The Impact of Population Variation in the Analysis of Microrna Target Sites

non-coding RNA

Article The Impact of Population Variation in the Analysis of microRNA Target Sites

Mohab Helmy †, Andrea Hatlen and Antonio Marco * School of Biological Sciences, University of Essex, Colchester CO4 3SQ, UK; [email protected] (M.H.); [email protected] (A.H.) * Correspondence: [email protected]; Tel.: +44-1206-873339 Current address: Faculty of Life Sciences & Medicine, King’s College London, London SE1 9RT, UK. † Received: 22 March 2019; Accepted: 20 June 2019; Published: 22 June 2019

Abstract: The impact of population variation in the analysis of regulatory interactions is an underdeveloped area. MicroRNA target recognition occurs via pairwise complementarity. Consequently, a number of computational prediction tools have been developed to identify potential target sites that can be further validated experimentally. However, as microRNA target predictions are done mostly considering a reference genome sequence, target sites showing variation among populations are neglected. Here, we studied the variation at microRNA target sites in human populations and quantified their impact in microRNA target prediction. We found that African populations carry a significant number of potential microRNA target sites that are not detectable in the current human reference genome sequence. Some of these targets are conserved in primates and only lost in Out-of-Africa populations. Indeed, we identified experimentally validated microRNA/transcript interactions that are not detected in standard microRNA target prediction programs, yet they have segregating target alleles abundant in non-European populations. In conclusion, we show that ignoring population diversity may leave out regulatory elements essential to understand disease and gene expression, particularly neglecting populations of African origin.

Keywords: miRNAs; human populations; gene regulation; microRNA target prediction; evolution

1. Introduction The study of gene regulatory iterations is at the heart of biomedical research. Many disease-causing mutations affect regulatory motifs that eventually lead to a miss-regulation of fine-tuned biological processes. In the last years, microRNAs have emerged as an important type of regulatory molecules involved in virtually all biological networks [1,2]. Both the deletion and overexpression of microRNAs have been associated to human diseases (see for instance the compilation by [3]). Notably, the role of microRNAs in cancer is a particularly active research field [4,5]. Unlike other regulatory molecules (such as transcription factors or RNA binding proteins), the interaction between regulator and target is mediated by pairwise nucleotide complementary [1]. Specifically, a microRNA is partially complementary to motifs in RNA transcripts and, after binding, translational repression or even RNA degradation mechanisms are triggered [1,6]. Indeed, changes at microRNA target sites are now associated to major human diseases including cancer, neurodegeneration and metabolic disorders (see [7,8] and references therein), among others. The study of microRNA target sites is becoming more important as we start to understand the complexities behind microRNA regulatory networks. From a computational point of view, the microRNA targeting mechanisms have a clear advantage: targets can be predicted by scanning genomic sequences for potential complementary sites [9–11]. However, microRNA target prediction algorithms have, in general, high rates of false positives [12,13]. Consequently, most research on microRNA targets complements a computational sieve of potential

Non-coding RNA 2019, 5, 42; doi:10.3390/ncrna5020042 www.mdpi.com/journal/ncrna Non-coding RNA 2019, 5, 42 2 of 8 target sites plus experimental validations, usually by luciferase assays or similar (e.g., [14]). Therefore, the ﬁrst step in microRNA target analysis is the prediction of binding sites from a primary sequence. In biomedical research, this primary sequence is often, if not always, the human reference genome sequence (currently hg38). However, a reference genome does not take into account existing variation within human populations and, therefore, any inference from such a sequence may be biased. Populations of African origin show a higher nucleotide diversity than other populations, probably as a consequence of a recent origin of Eurasian populations from a small group of African migrants (reviewed in [15]). There exist several projects aimed to account for all this nucleotide variability in human populations (i.e., [16]). Nevertheless, microRNA target prediction programs rely on reference genome sequences (as other regulatory interaction programs do), where this diversity is not truly represented. Variation at gene regulatory sites, particularly at transcription factor binding sites, has been studied in the past (e.g., [17,18]). In microRNA target sites, some studies have shown that, as expected, functional target sites are often conserved in populations [19,20]. On the other hand, the creation of novel, potentially harmful, microRNA target sites, is selected against in populations [21,22]. In summary, the literature clearly shows evidence that nucleotide variation occurs at microRNA target sites. The question is, how much of this variation is neglected in the human reference genome sequence, and therefore not accounted for in standard microRNA target prediction programs? In other words, what is the impact of nucleotide variation among humans in microRNA target prediction research? This is the question we are tackling in this paper.

