gray matter volume in adolescent anxiety: an impact of the brain-derived neurotrophic factor...

NEW RESEARCH|

Gray Matter Volume in Adolescent Anxiety: An Impactof the Brain-Derived Neurotrophic Factor

Val66Met Polymorphism?Sven C. Mueller, Ph.D., Aveline Aouidad, M.D.,Ph.D., Elena Gorodetsky, M.D., Ph.D., David Goldman, M.D.,

Daniel S. Pine, M.D., Monique Ernst, M.D., Ph.D.

Clin

184

Objective: Minimal research links anxiety disorders in adolescents to regional gray mattervolume (GMV) abnormalities and their modulation by genetic factors. Prior research suggeststhat a brain-derived neurotrophic factor (BNDF) Val66Met polymorphism may modulate suchbrain morphometry profiles. Method: Using voxel-based morphometry and magneticresonance imaging, associations of BDNF and clinical anxiety with regional GMVs of anteriorcingulate cortex, insula, amygdala, and hippocampus were examined in 39 affected (17Met allele carriers, 22 Val/Val homozygotes) and 63 nonaffected adolescents (33 Met allelecarriers, 41 Val/Val homozygotes). Results: Amygdala and anterior hippocampal GMVswere significantly smaller in patients than in healthy comparison adolescents, with a reversepattern for the insula. Post-hoc regression analyses indicated a specific contribution of socialphobia to the GMV reductions in the amygdala and hippocampus. In addition, insula anddorsal–anterior cingulate cortex (ACC) GMVs were modulated by BDNF genotype. In bothregions, and GMVs were larger in the Val/Val homozygote patients than in individuals carryingthe Met allele. Conclusions: These results implicate reduced GMV in the amygdala andhippocampus in pediatric anxiety, particularly social phobia. In addition, the data suggest thatgenetic factors may modulate differences in the insula and dorsal ACC. J. Am. Acad. ChildAdolesc. Psychiatry; 2013;52(2):184-195. Key Words: BDNF, voxel-based morphometry,adolescence, anxiety, insula

T o date, few studies have examined graymatter volume (GMV) abnormalities inpediatric anxiety disorders,1–3 in compari-

son to the larger literature in adults.4,5 Thesepediatric studies have mostly focused on themedial temporal lobe, including the amygdalaand hippocampus. Importantly, disagreementexists with regard to the directionality of effectsand diagnostic specificity. Although some studiesin pediatric generalized anxiety disorders (GAD)report enlarged amygdala volumes for the patientrelative to the healthy comparisons,1 other studiesin pediatric mixed samples, with GAD, separationanxiety disorder, and social phobia, reportreduced amygdala volume in the patient grouprelative to comparison subjects.3 In addition toalterations of medial temporal structures, more

ical guidance is available at the end of this article.

JOUR

www.jaacap.org

recent evidence implicates the anterior cingulatecortex (ACC) and the insula in adults with clinicalanxiety.5,6 These findings, however, are alsomixed.5 Critical to the present work, no reportshave yet documented alterations in these latterstructures in pediatric anxiety samples.

Some of the inconsistency in findings can berelated not only to unstable data in small samples,but also to genetic modulation. For instance,recent work highlights a modulation by thecommon brain-derived neurotrophic factor(BDNF) Val66Met polymorphism of brain struc-tures relevant to anxiety.7–9 In this single nucleo-tide polymorphism (SNP; rs 6265), a critical aminoacid valine-to-methionine substitution at codon66 (Val66Met) has been shown to reduce activity-dependent secretion of BDNF,9 a gene criticallycontributing to neuronal survival, migration,dendrite pruning, and axonal growth.10 AlthoughBDNF is expressed throughout the cortex, highexpression of this gene has been documented in

NAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 52 NUMBER 2 FEBRUARY 2013

dx.doi.org/10.1016/j.jaac.2012.11.016

dx.doi.org/10.1016/j.jaac.2012.11.016

dx.doi.org/10.1016/j.jaac.2012.11.016

dx.doi.org/10.1016/j.jaac.2012.11.016

MORPHOMETRY IN ADOLESCENT ANXIETY

the hippocampal formation, the prefrontal cortex,and the ACC.9,11,12 Notably, as a result of reducedBDNF expression in carriers with at least oneMet allele, previous research has reported struc-tural reductions of the amygdala, hippocampus,and ACC.8,13,14 Furthermore, a recent functionalimaging study in adolescents reported that theVal66Met polymorphism modulated amygdalaand hippocampal function during emotional pro-cessing differentially in healthy and clinicallyanxious adolescents.15 These preliminary findingscall for more work to further validate and to bettercharacterize associations among behavior, genet-ics, and brain structure in youth,16 given theknown role of the BDNF gene in development.10,17

Specifically in the context of anxiety, the potentialinfluence of the BDNF Val66Met polymorphismon brain structure has not yet been examinedin youth.

