330

1. To whom all correspondence should be addressed: Carl Johnson, Department of Biological Sciences, Box 1634-B,Vanderbilt University, Nashville, TN 37235; e-mail: carl.h.johnson@vanderbilt.edu.

JOURNAL OF BIOLOGICAL RHYTHMS, Vol. 23 No. 4, August 2008 330-340DOI: 10.1177/0748730408320284© 2008 Sage Publications

Genetic Differences in Human Circadian ClockGenes among Worldwide Populations

Christopher M. Ciarleglio,* Kelli K. Ryckman,† Stein V. Servick,‡ Akiko Hida,‡ Sam Robbins,§ NancyWells,§ Jennifer Hicks,§ Sydney A. Larson,‡ Joshua P. Wiedermann,‡ Krista Carver,§ Nalo Hamilton,§

Kenneth K. Kidd,çç Judith R. Kidd,çç Jeffrey R. Smith,† Jonathan Friedlaender,¶ Douglas G.McMahon,‡ Scott M. Williams,† Marshall L. Summar,# and Carl Hirschie Johnson‡,1

*Department of Biological Sciences, Neuroscience Graduate Program, Vanderbilt University, Nashville, TN,USA, †Department of Medicine and Center for Human Genetics Research, Vanderbilt University Medical

Center, Nashville, TN, USA, ‡Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA,§Vanderbilt University School of Nursing, Nashville, TN, USA, ççDepartment of Genetics, Yale UniversitySchool of Medicine, New Haven, CT, USA, ¶Department of Biological Anthropology (emeritus), Temple

University, Philadelphia, PA, USA, #Department of Pediatrics and Genetics, Vanderbilt University MedicalCenter, Nashville, TN, USA

Abstract The daily biological clock regulates the timing of sleep and physiolog-ical processes that are of fundamental importance to human health, perfor-mance, and well-being. Environmental parameters of relevance to biologicalclocks include (1) daily fluctuations in light intensity and temperature, and (2)seasonal changes in photoperiod (day length) and temperature; these parame-ters vary dramatically as a function of latitude and locale. In wide-rangingspecies other than humans, natural selection has genetically optimized adap-tiveness along latitudinal clines. Is there evidence for selection of clock genealleles along latitudinal/photoperiod clines in humans? A number of polymor-phisms in the human clock genes Per2, Per3, Clock, and AANAT have beenreported as alleles that could be subject to selection. In addition, this investiga-tion discovered several novel polymorphisms in the human Arntl and Arntl2genes that may have functional impact upon the expression of these clock tran-scriptional factors. The frequency distribution of these clock gene polymor-phisms is reported for diverse populations of African Americans, EuropeanAmericans, Ghanaians, Han Chinese, and Papua New Guineans (including 5subpopulations within Papua New Guinea). There are significant differences inthe frequency distribution of clock gene alleles among these populations.Population genetic analyses indicate that these differences are likely to arisefrom genetic drift rather than from natural selection.

Key words circadian, biological clock, sleep, FASPS, depression, Arntl, Bmal, Mop, Per,Clock, AANAT

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

Ciarleglio et al. / HUMAN CLOCK GENE POLYMORPHISMS IN DIFFERENT POPULATIONS 331

Eukaryotic and prokaryotic organisms manifestdaily (circadian) rhythms controlled by endogenousbiochemical oscillators (Dunlap et al., 2004). In mam-mals, this biological clock regulates the timing ofsleep and physiological processes—including feed-ing behavior, lipid and carbohydrate metabolism,sleep, and blood pressure control—that are of funda-mental importance to human health, performance,and well-being (Dunlap et al., 2004). These dailyoscillations are controlled by a circadian clock com-posed of autoregulatory transcription/translationfeedback loops of the expression of central clockgenes, especially Clock, Bmal1/Arntl, Bmal2/Arntl2,Per1, Per2, Per3, Cry1, and Cry2 (Dunlap et al., 2004;Lowrey and Takahashi, 2004). The clock-controlledhormone melatonin also plays an important role incircadian and sleep systems (Lewy et al., 2006).Perturbation of the central clock genes by mutationor knockout in mice elicits phenotypes ranging fromarrhythmicity (Bunger et al., 2000; van der Horstet al., 1999; Vitaterna et al., 1994) to metabolic syn-drome/obesity (Staels 2006), altered insulin/glucoseresponsiveness (Rudic et al., 2004), and increasedsusceptibility to cancer (Fu et al., 2002).

In humans, disorders of the biological clock havebeen strongly implicated in sleep disorders, shiftwork, alertness/performance, cardiovascular disease,and mental health (Dunlap et al., 2004; Kryger et al.,2005). For example, in depression, there is often adaily variation in mood (depression worse inmorning) with altered phasing of the cortisolsecretion rhythm (and other rhythms; Dunlap et al.,2004; McClung, 2007). Recent studies have shownthat clock gene polymorphisms are associated withseveral syndromes, including sleep and mooddisorders, for example: (1) variants in Per2 and CKIδwith “familial advanced sleep phase syndrome”(FASPS; Toh et al., 2001; Xu et al., 2005), (2) variantsof Per3 with delayed sleep phase syndrome (Ebisawaet al., 2001) and extreme diurnal preference (Archeret al., 2003), (3) a variable number of tandem repeatsin exon 18 of Per3 (VNTR) with sleep structure anddysfunction (Viola et al., 2007), (4) a single nucleotidepolymorphism (SNP) in the promoter of the humanarylalkylamine N-acetyl-transferase gene (AANAT)with short sleep duration (Wang et al., 2004), and (5)a polymorphism in the 3′UTR of Clock with diurnalpreference (Katzenberg et al., 1998), and sleepdisorders (Serretti et al., 2005).