2. Results and Discussion We compiled biallelic SNPs at predicted microRNA target sites (see Methods) for which one of the alleles is a target site and the other is not a target site. These are the target/near-target pairs as defined before [11,21]. Considering that populations of African origin host more variation, including ancestral alleles, than other populations, we hypothesize that ancestral target sites that are not present in the reference genome sequence were lost in Out-of-Africa populations. Thus, we first consider target sites whose allele in the reference genome sequence is not a target, yet the ancestral allele was a target site. The distribution of these alleles frequencies in five major groups (African, European, South Asian, Native American and East Asian) indicates that non-African populations tend to have the non-target allele whilst African populations show a wider variation, and a paucity of non-targets compared to the other populations (Figure1, top). Additionally, we consider target sites in the reference genome sequence that were not ancestral to human populations and, as expected, we found that African populations have lower frequencies of the target allele (Figure1, bottom). These results suggest that a significant loss and gain of microRNA target sites happened in Out-of-African populations. To illustrate the impact of variation at target sites we selected SNPs with a very high degree of population differentiation (Fst > 0.7; see Methods) that are not present in the human genome reference sequence. We found 26 target sites with such a degree of variation across populations (Table1) for highly expressed microRNAs. Many of these target sites are lost in Out-of-Africa populations, suggesting that losses rather than the gains of target sites are more likely to happen during differentiation between populations (see discussion in [22]). Non-coding RNA 2019, 5, 42 3 of 8 Non-Coding RNA 2019, 5, x 3 of 8

