d

Cs

i

ARCHIVAL REPORTS

High Dimensional Endophenotype Ranking in theSearch for Major Depression Risk GenesDavid C. Glahn, Joanne E. Curran, Anderson M. Winkler, Melanie A. Carless, Jack W. Kent Jr.,Jac C. Charlesworth, Matthew P. Johnson, Harald H.H. Göring, Shelley A. Cole, Thomas D. Dyer,Eric K. Moses, Rene L. Olvera, Peter Kochunov, Ravi Duggirala, Peter T. Fox, Laura Almasy, andJohn Blangero

Background: Despite overwhelming evidence that major depression is highly heritable, recent studies have localized only a singledepression-related locus reaching genome-wide significance and have yet to identify a causal gene. Focusing on family-based studies ofquantitative intermediate phenotypes or endophenotypes, in tandem with studies of unrelated individuals using categorical diagnoses,should improve the likelihood of identifying major depression genes. However, there is currently no empirically derived statistically rigorousmethod for selecting optimal endophentypes for mental illnesses. Here, we describe the endophenotype ranking value, a new objectiveindex of the genetic utility of endophenotypes for any heritable illness.

Methods: Applying endophenotype ranking value analysis to a high-dimensional set of over 11,000 traits drawn from behavioral/neurocognitive, neuroanatomic, and transcriptomic phenotypic domains, we identified a set of objective endophenotypes for recurrentmajor depression in a sample of Mexican American individuals (n � 1122) from large randomly selected extended pedigrees.

Results: Top-ranked endophenotypes included the Beck Depression Inventory, bilateral ventral diencephalon volume, and expressionlevels of the RNF123 transcript. To illustrate the utility of endophentypes in this context, each of these traits were utlized along with diseasestatus in bivariate linkage analysis. A genome-wide significant quantitative trait locus was localized on chromsome 4p15 (logarithm ofodds � 3.5) exhibiting pleiotropic effects on both the endophenotype (lymphocyte-derived expression levels of the RNF123 gene) and

isease risk.

onclusions: The wider use of quantitative endophenotypes, combined with unbiased methods for selecting among these measures,

hould spur new insights into the biological mechanisms that influence mental illnesses like major depression.

ctstrfcp1

catbmpfeqeeachdnnamt

Key Words: Endophenotype, endophenotype ranking, family stud-es, linkage, major depression, recurrent major depression

M ajor depression is a clinically heterogeneous and commonmental illness (1) with lifetime prevalence rates approach-ing 20% (2,3). Despite overwhelming evidence that major

depression is heritable (4), recent case-control association studieshave failed to identify a locus reaching genome-wide significance(5–9), leading some to conclude that common genetic variants withsubstantial odds ratios are unlikely to exist for the disease (7). Incontrast, recent family-based linkage studies of major depressionidentified a significant quantitative trait locus (QTL) on chromo-some 3p25-26 (logarithm of odds [LOD] � 4.0) in a large sample ofaffected sibling pairs (10). This effect replicated in a smaller sampleof individuals ascertained for heavy smoking (11). However, thecausal gene(s) for this QTL remain to be identified. Given our slowpace of discovery, new approaches may be necessary to improveunderstanding of specific causal genes influencing risk of mentalillness. One possible approach to speed gene localization/identifi-

From the Olin Neuropsychiatry Research Center (DCG, AMW), Institute ofLiving, Hartford Hospital, Hartford; Department of Psychiatry (DCG,AMW), Yale University School of Medicine, New Haven, Connecticut;Department of Genetics (JEC, MAC, JWK, JCC, MPJ, HHHG, SAC, TDD,EKM, RD, LA, JB), Texas Biomedical Research Institute; Department ofPsychiatry (RLO) and Research Imaging Institute (PK, PTF), University ofTexas Health Science Center San Antonio, San Antonio, Texas.

Address correspondence to David C. Glahn, Ph.D., Olin NeuropsychiatryResearch Center, Institute of Living, Whitehall Research Building, Hart-ford Hospital, 200 Retreat Avenue, Hartford, CT 06106; E-mail:david.glahn@yale.edu.

tReceived Oct 14, 2010; revised Aug 24, 2011; accepted Aug 25, 2011.

0006-3223/$36.00doi:10.1016/j.biopsych.2011.08.022

ation is the use of informative quantitative intermediate pheno-ypes or endophenotypes in families (12,13). Such an approach hastrategic benefits (e.g., simultaneous identification of endopheno-ypes, increased power to identify genes, increased power to detectare functional variants) over the more common paradigm that hasocused on collections of unrelated individuals and relied solely onategorical diagnoses. Endophenotype exploitation should im-rove the likelihood of identifying major depression genes (12,4 –16).

Although the application of allied phenotypes has been suc-essful in other complex illnesses (13,17–19), difficulties choosingppropriate endophenotypes for mental disorders have limitedheir use in psychiatry, where relatively less is known about theiological mechanisms that predispose illness than in other areas ofedicine. While the endophenotype concept is widely espoused in

sychiatric genetics (13,15,20), a formal or standardized approachor the identification of endophenotypes is lacking. Most studiesmploy purely phenotypic correlations between disease risk and auantitative risk factor to define putative endophenotypes. How-ver, the endophenotype concept fundamentally depends on thexistence of joint genetic determination of both endophenotypend disease risk (12,13). This obligatory pleiotropy is most effi-iently tested using family-based observations to assess both theeritability of the endophenotype and its genetic correlation withisease liability. To facilitate the identification of optimal endophe-otypes, we developed the endophenotype ranking value (ERV), aovel objective index of the genetic utility of endophenotypes forn illness. The ERV provides an unbiased and empirically derivedethod for choosing appropriate endophenotypes in a manner

hat balances the strength of the genetic signal for the endopheno-

ype and the strength of its relation to the disorder of interest. It is

BIOL PSYCHIATRY 2012;71:6–14© 2012 Society of Biological Psychiatry

t

hmnlnnApdsEp

idcoifttfcBrst

asiscmd

B

uTntoIdtn

I

mieaprFrflwwsa

pTmwh.

soo

T

R

PSGAH1345678

D.C. Glahn et al. BIOL PSYCHIATRY 2012;71:6–14 7

defined using the square root of the heritability of the illness (hi2),

he square root of the heritability of the endophenotype (he2), and

their genetic correlation (�g) and is expressed in the following for-mula:

ERVie � |�hi2�he

2�g|

Endophenotype ranking values vary between 0 and 1, whereigher values indicate that the endophenotype and the illness areore strongly influenced by shared genetic factors. This method

ecessitates that endophenotypes be heritable and have someevel of pleiotropy with the studied illness, reducing the heteroge-eity of the disease and focusing on the proportion of shared ge-etic factors influencing both the endophenotype and the illness.n advantage of the ERV approach is that very large numbers ofotential endophenotypes can be efficiently assessed before con-ucting molecular genetic analyses, analogous to high-throughputcreening methods developed for drug discovery. Furthermore, theRV approach is applicable to any heritable disease and any set ofotentially relevant traits.