The goals of the present study were to examine,in healthy and clinically anxious adolescents, first,GMVabnormalities in pediatric anxiety, includingdiagnostic specificity, and second, the influence ofBDNF genotype on the four brain structures mostcommonly implicated in anxiety disorders—theACC, hippocampus, amygdala, and insula. Withregard to neuroanatomical differences betweenpatients and healthy comparison subjects, wehypothesized that significant GMV alterationsof the amygdala and hippocampus would befound in adolescents with anxiety relative tocomparison adolescents, based on previous struc-tural findings in adolescents3 and adults.4 Withregard to the effects of genotype, consistent with arole of the BDNFVal66Met polymorphism inmodulating anxious behavior,7,15 we expectedvolumetric reductions in patients with the riskMet allele relative to those in patients with theVal/Val genotype. By contrast, based on a largerstudy in unaffected adolescents,18 we did notexpect such volumetric reduction in healthycomparison subjects carrying the Met allele rela-tive to the Val/Val homozygotes.

METHODParticipantsA total of 39 adolescents diagnosed with an anxietydisorder and 63 healthy comparison adolescents under-went structural neuroimaging (Table 1). None were onmedication at the time of testing. No morphometry datafor these samples have been reported previously.Patients were recruited at the National Institute of

JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


Mental Health. Both patients and healthy comparisonsubjects completed a medical, physical, and psychiatricassessment. Psychiatric status was assessed via astandardized, semi-structured psychiatric interview(Kiddie Schedule for Affective Disorders andSchizophrenia—Present and Lifetime version [K-SADS-PL]).19 This interview was performed by experi-enced clinicians who had been trained to exceed aminimally acceptable inter-rater reliability of k 4 0.75.Inclusion criteria for patients consisted of a diagnosis ofan anxiety disorder, and individual diagnoses weresubsequently confirmed by expert review by a seniorpsychiatrist. To minimize confounding from otherpsychopathology, exclusion criteria consisted of adiagnosis of current obsessive-compulsive disorder,lifetime history of posttraumatic stress disorder (PTSD),mania, psychosis, substance use, or pervasive devel-opmental disorder. Healthy comparison subjects wererecruited by advertisement in local newspapers, withinclusion criteria defined by the absence of psychiatricor neurological disorder, substance use, or medicationas determined by the K-SADS. The Full-Scale IQ scoreswere pro-rated based on the Vocabulary and BlockDesign subtests of the Wechsler Intelligence Scales forChildren20 and IQ 4 70 for both groups. The Institu-tional Review Board of the National Institute of MentalHealth approved the study. Parents signed consentforms, and adolescents signed assent forms, after beingexplained the study in detail.

Participants were characterized based on genotypeand diagnostic status. Among the 39 patients, 15 wereVal/Met heterozygotes, two Met/Met homozygotes,and 22 Val/Val homozygotes. Among the 63 healthycomparison subjects, 25 were Val/Met heterozygotes,two Met/Met homozygotes, and 36 Val/Val homozy-gotes (Table 1). Because of the rarity of Met homo-zygotes, and following the grouping strategy used byprevious investigators,13 the Met/Met homozygoteswere combined with the Val/Met heterozygotes to forma group of 17 Met carriers for the patient group and 27Met carriers for the healthy comparison group. Toassess whether the study sample size was sufficientlylarge to detect a reliable group-by-genotype interaction,a post-hoc power analysis using G*Power321 revealedan acceptable power level of 0.808. Power for the maingroup comparison was 0.894. Importantly, no differ-ences emerged in genotypic distribution across patientsand comparison subjects (w2

1 o 0.01, p ¼ .942), orgenotype frequencies (Hardy–Weinberg equilibriumtest, w2

1 ¼ 0.82, p ¼ .375). The between-group compar-isons of demographic variables revealed trendingdifferences on age (F1,100 ¼ 3.78, p ¼ .055) and IQ(F1,100 ¼ 3.30, p ¼ .072) between patients and compar-ison subjects, but no differences on sex distribution(w2

1 ¼ 1.38, p ¼ .240), or socioeconomic status (SES,Hollingshead; F1,92¼ 1.15, p¼ .286). Therefore, age andIQ were used as covariates of nuisance in all analyses. Inaddition, although total brain volume did not differ

www.jaacap.org 185

TABLE 1 Demographic Characteristics of the Study Sample Separately for Patients and Healthy Controls, by Brain-Derived Neurotrophic Factor (BDNF) Genotype