Figure 1. Target alleleallele frequenciesfrequencies at at microRNA microRNA target target sites sites where where the the target target is ancestralis ancestral and and not not in the in top thereference reference genome genome sequence sequence ( ) ( ortop the) or target the target is derived is derived and present and present in the reference in the reference genome sequencegenome (bottom). Super-populations are African (AFR), Mixed American (AMR), East Asian (EAS), European sequence (bottom). Super-populations are African (AFR), Mixed American (AMR), East Asian (EAS), (EUR) and South Asian (SAS). European (EUR) and South Asian (SAS). Table 1. MicroRNA target sites, not in the reference genome sequence, with a very high diﬀerentiation Toamong illustrate human the populations. impact of variation at target sites we selected SNPs with a very high degree of population differentiation (Fst > 0.7; see Methods) that are not present in the human genome a b b b b b referenceMicroRNA sequence. We Gene found 26 SNP target sitesEAS with AMRsuch a degreeAFR of variationEUR acrossSAS populationsFst (TablemiR-519a-5p 1) for highly1 MTAPexpressed rs7868374microRNAs.0.0079 Many 0.0476of these target 0.6838 sites 0.0050 are lost 0.1176 in Out-of-Africa 0.8239 populations,let-7a-5p suggesting2 MTAP that lossesrs7875199 rather than0.0536 the gains 0.0576 of target 0.7958 sites 0.0070 are more 0.1288 likely to 0.7997 happen 3 duringmiR-337-3p differentiationATP1A1 between populars1885802tions 0.0427(see discussion 0.1268 in [22]). 0.8109 0.0338 0.0450 0.7914 miR-185-3p 4 SLC5A10 rs1624825 0.9970 0.5159 0.9667 0.2028 0.7495 0.7859 miR-1180-3p MYEF2 rs2470102 0.7530 0.3386 0.9251 0.0060 0.2720 0.7726 Table 1. MicroRNA target sites, not in the reference genome sequence, with a very high miR-151a-5p 5 SCN2B rs624328 0.9405 0.8905 0.2519 0.9513 0.9673 0.7674 miR-4732-5pdifferentiation among AP5M1 human rs1889720populations. 0.9335 0.9107 0.1551 0.9433 0.9182 0.7637 miR-584-5p CYB5R4 rs6903739 0.0000 0.0749 0.7738 0.0477 0.0061 0.7514 MicroRNA Gene SNP a EAS b AMR b AFR b EUR b SAS b Fst miR-618 CYB5R4 rs9449733 1.0000 0.9251 0.2247 0.9523 0.9939 0.7397 1 miR-519a-5pmiR-18a-3p ANKRD65MTAP rs7868374 rs904589 0.0079 0.1131 0.0476 0.1945 0.6838 0.8222 0.0050 0.0964 0.1176 0.1881 0.8239 0.7332 miR-329-3plet-7a-5p2 6 MTAPCEP162 rs7875199rs6901546 0.05360.0139 0.0576 0.1628 0.7958 0.7670 0.0070 0.0706 0.1288 0.0900 0.7997 0.7326 miR-337-3pmiR-150-5p3 7 ATP1A1HM13 rs1885802rs6059873 0.04271.0000 0.1268 0.8156 0.8109 0.1725 0.0338 0.8429 0.0450 0.9622 0.7914 0.7298 miR-513a-3pmiR-185-3p4 8 SLC5A10TCERG1 rs1624825rs3822506 0.99700.7183 0.5159 0.1167 0.9667 0.0325 0.2028 0.0915 0.7495 0.2198 0.7859 0.7224 9 miR-128-1-5pmiR-1180-3p MYEF2BCL7C rs2470102rs11864054 0.75300.9117 0.3386 0.5447 0.9251 0.0242 0.0060 0.3817 0.2720 0.1595 0.7726 0.7144 10 miR-151a-5pmiR-192-5p 5 SCN2BC12orf65 rs624328rs1533703 0.94050.9980 0.8905 0.7262 0.2519 0.1490 0.9513 0.7694 0.9673 0.7945 0.7674 0.7074 a b miR-4732-5pSNP names areAP5M1 dbSNP accessionrs1889720 numbers. 0.9335Allele frequencies0.9107 refer0.1551 to the target0.9433 allele frequency0.9182 in each0.7637 of the ﬁve super-population groups. Superscript in microRNA names indicates other mature sequences with the miR-584-5psame target site: 1CYB5R4miR-520c-5p, rs6903739 miR-519b-5p, 0.0000 miR-526a-5p, 0.0749 miR-519c-5p, 0.7738 miR-518f-5p, 0.0477 miR-523-5p, 0.0061 miR-518d-5p, 0.7514 miR-522-5p,miR-618 miR-518e-5p;CYB5R4 2 let-7e-5p,rs9449733 let-7c-5p, 1.0000 let-7f-5p, 0.9251 let-7d-5p, 0.2247 let-7b-5p, miR-98-5p,0.9523 let-7i-5p,0.9939 let-7g-5p,0.7397 3 4 5 6 miR-18a-3pmiR-1294; miR-202-5p;ANKRD65 miR-128-1-5p;rs904589 0.1131miR-561-5p, 0.1945 miR-3605-5p, 0.8222 miR-24-3p, 0.0964 miR-151b; 0.1881miR-362-3p; 0.7332 7 miR-296-3p, miR-4446-3p, miR-23b-5p, miR-3158-3p; 8 miR-513c-3p; 9 miR-491-5p, miR-1296-5p; 10miR-215-5p. miR-329-3p6 CEP162 rs6901546 0.0139 0.1628 0.7670 0.0706 0.0900 0.7326 miR-150-5pWe observed7 thatHM13MTAP rs6059873have lost two1.0000 target 0.8156 sites for0.1725 two highly0.8429 expressed 0.9622 microRNAs 0.7298 in non-AfricanmiR-513a-3p populations.8 TCERG1 Thesers3822506 two target 0.7183 sites were0.1167 the ancestral0.0325 form0.0915 in human 0.2198 populations 0.7224 and conservedmiR-128-1-5p as targets9 BCL7C in other rs11864054 primates (Figure0.9117 2A).0.5447 Genome-wide 0.0242 association0.3817 studies0.1595 have0.7144 linked 10 variantsmiR-192-5p at MTAP C12orf65with the abundancers1533703 of0.9980 naevi (moles)0.7262 and0.1490 also with 0.7694 melanoma 0.7945 incidence 0.7074 [23 ,24], althougha SNP the names association are dbSNP between accessionMTAP numbers.and melanoma b Allele frequencies seems to berefer rather to the complex target allele [23]. frequency We speculate that thein each loss of of the two five target super-popula sites attion MTAP groups. transcripts Superscrip couldt in microRNA be associated names with indicates an elevated other mature expression 1 of thesequences gene in with Out-of-Africa the same populations,target site: perhapsmiR-520c-5p, by amiR-519b-5p, relaxationon miR-526a-5p, the selective miR-519c-5p, pressures that 2 maintainedmiR-518f-5p, the target miR-523-5p, sites under miR-518d-5p, purifying miR-522-5p, selection inmiR-518e-5p; African populations, let-7e-5p,where let-7c-5p, light let-7f-5p, exposure is let-7d-5p, let-7b-5p, miR-98-5p, let-7i-5p, let-7g-5p, miR-1294; 3 miR-202-5p; 4 miR-128-1-5p; 5 miR-561-5p, miR-3605-5p, miR-24-3p, miR-151b; 6 miR-362-3p; 7 miR-296-3p, miR-4446-3p, miR-23b-5p, miR-3158-3p; 8 miR-513c-3p; 9 miR-491-5p, miR-1296-5p; 10miR-215-5p.

