ORIGINAL RESEARCH published: 16 June 2021 doi: 10.3389/fgene.2021.683880 Admixture Has Shaped Romani Genetic Diversity in Clinically Relevant Variants Neus Font-Porterias 1 , Aaron Giménez 2 , Annabel Carballo-Mesa 3 , Francesc Calafell 1 and David Comas 1* 1 Departament de Ciències Experimentals i de la Salut, Institut de Biologia Evolutiva (UPF-CSIC), Universitat Pompeu Fabra, Barcelona, Spain, 2 Facultat de Sociologia, Universitat Autònoma de Barcelona, Barcelona, Spain, 3 Facultat de Geografia i Història, Universitat de Barcelona, Barcelona, Spain Genetic patterns of inter-population variation are a result of different demographic and adaptive histories, which gradually shape the frequency distribution of the variants. Edited by: However, the study of clinically relevant mutations has a Eurocentric bias. The Romani, Karoline Kuchenbaecker, University College London, the largest transnational minority ethnic group in Europe, originated in South Asia and United Kingdom received extensive gene flow from West Eurasia. Most medical genetic studies have only Reviewed by: explored founder mutations related to Mendelian disorders in this population. Here we Elizabeth Atkinson, analyze exome sequences and genome-wide array data of 89 healthy Spanish Roma Massachusetts General Hospital and Harvard Medical School, individuals to study complex traits and disease. We apply a different framework and United States focus on variants with both increased and decreased allele frequencies, taking into Olivia J. Veatch, University of Kansas Medical Center, account their local ancestry. We report several OMIM traits enriched for genes with United States deleterious variants showing increased frequencies in Roma or in non-Roma (e.g., *Correspondence: obesity is enriched in Roma, with an associated variant linked to South Asian ancestry; David Comas while non-insulin dependent diabetes is enriched in non-Roma Europeans). In addition, david.comas@upf.edu previously reported pathogenic variants also show differences among populations, Specialty section: where some variants segregating at low frequency in non-Roma are virtually absent This article was submitted to in the Roma. Lastly, we describe frequency changes in drug-response variation, where Genetics of Common and Rare Diseases, many of the variants increased in Roma are clinically associated with metabolic and a section of the journal cardiovascular-related drugs. These results suggest that clinically relevant variation in Frontiers in Genetics Roma cannot only be characterized in terms of founder mutations. Instead, we observe Received: 22 March 2021 Accepted: 13 May 2021 frequency differences compared to non-Roma: some variants are absent, while other Published: 16 June 2021 have drifted to higher frequencies. As a result of the admixture events, these clinically Citation: damaging variants can be traced back to both European and South Asian-related Font-Porterias N, Giménez A, ancestries. This can be attributed to a different prevalence of some genetic disorders Carballo-Mesa A, Calafell F and Comas D (2021) Admixture Has or to the fact that genetic susceptibility variants are mostly studied in populations of Shaped Romani Genetic Diversity European descent, and can differ in individuals with different ancestries. in Clinically Relevant Variants. Front. Genet. 12:683880. Keywords: Romani, whole-exome sequences, clinically relevant variants, drug-response variants, local ancestry doi: 10.3389/fgene.2021.683880 inference, Eurocentric bias Frontiers in Genetics | www.frontiersin.org 1 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma INTRODUCTION variants in this population has not been fully characterized. In addition, drug response-related genome-wide variation has Human genetic diversity is a continuum, which means that only been deeply examined in Croatian Roma, where variants there are no fixed, immutable or discrete boundaries between in ADME genes were found to have increased allele frequencies populations. Linguistic, geographic and social factors can lead to (Škarić-Jurić et al., 2018). different demographic histories and, in turn, to patterns of inter- To fill this gap, we examine whole exome sequences (WES) population variability (Lao et al., 2008; Barbujani et al., 2013; and genome-wide array data of 89 healthy Spanish Roma Batai et al., 2020; Heyer and Reynaud-Paligot, 2020). It has been individuals and characterize the functionally relevant genomic previously reported that this stratification has consequences in variants (i.e., associated to disease or to drug response) with either the genetics of complex traits and diseases [reviewed in Bentley increased or decreased allele frequencies in the Roma. Beyond et al. (2017); Sirugo et al. (2019)]. On the one hand, disease- frequency distribution differences, and taking into account that associated variants present allele frequency differences across Roma is an admixed population, we describe the ancestral origin populations. For example, 70% of the cystic fibrosis cases in of multiple variants by leveraging on the estimated local ancestry Europeans are due to 1F508 mutation in CFTR gene, while the of their background haplotypes. most common causal variant in South Africans with African ancestry is 3120 + 1G→A, and different mutations have different therapeutic targets (Padoa et al., 1999; Stewart and Pepper, MATERIALS AND METHODS 2017; Sirugo et al., 2019). On the other hand, genomic variation across populations is also observed for treatment response Data differences, especially in genes related to absorption, distribution, We used WES (mean depth of 54X), and genome-wide autosomal metabolism, and excretion (ADME) of drugs (Dopazo et al., 2016; SNP data (Affymetrix Axiom Genome-Wide Human Origins Škarić-Jurić et al., 2018; Sirugo et al., 2019). For example, the 1 array) for Spanish Roma individuals (89 and 62 samples, metabolism of the anticoagulant warfarin can differ due to several respectively) (Font-Porterias et al., 2021), deposited at EGA genetic polymorphisms; however, their frequencies are different (EGAS00001004599). The Spanish Roma WES were merged in European and African descent groups, which challenges the with previously published non-Roma WES from 1000G (mean correct dosage prescription (Bress et al., 2012; Johnson et al., depth of 65.7X): Iberian Population in Spain (IBS), Toscani 2017; Sirugo et al., 2019). in Italia (TSI), Punjabi from Lahore (PJL), Indian Telugu Thus, an accurate clinical assessment relies on the study from the United Kingdom (ITU), and Gujarati Indian from of clinically relevant genetic variants with different allele Houston (GIH) (Auton et al., 2015), resulting in a dataset frequencies across groups. Yet, there is an underrepresentation with 512 individuals and 410,225 variants. The genome-wide of human populations in the screening of these variants. SNP data was merged with IBS, TSI, PJL, ITU, and GIH from Particularly, genetic studies show a strong and systematic 1000G (1000 Genomes Project Consortium, 2012), with 474,632 Eurocentric bias (Need and Goldstein, 2009; Popejoy and genome-wide SNPs in 487 samples. Both datasets were then Fullerton, 2016; Bentley et al., 2017; Martin et al., 2019; combined to increase the covered genomic variants, building Sirugo et al., 2019). As a consequence, this bias prevents a merged WES-array dataset with 487 individuals and 878,162 to fully understand the genetic architecture of human SNPs. Variant annotation was performed using the Variant disease and leads to an incomplete genetic assessment of Effect Predictor tool (VEP) from Ensembl (McLaren et al., complex traits, and to an inaccurate disease diagnosis and 2016) focusing on three deleterious prediction scores: PolyPhen- treatment in under-represented groups (Martin et al., 2019; 2 (Adzhubei et al., 2010), GERP (Davydov et al., 2010) and Sirugo et al., 2019). CADD (Rentzsch et al., 2019), as previously explained (Font- The Roma population, also known by the misnomer Porterias et al., 2021). For each analysis, the corresponding of “Gypsies,” has been under-represented in these genome- dataset used is specified (WES dataset or merged WES- wide scans. They constitute the largest transnational minority array dataset). In addition, we have included a glossary of ethnic group in Europe (Council of Europe, 2012). Linguistic terms that may have ambiguous meanings in genetic studies and genetic evidence point to a South Asian origin and (Supplementary Note 1). subsequent diaspora toward Europe, with extensive non- Roma gene flow and multiple founder effects shaping their Local Ancestry Inference demographic history (Hancock, 2002; Matras, 2002; Mendizabal The phasing of the merged WES-array dataset, with 405,814 et al., 2012; Moorjani et al., 2013; Morar et al., 