Applying ERV analysis to a high-dimensional set of traits, wedentified a set of significant endophenotypes for recurrent majorepression (recurrent major depressive disorder [rMDD]). We fo-used on recurrent depression to reduce the clinical heterogeneityf the disorder and potentially increase the genetic control over the

llness (3,6). We performed an automated high-dimensional searchor endophenotypes via the ranking of 37 behavioral/neurocogni-ive, 85 neuroanatomic, and 11,337 lymphocyte-based transcrip-ional candidate endophenotypes for rMDD using data acquiredrom 1122 Mexican American individuals from large randomly as-ertained extended pedigrees who participated in the Genetics ofrain Structure and Function study. Finally, we employed the top-

anked endophentypes in bivariate linkage analysis, localizing aignificant QTL exhibiting pleiotropic effects on both endopheno-ype and disease risk.

Methods and Materials

ParticipantsA total of 1,122 Mexican American individuals from extended

pedigrees (71 families, average size 14.9 [1– 87] people) were in-cluded in the analysis. Participants were 64% female and ranged inage from 18 to 97 (mean � SD 47.11 � 14.2) years. Individuals in thiscohort have actively participated in research for over 18 years andwere randomly selected from the community with the constraintsthat they are of Mexican American ancestry, part of a large family,and live within the San Antonio region (see [21] for recruitmentdetails). No other inclusion or exclusion criteria were imposed in theinitial study. However, individuals were excluded from scanning formagnetic resonance imaging contraindications. In addition, indi-viduals were excluded from scanning and neurocognitive evalua-tion for history of neurological illnesses, stroke, or other majorneurological event. Reported pedigree relationships were empiri-cally verified with autosomal markers and intrafamilial relationshipswere edited if necessary (see Table 1 for familial relationships). Allparticipants provided written informed consent on forms approvedby the Institutional Review Boards at the University of Texas HealthScience Center San Antonio and at Yale University.

Diagnostic AssessmentAll participants received the Mini-International Neuropsychiat-

ric Interview (MINI) (22), a semistructured interview augmented toinclude items on lifetime diagnostic history. Masters- and doctor-

ate-level research staff, with established reliability (� � .85) for F

ffective disorders, conducted all interviews. All subjects with pos-ible psychopathology were discussed in case conferences thatncluded licensed psychologists or psychiatrists. Lifetime consen-us diagnoses were determined based on available medical re-ords, the MINI interview, and the interviewer’s narrative. Recurrentajor depression was defined as two or more distinct episodes of

epression meeting DSM-IV criteria.

ehavioral and Neurocognitive AssessmentEach participant received a 90-minute neuropsychological eval-

ation consisting of standard and computerized measures (23,24).hirty-five neurocognitive variables were derived from 17 separateeuropsychological tests, including measures of attention/concen-

ration, executive processing, working memory, declarative mem-ry, language processing, intelligence, and emotional processing.

n addition, participants completed two questionnaires indexingepressive mood: the Beck Depression Inventory-II (BDI-II) (25) and

he neuroticism questions of the Eysenck Personality Question-aire (26).

mage Acquisition and ProcessingMagnetic resonance imaging data were acquired on a 3T Sie-

ens (Erlangen, Germany) Trio scanner with an 8-channel head coiln the Research Imaging Institute, University of Texas Health Sci-nce Center San Antonio. Isotropic anatomic images (800 �m) werecquired for each subject using a retrospective motion-correctedrotocol (27). This protocol included the acquisition of seven full-

esolution volumes using a T1-weighted, three-dimensional Turbo-lash sequence with the following scan parameters: echo time [TE]/epetition time [TR]/inversion time � 3.04/2100/785 milliseconds,ip angle � 13°. Surface-based image analyses were conductedith FreeSurfer (28,29) following standardized protocols (30). T1-eighted images were segmented into gray matter thickness mea-

ures for 53 cortical regions and 21 subcortical volumes (averagedcross both hemispheres).

T2-weighted imaging data were acquired using a 1-mm isotro-ic, turbo spin echo FLAIR sequence with the following parameters:R/TE/inversion time/flip angle/echo train length � 5 seconds/353illiseconds/1.8 seconds/180°/221. White-matter hyperintensitiesere manually delineated in three-dimensional space using in-ouse software by experienced neuroanatomists with high (r2 �

90) test-retest reproducibility (31).Diffusion tensor imaging data acquisition used a single-shot

ingle spin echo, echo planar imaging sequence with a spatial res-lution of 1.7 � 1.7 � 3.0 mm (TR/TE � 8000/87 milliseconds, fieldf view � 200 mm, 55 directions, b � 0, and 800 seconds/mm2).

able 1. Pair-Wise Pedigree Relationships

elationship Number of Pairs

arent-Offspring 689iblings 784randparent-Grandchild 122vuncular 1248alf Siblings 135st Cousins 1602rd Degree 2128th Degree 2235th Degree 1341th Degree 584th Degree 309th Degree 36

ractional anisotropy values were estimated for each subject on 13

www.sobp.org/journal

|

aog

lt(l

ias

B

tlArrmetumIvpcggfi

R

H

(wm.bmttr

P

pnopmbPistwt

P

vmtapl

8 BIOL PSYCHIATRY 2012;71:6–14 D.C. Glahn et al.

tracts using Tract-Based Spatial Statistics software (32). Diffusiontensor imaging images provided fractional anisotropy indices for13 white-matter tracts (averaged across both hemispheres).

Transcriptional ProfilingTranscriptional profiling followed procedures described by

Göring et al. (33). Total RNA was isolated from lymphocytes andhybridized to Illumina (San Diego, California) Sentrix Human WholeGenome (WG-6) Series 1 BeadChips. These BeadChips simultane-ously probe �48,000 transcripts, representing more than 25,000annotated human genes. Although we previously identified 20,413quantitative transcripts in lymphocytes, we only examined thosewith heritabilities greater than or equal to .20 (n � 11,337) in thecurrent analysis.