A previous article reported that a biallelic tandemrepeat polymorphism (VNTR) in the human Per3 geneexhibited significantly different allelic frequencies

between Papua New Guineans (0.19) versus EastAsians (0.80-0.89), whereas European/American/African populations had intermediate frequencies(0.6~0.7; Nadkarni et al., 2005). We therefore testedwhether these geographical/ethnic differences inallelic frequency were common in other biologicalclock genes or whether this phenomenon was spe-cific to the VNTR polymorphism of Per3. This studyreports the frequency distribution of previouslyreported polymorphisms of potential functional sig-nificance in Per2, Per3, Clock, and AANAT in AfricanAmericans, European Americans, Han Chinese,Papua New Guineans, and Africans from Ghana.Because Papua New Guinean samples may be genet-ically diverse, we also performed analyses on sub-populations within Papua New Guinea. In addition,we performed systematic screening and characteriza-tion of common exonic and promoter polymorphismsin the ARNTL (aka hBmal1 or hMop3; Hogenesch et al.,1998) and ARNTL2 (aka hBmal2 or hMop9; Hogeneschet al., 2000) genes by single-stranded conformationpolymorphism (SSCP) analyses and sequencing, andreport the discovery of several novel polymorphisms.We use population genetic analyses on these clock genepolymorphisms (1) to understand population stratifi-cation, (2) to assess selection/drift, and (3) as a basis forfuture association analyses of clock gene polymor-phisms with phenotypes of behavioral and/or medicalsignificance.

MATERIALS AND METHODS

Study Subjects

DNA samples were collected from unrelatedhuman subjects from various global populations,including: African American (n = 48 from the CoriellCell Repository), European American (n = 422 col-lected at Vanderbilt), Han Chinese (n = 48 from theCoriell Cell Repository), New Guinean (n = 66), andGhanaian (n = 48, randomly selected from a cohort ofunrelated individuals of the Ga ethnic group thatwere collected as part of an ongoing study of thegenetic basis of hypertension and cardiovascular dis-ease; Williams et al., 2000). The Papua New Guinea(PNG) samples are further subdivided into separatesubpopulations as follows: Sepik (n = 18), Madang(n = 4), Gimi (n = 10), Goroka (n = 11), andNasioi/Bougainville (n = 23). The study wasapproved by all relevant institutional review boardsand participating individuals.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

332 JOURNAL OF BIOLOGICAL RHYTHMS / August 2008

DNA Isolation and Primer Design

For the European American (EA) samples collectedat Vanderbilt, 5 mL of whole blood was collected fromeach subject and DNA was isolated from lymphocytesby the Vanderbilt DNA core facility. For the Nasioi,Gimi, and Goroka PNG samples, DNA was isolatedfrom lymphoblast cell lines grown at Yale University.Previously reported primers were used for the Per2,Per3, and Clock variants (Table S1; additional data[Tables S1-S10 and Figs. S1-S2] are available athttp://jbr.sagepub.com/supplemental/). For Arntland Arntl2, forward and reverse primers yielding~200-bp amplicons were generated to cover the pro-moter regions, exons, and intronic regions immedi-ately flanking exons in a series of overlappingfragments (Table S1). These primers were designedwith an annealing temperature of ~55 °C.

SSCP Screening and Genotyping

Each amplicon was PCR amplified from the genomicDNA using 1 of 2 protocols: Platinum PCR Supermix(Invitrogen, Carlsbad, CA) was used for Per2, Per3, andClock. For Arntl and Arntl2, AmpliTaq Gold (AppliedBiosystems, Foster City, CA) was used with the pro-vided PCR Buffer II and MgCl2. Loading dye (4× = 40mM EDTA, 0.1% bromophenol blue, 0.1% xylenecyanol prepared in deionized formamide) was added toall PCR products. These products were then denaturedat 94 °C for 3 min, rapidly cooled to 4 °C, and separatedin MDE electrophoresis gels (Lonza, Basel, Switzerland)at 15 W in 0.6× TBE running buffer at 4 °C for 8 to 14 h.Gels were developed by silver staining to visualizebands (Nataraj et al., 1999). In the case of the Arntl andArntl2 genes, we used SSCP to screen for common poly-morphisms in the EA population (defined as thosepolymorphisms present in 5% or more of the popula-tion). As an initial screening cohort, we chose 20 ran-dom EA subjects (~ 5% of the total population) to lookfor novel variants and to verify the existence of knownpolymorphisms within each of the amplicons analyzed(80 amplicons for Arntl and 62 amplicons for Arntl2).The screening set was estimated to provide 98% powerto detect a polymorphism with a minor variant fre-quency of 0.10, and 87% power with a frequency of 0.05.Each amplicon was amplified by PCR and then run onan acrylamide gel for SSCP analysis. Samples that rep-resented the most common SSCP banding pattern andalso any samples that exhibited an uncommon band

shift pattern were sequenced as described below toconfirm the common sequence and to identify poly-morphisms (for the less common patterns).

In the cases of the Per2, Per3, Clock, and AANATgenes, we used previously reported primers orredesigned primers to produce ~200-bp ampliconsthat are roughly centered upon the site of SNPs thathave already been described (Table S1). For the Per2and Clock genes, we discovered novel SNPs that werelocated within the amplicons in addition to the tar-geted SNP. Finally, SNP allele frequencies in all geneswere determined on all the samples in the indicated5 populations using the SSCP patterns as confirmedby sequencing. Results from the genotyping analysesusing SSCP were concordant with the existing data indbSNP for allele frequencies in the different popula-tions, indicating the fidelity of this method.