2013; variants with minor allele frequency (MAF) > 1%, was Font-Porterias et al., 2019). Most medical genetic studies on performed using SHAPEIT (O’Connell et al., 2014), using this population have been focused on targeting the genetic the population-averaged genetic map from the HapMap phase variants responsible for the increased prevalence of certain II (International HapMap Consortium, 2003) and the 1000G genetic diseases [reviewed in Kalaydjieva et al. (2001); Morar dataset as a reference panel (1000 Genomes Project Consortium, et al. (2013)]. In this sense, several founder mutations have 2012). RFMix v1.5.4 (Maples et al., 2013) was run with one been identified: e.g., the p.R299X mutation in the LTBP2 expectation-maximization (EM) iteration to infer the local gene, which is responsible for congenital glaucoma (Azmanov ancestry of the phased haplotypes, using balanced reference et al., 2011). However, the distribution of disease-associated panels representing European (IBS and TSI populations) and Frontiers in Genetics | www.frontiersin.org 2 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma South Asian (PJL, GIH, and ITU populations) ancestries. As threshold, we decrease the number of false positive results in previously explained (Font-Porterias et al., 2021), the Roma the enrichment analyses, although we may lose some pathogenic individuals included in the present study show, on average, variants found at low frequencies (and for that reason, we then 68.4% and 31.6% of European and South Asian global ancestry focus on ClinVar pathogenic variants with less conservative proportions, with a standard deviation of 7%. Ancestry was thresholds). The same number of individuals were considered assigned when RFMix posterior probability was higher than for this analysis (70 individuals per population) to avoid biases 0.9, resulting in 96.3% of the variants with assigned ancestry. in MAC and MAF calculations. The over-representation analysis In order to match the local ancestry inference in heterozygous was performed with default parameters (Liao et al., 2019), variants and obtain the ancestry background of the allele, OMIM was selected as the functional disease database and all we adjusted the RFMix rephasing as previously performed the genes included in our variants set was used as background (Browning et al., 2018), since RFMix partially rephases the data gene set. Once the enriched pathways were identified, we when assigning local ancestry. However, when the variant was computed a chi-squared test to check whether the associated filtered out (MAF < 1%) in the phasing, only genotype ancestries genetic variants described in OMIM present in our dataset have can be retrieved. statistically different genotype frequencies between Roma and non-Roma groups. Genetic Portability We computed the allele sharing ratio (proportion of variants at Screening of Known Disease-Associated different frequency bins) from the Roma segregating in non- Variants Roma populations from WES variants dataset. In addition, We first identified previously reported Mendelian mutations we compared the linkage disequilibrium (LD) decay patterns in the Roma, annotated in Bianco et al. (2020) and checked between Roma and non-Roma from the genome-wide array the ancestry of their haplotypes to trace their putative origin. dataset using PopLDdecay (Zhang et al., 2018) with default However, this approach does not allow us to examine if parameters. We performed a two-sample Kolmogorov–Smirnov there is a different frequency spectrum of disease-associated test to check whether the decay distributions of Roma and non- variants comparing Roma and non-Roma. To that end, we Roma groups were statistically different. For both analyses, we then annotated the set of WES variants using ClinVar database used the same number of individuals per population to avoid (Landrum et al., 2014) and compared the frequency of clinically sample size biases (70 individuals per population for the allele validated variants among populations. We kept only variants sharing ratio and 62 individuals per population for the LD decay). with a clinical significance of “pathogenic,” which is the highest level of supported evidence. We selected those variants with Gene Enrichment Analyses a fold increase in risk allele frequency (RAF) > 1.5 between Using the WES dataset, we performed two different gene populations or 1.5% RAF for monomorphic variants. A chi- enrichment analyses using WEB-based GEne SeT AnaLysis squared test was performed to test whether the genotype Toolkit (Liao et al., 2019) to identify categories (or classes) of frequencies were significantly different across populations. genes that are over-represented in a particular set of genes, The same number of individuals were considered for this using a background gene set. First, we interrogated those genes analysis (70 individuals per population) to avoid biases in with more deleterious mutations (i.e., deleterious Nalleles or Nhom RAF calculations. per individual per gene). Although we expect that the most constrained genes (i.e., lower values of deleterious Nalleles or Nhom per individual per gene) will be shared across populations, we Screening Beyond Disease-Associated test whether there is a particular pathway enriched in the most Variants mutated genes in the Roma samples and the non-Roma groups, To examine pharmacogenetic variation in the WES dataset, we independently. To do so, we normalized Nalleles or Nhom by the studied mutations that disrupt drug binding domains without number of variants in each gene and we then examined the being deleterious to the protein. These variants are based on correlation between each pair of Roma to non-Roma populations function prediction: they might cause drug binding inhibition; per each gene. The over-representation analysis was performed however, not all variants have a reported association with drug with default parameters (Liao et al., 2019), Gene Ontology response (Hopkins and Groom, 2002; Dopazo et al., 2016). (GO) was selected as the functional database, and all the genes We also examined variants found in 31 core ADME genes1 included in our variants set was used as background gene set. (Škarić-Jurić et al., 2018). We selected those variants with a fold This analysis does not take into account the frequency of the increase in MAF > 1.5 between populations or 1.5% MAF for variants, since the calculation is performed per individual. We monomorphic variants. A chi-squared test was performed to test also performed a gene enrichment analysis to test whether the whether the genotype frequency was significantly different across genes with deleterious variants showing allele frequency increases populations. The same number of individuals were considered in Roma to non-Roma (or non-Roma to Roma) belong to for this analysis (70 individuals per population) to avoid biases specific genetic disease clusters. We included in the analysis in MAF calculations. The selected variants in both analyses were those genes with variants with a fold increase in minor allele searched in PharmGKB (Whirl-Carrillo et al., 2012) and the frequency (MAF) > 5 or fold increase in minor allele count (MAC) > 5 for monomorphic variants. By using this restrictive 1 www.pharmaadme.org Frontiers in Genetics | www.frontiersin.org 3 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma corresponding target drugs in DrugBank (Wishart et al., 2018) further show that the genomic distribution of accumulation of and PubChem (Kim et al., 2019). deleterious mutations (i.e., number of deleterious alleles per gene per individual) has comparable patterns between populations (Supplementary Note 2, Supplementary Figures 2, 3, and RESULTS Supplementary Tables 1, 2). Thus, the overall allele sharing and linkage disequilibrium patterns are comparable among Initial Assessment of Functional Variants populations. However, rare and private variants can present some as a First Evidence of Inter-Population challenges in the genetic characterization of Roma population, Variability especially for those variants with a frequency lower than 5%. In order to assess the genetic portability from non-Roma to As shown in a previous study, Roma and non-Roma groups Roma, we examined the allele sharing and linkage disequilibrium exhibit differences in the site frequency spectrum due to patterns between populations. We have previously shown that different demographic histories (Font-Porterias et al., 2021). An Roma exhibit a considerable amount of private variants; however, overrepresentation analysis including those genes with different their proportion is lower than in other populations: 15,287 allele frequency variants points to a differential OMIM trait population-specific variants in Roma; 25,060 in IBS; 24,158 