GenotypingDNA extracted from lymphocytes was used in polymerase chain

reactions (PCRs) for the amplification of individual DNA at 432 dinu-cleotide repeat microsatellite loci (short tandem repeats [STRs]),spaced approximately 10 cM intervals apart across the 22 auto-somes, with fluorescently labeled primers from the MapPairs Hu-man Screening Set, Versions 6 and 8 (Research Genetics, Huntsville,Alabama). Polymerase chain reactions were performed separatelyaccording to manufacturer specifications in Applied Biosystems9700 thermocyclers (Applied Biosystems, Foster City, California).For each individual, the products of separate PCRs were pooledusing the Robbins Hydra-96 Microdispenser (Robbins Scientific Cor-poration, Sunnyvale California), and a labeled size standard wasadded to each pool. The pooled PCR products were loaded into anABI PRISM 377 or 3100 Genetic Analyzer (Applied Biosystems) forlaser-based automated genotyping. The STRs and standards weredetected and quantified, and genotypes were scored using theGenotyper software package (Applied Biosystems).

Quantitative Genetic AnalysesAll analyses were conducted with SOLAR (34), which employs

maximum likelihood variance decomposition methods to deter-mine the relative importance of genetic and environmental influ-ences by modeling the covariance among family members as afunction of genetic proximity (see Supplement 1 for detail on vari-ance components methods).

The ERV represents the standardized genetic covariance be-tween the endophenotype (denoted by the subscript, e) and illness(denoted by the subscript, i) and is defined as ERVie �hi

2he2�g|. Heritability (h2) represents the portion of the pheno-

typic variance accounted for by additive genetic variance (h2 �2

g/2p). Genetic correlation represents the common genetic cova-

riance between two traits or pleiotropy (35). Bivariate quantitativegenetic analysis was used to estimate the genetic (�g) and environ-mental (�e) correlations between each potential endophenotype

nd rMDD. The phenotypic correlation (�p), which quantifies theverall relationship between the two traits, can be derived from theenetic and environmental correlations as �p � �g(h2

eh2i) �

�e[(1-h2e)(1-h2

i)]. These parameters are estimated by jointly uti-izing all available pedigree information with a multivariate normalhreshold model for combined dichotomous/continuous traits36,37). The significance of the ERV was tested by comparing the lnikelihood for the restricted null model (with �g constrained to equal

0) against the ln likelihood for the alternative model in which the �g

parameter is estimated. The resultant likelihood ratio test is asymp-totically distributed as a chi-square with a single degree of freedom.The corresponding p value is identical to that used for genetic

correlation. Before analysis, endophenotypes were normalized us- w

www.sobp.org/journal

ng an inverse Gaussian transformation. Age, sex, age � sex, age2,nd age2 � sex were included as covariates whose effects wereimultaneously estimated in all analyses.

ivariate Linkage AnalysisBivariate linkage analysis exploits the genetic covariance be-

ween the endophenotype and the illness to improve the power toocalize QTLs and to detect QTL-specific pleiotropic effects (36).fter addressing (by blanking, recalling, or retyping) mistyping er-

ors identified using Simwalk II (Savannah Simulations AG, Her-liberg, Switzerland) (38), genotype data were used to compute

aximum likelihood estimates of allele frequencies. Matrices ofmpirical estimates of identity-by-descent allele sharing at pointshroughout the genome for every relative pair were computedsing the Loki package (39). We used high-resolution chromosomalaps provided by deCODE genetics (deCODE genetics, Reykjavik

celand) (40). For the localization of QTLs, we performed both uni-ariate and bivariate variance components linkage analyses by em-loying the models for combined analysis of quantitative and di-hotomous phenotypes described by Williams et al. (36,37). Once aenome-wide significant localization was made, formal single de-ree of freedom likelihood ratio tests for pleiotropy were per-

ormed to test the specific hypothesis that a QTL at that locationnfluenced a given endophenotype/rMDD risk (35).

esults

eritability of Recurrent Major DepressionTwo hundred fifteen individuals met criteria for lifetime rMDD

19% of the sample; 73% female subjects). Eighty-six individualsere clinically depressed at the time of the assessment. The esti-ated heritability for lifetime rMDD was h2 � .463 (standard error �

12), p � 4.0 � 10�6. We previously demonstrated that this herita-ility estimate is not significantly influenced by common environ-ental factors as indexed by shared household (21). Additionally,

here was no evidence for dominance (p � .14) or additive � addi-ive epistasis (p � .18), suggesting that the heritability estimateeflects additive genetic factors.

otential Behavioral/Neurocognitive EndophenotypesThe 10 top-ranked behavioral/cognitive endophenotypes are

resented in Table 2. The BDI-II was the highest ranked endophe-otype in this class. Although the BDI-II was developed as an indexf mood state, the heritability of this measure was h2 � .254 (� .07),� 5.6 � 10�5, demonstrating that 25% of the variability on thiseasure is due to additive genetic factors. The genetic correlation

etween the BDI-II and the neuroticism questions from the Eysenckersonality Questionnaire, the second best ranked endophenotype

n this domain, was �G � .870 (� .09), p � 3.3 � 10�4, suggestingignificant pleiotropy and potential redundancy between thesewo measures. Top-ranked cognitive measures include indices oforking and declarative memory, attention, and emotion recogni-

ion.

otential Neuroimaging EndophenotypesThe top-ranked brain region was bilateral ventral diencephalon

olume (Table 2), a region primarily comprised of the hypothala-us. As part of the hypothalamic-pituitary-adrenal axis, the hypo-

halamus mediates neuroendocrine and neurovegetative functionsnd has been consistently implicated in the neurobiology of de-ression (41). Hypothalamic-pituitary-adrenal axis activity is regu-

ated, in part, by the hippocampus and amygdala (41), both regions

ith reasonably high ERV ranking (3rd and 13th ranked, respec-

tebrtdtp

GT

treugu(g

N

T

Sh

EPQ,f

D.C. Glahn et al. BIOL PSYCHIATRY 2012;71:6–14 9

tively). Our results suggest that the genetic factors that influencethe structure of these subcortical regions (Figure 1) also confer riskfor rMDD. Additionally, white-matter hyperintensity measures,which are associated with aging, cerebrovascular dysfunction,smoking, and a host of other depression-related pathologies (42),were highly ranked endophenotypes for rMDD. This result is consis-tent with and extends the vascular depression hypothesis (43), bysuggesting common genetic factors increase risk for rMDD andwhite-matter hyperintensities.