Demographic InformationPatients Met

(n ¼ 17)Patients Val

(n ¼ 22)Controls

Met (n ¼ 27)Controls

Val (n ¼ 36) p

Sex, female, n (%) 8 (47.1) 14 (63.6) 14 (51.5) 14 (41.5) .324Age 11.3 (2.6) 13.7 (2.5) 13.5 (3.1) 13.9 (2.5) o.05IQ 109.1 (12.6) 110.4 (11.63) 116.0 (10.3) 113.3 (14.8) .261SES Hollingshead 40.1 (23.7) 36.9 (16.1) 42.4 (17.8) 42.4 (16.1) .704STAI state 32.79 (5.32) 33.38 (4.61) .727STAI trait 37.85 (7.74) 41.50 (7.89) .200CDI 52.86 (13.55) 56.00 (12.93) .490

Ethnic ancestry factor, scores (SD)Europe 0.67 (0.40) 0.77 (0.30) 0.76 (0.27) 0.64 (0.39) .477Africa 0.00 (0.01) 0.10 (0.25) 0.03 (0.12) 0.15 (0.32) .094Middle East 0.11 (0.24) 0.07 (0.12) 0.09 (0.16) 0.08 (0.14) .894America 0.02 (0.03) 0.03 (0.07) 0.03 (0.10) 0.03 (0.08) .901Asia 0.13 (0.30) 0.02 (0.03) 0.04 (0.05) 0.04 (0.07) .084Oceania 0.00 (0.01) 0.00 (0.00) 0.00 (0.00) 0.01 (0.01) .688Far East Asia 0.06 (0.24) 0.00 (0.00) 0.05 (0.19) 0.05 (0.18) .730

Diagnosis, n (%)GAD 7 (50) 14 (64) .206a

Social Phobia 11 (65) 11 (50) .517a

Specific phobia 1 (6) 5 (23) .206a

Separation anxiety disorder 5 (29) 7 (32) 1.00a

Depression 2 (12) 6 (27) .426a

ODD 1 (6) 1 (5)ADHD 1 (6) 3 (14)Tic disorder 1 (6) —

Enuresis — 1 (5)

Note: Bold type indicates significant value. ADHD¼ attention-deficit/hyperactivity disorder; CDI¼Child Depression Inventory; GAD¼ generalized anxietydisorder; ODD ¼ oppositional defiant disorder; SES ¼ socioeconomic status; STAI ¼ State Trait Anxiety Inventory.aFisher’s exact test.

MUELLER et al.

between patients and comparison subjects (F1,100¼ 1.94,p¼ .167) or among the four genetic groups (F3,98¼ 0.83,p ¼ .482), this factor was included as a covariate toaccount for interindividual variability. Thus, indivi-dual variability in age, total brain volume, and IQwas controlled for in all analyses. In addition, themain effects of these variables as well as theirpotential influences on the findings were also assessed.Finally, patients in the Met and Val/Val group did notdiffer on any type of anxiety disorder (all p 4 .205)(Table 1).

To detect potential effects of population origin, werelied on objective, genetic ethnic ancestry factor scoresrather than self-report questionnaires. For each parti-cipant, seven ancestry factor scores were calculated inthe context of worldwide ancestry based on 1,052individuals and 51 populations represented in theHuman Genome Diversity Panel (HGDP) and geno-typed for the same panel of 186 ancestry informativemarkers (AIMs). This AIMs panel yields a seven-factorsolution for continental and certain subcontinental

JOUR

186 www.jaacap.org

populations that has been robustly observed acrossmultiple studies.22 No significant differences in ances-try factor scores emerged between the patient andcomparison group or the four BDNF genotype-by-diagnosis groups (all p 4 .05) (Table 1). In addition,full regression analyses were run with the Europeanand African ancestry factor scores included in themodel, because these were the two principal ancestrycomponents in this sample. Ethnic ancestry factorscores did not predict volumetric variation.

Single-Nucleotide Polymorphism GenotypingGenomic DNA was extracted from Epstein–Barr virus(EBV)–transformed lymphoblastoid cell lines oruntransformed white blood cells using Gentra Versa-gene Cell kits according to the manufacturer’s protocol.Genotyping was performed using the Illumina Gold-enGate SNP genotyping technology, determining hap-lotype coverage on the 1536 SNP National Institute onAlcohol Abuse and Alcoholism (NIAAA) Addictions




Array, with a 95% completion rate. Genotyping errorwas determined by replicating genotyping in 10% of thesample (error rate o0.005). BDNF Val66Met (rs6265)genotyping was confirmed using a Taqman Assay-on-Demand (Applied Biosystems, Foster City, CA). Geno-typing was performed according to the manufacturer’sprotocol, and genotype was determined at end pointusing an ABI 7900HT Sequence Detection System.Genotyping accuracy was determined empirically byduplicate genotyping of 25% of the samples selectedrandomly. The error rate was o0.005, and the comple-tion rate was 40.95. These analyses showed 100%concordance across genotyping samples.

