Journal of Human Genetics (2014) 59, 110–115 & 2014 The Japan Society of Human Genetics All rights reserved 1434-5161/14 OPEN www.nature.com/jhg

CORRESPONDENCE

Confounding by linkage disequilibrium

Journal of Human Genetics (2014) 59, 110–115; doi:10.1038/jhg.2013.130; published online 19 December 2013

Linkage disequilibrium (LD) and confound- association between height and uterine fitted model for the test of association may ing are two widely discussed concepts in fibroids) have been reported; these associa- become overspecified, more precisely over- Genetics and in Epidemiology, yet their tions could be explained by mere LD, more adjusted because of adjustment for a covari- relationship has received only intuitive or precisely they might result from confounding ate (BMI) influenced by a genetic no considerations. Taken in the narrow sense, by LD or GPD between loci influencing (locus 1) in LD with the causal locus LD refers to the nonrandom association of height and uterine fibroids. (locus 2). Most importantly, the covariate alleles at two or more linked genetic loci. The needs not to be on the causal pathway, bias degree to which these alleles are associated is toward the null can still be observed if CONFOUNDING BY LD the basis behind genetic association studies locus 1 contributes significantly to the var- To illustrate the notion of confounding by whereby the genetic variation across target iance of the obesity trait in the study sample. LD, let us consider the simple scenario of a chromosomal intervals or over the entire Furthermore, because of the population-spe- Mendelian disease with complete or quasi- genome is captured by a set of representative cific pattern of LD, confounding by LD is complete penetrance and a typed marker at genetic markers (tagging markers). The expected to vary across populations and thus gene locus 1 in LD with an untyped causal denomination of ‘linkage’ is somehow mis- could explain failures to replicate genetic gene variant at gene locus 2. Let us further leading because LD is a special case of the association findings in comparable studies assume locus 1 affects a rare Mendelian form more general gametic phase disequilibrium (those that controlled for the same confoun- of obesity and locus 2 affects cancer but none (GPD),1 which also occurs between variants ders) in different populations. of these gene functions is known. An epide- at the mitochondrial and nuclear loci in the miological study that has collected measure- form of cytonuclear disequilibrium.2 ments of potential confounders such as body UTERINE FIBROIDS AND OBESITY Confounding occurs when an extraneous mass index (BMI) may find this trait as an One of the illustrations of this phenomenon factor, or a set of factors, can at least partially important predictor of cancer in the studied is the complex association between obesity explain an apparent association or a lack of population, whereas the positive association and uterine leiomyomas (UL, also referred to an apparent association between a risk factor may actually be explained by LD between the as fibroids). Uterine fibroids are the most and the outcome. In the former case, the typed marker and the cancer-causing allele. common tumors in women of reproductive confounding variable, or confounder, causes Thus, LD can be a source of confounding as age (about 75% of women will develop this the association to appear, whereas in the BMI predicts the cancer outcome in the condition by the time they reach menopause second case the confounder masks a real absence of an apparent causal relationship. and 25% will have symptomatic fibroids). association.3 The classical example is the Fortunately, most of the genetic determinants Cumulative exposure to estrogen is believed apparent increase of risk for heart disease of the common form of obesity identified to be a major etiological factor5 and factors with alcohol consumption, an association and replicated so far explain only small that may influence the hormonal milieu, that might in fact be due to cigarette fractions of the risk and therefore the impact such as obesity, are believed to be smoking, a covariate highly correlated with of confounding by LD on association studies associated with risk.6 However, the clearly alcohol consumption.4 of common conditions may be limited but established risk factors are age (increasing With the flood of genetic association data not negligible. By contrast, if the linked risk with increasing premenopausal age), published over the last few years, and the marker tags a gene with a sizeable effect, menopause (risk decreases with menopause) many others to come, it is the expectation for example, a major gene, then confounding and African-American ethnicity (higher risk that an increased number of the reported by LD can significantly affect the results of compared with that of non-Hispanic associations will be false findings or association studies, be they genetic or not. In Whites). Epidemiological studies have considered as such. As false finding is the our example above, if BMI is controlled for shown mixed results with respect to the common explanation for unexpected (irrele- (that is, adjusted, stratified or restricted), the association between BMI and UL. Some vant) but statistically significant associations, genetic effect of locus 2 on the cancer out- studies reported positive or no associations, in most cases these findings are not followed- come cannot be accurately estimated and the whereas other more adequately designed or up. Nevertheless, associations appealing to observed association would typically be larger studies reported an inverse J-shaped common sense (association between marital biased toward the null depending on the association,7,8 with the association reaching a status and cutaneous lymphoma) or as variance explained by locus 1 and the degree peak in the overweight category. The intriguing as they may appear (validated of LD between the two loci. In this case, the independent replication of this atypical Correspondence 111 association of BMI with UL in two cohort most significant ones peaking in the RGS7- associated with a given trait and SNPs studies of UL7,8 ruled out possible detection FH genomic interval in European Americans associated