Sequence Analysis

Amplicons with band shifts were PCR amplifiedas stated above, cleaned with the QIAQuick PCRPurification Kit (Qiagen, Valencia, CA), andsequenced with an ABI 3730×l DNA Analyzer(Applied Biosystems). Control and shifted sequenceswere aligned using Clustal W, and variants werecompared to dbSNP build 126 for Homo sapiens(NCBI, Bethesda, MD).

Statistical Analyses

Allelic and genotypic distributions were com-pared between ethnic groups using Fisher’s exacttests. Hardy-Weinberg Equilibria (HWE) were calcu-lated for each marker using Fisher’s exact tests.Haplotype frequencies were estimated using anexpectation maximization (EM) algorithm that recon-structs haplotypes and determines frequencies thatmaximize the likelihood of the genotypic data(Excoffier and Slatkin, 1995; Hawley and Kidd, 1995).Differences in haplotype distributions between pop-ulations were determined using a regressionapproach (Zaykin et al., 2002). Wright’s FST estimateswere calculated between each population and withinthe PNG subpopulations using a method developedby Weir and Cockerham (1984). Wright’s FST is a mea-sure of population differentiation that ranges from 0to 1 with 0 indicating no differentiation and 1 indi-cating highly differentiated populations. The above

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

Ciarleglio et al. / HUMAN CLOCK GENE POLYMORPHISMS IN DIFFERENT POPULATIONS 333

analyses were performed with either Stata version 9,Powermarker (Liu and Muse, 2005), or Tools forPopulation Genetic Analysis (TFPGA, version 1.3available at http://www.marksgeneticssoftware.net/); all gave virtually the same results.

RESULTS

SSCP Screening of Polymorphismsin Arntl and Arntl2

Based on the previous study of the geographicalfrequency distributions of the Per3 VNTR, we assessedthe allele frequency of the clock gene SNPs in 5 popu-lations: EA, African Americans (AA), Ghanaians (GA),Han Chinese (HC), and PNG (Nadkarni et al., 2005).Because PNG populations are known to be geneticallyheterogeneous (Friedlaender et al., 2008), we assayedseveral subpopulations within PNG to determine ifthe study of Nadkarni and coworkers (2005) provideddata that reflected PNG as a whole or only certain sub-populations within PNG.

In addition to genotyping previously reportedSNPs within the clock genes Clock, Per2, Per3, andAANAT, we identified several novel SNPs in the pro-moter and exonic regions of the key clock genes Arntland Arntl2 using SSCP gel analyses (Nataraj et al.,1999). Our criterion for a “common” polymorphismwas frequency greater than 5% in the initial screeningcohort. This screening strategy identified 11 SNPs inArntl and Arntl2 that were chosen for further charac-terization (Table S1). Of these 11 SNPs, 3 were novel

variants in Arntl and 3 others were unreported vari-ants in Arntl2 (Table 1). Not only did our SSCPmethod reveal these 6 novel polymorphisms in theArntl and Arntl2 genes, it also revealed 1 new SNP inClock and 3 new SNPs in Per2 in the ampliconsdesigned to uncover the previously reported SNPs(Table 1). Two previously reported SNPs in Per2 andPer3 for which rs# have not been assigned arereferred herein as c.1984A>G (Toh et al., 2001) andc.2460A>G (Ebisawa et al., 2001; see Table S1 for thelocation of all SNPs discussed in this article alongwith the primer pairs we used).

Among the Arntl SNPs, ss95215855 was the onlyvariant found exclusively in the EA population(Table 2 and Fig. S1). The other two SNPs in Arntlwere both intronic (Table 1) and occurred in non-EApopulations at low frequency (Tables 2 and S2, Fig.S1). We screened 8 polymorphisms in the Arntl2 genein the EA population (Tables 1, 2, S2, Fig. S1), ofwhich 5 were polymorphic in all populations (Table2). The first (rs5797225) was a TTG repeat in the 5′region that is likely to act as promoter of Arntl2. Thispolymorphism consisted of either 7, 8, 9, or 10 TTGrepeats, with the 8 or 9 repeat structure being mostcommon in all populations (Table 2, Fig. S1). The next2 polymorphisms (rs7137588 and rs11048972) werewithin 19 bp of each other in the Arntl2 promoterregion and were variable in all populations (Table 2,Fig. S1). One other common SNP occurs within theputative promoter of Arntl2 (rs10548381) and anotherSNP (rs4964059) occurs in the intron between exons 2and 3 (ss95215857, ss95215858, and ss95215859 do notappear in Table 2 because they are very rare). One of

Table 1. Novel variants discovered in this study with their chromosomal locations.

Genes(OMIM) Location Chromosomal Position SNPs Placement Protein Change Base Change

Arntl 11p15.2 13286395 ss95215854 Intronic n/a T>G(602550) 13288144 ss95215855 5′UTR n/a A>C

13335041 ss95215856 Intronic n/a G>A

Arntl2 12p11.23 27375128 ss95215857 Promoter n/a A>G27420525 ss95215858 Intronic n/a T>C27464992 ss95215859 3′ flanking n/a A>G

Clock 4q12 55996140 ss95215860 3′UTR n/a A>G(601851)

Per2 2q37.3 238830546 ss95215861 Intronic n/a G>T(603426) 238830376 ss95215862 Exonic A>V C>T

238830273 ss95215863 Intronic n/a C>T

NOTE: Previously unreported single nucleotide polymorphisms (SNPs) are referred to by ss numbers; these have been submitted to dbSNPand will be publicly available as rs numbers in Build 130. OMIM references for each gene are shown below the gene names. Chromosomal loca-tions are based on the NCBI Build 36.1 for the human genome (release date March 2006). See Table S1 for complete list of variants studied andfurther information about reference sequences, flanking sequences, primer pairs, and so forth.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

334 JOURNAL OF BIOLOGICAL RHYTHMS / August 2008

the SNPs in Arntl2 departed from HWE at the p < 0.05level (rs4964059) and 2 SNPs departed from HWE atthe p < 0.01 level (rs10548381 and ss95215857; Tables2 and S2). Each of these departures was limited to asingle population, and for 2 of these 3 the popula-tions may represent either recent admixture amongdistinct populations (rs10548381 in AA) or stratification(rs4964059 in PNG).