Magnetic Resonance Imaging Data AcquisitionWhole-brain, high-resolution, T1-weighted anatomicalimages were acquired on a 3 Tesla. General ElectricSigna Scanner (Waukesha, WI). For reasons indepen-dent of the present work, the magnetic resonanceimaging (MRI) data were collected under two acquisi-tion sequences but on the same scanner. A first setof subjects (n ¼ 78) underwent scanning using anMPRAGE sequence consisting of 124 1.2-mm axialslices (no-gap), field of view ¼ 220 mm, matrix ¼256 � 192, time to inversion (TI)¼ 725 milliseconds, flipangle 61, and voxel size of 0.86 mm � 1.15 mm � 1.2 mm.A second set of subjects (n ¼ 24) underwent scanningusing an FSPGR sequence consisting of 124 1.2-mmaxial slices (no-gap), field of view ¼ 240 mm, matrix ¼256 � 256, flip angle 151, and voxel size 0.94 mm �0.94 mm � 1.2 mm. To avoid the potential confound ofusing two different acquisition sequences, there was anequal distribution of scan sequences among the fourgenotype groups (w2

3 ¼ 1.68, p ¼ .642) and betweenpatients and comparison subjects (w2

1 ¼ 0.76, p ¼ .381).Moreover, scan sequence was included as a covariate ofno interest in the analyses.

MRI ProcessingBefore processing, data were examined for motionartefacts (ghosting, blurring) and anatomical abnormal-ities. In the case of the presence of such artifacts, datawere excluded from the study sample. MRI data werepreprocessed and analyzed using SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and the Diffeomorphic Anatomi-cal Registration using Exponentiated Lie algebra (DAR-TEL) toolbox.23 The image origin was set at the anteriorcommissure (AC). Structural imaging data were seg-mented into white matter, gray matter, cerebral–spinalfluid, skull, and soft tissue classes outside the brain.These images were inspected for the quality of segmen-tation before a study-specific brain template was createdfrom all individual tissue-class images of each subjectusing iterative realignments. Data were then normalizedand warped to an isotropic voxel size of 1.5 mm �1.5 mm � 1.5 mm. Afterward, the data were modulated



by the Jacobian transformed tissue probability maps toobtain ‘‘volume’’ differences rather than ‘‘concentration’’differences in gray matter.24 Finally, data were checkedfor homogeneity across the sample, spatially normalizedto the Montreal Neurologic Institute coordinate referencesystem and smoothed with a 10-mm (full width at halfmaximum) isotropic Gaussian kernel.

Voxel-Based Morphometry AnalysisBased on strong a priori hypotheses of altered GMVvolume in four regions associated with anxiety (amyg-dala, hippocampus, ACC, and insula), an independentregion-of-interest (ROI) approach was adopted. Wecorrected for multiple comparisons using small-volume correction (SVC) with a Gaussian random fieldthreshold set at a ¼ 0.05, corrected, and an extent of atleast 10 contiguous voxels. Coordinates are reported inthe Montreal Neurological Institute (MNI) space inmillimeters [x, y, z]. All anatomical ROIs were createdusing standardized anatomical criteria25,26 on the MNItemplate, which were then applied to the normalizedbrain, at the group level. The ACC was hand traced bytwo experts with 100% agreement, whereas the amyg-dala and hippocampal ROI were drawn using a semi-automated computerized system, with intraclass corre-lations between two operators reaching 0.80 to 0.97.25

Finally, the insula ROI was created using the AnatomicalAutomatic Labeling (AAL) system.26 The main analysisconsisted of a full-factorial 2 � 2 Multivariate Analysisof Covariance (MANCOVA) with Group (healthy ver-sus patients) and BDNF Genotype (Val/Val versus Met)as the between-subject factors, and with age, IQ, scansequence, and whole-brain volume as covariates ofnuisance. Because of potential problems of nonstatio-narity in voxel-based morphometry analyses, use ofcluster-size p values is not recommended, and indivi-dual peak voxels are reported instead.24 These signifi-cant (p o .05, corrected) peak voxels were extracted andfurther analyzed with SPSS v.19 (Armonk, NY). How-ever, to prevent spurious findings associated withoutliers, we screened for and removed all outliers(2 standard deviations above or below the mean, range1–4 outliers) from the analyses. To account for multipletesting in these follow-up analyses, multiple compar-isons were corrected using the conservative step-downHolm procedure. Thus, only significant findings thatsurvived the additional adjustment for multiple com-parisons are presented. To assess the magnitude of asignificant between-group effect, effect sizes (Cohen’s d)were calculated. To examine diagnostic specificity onthe GMVreductions in the amygdala and hippocampus,a regression analysis was performed, which coded forpresence or absence of a specific anxiety disorder(generalized anxiety disorder [GAD], separation anxietydisorder [SAD], social phobia [SP]). To evaluate thepotential impact of comorbid depression on the find-ings, all analyses were re-run excluding those of the

www.jaacap.org 187

http://www.fil.ion.ucl.ac.uk/spm


MUELLER et al.

subjects with comorbid major depressive disorder(MDD). In a third set of analyses, the potential impactof age, IQ, and sex was examined. To assess the impact ofsymptom severity on the findings, regression analyseswere conducted between the significant voxels andanxiety scores in the patient group.