enrichment in non-Roma and Roma (Supplementary Note 3 in TSI; 21,160 in PJL; 22,040 in GIH; and 24,070 in ITU and Supplementary Tables 1, 3). Regarding rare conditions, (Font-Porterias et al., 2021). In addition, allele sharing is the genes related to non-Herlitz type junctional epidermolysis high for common variants (MAF > 5%) (over 86% of Roma bullosa are enriched when comparing deleterious variants with variants are present in non-Roma) (Figure 1). However, for higher frequency in Roma than non-Roma. The prevalence of this rare variants (MAF < 5%), the allele sharing is around 30– disease in Roma is higher than non-Roma in Spain (Martinez- 40% (Figure 1). Regarding linkage disequilibrium, decay patterns Frias and Bermejo, 1992), although none of the pathogenic are not statistically different between Roma and non-Roma variants described in OMIM are present in our dataset and (Supplementary Figure 1, p-value > 0.95). The total number only benign ones show increased frequencies (Supplementary of deleterious alleles per individual is similar among Roma and Table 4). In addition, the tetralogy of Fallot is enriched in non-Roma groups (Font-Porterias et al., 2021), and here we non-Roma (Table 1), with two pathogenic variants in GATA4 FIGURE 1 | Allele-sharing ratios among Roma and non-Roma groups. Proportion of Roma variants from each minor allele frequency bin (from >0% to 50%) which are also segregating at each non-Roma population. Frontiers in Genetics | www.frontiersin.org 4 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma TABLE 1 | Summary results of the overrepresentation analysis. Type Name Comparison Associated variants Rare Non-Herlitz Epidermolysis Bullosa junctional Roma > non-Roma – (OMIM 226650) Rare Tetralogy of Fallot (OMIM 187500) Non-Roma > Roma rs56208331, rs115099192 (pathogenic) (Tomita-Mitchell et al., 2007; Zhang et al., 2008) Cardiovascular and metabolic Obesity (OMIM 601665) Roma > non-Roma rs2282440 (associated) (Ha et al., 2006) Cardiovascular and metabolic Ischemic stroke (OMIM 601367) Roma > non-Roma rs6025 (risk factor) (Majerus, 1994; Casas et al., 2004) Cardiovascular and metabolic Insulin dependent diabetes (OMIM 125853) Roma > non-Roma – Cardiovascular and metabolic Non-insulin dependent diabetes (OMIM Non-Roma > Roma rs1800467 (likely benign) (Phani et al., 2014) 222100) Other Protection alcohol dependence (OMIM Roma > non-Roma rs1229984 (associated) (MacGregor et al., 2009) 103780) Other Breast cancer (OMIM 114480) Non-Roma > Roma – Conditions (traits or diseases) are grouped in three categories: rare, cardiovascular/metabolic and other. Name and Phenotype MIM number from OMIM database are shown. Comparison shows the first population as the one with increased allele frequency variants. Associated variants with increased allele frequencies are shown with its clinical annotation and corresponding reference study. More details can be found in Supplementary Tables 3–6. gene (rs56208331 and rs115099192) with increased frequencies higher prevalence of triple negative (TN) breast tumors (Reckova in South Asian groups (Supplementary Table 4). This is et al., 2017), and mutations in the enriched genes are mostly consistent with the higher prevalence of this condition in Asia linked to non-TN types (Peshkin et al., 2010; Lin et al., 2016; (Takkenberg and Roos-hesselink, 2011). Decker et al., 2017; Slavin et al., 2017; Sirisena et al., 2018). Multiple cardiovascular and metabolic disorders are also These results show multiple OMIM traits enriched for genes present in the gene overrepresentation analysis. Obesity is with a 5-fold increased allele frequency deleterious variants both enriched comparing Roma with non-Roma, especially with in Roma and in non-Roma. Although this is an exploratory Europeans. One variant in the SDC3 gene (rs2282440), which approach, it is consistent with previous literature and reports the is associated with obesity in Asians (Table 1) has a significantly first evidence that Roma do not show a systematically increased increased frequency in Roma (19%) and it is virtually absent in genetic susceptibility to disease. other European groups (Supplementary Tables 4, 5). All risk alleles of this variant in the Roma have South Asian ancestry (Supplementary Table 6), suggesting that this allele in the A Genome-Wide Screening Does Not Roma has a South Asian origin. In addition, genes related Support an Increased Susceptibility for to ischemic stroke are overrepresented in Roma (Table 1). Genetic Disorders One variant in the F5 gene, annotated as a risk factor in The presence of mutations responsible for particular Mendelian Europeans (rs6025), has significantly higher frequencies in Roma disorders in the Roma has been reported in several studies (see (8.6%) compared to all non-Roma (<1.5%) (Supplementary (Kalaydjieva et al., 2001; Álvarez et al., 2005; Bouwer et al., Tables 4, 5). This variant has a European ancestry background 2007; Gamella et al., 2013; Sevilla et al., 2013; Rocha et al., for 7 out of 8 risk alleles found (Supplementary Table 6), 2014; Cabrera-Serrano et al., 2018) among others). From these suggesting a European origin of this variant in the Roma. previously reported mutations, we found seven variants in our Lastly, both types of diabetes are enriched in our analysis: dataset (Table 2). Only in two out of seven variants, the risk allele non-insulin dependent diabetes is enriched in non-Roma, is present in non-Roma populations: rs1801968 (chr9:132580901) while insulin dependent diabetes is overrepresented in Roma and rs1126809 (chr11:89017961) (Table 2). The risk allele of the compared to non-Roma (Table 1). These results are consistent latter has a higher frequency in Europeans than South Asians with previous literature (Mendizabal et al., 2013; Werissa (Table 2) and, in the Roma, it has a European assigned ancestry et al., 2019), although our dataset does not include none in 17 out of 19 alleles (Supplementary Table 7), suggesting of the pathogenic or risk-factor variants (more details in that this variant (responsible for Oculocutaneous albinism) Supplementary Note 3). in the Roma has a European origin, as previously identified Other conditions such as protection to alcohol dependence (Bianco et al., 2020). In addition, two variants responsible are enriched in Roma compared to non-Roma (Table 1). One for Charcot–Marie–Tooth disease (rs119483085; chr8:134270617 variant in the ADH1B gene reported to be protective for and rs80338934; chr5:148389835) (Table 2) both have one risk alcohol dependence (rs1229984) shows a significantly higher allele with European ancestry (Supplementary Table 7). On the allele frequency in Roma without a clear ancestry origin contrary, rs77931234 (chr1:76226846; Acetyl-coA dehydrogenase (Supplementary Tables 5, 6). The enrichment for breast cancer deficiency) and rs104894396 (chr13:20763650; Deafness) appear in non-Roma (Table 1) is due to variants with increased allele to have a South Asian origin (Supplementary Table 7). In frequencies in ATM, BRCA2, and NQO2 genes. Roma show fact, rs104894396 risk allele is also present in PJL population Frontiers in Genetics | www.frontiersin.org 5 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma TABLE 2 | List of previously reported Mendelian mutations in Roma present in this study. Variant – allele Gene Roma IBS TSI PJL ITU GIH Disease rs77931234-G ACADM 0.037 0.000 0.000 0.000 0.000 0.000 Acetyl-coA dehydrogenase deficiency (Rocha et al., 2014) rs777176261-A BIN1 0.012 0.000 0.000 0.000 0.000 0.000 Centronuclear myopathy (Cabrera-Serrano et al., 2018) rs80338934-A SH3TC2 0.006 0.000 0.000 0.000 0.000 0.000 Charcot–Marie–Tooth disease (Claramunt et al., 2007; Sevilla et al., 2013) rs119483085-A NDRG1 0.006 0.000 0.000 0.000 0.000 0.000 Charcot–Marie–Tooth disease (Claramunt et al., 2007; Sevilla et al., 2013) rs1801968-G TOR1A 0.019 0.1 0.141 0.157 0.165 0.211 Dystonia (Kalaydjieva et al., 2001) rs1126809-A TYR 0.154 0.305 0.269 0.079 0.006 0.103 Oculocutaneous albinism (Gamella et al., 2013) rs104894396-T GJB2 0.019 0.000 0.000 0.007 0.000 0.000 Deafness (Álvarez et al., 2005; Bouwer et al., 2007) Variant rs ID and risk allele, risk allele frequency for each population and disease association are shown. (Table 2), which is consistent with a South Asian origin. This and 31 in ITU), while there are 84 and 74 segregating in IBS and variant, only present in Roma and Punjabi individuals in our TSI, respectively. dataset, is a non-synonymous mutation (W42X) responsible for autosomal recessive non-syndromic hearing loss. It has been previously reported in Spanish and Slovak Roma individuals and Many Drug-Response Variants in Roma an Indian origin has been suggested (Minárik et al., 2003; Álvarez Are Related to Metabolic and et al., 2005), which is