Population Analyses of Arntl, Arntl2,Clock, Per2, Per3, and AANAT Genes

We also analyzed previously reported SNPs in thecentral clock genes Clock, Per2, and Per3, and in theAANAT gene that encodes the rate-limiting enzyme ofmelatonin biosynthesis (Tables 1 and S1). Our rationalefor including these SNPs in the study was 3-fold. Thefirst was to compare our systematic analyses with thatof the only previous population genetic study of aclock gene SNP, namely of the VNTR SNP in Per3(Nadkarni et al., 2005). The second reason for studyingthese previously reported SNPs is that many of themhave been reported to have functional significance.Finally, our third rationale was that many of them are

sufficiently common in human populations that theymight have some value in explaining common syn-dromes that might be associated with clock gene poly-morphisms (in contrast, while the SNPs in Per2 andCKIδ have potent effects upon the timing of sleep andactivity [Toh et al., 2001; Xu et al., 2005], they are souncommon that they are unlikely to have any realexplanatory power for disorders in the general popu-lation). We again used the SSCP method to assess thefrequency of these SNPs by designing primers toamplify ~200-bp amplicons around the reported SNPs.We then sequenced representatives of every differentSSCP pattern we observed for every amplicon. Thismethod was successful in identifying the reportedSNPs, and in addition, we discovered novel SNPs inClock and Per2 within these amplicons (Table 1).

SNPs rs1801260 (in Clock) and rs4238989 (inAANAT) were polymorphic in all populations westudied (Table 2 and Fig. S2). SNP ss95215860 inClock, a previously unreported SNP we identified inthe PNG population, was unique to PNG (Table S2).The c.1984A>G SNP in Per2 that is associated withFASPS (Toh et al., 2001) was not detected in any ofour populations (Table S2). In the process of SSCP

Table 2. Allele frequencies of the common polymorphisms in all populations with deviations from Hardy-Weinberg Equilibrium (HWE).

Gene Allele African American European American Ghanaian African Han Chinese Papua New Guinea

Arntlss95215854 T 1.00 0.99 0.93 1.00 1.00

ss95215855 A 1.00 0.97 1.00 1.00 1.00

Arntl2rs5797225 GTT (7) 0.01 — — — —

GTT (8) 0.49 0.71 0.51 0.48 0.71GTT (9) 0.50 0.29 0.46 0.51 0.28GTT (10) — — 0.03 0.01 0.01

rs7137588 G 0.86 0.73 0.87 0.93 0.75rs11048972 A 0.98 0.94 0.87 0.98 1.00rs10548381 +/CTA 0.93* 0.90 0.98 0.98 0.99rs4964059 A 0.91 0.67 0.88 0.43 0.43**

Clockrs1801260 T 0.86 0.74** 0.83 0.92 0.83

Per2rs2304669 A 0.97 0.87 0.99 0.94 0.94rs2304670 G 0.90 0.93 0.87 0.94 1.00rs2304671 G 0.96 0.93 0.96 0.89 1.00

Per3rs228696 C 0.95** 0.97 0.98 1.00 1.00rs228697 C 0.98 0.99 1.00 0.95 0.83AB047536 4-repeat 0.59 0.66 0.64 0.81* 0.41rs10462021 A 0.93 0.83 0.96 1.00 0.98

AANATrs4238989 G 0.74 0.53 0.87 0.53 0.51

NOTE: See Table S2 for allele frequencies of a more complete list of polymorphisms studied.*Deviation from HWE at p < 0.01.**Deviation from HWE at p < 0.05.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

Ciarleglio et al. / HUMAN CLOCK GENE POLYMORPHISMS IN DIFFERENT POPULATIONS 335

screening for c.1984A>G, however, we identifiedother Per2 SNPs, including 3 novel SNPs, ss95215861,ss95215862, and ss95215863 with extremely low fre-quencies (<0.01), and 3 SNPs that have been previ-ously identified (rs2304669, rs2304670, andrs2304671; Table 2). For most of the SNPs in Per3,there was a low but detectable frequency of poly-morphisms in our populations (Table S2). The mostpolymorphic of the Per3 variants was the VNTRrepeat (AB047536) that has been associated withdelayed sleep phase syndrome and sleep structuredysfunction (Ebisawa et al., 2001; Viola et al., 2007)and that has been reported to exhibit significantlydifferent allelic frequencies among populations(Nadkarni et al., 2005). This polymorphism is a 4 or 5tandem repeat of a 54-bp sequence in exon 18 of Per3.We confirm that the 4-repeat allele is the most com-mon in all populations except for the PNG popula-tion, where the 5-repeat is most common (Table 2,Fig. 1A). The 5-repeat is least common in the HCpopulation (19%), and is approximately the same (34-41%) in the EA, AA, and Ghanaian populations.

Deviation from HWE for all the SNPs of sufficientfrequency are indicated in Tables 2 and S2 (4 SNPswere present in such low frequency that they werenot studied further: ss95215861, ss95215863,

rs35899625, and rs3760138). Includingthe Arntl2 SNPs that were out of HWEas noted above, there were 2 loci thatdeviated from HWE in the EA popula-tion, 1 of which had a nominal p value< 0.01 (Tables 2 and S2). In the HC andPNG populations there was 1 SNP ineach that deviated from HWE; how-ever, only the SNP in the Han popula-tion deviated with a nominal p < 0.01.In AA there were 3 SNPs that deviatedfrom HWE, 2 of which had a nominal pvalue < 0.01. None of these appears sig-nificant in the context of multiple tests.There were significant allele and geno-type differences between many of thepopulations we compared (Tables S3and S4). The populations that differedthe most in both allelic and genotypicdistributions were EA as comparedwith Ghanaians. In comparisons ofthese 2 populations, the common vari-ant was more frequent in EA (exceptfor the rs10462021 SNP in Per3). Asexpected, there were few allelic orgenotypic differences betweenGhanaians and AA (Tables S3 and S4).