Potential Transcriptional EndophenotypesEndophenotype ranking value analyses on 11,337 transcripts

are presented in Figure 2 and top-ranked transcriptional endophe-notypes for rMDD are presented in Table 2. The top-ranking tran-script, RNF123, is a member of the E3 ubiquitin-protein ligase family,which have diverse functions including protein degradation andmodulation of protein assembly, structure, function, and localiza-tion (44,45). Other members of the ubiquitin-proteosome systemwere previously implicated in anxiety, depression, and vulnerabilityto seizures (46,47). PDXK, an essential cofactor in the intermediate

Table 2. Ten Top-Ranked Endophenotypes per Domain for Recurrent Majo

Endophenotypes ERV

Behavioral/NeurocognitiveBeck Depression Inventory II .263EPQ Neuroticism .238Declarative Memory (CVLT Recognition) .136Working Memory (Digit Span Forward) .142Working Memory (Letter-Number Span) .135Penn Facial Memory (Immediate) .127Penn Facial Memory (Delayed) .134Attention (CPT hits) .119Attention (Trails A) .121Penn Emotion Recognition .117

euroimagingVentral diencephalon volume .240Parietal hyperintensity volume .282Hippocampus volume .204Pallidum volume .203Cerebellar white matter volume .218Frontal hyperintensity volume .255Corticospinal tract (FA) .208Subcortical hyperintensity volume .213Superior parietal gyrus thickness .178Thalamus proper volume .172

ranscriptionalRNF123 (3p24) .323PDXK (21q22) .331ZFP64 (20q13) .352RWDD2A (6q14) .260B4GALT7 (5q35) .276MARK2 (11q12) .180GADD45A (1p31) .344PIGN (18q21) .274HTT (4p16) .225ABHD12 (20p11) .269

Ten top-ranked endophenotypes for recurrent major depression in thupplement 1 for the complete rankings). Genetic correlations are betweeneritability estimates were estimated as part of bivariate models.

CPT, Continues Performance Test; CVLT, California Verbal Learning Test;ractional anisotropy.

metabolism of amino acids and neurotransmitters, including sero- r

onin and dopamine (48), confers risk for Parkinson disease (49) andpilepsy (48). Additionally, MARK2 and ABHD12 have previouslyeen implicated in neuronal migration (50), degeneration (51), and

egulation of endocannabinoid signaling pathways (52), respec-ively. Although other identified transcripts are less obvious candi-ates for rMDD risk, they may represent novel genes whose func-

ions are not fully understood and may extend to depressionhenotypes.

enome-Wide Bivariate Linkage Analyses Using rMDD andop-Ranked Endophenotypes

We performed a genome-wide linkage-based search for pleio-ropic quantitative trait loci influencing disease risk and the top-anked endophenotype from each class: BDI-II, bilateral ventral di-ncephalon volume, and the RNF123 transcript. First, standardnivariate linkage analyses were performed. Two traits exhibitedenome-wide or near genome-wide significance QTLs. The bestnivariate score for rMDD was found on chromosome 4 at 47 cM

LOD � 2.98, nominal p � .00011). While not reaching traditionalenome-wide significance, this result points to a potential disease-

ression

alue Genetic Correlation (�g) Heritability (h2)

10�5 .825 .25310�4 .739 .22810�2 �.338 .33810�2 �.295 .49010�2 �.267 .54110�2 �.319 .34410�2 �.295 .43910�2 �.387 .20210�2 .303 .34010�1 �.288 .347

10�3 �.425 .69410�3 .569 .57310�2 �.347 .77110�2 �.396 .56210�2 �.443 .52410�2 .483 .63510�2 �.900 .10110�2 .473 .45910�2 .363 .48010�2 �.294 .739

10�6 �.943 .20910�5 �.689 .48910�5 �.711 .47010�5 .666 .33710�5 �.732 .30910�5 �.399 .41210�5 .729 .43210�5 .646 .39910�5 �.546 .35810�4 .755 .272

egories of behavioral/cognitive, neuroimaging, and RNA transcripts (seerespective endophenotypes and lifetime affection status. Endophenotype

Eysenck Personality Questionnaire; ERV, endophenotype ranking value; FA,

r Dep

p V

1.9 �1.79 �5.49 �5.69 �6.39 �6.99 �8.19 �8.39 �9.69 �1.09 �

3.99 �7.89 �1.29 �1.39 �1.39 �1.39 �2.19 �4.19 �4.59 �4.89 �

5.29 �1.19 �2.09 �2.39 �3.69 �3.99 �4.09 �7.99 �7.99 �1.19 �

e catthe

elated QTL at chromosomal location 4p15. The bilateral ventral

www.sobp.org/journal

ef(

tufTdipQ

mtr[

ht.

10 BIOL PSYCHIATRY 2012;71:6–14 D.C. Glahn et al.

w

diencephalon exhibited an unequivocal genome-wide significantpeak on chromosome 7 at 131 cM (LOD � 3.40, nominal p � 3.8 �10�5). Neither BDI-II nor RNF123 expression levels showed strong

vidence for causal QTLs in univariate analysis. Suggestive evidenceor a QTL influencing BDI-II was found on chromosome 17 at 98 cMLOD � 2.57, nominal p � .0003). We found little evidence for a QTL

influencing quantitative RNF123 gene expression levels, with thesingle best univariate QTL location found on chromosome 6 at 53cM (LOD � 1.81).

Bivariate linkage analyses were performed to determine if QTLlocalization could be enhanced via simultaneous analysis withrMDD affection status. The most dramatic improvement in localiza-tion was seen for rMDD and RNF123 transcription levels. The bivari-

Figure 1. Endophenotype ranking value statistics for subcortical brain regiowere found to share genetic variance with liability for recurrent major deprnotype ranking value statistics, which provide an unbiased and empirically dventral diencephalon volume, a region primarily comprised of the hyposemi-transparent structure. ERV, endophenotype ranking value; L, left; R, rig

ate analysis of this endophenotype/disease combination substan- t

ww.sobp.org/journal

ially improved the evidence for a QTL located at 4p15 seen in thenivariate rMDD results. Figure 3 shows the QTL localization results

or the bivariate analysis and the two related univariate analyses.he peak bivariate LOD (scaled to a standard single degree of free-om LOD) was 3.51 (nominal p � 3.8 � 10�5) at 45 cM, a marked

mprovement over that seen for rMDD alone. No other rMDD/endo-henotype combination provided genome-wide evidence forTLs.

Table 3 shows the results of likelihood ratio statistic-based for-al tests of pleiotropy at the chromosome 4:45 cM location ob-

ained from the bivariate analysis of RNF123/rMDD. The marginalesults are from univariate analysis (technically co-incident linkage35]) and the strict test of pleiotropy that can be performed using

d recurrent major depression. Volume measurements of subcortical nuclein in extended pedigrees selected without regard to phenotype. Endophe-

method for choosing appropriate endophenotypes, were strongest for themus. For anatomical reference, in this image the cortex is shown as a

ns anessioerivedthala

he bivariate linkage model. The chromosome 4 locus significantly

c.of

ias

ct1tlip

D

mpddtrioindif

pTmgdiiusTctmtaBs

mgl

TT

T

RRVB

D.C. Glahn et al. BIOL PSYCHIATRY 2012;71:6–14 11

influences rMDD (p � 4.7 � 10�5), RNF123 (p � .0010), and dien-ephalon volume (p � .0290) and shows a trend for BDI-II (p �

1170). The fact that this QTL influences both risk of rMDD and twof our three best endophenotypes provides additional validation

or endophenotype identification, with evidence for rMDD increas-

Figure 2. Manhattan plot depicting whole transcriptomic search for expres-sion-based endophenotypes for recurrent major depression. Endopheno-type ranking values were calculated for 11,337 detected lymphocyte-basedtranscripts and recurrent major depression. Dashed lines reflect cutoffpoints for false discovery rate .10 (13 transcripts) and false discovery rate .25 (29 transcripts).