RESULTSThe MANCOVA revealed a significant main effectof group, and a significant group-by-genotypeinteraction.

Neuroanatomic Differences Between Patients andComparison SubjectsComparison Subjects 4 Patients. Significant maineffects of group revealed reduced GMV inadolescents with anxiety relative to healthycomparison adolescents in the right amygdalawith a large effect size [28, 0, �18] (F1,94 ¼ 9.84,p ¼ .002, d ¼ 0.876) and right anterior hippo-campus with a medium-to-large effect size

FIGURE 1 Main effect of group (comparison subjects 4 pvolume in the amygdala [28 0 18] and anterior hippocampus fobrain-derived neurotrophic factor (BDNF) Val66Met polymorphibrain template. Scatterplots for the amygdala cluster showing siggroup means. Units of y axis are in probability of gray matter

JOUR

188 www.jaacap.org

[27, �4, �18] (F1,95 ¼ 6.65, p ¼ .011, d ¼ 0.752)(Figure 1).

Patients 4 Comparison Subjects. Significant volumeincreases were seen in two clusters in the leftinsula [�46, 12, �14] (F1,94 ¼ 12.15, p ¼ .001,d ¼ 0.203) and [�48, 27, 3] (F1,92 ¼ 4.11, p ¼ .045,d ¼ 0.326) and two clusters in the right insula [52,11, �8] (F1,96 ¼ 8.98, p ¼ .003, d ¼ 0.19) and [40,30, �5] (F1,94 ¼ 8.79, p ¼ .004, d ¼ 0.372), all withsmaller effect sizes than for the medial temporallobe (Figure 2). No main effects in either directionwere found in the ACC. Importantly, the insulaand the ACC revealed significant group-by-genotype interactions.

Diagnostic SpecificityAmygdala and Hippocampus. To investigate towhat extent diagnostic specificity may havecontributed to the amygdala and hippocampalGMV reductions in adolescents with anxietyversus healthy comparison adolescents, post-hoc regression analyses were run on these

atients). Note: Figure displays the decreased gray matterr the patient relative to the comparison group regardless ofsm overlaid on the Montreal Neurological Institute (MNI) T1nificant differences in the post-hoc tests. Black lines indicatetissue per cubic millimeter (mm3). *p o .05.



FIGURE 2 Main effect of group (patients 4 comparison subjects). Note: Figure displays insula gray matter volumeincreases [52 11 �8] in the patients relative to the comparison group overlaid on the Montreal Neurological Institute (MNI)T1 brain template. Scatterplots showing significant differences in the post-hoc tests. Black lines indicate group means. Units ofy axis is in probability of gray matter tissue per cubic millimeter (mm3). *p o .05.


findings coding each type of anxiety disorder(GAD, SAD, SP) as a dichotomous variable(present, absent). Of these, 16 subjects werenon-SP GAD patients, 16 non-GAD SP patients,and six comorbid GAD/SP. Of the SADpatients, seven were comorbid with GAD, threewere comorbid with SP, one had all threecomorbidities, and one had no other diagnoses.These analyses revealed that only diagnosis ofSP was predictive of a smaller right amygdala[28, 0, �18] (t ¼ �2.96, p ¼ .004), with no effectsfor GAD (p ¼ .061) or SAD (p ¼ .859). A similarpicture emerged for the right anteriorhippocampus cluster with a significant effect forSP [27, �4, �18] (t ¼ 2.58, p ¼ .012), and noeffects for GAD (p ¼ .253) or SAD (p ¼ .657).

Insula. To examine to what extent the GMVincreases in the insula were driven by a specificdiagnosis, regression analyses revealed thatonly one cluster in the right insula GMV [40,30, �5] was associated with SP (t ¼ 2.25, p ¼.027) but not SAD (p ¼ .312) or GAD (p ¼ .464).



No other associations with insula volumewere found.

Impact of BDNF Genotype: Group-by-GenotypeInteractionSignificant group-by-genotype interactionsemerged in the right insula and right ACC(Figure 3, Table 2). Within the insula ROI, onecluster was significant in the posterior insula[48, �3, �12] (F3,90 ¼ 5.40, p ¼ .002) and one inthe anterior insula [45, 17, 12] (F3,91 ¼ 3.62, p ¼.016). Within the ACC, a significant clusteremerged in the dorsal ACC (dACC) [3, 30, 27](F3,92¼ 4.24, p¼ .007). To interpret these group-by-genotype interactions, we conducted two sets offollow-up analyses in these two ROIs. First, eachdiagnostic group was analyzed separately; thenanalyses were performed within genotype groups.