Loci were excluded from haplotype and FST analy-ses if 4 or more populations had only 1 of the 2 alle-les. Haplotypes for Arntl2, Per2, and Per3 differ infrequency among the populations (Tables S5 and S6).In particular, for Arntl2 the comparisons revealed sig-nificant differences (p < 0.001) for all of the popula-tion comparisons except when comparing Ghanaiansto AA. There were also significant differences in hap-lotype comparisons when comparing PNG to theother populations (Tables S5 and S6). FST averagedacross all 5 populations was below 0.1 for all poly-morphisms (Table S7 and Fig. 2A, black trace), withthe exception of Arntl2 (rs4964059) and Per3(rs228697) (FST = 0.139 for both SNPs). Comparing thePNG subpopulations among each other shows 2SNPs (rs7137588 and rs4964059 in Arntl2) for whichFST exceeds 0.1 and 1 SNP that nearly exceeds 0.1(ss95215860 in Clock), indicating greater genetic differ-entiation among PNG populations than across theglobal samples (Fig. 2A, gray trace).

Stratification within the PNG Subpopulations

Nadkarni and coworkers (2005) reported that theVNTR polymorphism of Per3 (AB047536) exhibited

Figure 1. Proportions of the VNTR SNP of Per3 (AB047536). (A) Proportions ofthe VNTR SNP in all 5 populations (AA == African American; EA == EuropeanAmericans; HC == Han Chinese; GA == Ghana Africans; and PNG == Papua NewGuineans). (B) Proportions of the VNTR SNP in PNG subpopulations. In bothpanels, gray is the 4-repeat and black is the 5-repeat. VNTR == variable numberof tandem repeats; SNP == single nucleotide polymorphism.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

336 JOURNAL OF BIOLOGICAL RHYTHMS / August 2008

significantly different allelic frequencies in PNG (0.19),as compared to East Asian (0.80-0.89) and European/American/African populations (intermediatefrequencies of 0.6~0.7). However, PNG populationscan be very diverse genetically (Friedlaender et al.,2008), and the previous study did not give theorigin(s) of its PNG study population. We thereforeperformed our analyses using 3 populations withinPNG (Sepik/Madang, Gimi, and Goroka) and 1population from the nearby island of Bougainville(Nasioi). We found that there is very significant

variation of clock gene polymorphisms among thePNG subpopulations. For example, the 4-repeatvariant of VNTR ranged in allele frequency from 25%(Gimi) to 54% (Nasioi; Fig. 1B). This result impliesthat Nadkarni and coworkers (2005) might havestudied a distinct subpopulation within PNG andthat their results might not be reflective of PNGpopulations in general. The VNTR was not the onlypolymorphism that showed significant variationamong PNG subpopulations, as implied by the allelefrequencies and FST values for PNG (Fig. 2Aand Tables 3, S7, and S8) and as is more explicitlyshown by the FST comparisons among the PNGsubpopulations (Fig. 2B and Table S9, includingpairwise comparisons between the PNG populationand the other populations, Table S10). There arenumerous FST values greater than 0.10 when the PNGsubpopulations are compared, for example, in thers5797225, rs7137588, rs4964059, rs2304669, andVNTR polymorphisms. In fact, the FST values formost of those SNPs exceed that of the VNTRpolymorphism when compared among the PNGsubpopulations (FST is only greater than 0.1 for theVNTR polymorphism when the Eastern Highlandssubpopulation is compared with the Nasioisubpopulation; Fig. 2B and Table S9). These datasupport the idea that genetic differentiation amongthe PNG populations is greater than that across muchof the rest of the world.

DISCUSSION

The function of biological clocks is to provide tem-poral information to the organism so that physiolog-ical and/or behavioral responses can be coordinatedduring the daily cycle to maximize adaptiveness.Environmental parameters of relevance to biologicalclocks include daily fluctuations in light intensityand temperature and seasonal changes in day lengthand temperature (Dunlap et al., 2004). These parame-ters vary dramatically as a function of latitude andlocale. Therefore, biological clocks within wide-rangingspecies may be genetically “tuned” by natural selec-tion to optimize adaptiveness along latitudinal clinesor other location-dependent factors. Indeed,Drosophila melanogaster exhibits such genetic poly-morphisms in circadian clock genes along latitudinalclines (Sawyer et al., 2006; Tauber et al., 2007). Homosapiens is a species that is distributed to practicallyevery terrestrial niche on earth. Therefore it is of

Figure 2. FST values for SNPs (theta P). Two comparisons areshown—the 5 populations together (panel A) and the PNG sub-populations (panel B). In panel A, the black line is an average FST

across all populations with all PNG subpopulations combinedinto 1; the gray line is the FST across all PNG subpopulationscompared simultaneously. Panel B is the pairwise comparison ofPNG subpopulations. Black solid traces are EH vs. NA, dashedtraces are EH vs. Coastal, gray solid traces are NA vs. Coastal.EH == Eastern Highlands population (Gimi and Goroka pooled);NA == Nasioi (Bougainville); and Coastal == Sepik and Madangpooled. SNP == single nucleotide polymorphism; PNG == PapuaNew Guinean.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

Ciarleglio et al. / HUMAN CLOCK GENE POLYMORPHISMS IN DIFFERENT POPULATIONS 337

interest to study levels and patterns of variation inhuman circadian clock genes to assess the evolution-ary mechanism that may have shaped these patterns.