Figure 3. Detection of a quantitative trait locus influencing recurrent majordepression and RNF123 expression levels on chromosome 4. Multipointlogarithm of odds (LOD) functions for chromosome 4 in 1122 individualsfrom large extended pedigrees from the Genetics of Brain Structure andFunction Study. The black line represents the univariate linkage analysis forRNF123 expression levels alone. The blue line represents the univariatelinkage analysis for recurrent major depression alone. The red line repre-sents the bivariate linkage analysis for recurrent major depression andRNF123 and reaches genome-wide significance (LOD � 3.5) at 45 cM (chro-

lmosomal band 4p15). The vertical axis is in LOD score units, and the hori-zontal axis is in units of genetic distance (cM) from the p arm telomere.

ng by nearly an order of magnitude. These results strongly supportQTL influencing rMDD and related endophenotypes at chromo-

ome 4p15.Given evidence for a QTL influencing diencephalon volume on

hromosome 7, we tested for pleiotropic effects. As expected, theseests revealed a major effect on diencephalon (pleiotropy p value �.6 � 10�5) and rMDD liability (pleiotropy p value � .0437). Both ofhese results are substantially improved over their univariate ana-ogues and only with bivariate analysis do we detect a significantnfluence of this QTL on rMDD liability. The other two leading endo-henotypes show no pleiotropic effects at this QTL.

iscussion

Our results demonstrate the utility of the ERV approach for for-ally identifying endophenotypes in high-dimensional data and

rovide a novel genome-wide significant QTL for recurrent majorepression. Bivariate genetic analyses including a quantitative en-ophenotype and disease risk significantly improved QTL detec-

ion over that observed utilizing diagnosis alone. These results mayeflect the improved statistical sensitivity of quantitative over qual-tative traits or that endophenotypes index a somewhat less heter-geneous aspect of the pathophysiology associated with mental

llnesses (53). In either case, quantitative endophenotypes can sig-ificantly improve the potential to localize loci for complex disor-ers like rMDD, where multiple genes with varying effects and

ncomplete penetrance are thought to interact with environmentalactors to determine illness susceptibility.

The present experiment demonstrates the utility of gene ex-ression measures in peripheral tissues for psychiatric phenotypes.ranscripts can be considered endophenotypes that, while re-oved from the phenomenology-based diagnosis, are close to

ene action and in the case of primary cis-regulation, provide evi-ence for a gene’s involvement in the illness. Although brain tissue

s ideal for gene expression studies in psychiatry, difficulty obtain-ng this tissue in genetically informative samples necessitates these of a surrogate marker and lymphocytes appear to be goodurrogates for detection of mental disease-relevant genes (54,55).he lymphocyte measures used in the present experiment wereollected 12 to 15 years before the current assessments, minimizinghe potential that these traits were influenced by acute variation in

ood or medication usage (56). It is notable that the top-rankedranscriptional endophenotype for rMDD was ranked higher thanny of the behavioral/cognitive or neuroimaging traits, includingDI-II, suggesting that transcripts may provide an important newet of markers for disease risk.

Our single strongest ERV result was observed for quantitativeessenger RNA levels of the RNF123 gene with risk for rMDD. This

ene (also known as KPC1) encodes ring finger protein 123, which isikely involved in the regulation of neurite outgrowth via its modu-

able 3. Tests of Pleiotropy at the Chromosome 4:45 cM Quantitativerait Locus

rait

Pleiotropy p Valuefrom Bivariate

Model

Co-Incident Linkagep Value from

Univariate Model

ecurrent MDD 4.7 � 10�5 1.1 � 10�4

NF123 Expression .0010 .0219entral Diencephalon Volume .0290 .0266DI-II Score .1170 .5000

BDI-II, Beck Depression Inventory-II; MDD, major depressive disorder.

ation of the degradation of the cyclin-dependent kinase inhibitor

www.sobp.org/journal

rt(vrdipodprbpd

irbaisiopaf

tDoiMcgt

GrrCipC

c

12 BIOL PSYCHIATRY 2012;71:6–14 D.C. Glahn et al.

w

p27(Kip1) (57,58). Cyclin-dependent kinase inhibitor p27(Kip1) isinvolved in increased hippocampal neuronal differentiation via aglucocorticoid receptor function that is observed upon administra-tion of the antidepressant sertraline (59). Thus, RNF123 appears tobe a novel candidate involved in hippocampal neurogenesis ofsignificant relevance to depression risk. We observed a significantnegative genetic correlation between RNF123 expression level anddisease risk consistent with evidence that RNF123 inhibits p27(Kip1)and depression amelioration. Thus, RNF123 represents a potentialdrug target for depression.

The dominant paradigm in psychiatric genetic studies focuseson a specific disease itself. However, as with most disease states,this end point is relatively distant from the causal anatomic orphysiological disruption. In contrast, we supplement disease statuswith quantitative endophenotypes, selected through an empiri-cally derived process, to identify and characterize genes that influ-ence rMDD. Since these endophenotypes vary within the normalpopulation, it is possible to localize genes influencing them in sam-ples ascertained without regard to a specific phenotype (illness).The endophenotype and normal variation strategy have been suc-cessfully applied to the study of other complex diseases such asheart disease (17,60,61) obesity (18,62,63), diabetes (64,65), hyper-tension (19, 66), and osteoporosis (67,68). However, this strategyhas not been effectively applied in the search for mental illnessgenes.

There is debate regarding the definition of a good endopheno-type or even if endophenotypes will benefit the search for mentalillness genes. We propose that endophenotypes that are heritableand genetically correlated with disease liability can facilitate geneidentification. Although both disease and endophenotype mustbe heritable for the ERV approach, there is no requirement that theendophenotype exhibit higher heritability than the disease itself.Higher heritability estimates do not imply a simpler genetic archi-tecture or improve the potential to localize genes (24). A quantita-tive endophenotype with a low but significant heritability estimatethat is genetically correlated with disease still allows one to rankindividuals along a continuous liability distribution (69), increasingpower to identify genes. The ERV index includes no assumptionabout the genetic architecture of an endophenotype. While endo-phenotypes that are closer to gene action are desirable, this is not arequisite of an endophenotype, as information about the geneticsimplicity of a particular endophenotype is generally not availableor easily quantified. A putative endophenotype with a high ERVvalue will reflect the genetic component of disease liability betterthan one with a low ERV. Therefore, even quantitative endopheno-types with complex genetic architectures (involving many genes)can offer major advantages in genetic dissection of disease liability.Indeed, the gold standard endophenotype for heart disease, low-density lipoprotein cholesterol levels, is a complex quantitative traitthat is not particularly close to gene action (given that it does notrepresent a single protein) that was successfully used to find cardio-vascular disease risk genes (70,71).