with another trait is different bias. In the absence of dose effect, plausible and in RGS7 and PLD5 ( among cases and controls. In our example, detection bias and homogeneity of the member 5) in African Americans vanishing associations with UL at given ‘UL’ association across populations, a possible (unpublished observation). However, it is SNPs after adjustment for BMI would be an explanation for this atypical association is still not clear which of FH, RGS7, PLD5 or indication for confounding due to LD with confounding by LD, that is, the presence of a another linked gene is the actual obesity gene ‘obesity’ SNPs. A direct demonstration of this major gene for obesity in the vicinity of a UL because the association pattern across the phenomenon is suggested by the data in locus. FH-linked region varied by the UL affection Figures 2a and b, which show a differential In a recent study, we have evaluated the status and race stratum. The emerging model LD pattern at the tested ‘obesity’ and ‘UL’ role of 1q43 in the develop- is not in disagreement with our postulate SNPs in UL cases and controls, respectively. ment of UL and we have identified two of linked obesity and fibroid . As can be seen, SNP375 (rs4391653) in RGS7 closely linked loci influencing differentially Interestingly, the variance explained by the and SNP1320 (rs7531009) in PLD5, the two the risk of UL in European and African- candidate 1q43 obesity gene is 20–30 times SNPs that remained significantly associated American women.9 The peak of association larger (r2 ¼ 2–3%) than those reported for with BMI after controlling for the UL affec- with UL overlapped a 100 kb-long genomic typical candidate obesity genes identified in tion status, show different levels of LD with region separating two head-to-tail genes meta-analyses of genome-wide association the ‘UL’ SNPs in cases and controls. Specifi- encoding RGS7 (regulator of G- study.14 Nevertheless, the role of FH or the cally, the SNPs at which the association was signaling 7) and FH (tricarboxylic acid cycle linked gene in non-syndromic UL is yet to be lost after adjustment for BMI (RGS7 SNPs fumarate hydratase enzyme; Figure 1). Muta- proven. As a proof-of-concept study for the 416–418 and RGS-FH intergenic SNPs 644 tions in FH have previously been associated phenomenon of confounding by LD, we and 651) exhibit opposite levels of LD with with two rare syndromic forms of UL, analyzed the LD pattern among Chr.1q43 the two ‘obesity’ SNPs 375 and 1320 in the hereditary leiomyomatosis and renal cell single nucleotide polymorphisms (SNPs) at UL cases compared with the controls. cancer and multiple cutaneous and uterine risk for UL and/or obesity in the European Although the current data support the leiomyomatosis.10,11 The current paradigm group of the National Institute of presence of colocalized susceptibility loci for associates FH to a tumor suppressor gene10– Environmental Health Sciences Uterine obesity13,15,16 and reproduction-related 12 but several observations argue against it Fibroid Study. Table 1 shows the list of traits,9–11,17,18 it remains to be seen whether and models for an alternative 1q43 fibroid 1q43 SNPs of interest, their relative position the present proof-of-concept study will be gene linked to FH and/or a pleiotropic FH and the significance of their association with strengthen or weakened following gene have been postulated.9 the UL and the BMI outcomes. To assess the identification and mutational analyses. In In an early study in the Quebec Family independent effect of SNPs on each of the UL our example, the use of tag SNPs as Study, we have mapped a quantitative trait and BMI outcomes, we evaluated the asso- proxies for the actual causal variants and locus for body fat to a chromosomal interval ciation with UL in models with and without BMI as proxy for body fat to assess genetic overlapping the FH locus.13 In the National adjustment for BMI, and the association with confounding by LD may mask the true Institute of Environmental Health Sciences BMI in models adjusted or not for the UL impact of this phenomenon. Consequently, Uterine Fibroid Study, we observed several affection status. Confounding by LD is sus- confirmation and validation of this signals for the association with BMI, with the pected if the LD pattern between SNPs hypothesis in a genomic context with known and validated genotype–phenotype relationships should be an important undertaking. In a meta-analysis study of age at natural menopause in populations of European des- cent, EXO1, a DNA repair gene that maps close to FH was highlighted as one of the 13 loci influencing this outcome.18 Mapping of age at natural menopause, one of the established risk factors for UL, to a region implicated in UL and obesity makes the potential of confounding by LD more plausible. Thus, correlated traits may provide clues to the location of disease Figure 1 Colocalization of human traits of potential relevance to uterine fibroids on chromosome genes but can also confound the 1q43. Genomic map of a 2 Mb-long interval spanning the fumarate hydratase (FH) gene locus and association. Actually, the UL example is not showing the position of specific microsatellite markers (vertical bars) and gene loci linked or a simple illustration of genetic confounding associated with diseases or traits of relevance to hormone-dependent tumors such as uterine by LD because UL is a hormonally dependent leiomyomas. Quebec Family Study (QFS); sex hormone-binding globulin (SHBG); HERITAGE cancer believed to be influenced by steroid (HERITAGE Family Study). The placement of the genes in the gene map shown above the plot and the bioavailability, a level of which is in turn coordinates show below were derived the assembly 19. Arrows indicate the orientation of the genes and are drawn proportionally to the size of the genes. RGS7 (regulator of G-protein 7); FH influenced by the body fat through (fumarate hydratase); KMO (kynurenine 3-monooxygenase); OPN3 (opsin 3); WDR64 (WD repeat downregulation of circulating sex hormone- 19 domain 64); EXO1 (exonuclease 1); MAP1LC3C (microtubule-associated protein 1 light chain 3 binding globulin. Coincidentally, linkage gamma); PLD5 (phospholipase D family, member 5). for the circulating level of sex hormone-