Our study is the first systematic report of the fre-quency distribution of polymorphisms in multipleclock genes. We chose 2 groups of polymorphisms tostudy: (1) previously reported polymorphisms ofpotential functional significance in Per2, Per3, Clock,and AANAT, and (2) polymorphisms in Arntl andArntl2 identified by a systematic SSCP screen of pro-moter and exonic regions. This screen revealed sev-eral novel polymorphisms in Arntl and Arntl2. Westudied 5 diverse populations (including subpopula-tions within PNG) from very different latitudes andenvironments. Significant differences between thepopulations were found for allelic and genotypic fre-quencies. For the reported clock gene polymor-phisms that also appear in HapMap, our allelefrequencies are nearly identical to the HapMap datafor most of the assays (Tables 2 and S2), confirmingthe accuracy of the SSCP method. There were 7 of 145SNP/population combinations that were out ofHWE—well within the range expected for a type 1error at the p < 0.05 level (7 of 145 = 4.8%). Moreover,none of the SNPs show more than 1 population devi-ating from HWE (Tables 2 and S2). Of note, 1 of theseis in the combined PNG population that is probablynot a single panmictic population (Friedlaender

et al., 2008), as shown by the large FST values amongthe discrete populations.

On the basis of the SNPs for which allelic andgenotypic frequency differences could be calculatedfor all 5 populations, EA and Ghanaians differedthe most. In contrast, there were very few allelic orgenotypic differences between the Ghanaian and AAsamples. As expected, haplotype comparisonsrevealed similar patterns (Tables S5 and S6). It isinteresting that the most common haplotype for allpopulations was the same for Per2 and Arntl, but thatthis was not the case for Arntl2 and Per3. For Arntl2,the most common haplotype in the Han populationwas exceedingly rare in all other populations exceptPNG, where it was the second most common haplo-type. In Per3, the most common haplotype inPNG was second most common in all of the otherpopulations (Table S5).

The PNG population was less different from theEA, AA, and Ghanaian populations than the EA-Ghanaian difference when the PNG samples werepooled, in contrast to expectations based on the pre-vious report (Nadkarni et al., 2005). As might beexpected in terms of geographic distance, the PNGpopulation as a whole was most similar to the HC interms of individual SNP allele frequencies (TableS3), but this difference was not as clear from thehaplotype analyses. This might indicate a very

Table 3. Allele frequencies for selected polymorphisms in the Papua New Guinea (PNG) subpopulations with deviations from Hardy-Weinberg Equilibrium (HWE).

Gene Allele Eastern Highlands Nasioi Coastal

Arntlss95215856 G 0.93 0.98 0.98

Arntl2rs5797225 GTT (7) — — —

GTT (8) 0.83 0.74 0.57*

GTT (9) 0.17 0.24 0.43GTT (10) — 0.02 —

rs7137588 G 0.57 0.89 0.78rs4964059 C 0.57 0.35 0.80

Clockrs1801260 T 0.88 0.85 0.77ss95215860 A 1.00 1.00 0.89

Per2rs2304669 A 0.95 1.00 0.86

Per3rs228697 C 0.76 0.89 0.84*

AB047536 5-repeat 0.74 0.46 0.57rs10462021 A 1.00 0.93 1.00

AANATrs4238989 G 0.55 0.48 0.50

NOTE: Eastern Highlands = Gimi and Goroka; Nasioi = Bougainville; Coastal = Sepik and Madang.*Deviation from HWE at p < 0.01.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

338 JOURNAL OF BIOLOGICAL RHYTHMS / August 2008

different history of the PNG founding chromo-somes. Upon examining subpopulations withinPNG, it became clear that even the VNTR SNP—studied by Nadkarni and coworkers (2005) andused to argue that PNG is a genetic outlier—hasvery different allele frequencies among subpopula-tions. For example, in the Nasioi subpopulation, theVNTR allele frequency is close to that in the non-PNG populations (EA/GA/AA). Therefore, thevery low VNTR frequency (19% for the 4-repeatallele) reported by Nadkarni and coworkers (2005)does not reflect PNG as a whole. Moreover, the FST

analyses that compare the three PNG subgroupsindicate much larger FST values among the PNGsubgroups (Fig. 2B and Table S9) than among any ofthe 5 global populations (Fig. 2A and Table S7).Because it is difficult to imagine that the environ-ment within the PNG region is more heterogeneousthan the environments inhabited by EA vs. HC vs.AA vs. GA vs. PNG, it is unlikely that the differ-ences in allele frequency we observe are due to nat-ural selection acting in response to environmentalpressures. Genetic similarities based on largenumbers of autosomal markers place the Nasioisubpopulation sample close to the East Asian popu-lations and far from the European and African pop-ulations (Tishkoff and Kidd, 2004). Some, but notall, of the markers we examined follow this pattern.