The present experiment establishes the value of randomly se-lected families in the search for common psychiatric illness genes.While we highlight the optimality of large families for the assess-ment of heritability, genetic correlations, and ERV calculations, wenote that modern high-density typing now allows empirical assess-ment of deep kinship between unrelated individuals that could inprinciple be used to estimate these parameters (albeit very ineffi-ciently due to the remoteness of relationships). Thus, very largepreviously collected data sets of unrelateds may be of some future

value in ERV estimation.

ww.sobp.org/journal

While we demonstrate the utlitiy of the ERV approach, the cur-ent experiment has several limitatons. For example, not all poten-ial candidate ednophentypes for affective disorders were included16), as this is impractial in large-scale genetic studies. In addition,erification of endophenotypes in independent samples is war-anted. However, when the goal of simultaneous evaluation ofisease liability and endophenotype is focused on gene discovery,

t may be folly to wait for such replication rather than immediatelyursuing an independent discovery avenue like deep sequencingf a gene whose expression level is genetically correlated withisease liability. The formal testing and rigorous defining of endo-henotypes for a given disease should speed the identification of

isk genes and improve our understanding of the underlying patho-iological processes. Endophenotypes identified by emperical ap-roaches like the ERV will likely outperform nonobjective expert-erived putative endophenotypes.

The endophenotype strategy has the potential to significantlymprove our understanding of the genetic architecture of psychiat-ic illnesses (13). However, choosing optimal endophenotypes forrain-related illness is difficult when relying on theoretical factorslone. The ERV approach provides an unbiased method for select-

ng endophenotypes that is applicable to any heritable disease andhould facilitate the use of endophentypes in the search for genesnfluencing brain-related illnesses. Objective formal identificationf endophenotypes using the ERV procedure led to improvedower to localize causal QTLs influencing risk of major depressionnd the identification of a novel potential player in depression riskocused on the RNF123 gene, its products, and its pathway.

Financial support for this study was provided by the National Insti-ute of Mental Health Grants MH0708143 (Principal Investigator [PI]:CG), MH078111 (PI: JB), and MH083824 (PI: DCG). Theoretical devel-pment of the endophenotype ranking value and its implementation

n SOLAR is supported by National Institute of Mental Health GrantH59490 (PI: JI). This investigation was conducted, in part, in facilities

onstructed with support from Research Facilities Improvement Pro-ram Grant Numbers C06 RR13556 and C06 RR017515 from the Na-

ional Center for Research Resources, National Institutes of Health.We thank the study participants, our research staffs, and Irving

ottesman for 50 years of championing endophenotypes in psychiat-ic genetics. Irving Gottesman is the true source of the endophenotypeanking value. We acknowledge the Azar and Shepperd families andhemGenex Pharmaceuticals for supporting the transcriptional profil-

ng, sequencing, genotyping, and statistical analysis. The supercom-uting facilities used for this work at the AT&T Genomics Computingenter were supported, in part, by a gift from the SBC Foundation.

All authors reported no biomedical financial interests or potentialonflicts of interest.

Supplementary material cited in this article is available online.

1. Belmaker R, Agam G (2008): Major depressive disorder. N Engl J Med358:55– 68.

2. Kessler R, Berglund P, Demler O, Jin R, Koretz D, Merikangas K, et al.(2003): The epidemiology of major depressive disorder: Results from theNational Comorbidity Survey Replication (NCS-R). JAMA 289:3095–3105.

3. McGuffin P, Katz R, Watkins S, Rutherford J (1996): A hospital-based twinregister of the heritability of DSM-IV unipolar depression. Arch GenPsychiatry 53:129 –136.

4. Sullivan P, Neale M, Kendler K (2000): Genetic epidemiology of majordepression: Review and meta-analysis. Am J Psychiatry 157:1552–1562.

5. Sullivan P, de Geus E, Willemsen G, James M, Smit J, Zandbelt T, et al.(2009): Genome-wide association for major depressive disorder: A pos-

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

D.C. Glahn et al. BIOL PSYCHIATRY 2012;71:6–14 13

sible role for the presynaptic protein piccolo. Mol Psychiatry 14:359 –375.

6. Shi J, Potash J, Knowles J, Weissman M, Coryell W, Scheftner W, et al.(2011): Genome-wide association study of recurrent early-onset majordepressive disorder. Mol Psychiatry 16:193–201.

7. Muglia P, Tozzi F, Galwey N, Francks C, Upmanyu R, Kong X, et al. (2010):Genome-wide association study of recurrent major depressive disorderin two European case-control cohorts. Mol Psychiatry 15:589 – 601.

8. Uher R, Perroud N, Ng M, Hauser J, Henigsberg N, Maier W, et al. (2010):Genome-wide pharmacogenetics of antidepressant response in theGENDEP Project. Am J Psychiatry 167:555–564.

9. Lewis C, Ng M, Butler A, Cohen-Woods S, Uher R, Pirlo K, et al. (2010):Genome-wide association study of major recurrent depression in theU.K. population. Am J Psychiatry 167:949 –957.

10. Breen G, Webb BT, Butler AW, van den Oord EJ, Tozzi F, Craddock N, et al.(2011): A genome-wide significant linkage for severe depression onchromosome 3: The Depression Network Study. Am J Psychiatry 168:840 – 847.

11. Pergadia ML, Glowinski AL, Wray NR, Agrawal A, Saccone SF, Loukola A,et al. (2011): A 3p26-3p25 genetic linkage finding for DSM-IV majordepression in heavy smoking families. Am J Psychiatry 168:848 – 852.

12. Blangero J, Williams JT, Almasy L (2003): Novel family-based approachesto genetic risk in thrombosis. J Thromb Haemost 1:1391–1397.

13. Gottesman II, Gould TD (2003): The endophenotype concept in psychi-atry: Etymology and strategic intentions. Am J Psychiatry 160:636 – 645.

14. Merikangas KR, Chakravarti A, Moldin SO, Araj H, Blangero JC, Burmeis-ter M, et al. (2002): Future of genetics of mood disorders research. BiolPsychiatry 52:457– 477.