congruent with the South Asian ancestry Cardiovascular Disorders assignation of the risk allele. Besides disease-associated variants, other functionally relevant We next investigated the 334 pathogenic variants described in mutations (e.g., pharmacogenomic variants) exhibit inter- the ClinVar database present in our dataset. In Roma, we found population genetic variation in the human genome, as mentioned 60 out of 334 variants segregating at low RAF in the population. above. Regarding drug binding domains, we identified 101 Only 27 variants have a RAF difference equal or higher than variants in our dataset that disrupt the domains without being 1.5 comparing Roma and non-Roma (Supplementary Table 8). deleterious for the protein. This set is less biased toward Although the RAF of these variants is low (below 5% in European genetic variation, since it is based on impact prediction, most cases), there are variants with increased frequency in rather than on previously discovered genetic associations Roma, but interestingly, we observe disease-associated variants (Dopazo et al., 2016). Only 26 variants were found to have with increased frequencies in European and South Asian non- a MAF fold increase ≥1.5 comparing Roma and non-Roma Roma populations. For example, rs1799807 (chr 3:165548529) (Supplementary Table 11). Variants with known association is a missense pathogenic variant causing the deficiency of drug phenotypes reported in PharmGKB (Whirl-Carrillo et al., butyrylcholine esterase and, consequently, postanesthetic apnea 2012) have higher MAF in European populations, showing the (McGuire et al., 1989; Jasiecki et al., 2019). This variant is only European-centric bias in biomedical genetic studies. For example, present in European populations and virtually absent in South the rs5918 variant, located in the ITGB3 gene (chr17:45360730), Asia and Roma (Supplementary Table 8). On the contrary, reduces the efficacy of aspirin and clopidogrel (Dropinski et al., rs137941190 (chr 11:126215441) is a missense pathogenic variant 2007; Motovska et al., 2009) (indicated for coronary artery disease for Al-Raqad syndrome, described in Pakistani patients (Ahmed and myocardial infarction) and it shows a higher allele frequency et al., 2014). The risk allele of this variant is absent in the in IBS and TSI than in Roma (Supplementary Table 11). European exomes, but it appears at low frequencies in the A variant found in the GLP1R gene (rs6923761; chr6:39034072) Roma and South Asian samples (Supplementary Table 8), reduces the treatment efficacy for obesity and type II diabetes (i.e., although the genotype frequencies are not statistically different sitagliptin, vildagliptin, and liraglutide) (Javorský et al., 2016) and (Supplementary Table 9). Regarding the ancestry inference, its frequency is significantly lower in Roma than in the tested European ancestry is assigned for the risk allele of most of these European populations (Supplementary Table 11). variants (Supplementary Table 10), except for rs104894396 (chr On the contrary, there are 13 variants with increased 13:20763650), as explained above. MAF in Roma (Supplementary Table 11) with significantly The screening of known disease-associated variants together different genotype frequencies compared to non-Roma groups with the local ancestry inference has allowed to report both (Supplementary Table 12). For example, there is a variant in the presence and the absence of particular mutations in the the CRYZ gene (chr1:75175886) with significantly higher MAF Roma and to trace their most likely ancestral origin. However, in Roma and South Asians than in Europeans (Supplementary most known clinically relevant variants have been discovered Table 11). Its protein contains a drug binding domain for in European populations, which leads to an ascertainment bias dicumarol, indicated for deep vein thrombosis (Supplementary that can weaken the results when studying the Roma population: Table 11) and 19 out of 27 minor alleles in Roma have out of the 334 pathogenic variants, less than 45 variants are South Asian ancestry (Supplementary Table 13). In the PTPRE segregating in South Asian populations (43 in PJL, 41 in GIH, gene drug binding domain for alendronate (indicated for Frontiers in Genetics | www.frontiersin.org 6 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma osteoporosis) (Supplementary Table 11), we identify a variant MAF increase in Roma than in IBS and TSI populations (chr10:129868686) in Roma (3.7%), virtually absent in non- (20%, 9%, and 11%, respectively) (Table 3) and it is reported Roma, except GIH (0.5%) (Supplementary Table 11). Four Roma to modify the metabolism of metformin, a drug used to individuals were heterozygotes for this variant: one with both treat type II diabetes (Table 3). Regarding ancestry inference, haplotypes assigned to South Asian ancestry and three with the minor alleles of these variants are almost exclusively of one European and one South Asian haplotype (Supplementary European ancestry (Supplementary Table 15). Lastly, three Table 13), which suggests that this variant in the Roma originated previously found variants with increased frequencies in Croatian in South Asia. Regarding the ancestry inference and besides the Roma (Škarić-Jurić et al., 2018) do not show significantly mentioned examples, many of the variants have the minor allele higher MAFs in Spanish Roma (rs10509681, rs8192709, and with European ancestry: 72% of the minor alleles are assigned rs34059508) (Table 3). to a European-related ancestry; slightly above the mean genome- The screening beyond disease-associated variants in the Roma wide ancestry (68.4%), but within the first SD of the distribution. population reveals that most of them can change the response Although experimental evidence suggesting these variants affect of drugs used for metabolic and cardiovascular disorders and the binding of these drugs is lacking, follow-up studies should be that they might have a European origin. However, this analysis performed to validate the functional impact of these variants. is based on the impact prediction and it is important to take into We next examined previously described variants in ADME account that the phenotype is also influenced by the environment, genes. In our dataset, 14 out of 95 of them show increased MAF non-coding variants and regulatory elements, among others. with a fold change equal or higher than 1.5 comparing Roma and non-Roma (Table 3). Some variants found in European groups are absent in the Roma exomes: e.g., rs34130495, which DISCUSSION modifies the metabolism of tramadol (indicated for mild-to- moderate pain) (Table 3). However, many of the variants with The underrepresentation of human populations in genetic increased frequencies in Spanish Roma are clinically associated studies impairs the understanding of genome architecture with metabolic and cardiovascular-related drugs (Table 3) and and exacerbates health differences. In order to overcome this some have significantly different genotype frequencies between limitation, the Eurocentric bias in the discovery of functional Roma and non-Roma (Supplementary Table 14). For example, variants has to be taken into account (Need and Goldstein, the rs4149056 variant shows a higher MAF in Roma than in IBS 2009; Popejoy and Fullerton, 2016; Bentley et al., 2017; (18% and 11%, respectively) (Table 3) and it increases the risk Martin et al., 2019; Sirugo et al., 2019). In the case of the of toxicity to simvastatin (indicated for hypercholesterolemia) Roma population, we found that the genetic portability with (Table 3). A previous study reports a frequency of 17.2 and European populations is overall high: allele sharing and LD 18.9% of this variant in Roma and non-Roma groups from decay patterns are comparable among groups. However, low Hungary, respectively (Nagy et al., 2015), suggesting that this frequency variants (MAF < 5%) can present some challenges variant in the Spanish Roma was present before the arrival into in the genetic characterization of this population, since only the Iberian Peninsula. rs316019 variant also shows a significant half of these variants in the Roma are also segregating TABLE 3 | List of variants in ADME genes found to have a fold increase in allele frequency equal or higher than 1.5 comparing Roma and non-Roma. Variant – allele Gene Roma IBS TSI PJL ITU GIH Clinical annotation rs12208357-T SLC22A1 0.019 (0.022) 0.058 0.058 0.021 0.024 0.010 Metabolism metformin (Shu et al., 2008; Sundelin et al., 2017) rs2282143-T SLC22A1 0.037 (0.085) 0.005 0.013 0.079 0.059 0.082 Metabolism metformin (Yoon et al., 2013) rs34130495-A SLC22A1 0.000 (0.023) 0.032 0.019 0.000 0.000 0.000 Metabolism tramadol (Tzvetkov et al., 2011) rs34059508-A SLC22A1 0.012 (0.048) 0.021 0.006 