Therefore, it seems likely that the differences in theclock gene allele distributions that we have observedare due to genetic drift (e.g., possibly a founders’effect). Most of the genetic heterogeneity amongPNG has likewise been attributed to genetic drift(Friedlaender et al., 2008). In particular, our data donot support selection of clock gene alleles on thebasis of latitude/photoperiod, which might be pre-dicted to affect the evolution of circadian clock genes(as in Drosophila; Sawyer et al., 2006; Tauber et al.,2007). Our Ghana and PNG samples are the bestdefined in terms of latitude and both are close to theequator where the annual photoperiod will be nearlyconstant at 12-h light/12-h dark (LD 12:12).Nevertheless, the allele frequencies are very differentbetween the Ghana and PNG populations. Our otherpopulations (EA, AA, and HC) are less well definedgeographically and have been subject to admixture.Nevertheless, these populations come from latitudesranging between 22°N and 70°N. At 22°N, the annualphotoperiod ranges between LD 13.2:10.8 (summer)and LD 10.4:13.6 (winter), whereas around 60°N pho-toperiods range between LD 18.4:5.6 (summer) andLD 5.4:18.6 (winter). (At 70°N there is no night in

midsummer and no day in midwinter.) Therefore, EAand HC populations will experience a broad range ofannual photoperiodic changes and there is no evi-dence from an EA/HC comparison to GA/PNG tosuggest a selective pressure for annual photoperiod.Our analysis is of central clock genes (and of a mela-tonin synthesis gene); perhaps the central clockmechanism has been mechanistically constrained inhumans and has therefore not been subject to naturalselection along latitudinal clines (in contrast to thesituation in Drosophila; Sawyer et al., 2006; Tauberet al., 2007). However, because the most importantinput signal for the entrainment of circadian pace-makers is light, it is very possible that the light inputpathway has been optimized by natural selection forspecific latitudes/photoperiods. Therefore, while wehave not found evidence for natural selection actingon central clock genes, our data do not exclude thepossibility that genes encoding input pathway com-ponents (e.g., melanopsin) will show genetic evi-dence for natural selection.

Finally, we discovered several new SNPs, includ-ing some relatively common polymorphisms in Arntland Arntl2. These polymorphisms may affect theexpression of these two critical clock genes. Forexample, because 5′UTR regions are often involvedin translational control, the novel ss95215855 SNP ofArntl could affect the translational efficiency of ArntlmRNA. In addition, common SNPs in the Arntl2gene could affect regulatory regions such as the pro-moter (rs5797225, rs7137588, rs11048972, ss95215857,and rs10548381) and intronic regions (rs4964059 andss95215858). For example, the rs7137588 SNP is in therecognition motif for the EBP-80 transcriptionalfactor, and therefore this SNP could affect transcrip-tional regulation of Arntl2. In addition to being keycomponents of the central circadian clockwork, theARNTL/BMAL1 and ARNTL2/BMAL2 transcrip-tional factors also mediate a large number of outputpathways controlled by the clock via E-box contain-ing promoters. Therefore, polymorphisms in the reg-ulatory regions of the Arntl and Arntl2 genes havethe potential to affect many clock-regulated outputprocesses.

Our study unveiled significant differentiationamong populations for clock gene allele, genotype,and haplotype frequencies that can serve as a basisfor future association analyses of clock gene poly-morphisms with phenotypes of behavioral and/ormedical significance. Although it is impossible basedon the present data to determine the motivating fac-tors in the patterns of differentiation, it appears

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

Ciarleglio et al. / HUMAN CLOCK GENE POLYMORPHISMS IN DIFFERENT POPULATIONS 339

unlikely that these differences are due to consistentpatterns of natural selection but rather are morelikely to be caused by random processes. This con-clusion is based on the large-scale differences amongthe Ghana and PNG populations that should havesimilar photoperiod and temperature regimens ascompared to the other populations studied.

ACKNOWLEDGMENTS

We thank Matthew Pullen and Kontip C. Mahautmrfor technical assistance, the Vanderbilt University DNACore for DNAstorage and sequencing, and Vanderbilt’sGeneral Clinical Research Center (supported in part bygrant M01 RR-00095 from the National Center forResearch Resources, NIH) for DNA isolation and stor-age. This research was supported by a Discovery Grantfrom Vanderbilt University, the Vanderbilt Joint Centerfor Nursing Research, U01 HL065962 (to Dr. DanRoden), and grants from the National Institute ofMental Health (R01 MH043836 to CHJ and R01MH063341 to DGM). CMC was supported by T32 andP50 awards from the National Institute of MentalHealth (T32 MH064913; P50 MH078028).

REFERENCES

Archer SN, Robilliard DL, Skene DJ, Smits M, Williams A,Arendt J, and von Schantz M (2003) A length polymor-phism in the circadian clock gene Per3 is linked todelayed sleep phase syndrome and extreme diurnalpreference. Sleep 26:413-415.

Bunger MK, Wilsbacher LD, Moran SM, Clendenin C,Radcliffe LA, Hogenesch JB, Simon MC, Takahashi JS,and Bradfield CA (2000) Mop3 is an essential compo-nent of the master circadian pacemaker in mammals.Cell 103:1009-1017.

Dunlap JC, Loros JJ, and DeCoursey PJ, eds. (2004)Chronobiology: Biological Timekeeping. Sunderland, MA:Sinauer.

Ebisawa T, Uchiyama M, Kajimura N, Mishima K, Kamei Y,Katoh M, Watanabe T, Sekimoto M, Shibui K, Kim K,et al. (2001) Association of structural polymorphisms inthe human period3 gene with delayed sleep phase syn-drome. EMBO Rep 2:342-346.

Excoffier L and Slatkin M (1995) Maximum-likelihood esti-mation of molecular haplotype frequencies in a diploidpopulation. Mol Biol Evol 12:921-927.

Friedlaender J, Friedlaender F, Reed F, Kidd K, Kidd J,Chambers G, Lea R, Koki G, Merriwether DA, andWeber J (2008) The genetic structure of Pacific Islanders.PLoS Genet 4:e19.

Fu L, Pelicano H, Liu J, Huang P, and Lee CC (2002) The cir-cadian gene Period2 plays an important role in tumorsuppression and DNA damage response in vivo. Cell111:41-50.

Hawley ME and Kidd KK (1995) HAPLO: A program usingthe EM algorithm to estimate the frequencies of multi-site haplotypes. J Hered 86:409-411.