15. Insel T, Cuthbert B (2009): Endophenotypes: Bridging genomic com-plexity and disorder heterogeneity. Biol Psychiatry 66:988 –989.

16. Hasler G, Drevets WC, Manji HK, Charney DS (2004): Discovering endo-phenotypes for major depression. Neuropsychopharmacology 29:1765–1781.

17. Kathiresan S, Willer C, Peloso G, Demissie S, Musunuru K, Schadt E, et al.(2009): Common variants at 30 loci contribute to polygenic dyslipide-mia. Nat Genet 41:56 – 65.

18. Comuzzie A, Hixson J, Almasy L, Mitchell B, Mahaney M, Dyer T, et al.(1997): A major quantitative trait locus determining serum leptin levelsand fat mass is located on human chromosome 2. Nat Genet 15:273–276.

19. Cho Y, Go M, Kim Y, Heo J, Oh J, Ban H, et al. (2009): A large-scalegenome-wide association study of Asian populations uncovers geneticfactors influencing eight quantitative traits. Nat Genet 41:527–534.

20. Ritsner M (2009): Handbook of Neuropsychiatric Biomarkers, Endopheno-types and Genes: Neuroanatomical and Neuroimaging Endophenotypesand Biomarkers. Dordrecht, The Netherlands: Springer.

21. Olvera RL, Bearden CE, Velligan DI, Almasy L, Carless MA, Curran JE, et al.(2011): Common genetic influences on depression, alcohol, and sub-stance use disorders in Mexican-American families. Am J Med Genet BNeuropsychiatr Genet 156:561–568.

22. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, etal. (1998): The Mini-International Neuropsychiatric Interview (M.I.N.I.):The development and validation of a structured diagnostic psychiatricinterview for DSM-IV and ICD-10. J Clin Psychiatry 59(suppl 20):22–33;quiz 34 –57.

23. Glahn D, Almasy L, Blangero J, Burk G, Estrada J, Peralta J, et al. (2007):Adjudicating neurocognitive endophenotypes for schizophrenia. Am JMed Genet B Neuropsychiatr Genet 144B:242–249.

24. Glahn D, Almasy L, Barguil M, Hare E, Peralta J, Kent JJ, et al. (2010):Neurocognitive endophenotypes for bipolar disorder identified in mul-tiplex multigenerational families. Arch Gen Psychiatry 67:168 –177.

25. Beck AT, Steer RA, GK B (1996): Manual for the Beck Depression Inventory-II. San Antonio, TX: Psychological Corporation.

26. Eysenck HJ, SBG E (1975): Manual of the Eysenck Personality Question-naire. San Diego: Educational and Industrial Testing Service.

27. Kochunov P, Lancaster J, Glahn D, Purdy D, Laird A, Gao F, Fox P (2006):Retrospective motion correction protocol for high-resolution anatomi-cal MRI. Hum Brain Mapp 27:957–962.

28. Dale AM, Fischl B, Sereno MI (1999): Cortical surface-based analysis. I.Segmentation and surface reconstruction. Neuroimage 9:179 –194.

29. Fischl B, Sereno MI, Dale AM (1999): Cortical surface-based analysis. II:

Inflation, flattening, and a surface-based coordinate system. Neuroim-age 9:195–207.

5

0. Winkler A, Kochunov P, Blangero J, Almasy L, Zilles K, Fox P, et al. (2010):Cortical thickness or grey matter volume? The importance of selectingthe phenotype for imaging genetics studies. Neuroimage 53:1135–1146.

1. Kochunov P, Glahn D, Winkler A, Duggirala R, Olvera R, Cole S, et al.(2009): Analysis of genetic variability and whole genome linkage ofwhole-brain, subcortical, and ependymal hyperintense white mattervolume. Stroke 40:3685–3690.

2. Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mac-kay CE, et al. (2006): Tract-based spatial statistics: Voxelwise analysis ofmulti-subject diffusion data. Neuroimage 31:1487–1505.

3. Goring HH, Curran JE, Johnson MP, Dyer TD, Charlesworth J, Cole SA, etal. (2007): Discovery of expression QTLs using large-scale transcriptionalprofiling in human lymphocytes. Nat Genet 39:1208 –1216.

4. Almasy L, Blangero J (1998): Multipoint quantitative-trait linkage analy-sis in general pedigrees. Am J Hum Genet 62:1198 –1211.

5. Almasy L, Dyer TD, Blangero J (1997): Bivariate quantitative trait linkageanalysis: Pleiotropy versus co-incident linkages. Genet Epidemiol 14:953–958.

6. Williams JT, Van Eerdewegh P, Almasy L, Blangero J (1999): Joint multi-point linkage analysis of multivariate qualitative and quantitative traits.I. Likelihood formulation and simulation results. Am J Hum Genet 65:1134 –1147.

7. Williams JT, Begleiter H, Porjesz B, Edenberg HJ, Foroud T, Reich T, et al.(1999): Joint multipoint linkage analysis of multivariate qualitative andquantitative traits. II. Alcoholism and event-related potentials. Am J HumGenet 65:1148 –1160.

8. Sobel E, Papp JC, Lange K (2002): Detection and integration of genotyp-ing errors in statistical genetics. Am J Hum Genet 70:496 –508.

9. Heath SC (1997): Markov chain Monte Carlo segregation and linkageanalysis for oligogenic models. Am J Hum Genet 61:748 –760.

0. Kong A, Gudbjartsson DF, Sainz J, Jonsdottir GM, Gudjonsson SA, Rich-ardsson B, et al. (2002): A high-resolution recombination map of thehuman genome. Nat Genet 31:241–247.

1. Nestler E, Barrot M, DiLeone R, Eisch A, Gold S, Monteggia L (2002):Neurobiology of depression. Neuron 34:13–25.

2. Fazekas F, Kleinert R, Offenbacher H, Schmidt R, Kleinert G, Payer F, et al.(1993): Pathologic correlates of incidental MRI white matter signal hy-perintensities. Neurology 43:1683–1689.

3. Alexopoulos G, Meyers B, Young R, Campbell S, Silbersweig D, CharlsonM (1997): ’Vascular depression’ hypothesis. Arch Gen Psychiatry 54:915–922.

4. Doolittle M, Ben-Zeev O, Bassilian S, Whitelegge J, Péterfy M, Wong H(2009): Hepatic lipase maturation: A partial proteome of interactingfactors. J Lipid Res 50:1173–1184.

5. Deshaies R, Joazeiro C (2009): RING domain E3 ubiquitin ligases. AnnuRev Biochem 78:399 – 434.

6. Kim S, Zhang S, Choi K, Reister R, Do C, Baykiz A, Gershenfeld HK (2009):An E3 ubiquitin ligase, Really Interesting New Gene (RING) Finger 41, is acandidate gene for anxiety-like behavior and beta-carboline-inducedseizures. Biol Psychiatry 65:425– 431.

