention studies, four trials have investigated the modulatory effect of MTHFR 677 C > T on serum or urinary folate levels in response to folate supplementation (95,191,400 or 800 μg/d of folate).No significant post-intervention (8 weeks to 3 months) differences were observed in serum folate among Colombian and Brazilian women (Arias et al. 2017;Lisboa et al. 2020).However, in two separate studies of Mexican women and men, the T alleles carriers of MTHFR 677 C > T had significantly lower serum folate levels after folate supplementation (12-14 weeks) compared to non-carriers (Guinotte et al. 2003;Solis et al. 2008).
Separately, UK women carrying the YY genotype of the Homeostatic Iron Regulator (HFE) C282Y allele had significantly higher serum ferritin with increased heme iron intake (Cade et al. 2005).Increased risk of anemia with higher consumption of coffee (≥4 cup/d) was reported in those carrying the C allele of Mitochondrial DNA rs28357984 (Kokaze et al. 2014).

Discussion
In this study we comprehensively reviewed the current literature on the effects of gene-diet interactions on nutritional status.This included examining the effect of genetic variants on undernutrition, overnutrition and micronutrient status (iron, zinc, folate and vitamin A), and in response to nutrient or dietary intakes.

GRS and dietary patterns on obesity
The vast majority of the identified studies focused on the impacts of gene-diet interactions on BMI or obesity risk, while only a limited number reported data related to under-nutrition and micronutrient status.Notably, most of the interaction findings have yet to be replicated in controlled trials or across diverse populations, with a particular paucity of data from Asian populations.Both observational studies and intervention trials consistently demonstrated the beneficial effects of avoiding diets that are high in total fat, SFA, TFA and n-6 PUFA; as well as increasing MUFA and n-3 PUFA intakes, for reducing obesity risk and obesity-related phenotypes and providing better weight loss outcomes among individuals genetically predisposed to obesity.
Most of the observational studies to date have focused on the investigation of individual nutrients, food items, or individual SNPs.However, people do not consume single nutrients or foods but rather a combination of many foods, and the complexity and multidimensionality of a normal diet can confound dietary intervention studies.To address this issue, analysis of food consumption patterns can be measured using both a priori and a posteriori approaches (Hu 2002).Prospective approaches measure either an individual's adherence to a specific diet e.g., Mediterranean, AMED, or DASH diets; or measure diet quality through scores e.g., AHEI-2010.For retrospective approaches, these typically use principal components analysis (PCA) and clusanalyses to assess diet patterns based on foods consumed, such as the Western, plant-based or meat-based diets (Stricker et al. 2013).
Genetic variants most commonly investigated in gene-diet interaction studies in relation to obesity, were SNPs in the FTO gene (rs9939609, rs1121980 and rs1421085), followed by MC4R (rs1778231), PPARG (rs1801282) and APOA5 (rs662799) genes (Livingstone, Celis-Morales, Papandonatos, et al. 2016;Razquin, Marti, and Martinez 2011;Tan, Mitra, and Amini 2018;Xiang et al. 2016).There were consistent evidence showing the associations between increased intake of total fat and SFA and increased BMI in the individuals carrying the risk allele of FTO rs9939609, rs1121980 and rs1421085, including those assessed by GRS with FTO SNPs inclusive (Alsulami, Nyakotey, et al. 2020 et al. 2015;Sonestedt et al. 2009;Sonestedt et al. 2011;Labayen et al. 2016;Park et al. 2013), and such associations were not observed in other macronutrients.Whereas increased intake of PUFA (Goni et al. 2015;Huang, Wang, Heianza, Wiggs, et al. 2019;Lemas et al. 2013;Riedel et al. 2013) or increased adherence to diets rich in vegetables, fruits and whole grains that assessed using the scoring systems such as MD, DASH, AMED, AHEI-2010, dietary diversity and healthy diet index scores (Ding et al. 2018;Goodarzi et al. 2021;Han et al. 2020;Seral-Cortes et al. 2020;Wang et al. 2018) showed better reduction in BMI and body fatness in the FTO risk allele carriers compared to the non-carriers.
Both FTO and MC4R genes are involved in the regulation of appetite and energy intake (Adan et al. 2006;Fawcett and Barroso 2010;Olszewski et al. 2009), while PPARG and APOA5 mediate adaptive thermogenesis (Wu, Cohen, and Spiegelman 2013) and lipoprotein metabolism (Su, Kong, and Peng 2018) respectively.Nonetheless, inconclusive findings were found on the interactions between dietary intake and the SNPs in MC4R, PPARG and APOA5 genes.However, it is established that obesity is a polygenic trait, and an individual's susceptibility to obesity is a result of the combined effect of many variants in many genes (Loos 2009).Researchers are addressing this through the analyses of GRS or polygenic risk scores (PRS), which are computed as a weighted sum of trait associated-risk alleles (Lewis and Vassos 2020).
Early work examining the hereditary basis of height demonstrated that GRS models that include large number of SNPs, each with effects too small to be detected individually, may better explain the molecular basis of complex traits and diseases, than using a smaller number of SNPs with confirmed associations (Yang et al. 2010).It is hoped that ultimately, GRS may both explain the variation observed between the populations and lead to improved disease prevention and treatment.In the studies examined within this review, the number of SNPs included in computed GRS for obesity ranged from 2 to 1,148,565 SNPs.This heterogeneity highlights the current challenge of how to utilize the limited data from genetic-association studies that identify individual SNPs with significant functional effects, while also determining the optimum number of SNPs to be included in the computation of GRS (Crouch and Bodmer 2020).
A consistent finding was for the beneficial effects of increased adherence to high-quality diets such as MD, AMED, and DASH diets in reducing obesity risk among individuals who had high GRS compared to those with low GRS (Ding et al. 2018;Han et al. 2020;Hosseini-Esfahani, Koochakpoor, Daneshpour, Sedaghati-khayat, et al. 2017;Seral-Cortes et al. 2020;Sotos-Prieto et al. 2020;Wang et al. 2018).Unfavorable effects of both Western diets and meal skipping on obesity indicators were also observed among those had high GRS compared to low GRS (Hosseini-Esfahani et al. 2019;Jaaskelainen et al. 2013;Masip et al. 2020;Nettleton et al. 2015).Individuals carrying higher GRS that had an increased adherence to MD were observed to have 0.67 times lower risk of obesity in Iranian populations (Hosseini-Esfahani, Koochakpoor, Daneshpour, Sedaghati-khayat, et al. 2017); as well as decreased adiposity in European and US populations (mean difference ± SD in BMI: −1.5 ± 0.67 kg/m 2 ) (Seral-Cortes et al. 2020;Sotos-Prieto et al. 2020).
In addition, increased adherence to AMED and DASH diets, or higher scores in the AHEI-2010, in individuals with higher GRS were also associated with reduced BMI in the US and European populations (Ding et al. 2018;Wang et al. 2018).Therefore, the results promisingly suggest for individuals with high GRS an even greater beneficial impact on reducing risk of obesity from the avoidance of dietary patterns high in red/processed meats and sugary, fried or fatty foods, and inclusion of a variety of vegetables, fruits and cereals in the diet.However, these findings were mainly obtained from observational studies and intervention trials are warranted to confirm these findings.