Journal of Human Genetics Correspondence 112

Table 1 Association of 1q43 SNPs with uterine fibroids and BMI in European Americans enrolled in the NIEHS study

Uterine fibroids (UL), Pa BMI, Pb

Obs SNP Inter-SNP distance, Gene Function BMI adjusted BMI unadjusted UL adjusted UL unadjusted

375 rs4391653 RGS7 Intronic 0.0063 0.0272 0.0245 416 rs4290050 57 222 RGS7 Intronic 0.0078 417 rs4378194 19 RGS7 Intronic 0.0153 418 rs2815842 229 RGS7 Intronic 0.0143 644 rs2994976 225 065 RGS7-FH Intergenic 0.0482 651 rs3014580 4859 RGS7-FH Intergenic 0.0457 656 rs2994979 4046 Downstream FH Intronic 0.0050 0.0800 658 rs11586971 1123 Downstream FH Intronic 0.0050 659 rs12754512 51 Downstream FH Intronic 0.0050 661 rs2994981 1967 downstream FH Intronic 0.0045 662 rs12566016 460 downstream FH Intronic 0.0045 663 rs10926494 204 downstream FH Intronic 0.0017 673 rs2144 7176 Downstream FH Intronic 0.0014 943 rs10802996 365 915 EXO1 50-UTR 0.0075 948 rs4149864 58 417 EXO1-MAP1LC3C Intergenic 0.0724 0.0274 987 rs4150028 29 371 EXO1-MAP1LC3C Intergenic 0.0193 1037 rs3845563 545 213 PLD5 Intronic 0.0016 1292 rs10926736 10 854 PLD5 Intronic 0.0019 1320 rs7531009 7558 PLD5 Intronic 0.0064 0.0098 0.0099 1335 rs10926756 208 978 Far upstream PLD5 Intergenic 0.0015 1340 rs1039534 1531 Far upstream PLD5 Intergenic 0.0047 1587 rs10754775 160 Far upstream PLD5 Intergenic 0.0047 1600 rs6673294 1757 Far upstream PLD5 Intergenic 0.0043 1602 rs10803075 493 Far upstream PLD5 Intergenic 0.0053 1603 rs6679445 494 Far upstream PLD5 Intergenic 0.0014

Abbreviations: BMI, body mass index; FH, fumarate hydratase; PDL5, phospholipase D member 5; SNP, single nucleotide polymorphism; UL, uterine leiomyoma. aLogistic regression models adjusted for covariates (age, age at menarche, parity and physical activity). bGeneral linear models adjusted for the effects of age and physical activity. Note that only SNPs that showed significant levels of association with either the uterine fibroid or the BMI outcome or with both outcomes are shown.