0.000 0.000 0.000 Metabolism metformin (Shu et al., 2008) rs316019-A SLC22A2 0.204 (0.066) 0.089 0.109 0.107 0.165 0.134 Metabolism metformin; Toxicity cisplatin-anthracyclines (Visscher et al., 2012; Yoon et al., 2013) rs1800460-T TPMT 0.025 (0.002) 0.037 0.019 0.000 0.000 0.000 Toxicity azathioprine and mercaptopurine (Stocco et al., 2012; Steponaitiene et al., 2016) rs717620-T ABCC2 0.185 (0.253) 0.211 0.186 0.071 0.047 0.077 Efficacy and dosage atorvastatin; Toxicity fluorouracil (Cecchin et al., 2013; Prado et al., 2018) rs4244285-A CYP2C19 0.191 (0.155) 0.133 0.090 0.314 0.388 0.330 Efficacy and toxicity clopidogrel; Efficacy amitriptyline rs10509681-C CYP2C8 0.099 (0.165) 0.168 0.141 0.050 0.024 0.041 Metabolism rosiglitazone (Dawed et al., 2016) rs1058930-C CYP2C8 0.105 (0.056) 0.042 0.051 0.014 0.006 0.010 Metabolism diclofenac (Dorado et al., 2008) rs4149056-C SLCO1B1 0.179 (0.102) 0.111 0.224 0.050 0.076 0.021 Toxicity simvastatin (Shek et al., 2017) rs1048943-C CYP1A1 0.037 (0.033) 0.016 0.045 0.107 0.118 0.113 Efficacy capecitabine and docetaxel (Dong et al., 2012) rs1799814-T CYP1A1 0.043 (0.033) 0.089 0.013 0.014 0.018 0.005 Metabolism dacarbazine (Lewis et al., 2016) rs8192709-T CYP2B6 0.093 (0.128) 0.047 0.064 0.043 0.035 0.036 Toxicity efavirenz (Dhoro et al., 2014; Dickinson et al., 2016) Variant rs ID and risk allele, gene, risk allele frequency for each population and PharmGKB main clinical annotations are shown. Croatian Roma frequencies from Škarić-Jurić et al., 2018 are also included within brackets in the Roma column (those variants with significantly higher frequencies are in bold). Frontiers in Genetics | www.frontiersin.org 7 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma in non-Roma populations. This is consistent with the fact polymorphism (rs4244285) in Hungarian and Portuguese Roma that low frequency and rare variants are more population groups (Sipeky et al., 2013; Teixeira et al., 2015) is not observed structured than common variants (Casals and Bertranpetit, 2012; in Spanish Roma. Bomba et al., 2017). For complex diseases, polygenic risk scores (PRSs) are The overrepresentation analysis shows enrichment of some designed to predict the phenotype from genetic data, combining gene sets with increased MAF deleterious variants for genetic the effect sizes of multiple variants and their frequency (Wray disorders. Interestingly, this enrichment occurs both in Roma et al., 2007). However, the pre-computed effect sizes usually and in non-Roma, consistent with previous literature. For derive from genome-wide association studies (GWAS), and given example, we identify an enrichment for non-triple negative its Eurocentric bias, PRSs have a greater predictive accuracy breast cancer in non-Roma, in agreement with a lower incidence in populations with European ancestry (Martin et al., 2017, in Roma, which show more triple negative cases (Reckova 2019; De La Vega and Bustamante, 2018; Kim et al., 2018; et al., 2017). Triple negative refers to the overexpression of Gurdasani et al., 2019). Particularly, they show a systematic bias three common markers (i.e., estrogen receptor, progesterone when applied to other populations due to several factors: (i) receptor or HER2 oncogene) (Foulkes et al., 2010). However, GWAS are biased toward those variants segregating in the study the overexpression of these markers is common in breast cancer population; (ii) when two populations have different LD patterns, patients of European descent, but not in other groups, such as tagSNPs and causal variants can differ; and (iii) environmental African-descent (Brewster et al., 2014) or Roma (Reckova et al., and genetic factors can be confounded when phenotypes are 2017) patients. geographically stratified (Martin et al., 2019; Sirugo et al., In the present study, we also examine previously 2019). Although new methods are emerging to overcome these defined disease-associated variants. Besides confirming limitations (Márquez-Luna et al., 2017; Marnetto et al., 2020), the ancestry origin of some Mendelian mutations in an in-depth analysis of PRS accuracy in non-European-descent the Roma, we provide new evidence: the risk allele populations is needed before implementing them to other under- responsible for Acetyl-coA dehydrogenase deficiency represented populations (Sirugo et al., 2019). (rs77931234) is traced to a South Asian-like ancestry, Here, we provide new evidence of a different frequency while the risk alleles for Charcot–Marie–Tooth disease spectrum of clinically relevant variants across populations: while variants (rs119483085, rs80338934) show European-related some have increased allele frequencies in the Roma, others ancestry. In addition, we perform a comprehensive study are virtually absent. This was possible due to the availability of the pathogenic variation reported in ClinVar database of a substantial number of whole-exome sequences at high (Landrum et al., 2014). The results show a different coverage, which allows the study of clinically relevant genetic frequency spectrum of previously identified variants when variation, enriched in low frequency variants (Bomba et al., comparing Roma and non-Roma groups. However, most 2017). However, this different frequency spectrum cannot be of the variants in the Roma are traced to a European directly attributed to a lower prevalence of some diseases; origin, evidencing the Eurocentric bias in public databases instead, other unstudied variants might be responsible for (Kessler et al., 2016). disease susceptibility and drug response in this population. Regarding pharmacogenetic variation, we identify variants Although it is an exploratory approach without functional with increased and decreased MAF in Roma. Many of the validation, this study aims to shift the traditional paradigm of variants in drug-binding domains that we found with increased focusing only on the increased genetic risk for some diseases frequencies in non-Roma European populations have been in the so-called “isolated” populations. In fact, these results previously associated with a drug response trait, while those with further confirm that Roma are not so genetically isolated. increased frequencies in Roma or South Asian groups are still Gene flow with European groups accounts for 65% of their not specifically characterized. This is particularly relevant, since genetic ancestry (Font-Porterias et al., 2019); thus, clinically many of them are related to metabolic and cardiovascular drugs damaging variants are traced both to South Asian and European- and previous studies suggest that Roma show higher prevalence related haplotypes. Lastly, we caution that these results are of these diseases (Vozarova De Courten et al., 2003; Živković geographically limited to the Roma population from Spain, and et al., 2010). Following the expectations of European ancestry further characterization should be performed in other groups proportions, most drug-response variants in Roma are traced with different demographic trajectories and with increasing to a European-related origin, which is not a signal of bias, sample sizes. because the list of variants that disrupt drug binding domains We would like to remark that this study does not aim is not based on previously known associations discovered in to exacerbate the importance of inter-population variability, populations of European descent (Dopazo et al., 2016). Thus, to justify health differences in minority ethnic groups, or to these variants show inter-population variability in the Roma, advocate for racialized medicine. In fact, genetic ancestry is since a differential European gene flow among Roma groups not the only determinant of ethnicity or health, and social has been previously described (Mendizabal et al., 2012; Font- factors should be considered. Given that genetic diversity is Porterias et al., 2019). In this sense, the variants reported with a continuum, large scale genome-wide studies are needed to increased frequency in Spanish Roma do not completely overlap fully capture and represent human variation, without excluding with a similar study in Croatian Roma (Škarić-Jurić et al., 2018). any population while respecting their rights and interests and Moreover, the significantly increased frequency of a CYP2C19 properly accounting for demographic differences. This would Frontiers in Genetics | www.frontiersin.org 8 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma prevent the current overgeneralization of the results obtained FUNDING from genetic studies on populations with only European ancestry in the assessment of disease risk testing and treatment response This study was supported by the Spanish Ministry of (Martin et al., 2019; Batai et al., 2020; Hudson et al., 2020). Science, Innovation and Universities (MCIU), Agencia Estatal de Investigación (AEI) grant number PID2019-106485GB- I00/AEI/10.13039/501100011033, and “Unidad de Excelencia DATA AVAILABILITY STATEMENT María de Maeztu” (AEI, CEX2018-000792-M). NF-P was supported by a FPU17/03501 fellowship. The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s. ACKNOWLEDGMENTS We would like to thank the invaluable collaboration of the ETHICS STATEMENT Federacioì d’Associacions Gitanes de Catalunya (FAGiC). The studies involving human participants were reviewed and We would also like to thank the members of the Human approved by the CEIC-Parc de Salut Mar 2019/8900/I. The Genome Diversity Group and Genomics of Individuality patients/participants provided their written informed consent to Lab (Institut de Biologia Evolutiva, UPF-CSIC) for the participate in this study. insightful discussions about the results of this study and the computational support from the Scientific IT Core Facility from Universitat Pompeu Fabra. AUTHOR CONTRIBUTIONS NF-P and DC contributed to the design and conception of SUPPLEMENTARY MATERIAL the study. NF-P performed and implemented the data analysis. All authors contributed to the interpretation and discussion of The Supplementary Material for this article can be found the results and writing of the manuscript, and approved the online at: https://www.frontiersin.org/articles/10.3389/fgene. submitted version. 2021.683880/full#supplementary-material REFERENCES marked admixture have shaped European Roma genomes. Mol. Biol. Evol. 37, 3175–3187. doi: 10.1093/molbev/msaa156 1000 Genomes Project Consortium, (2012). An integrated map of genetic variation Bomba, L., Walter, K., and Soranzo, N. (2017). The impact of rare and low- from 1,092 human genomes. Nature 491, 56–65. doi: 10.1038/nature11632 frequency genetic variants in common disease. Genome Biol. 18, 1–17. Adzhubei, I. A., Schmidt, S., Peshkin, L., Ramensky, V. E., Gerasimova, A., Bork, P., Bouwer, S., Angelicheva, D., Chandler, D., Seeman, P., Tournev, I., and Kalaydjieva, et al. (2010). A method and server for predicting damaging missense mutations. L. (2007). Carrier rates of the ancestral Indian W24X mutation in GJB2 in the Nat. Methods 7, 248–249. doi: 10.1038/nmeth0410-248 general Gypsy population and individual subisolates. Genet. Test. 11, 455–458. Ahmed, I., Buchert, R., Zhou, M., Jiao, X., Mittal, K., Sheikh, T. I., et al. (2014). doi: 10.1089/gte.2007.0048 Mutations in DCPS and EDC3 in autosomal recessive intellectual disability Bress, A., Patel, S. R., Perera, M. A., Campbell, R. T., Kittles, R. A., indicate a crucial role for mRNA decapping in neurodevelopment. Hum. Mol. and Cavallari, L. H. (2012). Effect of NQO1 and CYP4F2 genotypes on Genet. 24, 3172–3180. doi: 10.1093/hmg/ddv069 warfarin dose requirements in Hispanic-Americans and African-Americans. Álvarez, A., Castillo, I., Villamar, M., Aguirre, L. A., González-Neira, A., López- Pharmacogenomics 13, 1925–1935. doi: 10.2217/pgs.12.164 Nevot, A., et al. (2005). High Prevalence of the W24X mutation in the gene Brewster, A. M., Chavez-MacGregor, M., and Brown, P. (2014). Epidemiology, encoding Connexin-26 (GJB2) in Spanish Romani (Gypsies) with autosomal biology, and treatment of triple-negative breast cancer in women of African recessive non-syndromic hearing loss. Am. J. Med. Genet. 137A, 255–258. doi: ancestry. Lancet Oncol. 15, e625–e634. 10.1002/ajmg.a.30884 Browning, S. R., Browning, B. L., Daviglus, M. L., Durazo, R. A., Schneiderman, N., Auton, A., Abecasis, G. R., Altshuler, D. M., Durbin, R. M., Bentley, D. R., Kaplan, R. C., et al. (2018). Ancestry-specific recent effective population size in Chakravarti, A., et al. (2015). A global reference for human genetic variation. the Americas. PLoS Genet. 14:e1007385. doi: 10.1371/journal.pgen.1007385 Nature 526, 68–74. doi: 10.1038/nature15393 Cabrera-Serrano, M., Mavillard, F., Biancalana, V., Rivas, E., Morar, B., Hernández- Azmanov, D. N., Dimitrova, S., Florez, L., Cherninkova, S., Draganov, D., Morar, B., Laín, A., et al. (2018). A Roma founder BIN1 mutation causes a novel phenotype et al. (2011). LTBP2 and CYP1B1 mutations and associated ocular phenotypes of centronuclear myopathy with rigid spine. Neurology 91, e339–e348. in the Roma/Gypsy founder population. Eur. J. Hum. Genet. 19, 326–333. Casals, F., and Bertranpetit, J. (2012). Human genetic variation, shared and private. doi: 10.1038/ejhg.2010.181 Science 336, 39–40. doi: 10.1126/science.1224528 Barbujani, G., Ghirotto, S., and Tassi, F. (2013). Nine things to remember about Casas, J. P., Hingorani, A. D., Bautista, L. E., and Sharma, P. (2004). Meta-analysis human genome diversity. Tissue Antigens82, 155–164. doi: 10.1111/tan.12165 of genetic studies in ischemic stroke: thirty-two genes involving approximately Batai, K., Hooker, S., and Kittles, R. A. (2020). Leveraging genetic ancestry to study 18 000 cases and 58 000 controls. Arch. Neurol. 61, 1652–1661. doi: 10.1001/ health disparities. Am. J. Phys. Anthropol. 175, 1–13. archneur.61.11.1652 Bentley, A. R., Callier, S., and Rotimi, C. N. (2017). Diversity and inclusion in Cecchin, E., D’Andrea, M., Lonardi, S., Zanusso, C., Pella, N., Errante, genomic research: why the uneven progress? J. Commun. Genet. 8, 255–266. D., et al. (2013). A prospective validation pharmacogenomic study in doi: 10.1007/s12687-017-0316-6 the adjuvant setting of colorectal cancer patients treated with the 5- Bianco, E., Laval, G., Font-Porterias, N., García-Fernandez, C., Dobon, B., Sabido- fluorouracil/leucovorin/oxaliplatin (FOLFOX4) regimen. Pharmacogenomics J. Vera, R., et al. (2020). Recent common origin, reduced population size, and 13, 403–409. doi: 10.1038/tpj.2012.31 Frontiers in Genetics | www.frontiersin.org 9 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma Claramunt, R., Sevilla, T., Lupo, V., Cuesta, A., Millán, J. M., Vílchez, J. J., et al. Hopkins, A. L., and Groom, C. R. (2002). The druggable genome. Nat. Rev. Drug (2007). The p.R1109X mutation in SH3TC2 gene is predominant in Spanish Discov. 1, 727–730. Gypsies with Charcot-Marie-Tooth disease type 4. Clin. Genet. 71, 343–349. Hudson, M., Garrison, N. A., Sterling, R., Caron, N. R., Fox, K., Yracheta, J., doi: 10.1111/j.1399-0004.2007.00774.x et al. (2020). Rights, interests and expectations: indigenous. Nat. Rev. Genet. Council of Europe, (2012). Human Rights of Roma and Travellers in Europe. 21, 377–384. Strasbourg: Council of Europe. International HapMap Consortium, (2003). The international HapMap project. Davydov, E. V., Goode, D. L., Sirota, M., Cooper, G. M., Sidow, A., and Batzoglou, Nature 426, 789–796. doi: 10.1038/nature02168 S. (2010). Identifying a high fraction of the human genome to be under selective Jasiecki, J., Żuk, M., Krawczyńska, N., Jońca, J., Szczoczarz, A., Lewandowski, K., constraint using GERP++. PLoS Comput. Biol. 6:e1001025. doi: 10.1371/journal. et al. (2019). Haplotypes of butyrylcholinesterase K-variant and their influence pcbi.1001025 on the enzyme activity. Chem. Biol. Interact. 307, 154–157. doi: 10.1016/j.cbi. Dawed, A. Y., Donnelly, L., Tavendale, R., Carr, F., Leese, G., Palmer, C. N. A., 2019.05.007 et al. (2016). CYP2C8 and SLCO1B1 variants and therapeutic response to Javorský, M., Gotthardová, I., Klimčáková, L., Kvapil, M., Židzik, J., Schroner, Z., thiazolidinediones in patients with Type 2 diabetes. Diabetes Care 39, 1902– et al. (2016). A missense variant in GLP1R gene is associated with the glycaemic 1908. doi: 10.2337/dc15-2464 response to treatment with gliptins. Diabetes Obes Metab. 18, 941–944. doi: De La Vega, F. M., and Bustamante, C. D. (2018). Polygenic risk scores: a biased 10.1111/dom.12682 prediction? Genome Med. 10:100. Johnson, J., Caudle, K., Gong, L., Whirl-Carrillo, M., Stein, C., Scott, S., Decker, B., Allen, J., Luccarini, C., Pooley, K. A., Shah, M., Bolla, M. K., et al. (2017). et al. (2017). Clinical pharmacogenetics implementation consortium (CPIC) Rare, protein-truncating variants in ATM, CHEK2 and PALB2, but not XRCC2, guideline for pharmacogenetics-guided warfarin dosing: 2017 update. Clin. are associated with increased breast cancer risks. J. Med. Genet. 54, 732–741. Pharmacol. Ther. 102, 397–404. doi: 10.1002/cpt.668 Dhoro, M., Ngara, B., Kadzirange, G., Nhachi, C., and Masimirembwa, C. (2014). Kalaydjieva, L., Gresham, D., and Calafell, F. (2001). Genetic studies of the Roma Genetic variants of drug metabolizing enzymes and drug transporter (ABCB1) (Gypsies): a review. BMC Med. Genet. 2:5. doi: 10.1186/1471-2350-2-5 as possible biomarkers for adverse drug reactions in an HIV/AIDS Cohort in Kessler, M. D., Yerges-Armstrong, L., Taub, M. A., Shetty, A. C., Maloney, K., Jeng, Zimbabwe. Curr. HIV Res. 11, 481–490. doi: 10.2174/1570162x113119990048 L. J. B., et al. (2016). Challenges and disparities in the application of personalized Dickinson, L., Amin, J., Else, L., Boffito, M., Egan, D., Owen, A., et al. (2016). genomic medicine to populations with African ancestry. Nat. Commun. 