Insula ROI. Within the anterior insula cluster[45, 17, 12], Met subjects with anxiety showedsmaller GMV than Val/Val subjects withanxiety (F1,32 ¼ 9.40, p ¼ .004, d ¼ 0.787)

www.jaacap.org 189

FIGURE 3 Interaction of brain-derived neurotrophic factor polymorphism by group. Note: (Upper panel) Significant graymatter volume changes in the dACC [30 3 27] overlaid on T1 MNI template with adjacent scatterplot for the significantinteractions in the post-hoc tests. (Lower panel) Significant gray matter volume changes in the anterior insula [45 17 12]overlaid on T1 Montreal Neurological Institute (MNI) template with adjacent scatterplot, illustrating the significantinteractions in the post-hoc tests. Black lines indicate group means. Units of y axis is in probability of gray matter tissue percubic millimeter (mm3). *p o .0

MUELLER et al.

(Figure 3). In the posterior insula cluster[48, �3, �12], GMV differed significantlybetween Met with anxiety and Val/Val patientswith anxiety (F1,32 ¼ 7.93, p ¼ .008, d ¼ 0.569).When GMV was compared within genotype,healthy Met carriers exhibited greater GMVthan patient Met carriers (F1,37 ¼ 10.53, p ¼.002, d ¼ 0.869).

ACC ROI. GMV was larger in Val/Val homo-zygotes with anxiety when compared withhealthy Val/Val homozygotes in the dorsalACC (F1,52 ¼ 10.77, p ¼ .002, d ¼ 0.455). Therewere no significant between-group effects withinthe Met genotype (Figure 3).

In summary, the interaction of BDNF genotypeand group reflected that, in the group withanxiety, there were smaller anterior and posteriorinsula volumes in the Met carriers relative to Val/Val homozygotes, and in the healthy group therewere no gene-related differences. Furthermore,within genotype, healthy Met allele carriers hadlarger GMV in the posterior insula than didpatients with the Met allele.

JOUR

190 www.jaacap.org

Additional Exploratory AnalysesEffects of Age, IQ, or Sex, and Anxiety SymptomSeverity. To examine the potential influence ofage, sex, IQ, or symptom severity on the currentfindings, main effects as well as interactions ofthese variables were examined in relation togroup status and BDNF genotype. Nosignificant main effects of any of these variableswere detected (all p 4 .05). Importantly, nointeractions with group or genotype weresignificant (all p 4 .05). Finally, regressionanalyses did not reveal a significant impact ofseverity of anxiety on the findings (all p 4 .05).

Psychiatric Comorbidity with MDD. To examinewhether findings were driven by the patientswith comorbid MDD (n ¼ 8), the data werereanalyzed for all ROIs without these subjects.Findings remained significant for allcomparisons, that is, group effects on amygdala(F1,87 ¼ 7.03, p ¼ .010), hippocampus (F1,88 ¼

4.64, p ¼ .034), and insula (all clusters:F1,84 4 5.91, all p o .017); and gene-by-group



TABLE 2 Statistical Parametric Mapping Results of Gray Matter Volume (GMV) Differences in Adolescents With AnxietyVersus Healthy Comparison Adolescents, by Main Effect of Group and Brain-Derived Neurotrophic Factor (BDNF)Genotype

ClusterSize (mm3)

MNI coordinates Talairach coordinates

Region Side t Value z Value x y z x y z

Main effect of groupComparison subjects 4 patients

Amygdala R 243 2.20 2.17 28 0 �18 28 �1 �152.07 2.05 25 �9 �8 25 �9 �6

Anterior hippocampus R 48 1.89 1.88 27 �4 �18 27 �5 �151.81 1.80 22 �7 �14 22 �7 �11

Patients 4 comparison subjectsInsula L 219 3.04 2.96 �46 12 �14 �46 11 �12

L 31 2.09 2.06 �48 27 3 �48 26 1R 438 3.00 2.93 52 11 �8 51 10 �7R 256 2.98 2.91 40 30 �5 40 29 �6

Interaction (Val [comparison subjects – patients] –Met [comparison subjects – patients])Dorsal ACC R 238 2.65 2.60 3 30 27 3 30 23Posterior insula R 92 3.90 3.75 48 �3 �12 48 �3 �10

3.16 3.07 46 �9 �8 46 �9 �62.46 2.42 45 �15 �6 45 �15 �4

Anterior insula R 135 3.21 3.11 45 17 12 45 17 102.46 2.42 42 20 13 42 20 112.35 2.32 43 27 4 43 26 2

Note: The printout was obtained using SPM8 (Wellcome Department of Cognitive Neurology, London, England; http://www.fil.ion.ucl.ac.uk/spm), smallvolume correction (SVC) applied for multiple comparisons at p o.05. Montreal Neurological Institute (MNI) coordinates were converted to Talairachcoordinates using the Wake Forest University (WFU) Pickatlas tool. ACC ¼ anterior cingulate cortex; L ¼ left; R ¼ right.


effects for all regions (all F3,82 4 5.56, allp o .010). This suggests that comorbiddepression did not significantly impact thefindings.