binding globulin in the HERITAGE have amply been discussed in the past;22 whether the atypical association of BMI and study peaked at D1S321, a marker within here I propose a thrifty phenotype model UL is real or artifactual, and whether it is 150–200 kb distance from FH.20 Thus, for the high incidence of UL in women. related to the candidate 1q43 region or not, further complications arise as the linked Reproductive suppression as an adaptive confounding by LD should be real and a candidate obesity gene may affect the response to low-energy availability23 would valid notion potentially explaining the association with fibroids in several ways: be an attractive evolutionary model for UL. growing reports on shared genetic through the causal pathway by decreasing With reproduction and metabolism (fat polymorphisms between obesity- and the serum level of sex hormone-binding storage) representing the most selectively reproduction-related traits28–33 including globulin and consequent increase in constrained traits and functions, occurrence those implicating chromosome 1q43.9,18 bioavailability of free steroids, confounding and conservation of LD between their genetic Similarly, follow-up of genome-wide linkage by LD or through combination of both. determinants in populations are alike. In this and admixture signals for UL in the NIEHS line of thought, a large-scale study (CARe— cohort pointed to several known obesity- LD AND SELECTION the National Heart, Lung, and Blood related genes (Aissani et al.,unpublished On the other hand, confounding by LD can Institute Candidate Gene Association observation). be useful for association studies (gain of Resource) of ultraconserved polymorphisms statistical power in bivariate models) because in the human genome, which are believed CONFOUNDING BY GPD genes that have similar and/or coordinated to affect reproductive (age at natural The occurrence of GPD may reflect demo- functions tend to be clustered in the genome menopause, number of children, age at first graphic patterns or can be the result of of eukaryotes.21 In the obesity and UL child and age at last child) and overall evolutionary processes that favored inter- example, it is tempting to think that the (longevity, BMI and height) fitness,24,25 has actions between genes (epistasis) contribut- atypical association between obesity and shown an excess of associations with BMI.26 ing to the expression of specific traits in fibroids may actually reflect evolutionary Interestingly, the most strongly associated populations under specific environments. selective constraints on ‘thrifty’ obesity and SNP, rs10818872, occurred in DENND1A,a Similarly, GPD can confound association reproduction-associated genes that have locus previously shown to be associated with tests and may account for a number of the evolved different mutation patterns in the polycystic ovary syndrome and detected claimed associations with common diseases history of human populations. The selective by a SNP in high LD (r2 ¼ 0.826) with (for example, 349 hits with genome-wide advantages of the overweight-to-obese traits rs10818872.27 Independently, however, distribution were found in a query of the

Journal of Human Genetics 22ad13 sls fe otoln o oyms ne BI.Afl oo eso fti gr saalbea the at available is figure this of version color full A (BMI). index mass body for controlling after lost is downstream 1335 region and the 1292 in SNPs (UL)’ leiomyoma oto (panel control iue2 Figure atr flnaedsqiiru L)btensnl uloieplmrhss(Ns trs o trn bod n/roeiyi ae(panel case a in obesity and/or fibroids uterine for risk at (SNPs) polymorphisms nucleotide single between (LD) disequilibrium linkage of Pattern 7 1 1 1 4 5 5 5 5 6 6 6 7 4 0719 3013 5710 621603 1602 1600 1587 1335 1320 1292 1037 943 673 663 662 661 659 658 656 651 644 418 417 416 375 1603 1602 1600 1587 1335 1320 1292 1037 943 673 663 662 661 659 658 656 651 644 418 417 416 375 rs4391653 rs4391653 14 7 b td fueiefirisi h ainlIsiueo niomna elhSine trn iri td.Nt hti otatt h ‘uterine the to contrast in that Note Study. Fibroid Uterine Sciences Health Environmental of Institute National the in fibroids uterine of study ) 89 71 rs4290050 rs4290050 Block 1(0kb) Block 1(0kb) 71 89 44 63 rs4378194 rs4378194 44 39 63 9 39 49 12 7 rs2815842 6 4 rs2815842 49 49 17 12 6 5 4 5 Block 2(4kb) 15 49 10 17 5 2 rs2994976 6 5 rs2994976 16 10 10 5 2 5 5 9 10 10 11 9 rs3014580 6 7 8 9 rs3014580 12 10 11 4 4 5 2 8 FH 11 81 79 11 6 7 4 rs2994979 9 7 9 rs2994979 SP6663,wihso o eeso rn Dwt h w oeiy Ns h soito ihSP1037, SNP with association the SNPs, ‘obesity’ two the with LD no or of levels low show which 656–673), (SNP 97 11 64 12 7 4 5 9 8 9 97 15 11 97 4 1 9 1 6 1 rs11586971 rs11586971 98 15 95 6 4 9 2 3 3 1 96 69 11 2 4 1 9 9 1