Limited research on micronutrient status and undernutrition
Although the genetic susceptibility to obesity and its interactions with dietary components have been extensively investigated, there are scarce data on the influence of gene-diet interactions on thinness or under-nutrition and micronutrient status.With respect to blood micronutrient status, our review finds that MTHFR 677 C > T were the most common SNPs associated with serum folate levels, with individuals carrying the T allele had significantly lower serum folate levels compared to the non-carriers after intervening with folate supplementation or diets high in folate intake among the Mexican (Guinotte et al. 2003;Solis et al. 2008) and Japanese women (Hiraoka 2004), although this effect was not observed in Colombia (Arias et al. 2017) and Brazilian women (Lisboa et al. 2020).On contrary, observational studies reported that polymorphism in ALDH2 rs671 (Tao et al. 2019) was associated with significantly higher folate levels with higher intake of alcohol, whereas HFE C282Y (Cade et al. 2005) and FOLH1 T484C (Cummings et al. 2017) were associated significantly higher ferritin levels with higher intakes of heme iron and folate, respectively.
In this review, we only found one study reported the interaction between IGF rs680 and dairy products on body height in the children (Dedoussis et al. 2010).However, this is understandable as the low prevalence of extreme thinness may have posed challenges to the recruitment of healthy thin individuals to study the effects of gene-diet interactions on undernutrition, and hence contributes to the data scarcity.A recent study suggested that thinness is, like obesity, a heritable trait.The authors identified 10 loci, previously found to be associated with obesity were also influencing thinness.These SNPs included FTO rs9930333, MC4R rs2168711 and Transmembrane Protein 18 (TMEM18) rs6748821 (Riveros-McKay et al. 2019).In a separate study, Apolipoprotein H (APOH) rs52797880 was reported as an obesity-resistance gene that interacted with FTO rs9939609 and doubled the odds of thinness (Hasstedt et al. 2016).It has been proposed that the inheritance of thinness may exert mild protective effect in mitigating against the development of obesity caused by the environmental factors such as dietary components and physical activity levels (Costanzo and Schiffman 1989), which deser ves further investigation.