DISCUSSIONThis structural MRI study of adolescents withanxiety and healthy comparison adolescentsexamined GMV in four regions (insula, ACC,amygdala, and hippocampus), all previouslyimplicated in anxiety disorders. In addition,preliminary evidence for a role of the BDNFVal66Met polymorphism in the modulation ofanxiety-related structural differences wasexplored. Four main findings emerged. First,GMVs of the right amygdala and right anteriorhippocampus were smaller in patients relativeto comparison subjects, regardless of genotype.Second, additional regression analyses sug-gested some specificity of these GMV volumealterations to social phobia. Third, in contrast tothe amygdala and hippocampus findings, insula



volume was larger in the patient group relativeto the comparison group. Fourth, right-lateralized dACC and insula volume appearedto be modulated by the BDNF Val66Met poly-morphism differentially in adolescents withanxiety and in comparison subjects. GMVs ofthe insula and ACC were smaller in the riskMet allele carriers relative to the Val/Val homo-zygotes in the patient group, whereas thereverse effect was observed in the comparisongroup.

To date, the few studies examining GMVabnormalities in pediatric anxiety have reportedmixed results showing either GMV increases1 ordecreases3 in patient groups relative to compar-ison groups. Moreover, these studies have alsodiffered in methodology, using either voxel-basedmorphometry3 or a manual tracing approach1 tomeasure volumetric changes. The first mainfinding of the present study was the reducedvolumes of the amygdala and hippocampus in thegroup with anxiety relative to the comparisongroup. These data are consistent with, and form

www.jaacap.org 191


MUELLER et al.

an independent replication of, an earlier study in asmaller group of clinically anxious adolescents(N ¼ 17; mean age 13 years),3 extending priorwork in adults.27–29 The current data also demon-strate reduced GMV of the hippocampus, inaddition to the amygdala, a finding not reportedpreviously in adolescents. One possible factor thatmay have contributed to the mixed findings inprevious studies is the potential impact of diag-nostic specificity on GMV alterations. Indeed,whereas amygdala GMV reductions were foundin a mixed sample of pediatric anxiety disorders(GAD, SAD, SP),3 amygdala and superior tem-poral gyrus volumetric increases were found inpediatric posttraumatic stress disorder30 andGAD.1,2 When diagnostic specificity was exam-ined in relation to the current findings, the datasuggested a specific relationship of the amygdalaand hippocampal GMV reductions with SP butnot with SAD or GAD. Although these findingsshould be considered tentative, they suggest acontribution of SP to GMV reductions in theamygdala and hippocampus, at least in adoles-cents, and warrant replication in much largersamples of individuals with specific anxietydisorders.

The mechanisms that might produce structuralabnormalities in the present study remain unclear.However, prior findings of functional changes inadolescent mood and anxiety disorders mightprovide clues.31,32 For example, a recent study inadolescent bipolar disorder reported an inverserelationship between amygdala function duringemotion processing and amygdala structure.33

Such findings are consistent with the suggestionthat over-activation of the amygdala reflectsglutamatergic excitotoxicity, which would leadto volume loss over time.34 However, no directevidence of excitotoxicity exists in availableresearch on pediatric anxiety disorders, and morework is needed in this area. Moreover, the existingwork focuses selectively on the amygdala, raisingquestions as to the degree to which similar ordifferent mechanisms might also explain anxiety-related increases in gray matter volume in otherregions.2,35 Enhanced interoceptive awarenesshas been postulated in anxiety6 and is a keyfunction of the anterior insula.36 Underlyingmechanisms of such disturbance in interoceptiveawareness could manifest as increases in insulavolume. Of note, current findings in the amygdalaand hippocampus were right lateralized. Such aneffect is certainly consistent with prior reports of

JOUR

192 www.jaacap.org

structural1 and functional32 perturbations in theright but not the left amygdala associated withanxiety. Although left-lateral alterations in thecurrent study were present at a sub-thresholdlevel, the current data are also in line withprevious reports of larger perturbations involume on the right relative to the left side inother structures in pediatric anxiety.2 These find-ings lend further support to laterality theories ofemotion, which posit a special role of the rightside of the brain in the processing of negativeemotions.37