Non-Coding RNA 2019, 5, x 4 of 8

We observed that MTAP have lost two target sites for two highly expressed microRNAs in non-African populations. These two target sites were the ancestral form in human populations and conserved as targets in other primates (Figure 2A). Genome-wide association studies have linked variants at MTAP with the abundance of naevi (moles) and also with melanoma incidence [23,24], although the association between MTAP and melanoma seems to be rather complex [23]. We speculate that the loss of two target sites at MTAP transcripts could be associated with an elevated Non-codingexpression RNA of2019 the, 5 , 42gene in Out-of-Africa populations, perhaps by a relaxation on the selective4 of 8 pressures that maintained the target sites under purifying selection in African populations, where higher,light exposure and an overproduction is higher, and ofan MTAPoverproduction may be linked of MTAP to a higher may be incidence linked to of skina higher cancer. incidence Under thisof skin cancer. Under this hypothesis, we expect the target allele frequency to be also negatively hypothesis, we expect the target allele frequency to be also negatively correlated with the latitude correlated with the latitude of sampling of non-African human populations. We compiled sampling of sampling of non-African human populations. We compiled sampling information for human information for human populations (see Methods) and build a linear model to predict the target populations (see Methods) and build a linear model to predict the target allele frequency as a function allele frequency as a function of the latitude. We observed that populations from higher latitudes of the latitude. We observed that populations from higher latitudes have lower target allele frequencies have lower target allele frequencies at both of the studied target sites (Figure 2B), although the at both of the studied target sites (Figure2B), although the statistical support was weak (rs7875199: statistical support was weak (rs7875199: R2 (adjusted) = 0.18, p = 0.06; rs7868374: R2 (adjusted) = 0.14, R2 (adjusted) = 0.18, p = 0.06; rs7868374: R2 (adjusted) = 0.14, p = 0.09). We would like to emphasize p = 0.09). We would like to emphasize that the association between latitude and MTAP target site that the association between latitude and MTAP target site allele frequencies is still speculative, but the allele frequencies is still speculative, but the fact that other similar associations has been found at fact that other similar associations has been found at microRNA target sites (e.g., [25]) suggest that microRNA target sites (e.g., [25]) suggest that other latitude depending microRNA function remain other latitude depending microRNA function remain to be discovered. to be discovered. Likewise, we found a similar pattern in two target sites at CYB5R4, a gene associated to Likewise, we found a similar pattern in two target sites at CYB5R4, a gene associated to gene-environment interactions [26,27]. In particular, CYB5R4 has been associated to the risk of diabetes, gene-environment interactions [26,27]. In particular, CYB5R4 has been associated to the risk of anddiabetes, variants and at variants its 50 region at its have 5′ region been have found been to be found speciﬁc to be of specific African of populations African populations [26]. The association [26]. The betweenassociation population between variantspopulation and variants disease-related and disease-related genes such as genesMTAP suchand asCYB5R4 MTAPremains and CYB5R4 highly speculative,remains highly yet itspeculative, is very suggestive. yet it is Invery any sugges case, thetive. identiﬁed In any case, target the sites identified at MTAP targetand CYB5R4sites at MTAPare not presentand CYB5R4 in the are human not present genome in reference the human genome genome yet theyrefere arence highly genome abundant yet they in are African highly populations abundant andin African conserved populations in primates, and indicating conserved that in primates, target prediction indicating programs that target often prediction neglect an programs important often part ofneglect human an variability. important part of human variability.

FigureFigure 2.2. MicroRNAMicroRNA targettarget sites variation at transcript from the the MTAP MTAP gene: gene: ( (AA)) alignment alignment of of 3 3′UTR0UTR fragmentsfragments fromfrom 1010 primateprimate species,species, including human, human, and and the the location location of of th thee target target sites sites and and the the polymorphicpolymorphic nucleotides;nucleotides; ( B(B) scatter-plot) scatter-plot between between the the latitude latitude (in (in absolute absolute value) value) and and the targetthe target allele frequencyallele frequency for target for sitestarget for sites miR-519a-5p for miR-519a-5p (red) and(red) let-7a-5p and let-7a-5p (blue). (blue). Straight Straight lines lines are ﬁtted are fitted linear modelslinear models (see main (see text). main text).