rs12754512 Block 3(10kb) rs12754512 15 98 96 10 36 37 95 98 11 0 4 8 3 Block 2(10kb) 15 98 96 11 58 25 28 95 98 25 15 53 5 0 6 8 rs2994981 rs2994981 14 58 16 15 16 15 38 53 11 96 96 5 4 7 2 9 25 13 14 18 16 21 23 38 19 96 5 1 4 5 0 rs12566016 3 rs12566016 11 13 98 26 19 24 23 98 13 4 5 9 1 4 0 5 4 2 3 15 11 45 51 21 27 23 50 7 6 3 3 1 0 1 4 3 4 rs10926494 rs10926494 Correspondence 15 45 17 10 98 24 27 49 12 10 10 98 11 1 4 9 2 1 0 0 11 24 49 3 6 0 9 8 1 3 3 8 1 1 0 6 1 3 rs2144 rs2144 48 12 35 10 38 11 6 1 9 3 9 2 8 6 4 1 4 1 52 10 17 10 14 3 3 1 7 9 1 5 2 3 4 5 rs10802996 rs10802996 35 16 38 18 0 1 7 1 5 3 6 4 5 1 3 3 28 21 17 40 1 0 9 2 7 6 6 5 1 9 rs3845563 rs3845563 41 17 17 17 20 0 2 1 1 4 8 7 0 2 10 11 22 19 1 1 9 3 6 0 1 rs10926736 rs10926736 26 15 19 19 29 14 2 2 1 4 12 19 10 1 9 0 7 9 3 8 rs7531009 rs7531009 13 32 12 16 20 57 4 1 1 4 ora fHmnGenetics Human of Journal 39 25 10 32 21 rs10926756 8 5 3 rs10926756 15 33 12 10 14 3 9 6 16 34 39 10 1 2 rs10754775 rs10754775 32 27 61 20 Block 4(4kb) Block 3(4kb) 12 15 rs6673294 rs6673294 97 93 97