Quality of available gene-diet interaction studies
An adequate sample size is a critical component in gene-diet interaction studies to avoid underpowered statistical analyses (Gauderman 2002).Low statistical power reduces the chance of detecting a true interaction and may produce false negative findings.It has been evidenced that a minimum sample size of ∼6500 is needed in a case-control study design to achieve a 80% power to detect a gene-diet interaction (with an OR of ∼1.5) with a 50% allele frequency in the population (García-Closas and Lubin 1999).Based on the studies included in this review, the 15 th and 75 th percentile of the total sample size of the intervention studies ranged from 75 and 365, respectively (Table S2), and almost 75% of the case-control studies had a sample size < 5000.
Sample size issues may be further exaggerated in studies investigating multiple genetic variants for interactions with dietary components, which require adjustment for multiple testing to avoid false positive findings (Bouaziz, Jeanmougin, and Guedj 2012).In fact, if the p values for the significance of the gene-diet interactions were to be adjusted for all the tested genetic variants, such interactions are unlikely to remain significant.Therefore, computation of GRS which combines all the SNPs tested may be able to address this issue by avoiding power loss due to the multiple-testing correction (Lin et al. 2019).Other factors such as adjustment for multiple ethnicities or populations, calculation for allele frequencies using Hardy-Weinberg Equilibrium, genetic model used, genotype relative risk (effect size) genotyping errors, accuracy of the measurements of exposures and outcomes may affect the sample size and limit the statistical power to evaluate gene-diet interaction (Gordon and Finch 2005).

Strengths and limitations
There were some limitations to this current review.Studies varied dramatically in terms of their reported dietary components, dietary assessment methods and genetic variants.The high heterogeneity of the studies done to date increases the complexity of interpretation.Moreover, the majority of gene-diet interaction findings have not been replicated in multiple studies and were mainly reported from the US and European populations.Therefore, their findings may not be generalizable to other ethnicities or populations.In this review, only the interactions between genetic variations and dietary factors were evaluated, although we know that other factors such as physical activity, epigenetic and gut microbiome play a critical role in modulating nutritional status.These parameters are important determinants of obesity and potential confounding factors that should not be ignored.
Nonetheless, our study had some notable strengths.In particular, both observational studies and intervention trials were included to comprehensively examine the effects of gene-diet interactions on under and overnutrition and micronutrient deficiency.In addition, quality evaluation was conducted rigorously whereby we utilized an additional methodology quality evaluation tool specifically tailored to gene-diet interaction research.The majority of included studies were assessed as having either low or medium risk of bias.As the evaluation of dietary patterns and the combined effect of multiple genetic variants using GRS may provide better understanding on the complex gene-diet interaction, this thorough synthesis of the literature to date provides a useful tool for future research design.

Conclusion
This systematic review reveals that most of the gene-diet interaction studies to date have focused on overnutrition.The findings suggest that healthy dietary patterns, characterized by the high intakes of whole grains, vegetables and fruits, and low intakes of total fat and SFA, may benefit individuals who had higher GRS compared to lower GRS, particularly those carrying the risk alleles of FTO SNPs (rs9939609, rs1121980 and rs1421085) in reducing or managing their body weight.Other SNPs in MC4R, PPARG and APOA5 were also commonly studied for interactions with nutrients or diet in overnutrition though findings were inconclusive.However, most of the interaction findings identified to date have yet to be replicated in trials across multiple populations, more data from Asian populations are warranted.Notably, there are insufficient data available for drawing conclusions about gen-diet interactions and effects on undernutrition and micronutrient status.Although MTHFR 677 C > T was commonly found to be associated with serum folate levels, inconsistent findings were observed across different populations.Future gene-diet interaction research should focus on the investigation of both under and overnutrition, to better identify the most effective dietary patterns for personalized nutrition strategies to improve the human health.

Table 2 .
Healthy eating index; aMed, alternate Mediterranean diet; BMi, body mass index; dasH, dietary approaches to stop Hypertension; Grs, genetic risk score; ns, not stated; sd, standard deviation.

Table 3 .
summary of intervention trials (n = 56) examining gene-diet interactions and weight loss outcomes.

Table 4 .
Human studies examining the effects of gene-diet interactions on micronutrient status.(general) 1 was assessed by Quality Criteria Checklist and risk of bias (genetic) 2 was assessed by a methodology quality evaluation method specifically tailored to gene-diet interaction research.↓ indicates significant lower or greater reduction among the risk alleles carriers; ↑ indicates significant lower or smaller reduction among the risk alleles carriers; ↔ indicates no significant differences or post-intervention changes between the risk alleles and non-risk alleles carriers.ALDH2; aldehyde dehydrogenase 2 Family Member, BMi; body mass index, dFe; dietary Folate equivalent, FOLH1; Folate Hydrolase 1, HFE; Homeostatic iron regulator, MTHFR; Methylene tetrahydrofolate reductase; ns, not stated; rda; recommended daily allowance, tHcy; total homocysteine.