WeWe identifiedidentified microRNAmicroRNA/gene/gene interactionsinteractions that have been been experimentally experimentally detected detected in in high-throughput experiments with a significant population differentiation (P < 0.05, see Methods) high-throughput experiments with a significant population differentiation (PFst < 0.05, see Methods) thatthat were were not not detected detected as targetsas targets in the in reference the reference genome genome sequence sequence (Table2 ).(Table For instance, 2). For miR-24-3p,instance, amiR-24-3p, tumour-suppressor a tumour-suppressor microRNA [microRNA28], has a [28], polymorphic has a polymorphic target site target in Caspase site in 10 Caspase transcripts, 10 whosetranscripts, product whose is anproduct component is an component of the apoptotic of the apoptotic machinery. machinery. This target This sitetarget is site present is present in about in 26%about or 26% African or African sampled sampled alleles whilstalleles itwhilst is not it detectedis not detected in most in other most populations. other populations. Indeed, Indeed, neither TargetScanneither TargetScan nor DIANA-microT nor DIANA-microT [10] predict [10] this predict interaction. this interaction. This result This further result supports further supports the view the that reference genome sequences used to predict microRNA target sites often ignore population variability that may indicate functional diversity. As this work is a computational evaluation of potential microRNA target sites, we have not validated any of the targets. However, we wanted to explore whether experimentally validated polymorphic target sites showed some population differentiation. Hence, we explored the 111 variants tested by PASSPORT-seq [29] of which 59 were listed in our database. The distribution of Fst values shows population differentiation for some targets (Figure3). For instance, eight target sites (over 13% of the total) had an associated Fst value of 0.3 or more. Non-Coding RNA 2019, 5, x 5 of 8 view that reference genome sequences used to predict microRNA target sites often ignore population variability that may indicate functional diversity.

Table 2. Target sites identified in high-throughput experiments with high population differentiation.

MicroRNA Gene SNP a EAS b AMR b AFR b EUR b SAS b Fst miR-1307-3p VSTM4 rs4240499 0.2361 0.4971 0.9017 0.3976 0.5031 0.5352 miR-512-3p EIF2B2 rs4556 0.8393 0.6441 0.1884 0.5388 0.5194 0.4327 miR-197-3p DBT rs6701655 0.9296 0.8905 0.4168 0.8648 0.8978 0.4186 Non-codingmiR-17-5p RNA 2019, 5,MRPS10 42 rs3199638 0.7937 0.6369 0.2148 0.6372 0.7679 0.39465 of 8 miR-326 SLC27A4 rs7048106 0.3185 0.2176 0.7617 0.2386 0.3476 0.3782 miR-24-3p CASP10 rs13432040 0.000 0.0259 0.2587 0.000 0.000 0.2943 miR-342-3pTable 2. Target sitesLRPAP1 identified inrs3468 high-throughput 0.5089 experiments 0.2507 with0.1611 high population0.3489 diff0.2669erentiation. 0.238 a SNP namesMicroRNA are dbSNP accession Gene numbers. SNP a b AlleleEAS b frequenciesAMR b referAFR bto theEUR targetb alleleSAS bfrequencyFst in each of the five super-pomiR-1307-3ppulation VSTM4 groups. rs4240499 0.2361 0.4971 0.9017 0.3976 0.5031 0.5352 miR-512-3p EIF2B2 rs4556 0.8393 0.6441 0.1884 0.5388 0.5194 0.4327 As this work is a computational evaluation of potential microRNA target sites, we have not miR-197-3p DBT rs6701655 0.9296 0.8905 0.4168 0.8648 0.8978 0.4186 validated miR-17-5pany of the targets. MRPS10 However,rs3199638 we0.7937 wanted 0.6369 to explore 0.2148 whether 0.6372 experimentally 0.7679 0.3946 validated polymorphicmiR-326 target sites SLC27A4 showedrs7048106 some population0.3185 0.2176 differentiation. 0.7617 0.2386 Hence, 0.3476we explored 0.3782 the 111 variants testedmiR-24-3p by PASSPORT-seq CASP10 rs13432040 [29] of which0.000 59 0.0259were listed 0.2587 in our 0.000 database. 0.000 The distribution 0.2943 of miR-342-3p LRPAP1 rs3468 0.5089 0.2507 0.1611 0.3489 0.2669 0.238 Fst values shows population differentiation for some targets (Figure 3). For instance, eight target a SNP names are dbSNP accession numbers. b Allele frequencies refer to the target allele frequency in each of the sites (overfive 13% super-population of the total) groups. had an associated Fst value of 0.3 or more.