rs10803075 rs10803075 ora fHmnGenetics Human of Journal 98

ora online. journal rs6679445 rs6679445 a )and 113 Correspondence 114

Map viewer database—NCBI—with the key- variation on the combined outcomes studies, I extended the reasoning to convey word ‘obesity’). Studies exploring the genetic (bivariate or multivariate analyses) new thoughts on the possible existence of basis of correlated traits and clinical pheno- theoretically should increase the power to thrifty phenotypes to account for the high types in relationship to coordinated epistatic capture the underlying pleiotropic loci. incidence of uterine fibroids and to some gene expression are still in their infancy and extent of reproductive cancer in general in new integrated approaches to tackle this CONFOUNDING BY CYTONUCLEAR LD human populations. Confounding by LD or complexity are needed. The existence of Confounding by yet another type of LD, GPD is a novel concept that may in part trans-regulation of expression phenotypes at cytonuclear LD (for LD between DNA var- explain failures to replicate findings in the genome level34 further suggests that GPD iants present in organelles and in the genetic associations that incorporate covari- may underlie the correlation among human nucleus), is more challenging and remains ates differentially correlated in different traits through coordinated gene expressions largely unexplored. A priori,anypositive populations. Finally, while perplexing but In this context, the emerging field of associations with nuclear variants may actu- legitimate interrogations have been raised in phenomics and its combination with ally be proxies for true causal mitochondrial this paper on the meaning of genome– genome-wide association study in the variants and vice versa.Confoundingby phenome associations in the context of so-called phenome-wide association studies cytonuclear LD can be more subtle, given dynamic genomes, coordinated gene expres- is an important advance in the design of that as many as 966 or more nuclear-encoded sions and correlated traits and diseases, a the next generation of genetic association factors are involved in the maintenance and framework of thought for genetically and studies of common diseases.35 Furthermore, function of the mitochondrion. Thus, mito- evolutionary determined ‘thrifty correlated the development of resources such as chondrial diseases can have diverse modes of phenotypes’ is suggested to account for the Population Architecture using Genomics inheritance (maternal, Mendelian and a com- high incidence of uterine fibroids in the and Epidemiology—National Human Gen- bination of the two) with variable expression contemporary obesogenic environment. ome Research Institute to characterize well- of their phenotypes owing to the stochastic replicated genome-wide association study and quantitative nature of inheritance of CONFLICT OF INTEREST variants in relationship to many traits and heteroplasmic mutations.41 As a result, The author declares that the research was disease phenotypes36 will provide current knowledge and methodological conducted in the absence of any commercial opportunities to test new hypotheses on the approaches to evaluate the joint effects of or financial relationships that could be con- genetic basis of correlated traits and co-inherited mitochondrial and nuclear strued as a potential conflict of interest. comorbidities. variants on human diseases are limited. For diseases such as cancer, infection and ACKNOWLEDGEMENTS CONFOUNDING BY PLEIOTROPY immune inflammatory diseases, the I am indebted to Dr Donna Baird for sharing The other source of genetic confounding is interaction effects of specific nuclear and the data and the biospecimens collected for the obviously pleiotropy (one gene or one muta- apoptosis-related mitochondrial variants are NIEHS Uterine Fibroid Study and would like to tion affecting more than one trait or pheno- likely to be important determinants of thank the participants to this study. This study type), a concept known for exactly a disease risk or progression. One possible was funded by NIH Grant R01-HD064398 to BA century37 but the interference of this way to control for the effects of from the National Institute of Child and Health phenomenon in genetic association studies mitochondrial variants in genome-wide Development (NIH-NICHD) and National has not been fully investigated. Empirical association study is to type diagnostic Institute for Environmental and Health Science (NIH-NIEHS). data essentially from model organisms mitochondrial polymorphisms to allow for indicated that pleiotropy is a common the mitochondrial haplogroup membership Brahim Aissani phenomenon.38 At least in yeast, pleiotropy to be inferred and used as a covariate.42 is believed to be mostly of type 2, that is, a The interplay between the nuclear and the Department of Epidemiology, School of single molecular function resulting in mitochondrial genomes is likely to be central Public Health, University of Alabama at multiple effects (as opposed to type 1 in the regulation of energy homeostasis and Birmingham, Birmingham, AL, USA pleiotropy, which is one gene, multiple perturbations in the coordinated expression E-mail: [email protected] functions).39 As outlined in the UL of the underlying nuclear and mitochondrial example, it is not yet clear whether the UL genes may lead to poor cellular performances gene on 1q43 is pleiotropic (an obesity gene and diseases. Conceivably, recent population affecting UL through change in hormonal admixtures may increase the risk for certain 1 Wang, X., Elston, R. C. & Zhu, X. The meaning of interaction. Hum. Hered. 70, 269–277 (2010). milieu), a hypothesis supported by the conditions that arise from selectively con- 2 Asmussen, M. A., Arnold, J. & Avise, J. C. Definition linkage of the FH region to the level of strained human traits such as obesity, repro- and properties of disequilibrium statistics for associa- circulating sex hormone-binding globulin in duction and resistance to infection, hence tions between nuclear and cytoplasmic genotypes. Genetics 115, 755–768 (1987). the HERITAGE study, or a UL gene reaching genetic fitness in the parental popu- 3 Woodward, M. Epidemiology: Study Design and Data confounded by a linked obesity gene. In lations of recently admixed populations. Analysis, 2nd ed. 2nd edn (Chapmanand Hall/CRC, either scenario, FH remains a strong Boca Raton, FL, USA, 2004). 4 Dales, L. G. & Ury, H. K. An improper use of statistical candidate pleiotropic gene. Indeed, a model CONCLUSION significance testing in studying covariables. Int. J. has been proposed whereby a single The correlations among traits and conse- Epidemiol. 7, 373–375 (1978). 5 Andersen, J. Growth factors and cytokines in uterine inactivating mutation can affect distinct quently among clinical phenotypes undoubt- leiomyomas. Semin. Reprod. Endocrinol. 14, 269–282 9 functions encoded by FH, with cytosolic edly have clouded the interpretation of many (1996). FH possibly acting in DNA repair activity40 genetic association data. Beyond the initial 6 Schwartz, S. M., Marshall, L. M. & Baird, D. D. Epidemiologic contributions to understanding the etiol- and mitochondrial FH in metabolism. quest to highlight aspects of LD that may ogy of uterine leiomyomata. Environ. Health Perspect. Models that analyze the effects of genetic affect the interpretation of association 108(Suppl 5), 821–827 (2000).