7, 1–8. Comprehensive pharmacokinetic, pharmacodynamic and pharmacogenetic Kim, M. S., Patel, K. P., Teng, A. K., Berens, A. J., and Lachance, J. (2018). evaluation of once-daily Efavirenz 400 and 600 mg in Treatment-Naïve Genetic disease risks can be misestimated across global populations. Genome HIV-Infected Patients at 96 weeks: results of the ENCORE1 study. Clin. Biol. 19:179. Pharmacokinet. 55, 861–873. doi: 10.1007/s40262-015-0360-5 Kim, S., Chen, J., Cheng, T., Gindulyte, A., He, J., He, S., et al. (2019). PubChem Dong, N., Yu, J., Wang, C., Zheng, X., Wang, Z., Di, L., et al. (2012). 2019 update: improved access to chemical data. Nucleic Acids Res. 47, D1102– Pharmacogenetic assessment of clinical outcome in patients with metastatic D1109. breast cancer treated with docetaxel plus capecitabine. J. Cancer Res. Clin. Landrum, M. J., Lee, J. M., Riley, G. R., Jang, W., Rubinstein, W. S., Church, D. M., Oncol. 138, 1197–1203. doi: 10.1007/s00432-012-1183-5 et al. (2014). ClinVar: public archive of relationships among sequence variation Dopazo, J., Amadoz, A., Bleda, M., Garcia-Alonso, L., Alemán, A., García-García, and human phenotype. Nucleic Acids Res. 42, 980–985. F., et al. (2016). 267 spanish exomes reveal population-specific differences in Lao, O., Lu, T. T., Nothnagel, M., Junge, O., Freitag-Wolf, S., Caliebe, A., et al. disease-related genetic variation. Mol. Biol. Evol. 33, 1205–1218. doi: 10.1093/ (2008). Correlation between genetic and geographic structure in Europe. Curr. molbev/msw005 Biol. 18, 1241–1248. Dorado, P., Cavaco, I., Cáceres, M. C., Piedade, R., Ribeiro, V., and LLerena, Lewis, B. C., Korprasertthaworn, P., and Miners, J. O. (2016). Impaired A. (2008). Relationship between CYP2C8 genotypes and diclofenac 5- dacarbazine activation and 7-ethoxyresorufin deethylation in vitro by hydroxylation in healthy Spanish volunteers. Eur. J. Clin. Pharmacol. 64, polymorphic variants of CYP1A1 and CYP1A2: implications for cancer 967–970. doi: 10.1007/s00228-008-0508-4 therapy. Pharmacogenet. Genomics 26, 453–461. doi: 10.1097/fpc.0000000000 Dropinski, J., Musial, J., Sanak, M., Wegrzyn, W., Nizankowski, R., and Szczeklik, 000236 A. (2007). Antithrombotic effects of aspirin based on PLA1/A2 glycoprotein Liao, Y., Wang, J., Jaehnig, E. J., Shi, Z., and Zhang, B. (2019). WebGestalt 2019: IIIa polymorphism in patients with coronary artery disease. Thromb. Res. 119, gene set analysis toolkit with revamped UIs and APIs. Nucleic Acids Res. 47, 301–303. doi: 10.1016/j.thromres.2006.03.005 W199–W205. Font-Porterias, N., Caro-Consuegra, R., Lucas-Saìnchez, M., Lopez, M., Gimeìnez, Lin, P. H., Kuo, W. H., Huang, A. C., Lu, Y. S., Lin, C. H., Kuo, S. H., et al. (2016). A., Carballo, A., et al. (2021). The counteracting effects of demography on Multiple gene sequencing for risk assessment in patients with early-onset or functional genomic variation: the Roma paradigm. Mol. Biol. Evol. msab070. familial breast cancer. Oncotarget 7, 8310–8320. Font-Porterias, N., Arauna, L. R., Poveda, A., Bianco, E., Rebato, E., Prata, M. J., MacGregor, S., Lind, P. A., Bucholz, K. K., Hansell, N. K., Madden, P. A. F., Richter, et al. (2019). European Roma groups show complex West Eurasian admixture M. M., et al. (2009). Associations of ADH and ALDH2 gene variation with self footprints and a common South Asian genetic origin. PLoS Genet. 15:e1008417. report alcohol reactions, consumption and dependence: an integrated analysis. doi: 10.1371/journal.pgen.1008417 Hum. Mol. Genet. 18, 580–593. doi: 10.1093/hmg/ddn372 Foulkes, W. D., Smith, I. E., and Reis-Filho, J. S. (2010). Triple-Negative Breast Majerus, P. W. (1994). Bad blood by mutation. Nature 369, 14–15. doi: 10.1038/ Cancer. N. Engl. J. Med. 363, 1938–1948. 369014a0 Gamella, J. F., Carrasco-Muñoz, E. M., and Núñez Negrillo, A. M. (2013). Maples, B. K., Gravel, S., Kenny, E. E., and Bustamante, C. D. (2013). Oculocutaneous albinism and consanguineous marriage among Spanish RFMix?: a discriminative modeling approach for rapid and robust local- Gitanos or Calé-a study of 83 cases. Coll Antropol. 37, 723–734. ancestry inference. Am. J. Hum. Genet. 93, 278–288. doi: 10.1016/j.ajhg.2013. Gurdasani, D., Barroso, I., Zeggini, E., and Sandhu, M. S. (2019). Genomics of 06.020 disease risk in globally diverse populations. Nat. Rev. Genet. 20, 520–535. doi: Marnetto, D., Pärna, K., Läll, K., Molinaro, L., Montinaro, F., Haller, T., et al. 10.1038/s41576-019-0144-0 (2020). Ancestry deconvolution and partial polygenic score can improve Ha, E., Kim, M. J., Choi, B. K., Rho, J. J., Oh, D. J., Rho, T. H., et al. (2006). Positive susceptibility predictions in recently admixed individuals. Nat. Commun. association of obesity with single nucleotide polymorphisms of syndecan 3 in 11, 1–9. the Korean population. J. Clin. Endocrinol. Metab. 91, 5095–5099. doi: 10.1210/ Márquez-Luna, C., Loh, P. R., Price, A. L., Kooner, J. S., Saleheen, D., Sim, X., et al. jc.2005-2086 (2017). Multiethnic polygenic risk scores improve risk prediction in diverse Hancock, I. F. (2002). We are the Romani People. Hatfield: University of populations. Genet. Epidemiol. 41, 811–823. doi: 10.1002/gepi.22083 Hertfordshire Press. Martin, A. R., Gignoux, C. R., Walters, R. K., Wojcik, G. L., Neale, B. M., Gravel, Heyer, E., and Reynaud-Paligot, C. (2020). Us and them: from prejudice to racism. S., et al. (2017). Human demographic history impacts genetic risk prediction An original analysis of race and racism. Am. J. Phys. Anthropol. 175,1–9. doi: across diverse populations. Am. J. Hum. Genet. 100, 635–649. doi: 10.1016/j. 10.4135/9781446218877.n1 ajhg.2017.03.004 Frontiers in Genetics | www.frontiersin.org 10 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma Martin, A. R., Kanai, M., Kamatani, Y., Okada, Y., Neale, B. M., and Daly, M. J. Sevilla, T., Martínez-Rubio, D., Márquez, C., Paradas, C., Colomer, J., Jaijo, T., (2019). Clinical use of current polygenic risk scores may exacerbate health et al. (2013). Genetics of the charcot-marie-tooth disease in the spanish gypsy disparities. Nat. Genet. 51, 584–591. doi: 10.1038/s41588-019-0379-x population: the hereditary motor and sensory neuropathy-Russe in depth. Clin. Martinez-Frias, M. L., and Bermejo, E. (1992). Prevalence of congenital anomaly Genet. 83, 565–570. doi: 10.1111/cge.12015 syndromes in a Spanish gypsy population. J. Med. Genet. 29, 483–486. Shek, A. B., Kurbanov, R. D., Abdullaeva, G. J., Nagay, A. V., Hoshimov, S. U., Matras, Y. (2002). Romani: A Linguistic Introduction. Cambridge, MA: Cambridge Nizamov, U. I., et al. (2017). Simvastatin intolerance genetic determinants: some University Press. features in ethnic Uzbek patients with coronary artery disease. Arch. Med. Sci. McGuire, M. C., Nogueira, C. P., Bartels, C. F., Lightstone, H., Hajra, A., Van Atheroscler. Dis. 2, 68–75. doi: 10.5114/amsad.2017.70597 der Spek, A. F. L., et al. (1989). Identification of the structural mutation Shu, Y., Brown, C., Castro, R. A., Shi, R. J., Lin, E. T., Owen, R. P., et al. (2008). Effect responsible for the dibucaine-resistant (atypical) variant form of human serum of genetic variation in the organic cation transporter 1, OCT1, on metformin cholinesterase. Proc. Natl. Acad. Sci. U.S.A. 86, 953–957. doi: 10.1073/pnas.86. pharmacokinetics. Clin. Pharmacol. Ther. 83, 273–280. doi: 10.1038/sj.clpt. 3.953 6100275 McLaren, W., Gil, L., Hunt, S.E., et al. (2016). The Ensembl Variant Effect Predictor. Sipeky, C., Weber, A., Szabo, M., Melegh, B. I., Janicsek, I., Tarlos, G., et al. (2013). Genome Biol. 17, 122. doi: 10.1186/s13059-016-0974-4 High prevalence of CYP2C19∗ 2 allele in Roma samples: study on Roma and Mendizabal, I., Lao, O., Marigorta, U. M., Kayser, M., and Comas, D. Hungarian population samples with review of the literature. Mol. Biol. Rep. 40, (2013). Implications of population history of European Romani on genetic 4727–4735. doi: 10.1007/s11033-013-2569-4 susceptibility to disease. Hum. Hered. 76, 194–200. doi: 10.1159/000360762 Sirisena, N. D., Adeyemo, A., Kuruppu, A. I., Samaranayake, N., and Dissanayake, Mendizabal, I., Lao, O., Marigorta, U. M., Wollstein, A., Gusmão, L., Ferak, V., V. H. W. (2018). Genetic variants associated with clinicopathological profiles et al. (2012). Reconstructing the population history of European Romani from in sporadic breast cancer in Sri Lankan women. J. Breast Cancer 21, 165–172. genome-wide data. Curr. Biol. 22, 2342–2349. doi: 10.1016/j.cub.2012.10.039 doi: 10.4048/jbc.2018.21.2.165 Minárik, G., Ferák, V., Feráková, E., Ficek, A., Poláková, H., and Kádasi, L. (2003). Sirugo, G., Williams, S. M., and Tishkoff, S. A. (2019). The missing diversity in High frequency of GJB2 Mutation W24X among Slovak Romany (Gypsy) human genetic studies. Cell 177, 26–31. doi: 10.1016/j.cell.2019.02.048 patients with non-syndromic hearing loss (NSHL). Gen. Physiol. Biophys. 