Figure 3. Fst values for polymorphic microRNA target sites experimentally evaluated with FigurePASSPORT-seq 3. Fst values (see mainfor polymorphic text). microRNA target sites experimentally evaluated with PASSPORT-seq (see main text). Last, we investigated polymorphic target sites identified by the popular software TargetScan. WeLast, downloaded we investigated the whole polymorphic conserved and target non-conserved sites identified target datasetsby the popula and identifiedr software polymorphic TargetScan. We targetdownloaded sites also presentthe whole in our databaseconserved PopTargs. and Wenon-conserved compared the targettarget allele datasets frequencies and between identified polymorphicpopulations target and concludedsites also that,present for both in our conserved database and nonconservedPopTargs. We targets, compared the target the allele target was allele frequenciesmore common between in Europeanpopulations than and in African concluded populations that, for (Table both3 ).conserved Indeed, for and nonconserved nonconserved targets, targets, the differences are larger. the target allele was more common in European than in African populations (Table 3). Indeed, for nonconservedTable 3.targets,Wilcoxon the signed-rank differences (paired) are testlarger. between African and European target allele frequencies in TargetScan predicted microRNA target sites. Table 3. Wilcoxon signed-rank (paired) test between African and European target allele frequencies 1 2 in TargetScanTargetScan predicted Dataset microRNA Sample Size target (pairs) sites. p Shift Median Shift 95% CI 16 4 4 conserved 9245 <2.2 10− 7.976 10− (7.833–7.878) 10− × 16 × 3 1 3 2 TargetScannonconserved Dataset Sample Size 1041868 (pairs) <2.2p 10− Shift1.174 Median10− (1.151–1.210)Shift 95% 10− CI × −16 × −4 −4 conserved1 Estimated shift median for the9245 difference: European< 2.2 minus × 10 African target7.976 allele frequency.× 10 2 Confidence(7.833–7.878) Interval. 10 nonconserved 1041868 < 2.2 × 10−16 1.174 × 10−3 (1.151–1.210) 10−3 1 EstimatedIn conclusion, shift median here, we for showed the difference: that the studyEuropean of variation minus Afri in humancan target populations allele frequency. reveals the 2 presenceConfidence of Interval. microRNA target sites that are not detected with current methodological frameworks. We suggest that variation at target sites must be taken into account in biomedical studies as some

populations (mostly of African origin) may be neglected. Further research is needed to quantify the impact of population variation in the function of microRNA target sites. Non-coding RNA 2019, 5, 42 6 of 8

3. Methods Polymorphic sites at predicted microRNA target sites were retrieved from the PopTargs database (version 2; Hatlen, Helmy and Marco, under review; https://poptargs.essex.ac.uk). Only target sites for highly expressed microRNAs were considered. Population frequencies, Fst values and ancestral alleles were also retrieved in PopTargs, as provided in [30]. We only consider segregating alleles with a frequency between 0.01 and 0.99. Sequence alignment of primate sequences was extracted using the UCSC table browser [31]. The geographical location (latitude) of the genome samples sequenced in the 1000 genomes project was obtained from the sample descriptions at https: //www.coriell.org/1/NHGRI/Collections/1000-Genomes-Collections/1000-Genomes-Project. When the speciﬁc location was ambiguous, we used the location of the capital city of the country of origin. For multiple collection points we computed the centroid middle point. Recently migrated populations were discarded (CHD, CEU, ASW, ACB, GIH, STU and ITU as described at IGSR: The International Genome Sample Resource (http://www.internationalgenome.org/category/population/)) were discarded from the analysis in Figure2B. We extracted microRNA targets from the miRTarBase database [ 32], which were detected only in high-throughput experiments (PAR-CLIP and/or HITS-CLIP).