Journal of Human Genetics Correspondence 115

7 Baird, D. D., Dunson, D. B., Hill, M. C., Cousins, D. & repair and immune pathways. Nat. Genet. 44, 32 den Hoed, M., Ekelund, U., Brage, S., Grontved, A., Schectman, J. M. Association of physical activity with 260–268 (2012). Zhao, J. H., Sharp, S. J. et al. Genetic susceptibility to development of uterine leiomyoma. Am. J. Epidemiol. 19 Dorgan, J. F., Reichman, M. E., Judd, J. T., Brown, C., obesity and related traits in childhood and adolescence: 165, 157–163 (2007). Longcope, C., Schatzkin, A. et al. Relation of energy, influence of loci identified by genome-wide association 8 Wise, L. A., Palmer, J. R., Spiegelman, D., Harlow, B. L., fat, and fiber intakes to plasma concentrations of studies. Diabetes 59, 2980–2988 (2010). Stewart, E. A., Adams-Campbell, L. L. et al. Influence of estrogens and androgens in premenopausal women. 33 Zhao, J., Bradfield, J. P., Li, M., Wang, K., Zhang, H., body size and body fat distribution on risk of uterine Am. J. Clin. Nutr. 64, 25–31 (1996). Kim, C. E. et al. The role of obesity-associated loci leiomyomata in U.S. black women. Epidemiology 16, 20 Ukkola, O., Rankinen, T., Gagnon, J., Leon, A. S., identified in genome-wide association studies in the 346–354 (2005). Skinner, J. S., Wilmore, J. H. et al. Agenome-wide determination of pediatric BMI. Obesity 17, 9 Aissani, B., Wiener, H. & Zhang, K. Multiple hits for the linkage scan for steroids and SHBG levels in black and 2254–2257 (2009). association of uterine fibroids on human chromosome white families: the HERITAGE Family Study. J. Clin. 34 Goring, H. H., Curran, J. E., Johnson, M. P., Dyer, T. D., 1q43. PLoS One 8, e58399 (2013). Endocrinol. Metab. 87, 3708–3720 (2002). Charlesworth, J., Cole, S. A. et al. Discovery of expres- 10 Alam, N. A., Bevan, S., Churchman, M., Barclay, E., 21 Hurst, L. D., Pal, C. & Lercher, M. J. The evolutionary sion QTLs using large-scale transcriptional profiling in Barker, K., Jaeger, E. E. et al. Localization of a gene dynamics of eukaryotic gene order. Nat. Rev. Genet. 5, human lymphocytes. Nat. Genet. 39, 1208–1216 (MCUL1) for multiple cutaneous leiomyomata and 299–310 (2004). (2007). uterine fibroids to chromosome 1q42.3-q43. Am. J. 22 Neel, J. V. Diabetes mellitus: a ‘thrifty’ genotype 35 Denny, J. C., Ritchie, M. D., Basford, M. A., Pulley, J. Hum. Genet. 68, 1264–1269 (2001). rendered detrimental by ‘progress’? Am. J. Hum. M., Bastarache, L., Brown-Gentry, K. et al. PheWAS: 11 Tomlinson, I. P., Alam, N. A., Rowan, A. J., Barclay, E., Genet. 14, 353–362 (1962). demonstrating the feasibility of a phenome-wide scan Jaeger, E. E., Kelsell, D. et al. Germline mutations in 23 Jasienska, G. Energy metabolism and the evolution of to discover gene-disease associations. Bioinformatics FH predispose to dominantly inherited uterine fibroids, reproductive suppression in the human female. Acta 26, 1205–1210 (2010). skin leiomyomata and papillary renal cell cancer. Nat. Biotheor. 51, 1–18 (2003). 36 Matise, T. C., Ambite, J. L., Buyske, S., Carlson, C. S., Genet. (2002)30, 406–410. 24 Derti, A., Roth, F. P., Church, G. M. & Wu, C. T. Cole, S. A., Crawford, D. C. et al. The Next PAGE in 12 Lehtonen, R., Kiuru, M., Vanharanta, S., Sjo¨berg, J., Mammalian ultraconserved elements are strongly understanding complex traits: design for the analysis of Aaltonen, L. M., Aittoma¨ki, K. et al. Biallelic inactiva- depleted among segmental duplications and copy Population Architecture Using Genetics and Epidemiol- tion of fumarate hydratase (FH) occurs in nonsyndro- number variants. Nat. Genet. 38, 1216–1220 (2006). ogy (PAGE) Study. Am. J. Epidemiol. 174, 849–859 mic uterine leiomyomas but is rare in other tumors. 25 Ahituv, N., Zhu, Y., Visel, A., Holt, A., Afzal, V., (2011). Am. J. Pathol. 