Hogenesch JB, Gu YZ, Jain S, and Bradfield CA (1998) Thebasic-helix-loop-helix-PAS orphan MOP3 forms tran-scriptionally active complexes with circadian andhypoxia factors. Proc Natl Acad Sci U S A 95:5474-5479.

Hogenesch JB, Gu YZ, Moran SM, Shimomura K, RadcliffeLA, Takahashi JS, and Bradfield CA (2000) The basichelix-loop-helix-PAS protein MOP9 is a brain-specificheterodimeric partner of circadian and hypoxia factors.J Neurosci 20:RC83.

Katzenberg D, Young T, Finn L, Lin L, King DP, TakahashiJS, and Mignot E (1998) A CLOCK polymorphism asso-ciated with human diurnal preference. Sleep 21:569-576.

Kryger MH, Roth T, and Dement WC, eds. (2005) Principlesand Practice of Sleep Medicine. Philadelphia: Elsevier/Saunders.

Lewy AJ, Emens J, Jackman A, and Yuhas K (2006)Circadian uses of melatonin in humans. Chronobiol Int23:403-412.

Liu K and Muse S (2005) PowerMarker: An integratedanalysis environment for genetic marker analysis.Bioinformatics 21:2128-2129.

Lowrey PL and Takahashi JS (2004) Mammalian circadianbiology: Elucidating genome-wide levels of temporalorganization. Annu Rev Genomics Hum Genet 5:407-441.

McClung CA (2007) Circadian genes, rhythms and the biol-ogy of mood disorders. Pharmacol Ther 114:222-232.

Nadkarni NA, Weale ME, von Schantz M, and Thomas MG(2005) Evolution of a length polymorphism in thehuman PER3 gene, a component of the circadian sys-tem. J Biol Rhythms 20:490-499.

Nataraj AJ, Olivos-Glander I, Kusukawa N, and HighsmithWE Jr (1999) Single-strand conformation polymorphismand heteroduplex analysis for gel-based mutation detec-tion. Electrophoresis 20:1177-1185.

Rudic RD, McNamara P, Curtis AM, Boston RC, Panda S,Hogenesch JB, and Fitzgerald GA (2004) BMAL1 andCLOCK, two essential components of the circadianclock, are involved in glucose homeostasis. PLoS Biol2:e377.

Sawyer LA, Sandrelli F, Pasetto C, Peixoto AA, Rosato E,Costa R, and Kyriacou CP (2006) The period gene Thr-Gly polymorphism in Australian and African Drosophilamelanogaster populations: Implications for selection.Genetics 174:465-480.

Serretti A, Cusin C, Benedetti F, Mandelli L, Pirovano A,Zanardi R, Colombo C, and Smeraldi E (2005) Insomniaimprovement during antidepressant treatment and CLOCKgene polymorphism. Am J Med Genet B NeuropsychiatrGenet 137:36-39.

Staels B (2006) When the clock stops ticking, metabolicsyndrome explodes. Nat Med 12:54-55.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from

340 JOURNAL OF BIOLOGICAL RHYTHMS / August 2008

Tauber E, Zordan M, Sandrelli F, Pegoraro M, OsterwalderN, Breda C, Daga A, Selmin A, Monger K, Benna C, et al.(2007) Natural selection favors a newly derived timelessallele in Drosophila melanogaster. Science 316:1895-1898.

Tishkoff SA and Kidd KK (2004) Implications of biogeogra-phy of human populations for ‘race’ and medicine. NatGenet 36:S21-S27.

Toh KL, Jones CR, He Y, Eide EJ, Hinz WA, Virshup DM,Ptacek LJ, and Fu YH (2001) An hPer2 phosphorylationsite mutation in familial advanced sleep phase syn-drome. Science 291:1040-1043.

van der Horst GT, Muijtjens M, Kobayashi K, Takano R,Kanno S, Takao M, de Wit J, Verkerk A, Eker AP, vanLeenen D, et al. (1999) Mammalian Cry1 and Cry2 areessential for maintenance of circadian rhythms. Nature398:627-630.

Viola AU, Archer SN, James LM, Groeger JA, Lo JC, SkeneDJ, von Schantz M, and Dijk DJ (2007) PER3 polymor-phism predicts sleep structure and waking perfor-mance. Curr Biol 17:613-618.

Vitaterna MH, King DP, Chang AM, Kornhauser JM,Lowrey PL, McDonald JD, Dove WF, Pinto LH, Turek

FW, and Takahashi JS (1994) Mutagenesis and mappingof a mouse gene, Clock, essential for circadian behavior.Science 264:719-725.

Wang GY, Lee CGL, and Lee EJD (2004) Genetic variabilityof arylalkylamine-Nacetyl-transferase (AA-NAT) geneand human sleep/wake pattern. Chronobiol Int21:229–237.

Weir BS and Cockerham CC (1984) Estimating F-statisticsfor the analysis of population structure. Evolution38:1358-1370.

Williams SM, Addy JH, Phillips JA 3rd, Dai M, Kpodonu J,Afful J, Jackson H, Joseph K, Eason F, Murray MM, et al.(2000) Combinations of variations in multiple genes areassociated with hypertension. Hypertension 36:2-6.

Xu Y, Padiath QS, Shapiro RE, Jones CR, Wu SC, Saigoh N,Saigoh K, Ptacek LJ, and Fu YH (2005) Functional conse-quences of a CKIδ mutation causing familial advancedsleep phase syndrome. Nature 434:640-644.

Zaykin DV, Westfall PH, Young SS, Karnoub MA, WagnerMJ, and Ehm MG (2002) Testing association of statisticallyinferred haplotypes with discrete and continuous traits insamples of unrelated individuals. Hum Hered 53:79-91.

at BROWN UNIVERSITY on March 14, 2013jbr.sagepub.comDownloaded from