Author Contributions: Conceptualization, A.M.; methodology, A.H. and A.M.; formal analysis, M.H. and A.M.; investigation, M.H., A.H. and A.M.; data curation, M.H., A.H. and A.M.; writing—original draft preparation, A.M.; writing—review and editing, M.H. and A.M.; supervision, A.M.; project administration, A.M..; funding acquisition, A.M. Funding: This research was funded by Wellcome Trust, grant number 200585/Z/16/Z. Acknowledgments: The authors acknowledge the use of the High Performance Computing Facility (Ceres) and its associated support services at the University of Essex in the completion of this work. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Bartel, D.P. MicroRNAs: Target Recognition and Regulatory Functions. Cell 2009, 136, 215–233. [CrossRef] [PubMed] 2. Axtell, M.J.; Westholm, J.O.; Lai, E.C. Vive la différence: Biogenesis and evolution of microRNAs in plants and animals. Genome Biol. 2011, 12, 221. [CrossRef][PubMed] 3. Jiang, Q.; Wang, Y.; Hao, Y.; Juan, L.; Teng, M.; Zhang, X.; Li, M.; Wang, G.; Liu, Y. miR2Disease: A manually curated database for microRNA deregulation in human disease. Nucleic Acids Res. 2009, 37, D98–D104. [CrossRef][PubMed] 4. Di Leva, G.; Garofalo, M.; Croce, C.M. microRNAs in cancer. Annu. Rev. Pathol. 2014, 9, 287–314. [CrossRef] [PubMed] 5. Croce, C.M. Causes and consequences of microRNA dysregulation in cancer. Nat Rev Genet 2009, 10, 704–714. [CrossRef][PubMed] 6. Ghildiyal, M.; Zamore, P.D. Small silencing RNAs: An expanding universe. Nat. Rev. Genet. 2009, 10, 94–108. [CrossRef] 7. Bhattacharya, A.; Ziebarth, J.D.; Cui, Y. PolymiRTS Database 3.0: Linking polymorphisms in microRNAs and their target sites with human diseases and biological pathways. Nucleic Acids Res. 2014, 42, D86–D91. [CrossRef][PubMed] 8. Hiard, S.; Charlier, C.; Coppieters, W.; Georges, M.; Baurain, D. Patrocles: A database of polymorphic miRNA-mediated gene regulation in vertebrates. Nucleic Acids Res. 2010, 38, D640–D651. [CrossRef] 9. Agarwal, V.; Bell, G.W.; Nam, J.-W.; Bartel, D.P. Predicting effective microRNA target sites in mammalian mRNAs. eLife 2015, 4, e05005. [CrossRef] 10. Paraskevopoulou, M.D.; Georgakilas, G.; Kostoulas, N.; Vlachos, I.S.; Vergoulis, T.; Reczko, M.; Filippidis, C.; Dalamagas, T.; Hatzigeorgiou, A.G. DIANA-microT web server v5.0: Service integration into miRNA functional analysis workflows. Nucleic Acids Res. 2013, 41, W169–W173. [CrossRef] Non-coding RNA 2019, 5, 42 7 of 8