164, 17–22 (2004). Pennacchio, L. A. et al. Deletion of ultraconserved 37 Stearns, F. W. One hundred years of pleiotropy: a 13 Aissani, B., Perusse, L., Lapointe, G., Chagnon, Y. C., elements yields viable mice. PLoS Biol. 5, e234 retrospective. Genetics 186, 767–773 (2010). Bouchard, L., Walts, B. et al. A quantitative trait locus (2007). 38 Wagner, G. P. & Zhang, J. The pleiotropic structure of for body fat on chromosome 1q43 in French 26 Chiang, C. W., Liu, C. T., Lettre, G., Lange, L. A., the genotype-phenotype map: the evolvability of com- Canadians: linkage and association studies. Obesity Jorgensen, N. W., Keating, B. J. et al. Ultraconserved plex organisms. Nat. Rev. Genet. 12, 204–213 14, 1605–1615 (2006). elements in the human genome: association and (2011). 14 Kilpelainen, T. O., Zillikens, M. C., Stancakova, A., transmission analyses of highly constrained single- 39 He, X. & Zhang, J. Toward a molecular understanding of Finucane, F. M., Ried, J. S., Langenberg, C. et al. nucleotide polymorphisms. Genetics 192, 253–266 pleiotropy. Genetics 173, 1885–1891 (2006). Genetic variation near IRS1 associates with reduced (2012). 40Yogev,O.,Yogev,O.,Singer,E.,Shaulian,E.,Golberg,M., adiposity and an impaired metabolic profile. 27 Chen, Z. J., Zhao, H., He, L., Shi, Y., Qin, Y., Shi, Y. Fox, T. D. et al. Fumarase: a mitochondrial metabolic Nat. Genet. 43, 753–760 (2011). et al. Genome-wide association study identifies sus- enzyme and a cytosolic/nuclear component of the 15 Bosse, Y., Chagnon, Y. C., Despres, J. P., Rice, T., ceptibility loci for polycystic ovary syndrome on chro- DNA damage response. PLoS Biol. 8, e1000328 Rao, D. C., Bouchard, C. et al. Genome-wide linkage mosome 2p16.3, 2p21 and 9q33.3. Nat. Genet. 43, (2010). scan reveals multiple susceptibility loci influencing 55–59 (2011). 41 Wallace, D. C. Mitochondrial diseases in man and lipid and lipoprotein levels in the Quebec Family Study. 28 Elks, C. E., Perry, J. R., Sulem, P., Chasman, D. I., mouse. Science 283, 1482–1488 (1999). J. Lipid Res. 45, 419–426 (2004). Franceschini, N., He, C. et al. Thirty new loci for age at 42 Johnson, D. C., Shrestha, S., Wiener, H. W., 16 Dong, C., Beecham, A., Slifer, S., Wang, L., menarche identified by a meta-analysis of genome-wide Makowsky, R., Kurundkar, A., Wilson, C. M. et al. McClendon, M. S., Blanton, S. H. et al. Genome-wide association studies. Nat. Genet. 42, 1077–1085 Mitochondrial DNA diversity in the African American linkage and peak-wide association study of obesity- (2010). population. Mitochondrial DNA 24, 613–615 (2013). related quantitative traits in Caribbean Hispanics. 29 Fernandez-Rhodes, L., Demerath, E. W., Cousminer, D. L., Hum. Genet. 129, 209–219 (2011). Tao, R., Dreyfus, J. G., Esko, T. et al. Association of 17 Gross, K. L., Panhuysen, C. I., Kleinman, M. S., Gold- adiposity genetic variants with menarche timing in This work is licensed hammer, H., Jones, E. S., Nassery, N. et al. Involvement 92,105 women of European descent. Am.J.Epidemiol. under a Creative Commons of fumarate hydratase in nonsyndromic uterine leiomyo- 178, 451–460 (2013). mas: genetic linkage analysis and FISH studies. Genes 30 Ahmed, M. L., Ong, K. K. & Dunger, D. B. Childhood Attribution-NonCommercial-NoDerivs 3.0 Cancer 41, 183–190 (2004). obesity and the timing of puberty. Trends Endocrinol. Unported License. To view a copy of 18 Stolk, L., Perry, J. R., Chasman, D. I., He, C., Mangino, Metab. 20, 237–242 (2009). M., Sulem, P. et al. Meta-analyses identify 13 loci 31 Frayling, T. M. & Ong, K. Piecing together the FTO this license, visit http://creativecommons. associated with age at menopause and highlight DNA jigsaw. Genome Biol. 12, 104 (2011). org/licenses/by-nc-nd/3.0/

Journal of Human Genetics