7. Nishioka G, Yamada M, Kudo K, Takahashi K, Kiuchi Y, Higuchi T, et al.(2003): Induction of kf-1 after repeated electroconvulsive treatment andchronic antidepressant treatment in rat frontal cortex and hippocam-pus. J Neural Transm 110:277–285.

8. Gachon F, Fonjallaz P, Damiola F, Gos P, Kodama T, Zakany J, et al. (2004):The loss of circadian PAR bZip transcription factors results in epilepsy.Genes Dev 18:1397–1412.

9. Elstner M, Morris C, Heim K, Lichtner P, Bender A, Mehta D, et al. (2009):Single-cell expression profiling of dopaminergic neurons combinedwith association analysis identifies pyridoxal kinase as Parkinson’s dis-ease gene. Ann Neurol 66:792–798.

0. Sapir T, Sapoznik S, Levy T, Finkelshtein D, Shmueli A, Timm T, et al.(2008): Accurate balance of the polarity kinase MARK2/Par-1 is requiredfor proper cortical neuronal migration. J Neurosci 28:5710 –5720.

1. Nishimura I, Yang Y, Lu B (2004): PAR-1 kinase plays an initiator role in atemporally ordered phosphorylation process that confers tau toxicity inDrosophila. Cell 116:671– 682.

2. Blankman J, Simon G, Cravatt B (2007): A comprehensive profile of brainenzymes that hydrolyze the endocannabinoid 2-arachidonoylglycerol.Chem Biol 14:1347–1356.

3. Blangero J (2004): Localization and identification of human quantitativetrait loci: King harvest has surely come. Curr Opin Genet Dev 14:233–240.

www.sobp.org/journal

5

5

5

5

5

6

6

6

6

6

6

6

6

6

6

7

7

14 BIOL PSYCHIATRY 2012;71:6–14 D.C. Glahn et al.

w

54. Tsuang MT, Nossova N, Yager T, Tsuang MM, Guo SC, Shyu KG, et al.(2005): Assessing the validity of blood-based gene expression profilesfor the classification of schizophrenia and bipolar disorder: A prelimi-nary report. Am J Med Genet B Neuropsychiatr Genet 133:1–5.

5. Borovecki F, Lovrecic L, Zhou J, Jeong H, Then F, Rosas HD, et al. (2005):Genome-wide expression profiling of human blood reveals biomarkersfor Huntington’s disease. Proc Natl Acad Sci U S A 102:11023–11028.

6. Tsankova N, Berton O, Renthal W, Kumar A, Neve R, Nestler E (2006):Sustained hippocampal chromatin regulation in a mouse model ofdepression and antidepressant action. Nat Neurosci 9:519 –525.

7. Zheng YL, Li BS, Rudrabhatla P, Shukla V, Amin ND, Maric D, et al. (2010):Phosphorylation of p27Kip1 at Thr187 by cyclin-dependent kinase 5modulates neural stem cell differentiation. Mol Biol Cell 21:3601–3614.

8. Zhao J, Zhang S, Wu X, Huan W, Liu Z, Wei H, et al. (2011): KPC1 expres-sion and essential role after acute spinal cord injury in adult rat. Neuro-chem Res 36:549 –558.

9. Anacker C, Zunszain PA, Cattaneo A, Carvalho LA, Garabedian MJ,Thuret S, et al. (2011): Antidepressants increase human hippocampalneurogenesis by activating the glucocorticoid receptor. Mol Psychiatry16:738 –750.

0. Almasy L, Hixson JE, Rainwater DL, Cole S, Williams JT, Mahaney MC, etal. (1999): Human pedigree-based quantitative-trait-locus mapping: Lo-calization of two genes influencing HDL-cholesterol metabolism. Am JHum Genet 64:1686 –1693.

1. Ayra R, Blangero J, Williams K, Almasy L, Dyer T, Leach R, et al. (2002):Factors of insulin resistance syndrome (IRS)-related phenotypes arelinked to genetic locations on chromosomes 6 and 7 in nondiabeticMexican Americans. Diabetes 51:841– 847.

2. Cai G, Cole SA, Freeland-Graves JH, MacCluer JW, Blangero J, Comuzzie

AG (2004): Genome-wide scans reveal quantitative trait loci on 8p and

ww.sobp.org/journal

13q related to insulin action and glucose metabolism: The San AntonioFamily Heart Study. Diabetes 53:1369 –1374.

3. Willer C, Speliotes E, Loos R, Li S, Lindgren C, Heid I, et al. (2009): Six newloci associated with body mass index highlight a neuronal influence onbody weight regulation. Nat Genet 41:25–34.

4. Cai G, Cole SA, Bastarrachea-Sosa RA, Maccluer JW, Blangero J, Co-muzzie AG (2004): Quantitative trait locus determining dietary macro-nutrient intakes is located on human chromosome 2p22. Am J Clin Nutr80:1410 –1414.

5. Mitchell BD, Cole SA, Hsueh WC, Comuzzie AG, Blangero J, MacCluer JW,Hixson JE (2000): Linkage of serum insulin concentrations to chromo-some 3p in Mexican Americans. Diabetes 49:513–516.

6. Kammerer CM, Gouin N, Samollow PB, VandeBerg JF, Hixson JE, Cole SA,et al. (2004): Two quantitative trait loci affect ACE activities in Mexican-Americans. Hypertension 43:466 – 470.

7. Kammerer CM, Schneider JL, Cole SA, Hixson JE, Samollow PB, O’ConnellJR, et al. (2003): Quantitative trait loci on chromosomes 2p, 4p, and 13qinfluence bone mineral density of the forearm and hip in Mexican Amer-icans. J Bone Miner Res 18:2245–2252.

8. Kiel D, Demissie S, Dupuis J, Lunetta K, Murabito J, Karasik D (2007):Genome-wide association with bone mass and geometry in the Fra-mingham Heart Study. BMC Med Genet 8(suppl 1):S14.

9. Williams J, Blangero J (2004): Power of variance component linkageanalysis-II. Discrete traits. Ann Hum Genet 68:620 – 632.

0. Brown MS, Goldstein JL (1976): Familial hypercholesterolemia: A ge-netic defect in the low-density lipoprotein receptor. N Engl J Med 294:1386 –1390.

1. Cohen J, Pertsemlidis A, Kotowski IK, Graham R, Garcia CK, Hobbs HH(2005): Low LDL cholesterol in individuals of African descent resulting

from frequent nonsense mutations in PCSK9. Nat Genet 37:161–165.