11. Marco, A. SeedVicious: Analysis of microRNA target and near-target sites. PLoS ONE 2018, 13, e0195532. [CrossRef][PubMed] 12. Alexiou, P.; Maragkakis, M.; Papadopoulos, G.L.; Reczko, M.; Hatzigeorgiou, A.G. Lost in translation: An assessment and perspective for computational microRNA target identification. Bioinformatics 2009, 25, 3049–3055. [CrossRef][PubMed] 13. Pinzón, N.; Li, B.; Martinez, L.; Sergeeva, A.; Presumey, J.; Apparailly, F.; Seitz, H. microRNA target prediction programs predict many false positives. Genome Res. 2017, 27, 234–245. [CrossRef][PubMed] 14. Kuhn, D.E.; Martin, M.M.; Feldman, D.S.; Terry, A.V.; Nuovo, G.J.; Elton, T.S. Experimental validation of miRNA targets. Methods 2008, 44, 47–54. [CrossRef][PubMed] 15. Barbujani, G.; Colonna, V. Human genome diversity: Frequently asked questions. Trends Genet. Tig. 2010, 26, 285–295. [CrossRef][PubMed] 16. 1000 Genomes Project Consortium; Abecasis, G.R.; Altshuler, D.; Auton, A.; Brooks, L.D.; Durbin, R.M.; Gibbs, R.A.; Hurles, M.E.; McVean,G.A. A map of human genome variation from population-scale sequencing. Nature 2010, 467, 1061–1073. [CrossRef][PubMed] 17. Garfield, D.; Haygood, R.; Nielsen, W.J.; Wray, G.A. Population genetics of cis-regulatory sequences that operate during embryonic development in the sea urchin Strongylocentrotus purpuratus. Evol. Dev. 2012, 14, 152–167. [CrossRef] 18. Kasowski, M.; Grubert, F.; Heffelfinger, C.; Hariharan, M.; Asabere, A.; Waszak, S.M.; Habegger, L.; Rozowsky, J.; Shi, M.; Urban, A.E.; et al. Variation in transcription factor binding among humans. Science 2010, 328, 232–235. [CrossRef] 19. Chen, K.; Rajewsky, N. Natural selection on human microRNA binding sites inferred from SNP data. Nat. Genet. 2006, 38, 1452–1456. [CrossRef] 20. Saunders, M.A.; Liang, H.; Li, W.-H. Human polymorphism at microRNAs and microRNA target sites. Proc. Natl. Acad. Sci. USA 2007, 104, 3300–3305. [CrossRef] 21. Marco, A. Selection Against Maternal microRNA Target Sites in Maternal Transcripts. G3 Genesgenomesgenetics 2015, 5, 2199–2207. [CrossRef][PubMed] 22. Hatlen, A.; Marco, A. Pervasive selection against microRNA target sites in human populations. bioRxiv 2018, 420646. 23. Kvaskoff, M.; Whiteman, D.C.; Zhao, Z.Z.; Montgomery, G.W.; Martin, N.G.; Hayward, N.K.; Duffy, D.L. Polymorphisms in naevus-associated genes MTAP, PLA2G6, and IRF4 and the risk of invasive cutaneous melanoma. Twin Res. Hum. Genet. 2011, 14, 422–432. [CrossRef][PubMed] 24. Bishop, D.T.; Demenais, F.; Iles, M.M.; Harland, M.; Taylor, J.C.; Corda, E.; Randerson-Moor, J.; Aitken, J.F.; Avril, M.-F.; Azizi, E.; et al. Genome-wide association study identifies three loci associated with melanoma risk. Nat. Genet. 2009, 41, 920–925. [CrossRef][PubMed] 25. Li, J.; Liu, Y.; Xin, X.; Kim, T.S.; Cabeza, E.A.; Ren, J.; Nielsen, R.; Wrana, J.L.; Zhang, Z. Evidence for Positive Selection on a Number of MicroRNA Regulatory Interactions during Recent Human Evolution. PLoS Genet. 2012, 8, e1002578. [CrossRef] 26. Chang, C.L.; Cai, J.J.; Huang, S.Y.; Cheng, P.J.; Chueh, H.Y.; Hsu, S.Y.T. Adaptive Human CDKAL1 Variants Underlie Hormonal Response Variations at the Enteroinsular Axis. PLoS ONE 2014, 9, e105410. [CrossRef] 27. Chang, C.L.; Cai, J.J.; Cheng, P.J.; Chueh, H.Y.; Hsu, S.Y.T. Identification of Metabolic Modifiers That Underlie Phenotypic Variations in Energy-Balance Regulation. Diabetes 2011, 60, 726–734. [CrossRef] 28. Duan, Y.; Hu, L.; Liu, B.; Yu, B.; Li, J.; Yan, M.; Yu, Y.; Li, C.; Su, L.; Zhu, Z.; et al. Tumor suppressor miR-24 restrains gastric cancer progression by downregulating RegIV. Mol. Cancer 2014, 13, 127. [CrossRef] 29. Ipe, J.; Collins, K.S.; Hao, Y.; Gao, H.; Bhatia, P.; Gaedigk, A.; Liu, Y.; Skaar, T.C. PASSPORT-seq: A Novel High-Throughput Bioassay to Functionally Test Polymorphisms in Micro-RNA Target Sites. Front. Genet. 2018, 9, 219. [CrossRef] 30. Pybus, M.; Dall’Olio, G.M.; Luisi, P.; Uzkudun, M.; Carreño-Torres, A.; Pavlidis, P.; Laayouni, H.; Bertranpetit, J.; Engelken, J. 1000 Genomes Selection Browser 1.0: A genome browser dedicated to signatures of natural selection in modern humans. Nucleic Acids Res. 2014, 42, D903–D909. [CrossRef] Non-coding RNA 2019, 5, 42 8 of 8

31. Karolchik, D.; Hinrichs, A.S.; Furey, T.S.; Roskin, K.M.; Sugnet, C.W.; Haussler, D.; Kent, W.J. The UCSC Table Browser data retrieval tool. Nucleic Acids Res. 2004, 32, D493–D496. [CrossRef][PubMed] 32. Chou, C.-H.; Shrestha, S.; Yang, C.-D.; Chang, N.-W.; Lin, Y.-L.; Liao, K.-W.; Huang, W.-C.; Sun, T.-H.; Tu, S.-J.; Lee, W.-H.; et al. miRTarBase update 2018: A resource for experimentally validated microRNA-target interactions. Nucleic Acids Res. 2018, 46, D296–D302. [CrossRef][PubMed]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).