22, Škarić-Jurić, T., Tomas, Ž, Petranović, M. Z., Božina, N., Naranćić, N. S., Janićijević, 549–556. B., et al. (2018). Characterization of ADME genes variation in Roma and Moorjani, P., Patterson, N., Loh, P.-R., Lipson, M., Kisfali, P., Melegh, B. I., 20 populations worldwide. PLoS One 13:e0207671. doi: 10.1371/journal.pone. et al. (2013). Reconstructing Roma history from genome-wide data. PLoS One 0207671 8:e58633. doi: 10.1371/journal.pone.0058633 Slavin, T. P., Maxwell, K. N., Lilyquist, J., Vijai, J., Neuhausen, S. L., Hart, S. N., et al. Morar, B., Azmanov, D. N., and Kalaydjieva, L. (2013). “Roma (Gypsies): genetic (2017). The contribution of pathogenic variants in breast cancer susceptibility Studies,” in eLS (Chichester: Wiley). Eds., John Wiley & Sons, Ltd genes to familial breast cancer risk. npj Breast Cancer 3:22. Motovska, Z., Widimsky, P., Kvasnicka, J., Petr, R., Bilkova, D., Hajkova, J., et al. Steponaitiene, R., Kupcinskas, J., Survilaite, S., Varkalaite, G., Jonaitis, L., Kiudelis, (2009). High loading dose of clopidogrel is unable to satisfactorily inhibit G., et al. (2016). TPMT and ITPA genetic variants in Lithuanian inflammatory platelet reactivity in patients with glycoprotein IMA gene polymorphism: a bowel disease patients: prevalence and azathioprine-related side effects. Adv. genetic substudy of PRAGUE-8 trial. Blood Coagul. Fibrinolysis 20, 257–262. Med. Sci. 61, 135–140. doi: 10.1016/j.advms.2015.09.008 doi: 10.1097/mbc.0b013e328325455b Stewart, C., and Pepper, M. S. (2017). Cystic fibrosis in the African diaspora. Ann. Nagy, A., Sipeky, C., Szalai, R., Melegh, B. I., Matyas, P., Ganczer, A., et al. (2015). Am. Thorac. Soc. 14, 1–7. doi: 10.1513/annalsats.201606-481fr Marked differences in frequencies of statin therapy relevant SLCO1B1 variants Stocco, G., Yang, W., Crews, K. R., Thierfelder, W. E., Decorti, G., Londero, and haplotypes between Roma and Hungarian populations. BMC Genet. 16:108. M., et al. (2012). PACSIN2 polymorphism influences TPMT activity and doi: 10.1186/s12863-015-0262-4 mercaptopurine-related gastrointestinal toxicity. Hum. Mol. Genet. 21, 4793– Need, A. C., and Goldstein, D. B. (2009). Next generation disparities in human 4804. doi: 10.1093/hmg/dds302 genomics: concerns and remedies. Trends Genet. 25, 489–494. doi: 10.1016/j. Sundelin, E. I. O., Gormsen, L. C., Jensen, J. B., Vendelbo, M. H., Jakobsen, S., tig.2009.09.012 Munk, O. L., et al. (2017). Genetic polymorphisms in organic cation transporter O’Connell, J., Gurdasani, D., Delaneau, O., Pirastu, N., Ulivi, S., Cocca, M., et al. 1 attenuates hepatic metformin exposure in humans. Clin. Pharmacol. Ther. (2014). A general approach for haplotype phasing across the full spectrum of 102, 841–848. doi: 10.1002/cpt.701 relatedness. PLoS Genet. 10:e1004234. doi: 10.1371/journal.pgen.1004234 Takkenberg, J. J. M., and Roos-hesselink, J. W. (2011). Birth prevalence of Padoa, C., Goldman, A., Jenkins, T., and Ramsay, M. (1999). Cystic fibrosis carrier congenital heart disease worldwide a systematic review and meta-analysis. JAC frequencies in populations of African origin. J. Med. Genet. 36, 41–44. 58, 2241–2247. Peshkin, B. N., Alabek, M. L., and Isaacs, C. (2010). BRCA1/2 mutations and triple Teixeira, J., Amorim, A., Prata, J. M., and Quental, S. (2015). Pharmacogenetic negative breast cancers. Breast Dis. 32, 25–33. doi: 10.3233/bd-2010-0306 polymorphisms in a portuguese gypsy population. Curr. Pharmacogenomics Phani, N. M., Guddattu, V., Bellampalli, R., Seenappa, V., Adhikari, P., Nagri, S. K., Person. Med. 13, 36–40. doi: 10.2174/1875692113666150703180101 et al. (2014). Population specific impact of genetic variants in KCNJ11 gene to Tomita-Mitchell, A., Maslen, C. L., Morris, C. D., Garg, V., and Goldmuntz, E. type 2 diabetes: a case-control and meta-analysis study. PLoS One 9:e107021. (2007). GATA4 sequence variants in patients with congenital heart disease. doi: 10.1371/journal.pone.0107021 J. Med. Genet. 44, 779–783. doi: 10.1136/jmg.2007.052183 Popejoy, A. B., and Fullerton, S. M. (2016). Genomics is failing on diversity. Nature Tzvetkov, M. V., Saadatmand, A. R., Lötsch, J., Tegeder, I., Stingl, J. C., and 538, 161–164. doi: 10.1038/538161a Brockmöller, J. (2011). Genetically polymorphic OCT1: another piece in Prado, Y., Arencibia, A., Zambrano, T., and Salazar, L. A. (2018). Gender-specific the puzzle of the variable pharmacokinetics and pharmacodynamics of the association between ABCC2 -24C>T SNP and reduction in triglycerides in opioidergic drug tramadol. Clin. Pharmacol. Ther. 90, 143–150. doi: 10.1038/ chilean patients treated with atorvastatin. Basic Clin. Pharmacol. Toxicol. 122, clpt.2011.56 517–522. doi: 10.1111/bcpt.12943 Visscher, H., Ross, C. J. D., Rassekh, S. R., Barhdadi, A., Dubé, M. P., Al- Reckova, M., Mardiak, J., Plank, L., Vulevova, M., Cingelova, S., and Mego, M. Saloos, H., et al. (2012). Pharmacogenomic prediction of anthracycline-induced (2017). Differences in incidence and biological characteristics of breast cancer cardiotoxicity in children. J. Clin. Oncol. 30, 1422–1428. between roma and non-roma patients in Slovakia. Klin Onkol. 30, 48–54. doi: Vozarova De Courten, B., De Courten, M., Hanson, R. L., Zahorakova, A., Egyenes, 10.14735/amko201748 H. P., Tataranni, P. A., et al. (2003). Higher prevalence of type 2 diabetes, Rentzsch, P., Witten, D., Cooper, G. M., Shendure, J., and Kircher, M. (2019). metabolic syndrome and cardiovascular diseases in gypsies than in non-gypsies CADD: predicting the deleteriousness of variants throughout the human in Slovakia. Diabetes Res. Clin. Pract. 62, 95–103. doi: 10.1016/s0168-8227(03) genome. Nucleic Acids Res. 47, D886–D894. 00162-1 Rocha, H., Castiñeiras, D., Delgado, C., Egea, J., Yahyaoui, R., González, Y., et al. Werissa, N. A., Piko, P., Fiatal, S., Kosa, Z., Sandor, J., and Adany, R. (2019). SNP- (2014). Birth prevalence of fatty acid β-Oxidation disorders in iberia. JIMD Rep. based genetic risk score modeling suggests no increased genetic susceptibility of 16, 89–94. doi: 10.1007/8904_2014_324 the roma population to Type 2 diabetes mellitus. Genes 10, 1–16. Frontiers in Genetics | www.frontiersin.org 11 June 2021 | Volume 12 | Article 683880 Font-Porterias et al. Clinically Relevant Variants in Roma Whirl-Carrillo, M., McDonagh, E. M., Hebert, J. M., Gong, L., Sangkuhl, K., Thorn, Zhang, W., Li, X., Shen, A., Jiao, W., Guan, X., and Li, Z. (2008). GATA4 mutations C. F., et al. (2012). Pharmacogenomics knowledge for personalized medicine. in 486 Chinese patients with congenital heart disease. Eur. J. Med. Genet. 51, Clin. Pharmacol. Ther. 92, 414–417. 527–535. doi: 10.1016/j.ejmg.2008.06.005 Wishart, D. S., Feunang, Y. D., Guo, A. C., Lo, E. J., Marcu, A., Grant, J. R., et al. Živković, T. B., Marjanović, M., Prgomelja, S., Soldatović, I., Koprivica, B., Acković, (2018). DrugBank 5.0: a major update to the DrugBank database for 2018. D., et al. (2010). Screening for diabetes among roma people living in Serbia. Nucleic Acids Res. 46, D1074–D1082. Croat Med. J. 51, 144–150. doi: 10.3325/cmj.2010.51.144 Wray, N. R., Goddard, M. E., and Visscher, P. M. (2007). Prediction of individual genetic risk to disease from genome-wide association Conflict of Interest: The authors declare that the research was conducted in the studies. Genome Res. 17, 1520–1528. doi: 10.1101/gr.666 absence of any commercial or financial relationships that could be construed as a 5407 potential conflict of interest. Yoon, H., Cho, H. Y., Yoo, H. D., Kim, S. M., and Lee, Y. B. (2013). Influences of organic cation transporter polymorphisms on the population pharmacokinetics Copyright © 2021 Font-Porterias, Giménez, Carballo-Mesa, Calafell and Comas. of metformin in healthy subjects. AAPS J. 15, 571–580. doi: 10.1208/s12248- This is an open-access article distributed under the terms of the Creative Commons 013-9460-z Attribution License (CC BY). The use, distribution or reproduction in other forums Zhang, C., Dong, S., Xu, J., He, W., and Yang, T. (2018). PopLDdecay?: a fast is permitted, provided the original author(s) and the copyright owner(s) are credited and effective tool for linkage disequilibrium decay analysis based on variant and that the original publication in this journal is cited, in accordance with accepted call format files. Bioinformatics 35, 1786–1788. doi: 10.1093/bioinformatics/ academic practice. No use, distribution or reproduction is permitted which does not bty875 comply with these terms. Frontiers in Genetics | www.frontiersin.org 12 June 2021 | Volume 12 | Article 683880