The third main finding suggests that BDNFgenotype modulates GMV in the dorsal portionof the ACC (dACC) and insula differently inclinically anxious and healthy youth. Specifically,the regional selectivity of the dACC and insulasuggest these regions may be particularly sensitiveto the action of BDNF in adolescence, in a way thatmay harbor risk for anxiety. Regarding the dACC,studies in several cohorts of anxiety disorders (SP,specific phobia, GAD) document abnormalities indACC function.32,38,39 A previous pediatric func-tional MRI study reports elevated responses in thisregion [Talairach: 4 6 40] to fearful expression inyouths with GAD relative to comparisons.32 Inaddition, adult patients with SP show abnormaldACC responses to negative social conditions, suchas negative comments38 [Talairach: �13 3 44] or inanticipation of public speaking [Talairach: �24 24 24or �20 4 28].39 Finally, dACC response [Talairach: 016 47 (for easier comparison, MNI coordinates wereconverted to Talairach coordinates using WakeForest University Pickatlas)] to implicit fear ismodulated by BDNF polymorphism, with greateractivation in Met carriers.40 Our findings resonatewith these studies, and provide further support fora role of dACC in anxiety disorders in adolescence,as well as a potential to be uniquely modulated bythe BDNF Val66Met polymorphism in adolescentswith anxiety.

The anterior insula is considered to be a primarycontributor to anxiety in adults.6 By comparison,functional neuroimaging studies of anxiety inadolescents have, to date, focused primarily onthe amygdala.15 The present finding further high-lights the relevance of extending this focus to theinsula in pediatric studies. It is noteworthy that theVal66Met polymorphism modulated the volume ofboth the anterior and posterior insula regions inadolescents with anxiety but not in healthy com-parison adolescents. This functional link betweenthe two insula regions is particularly relevant to




anxiety, based on the hypersensitivity to bodilysignals of individuals with anxiety.6 The posteriorinsula integrates interoceptive signals,41,42 which inturn are carried to the anterior insula for integra-tion, a process that may be excessively reinforced inanxiety.36

Collectively, these preliminary findings impli-cate BDNF in modulation of dACC and insulaGMV in adolescents with anxiety. Such findingsraise questions regarding the specific aspects ofanxiety that are influenced by this genetic mod-ulation, and the extent to which this modulation isspecific to the adolescent period. Such geneticeffect might be exploited therapeutically in thefuture, given the recent evidence that a variant ofthe BDNF gene is associated with improvedtherapeutic outcome with serotonin reuptakeinhibitors in adult patients with anxiety.43

Finally, this pediatric study may inform ondevelopmental aspects of gene–brain associations.Our analyses failed to detect a genotypic modula-tion of amygdala or hippocampal volumes inhealthy comparison adolescents, in contrast toprevious work in healthy adults.8,13 Yet, thisdiscrepant finding is congruent with the studyby Toro et al.18 who found no significant impact ofthe Val66Met polymorphism on amygdala orhippocampal volume in 331 healthy adolescents.Similarly, Casey et al.44 reported, in healthy chil-dren, the absence of BDNF genotypic effect onamygdala volume, but reported a trend towardsmaller hippocampal volume in the Met allelecarriers compared with the Val/Val homozygotes.Taken together, these studies suggest that themorphometry of the amygdala and hippocampusmight start to be differentially affected by BDNFgene variants only in adulthood. This matura-tional change might be related to the ontogenictrajectory of BDNF levels,35 or to changes incellular receptivity to BDNF action with age.The fact that BDNF genotype affects brain struc-ture in adolescents with psychopathology but notin healthy youths may be related to neurochemicalchanges associated with the pathology. For exam-ple, perturbations in BDNF levels may contributeto structural abnormalities in psychopathology byinfluencing critical protein (reelin) or receptor(TrkB) activity fluctuations in anxiety that furtherinfluence brain developmental processes such asneuronal differentiation or cortical lamination.12

As a result of altered BDNF expression, possiblybecause of change in glutamate signaling,12 indi-viduals with anxiety or other psychopathology



would then demonstrate perturbed structuralneural development during the adolescent period.More research is needed to determine the influ-ence of these molecular processes on brain matura-tion. However, given the absence of findings inseveral independent studies,18,44 an alternativeexplanation for the absence of an influence ofthe Val66Met polymorphism on healthy youthscould be the potential presence of false-positiveresults in this prior line of work in adults. In anycase, this study stresses the importance of under-standing the functional and clinical significance ofthe developmental effect of BDNF variant on brainmorphometry, particularly in relation to the onsetof psychiatric disorders.

Some limitations require consideration. Onelimitation concerns the relatively small numberof subjects in the patient BDNF Met group (N ¼17). This limitation is mitigated by a few factors.Importantly, a power analysis revealed an accep-table level of power to detect reliable differencesin the group-by-genotype interaction. In addi-tion, the current study includes a relatively largenumber of patients, coupled with the fact that allpatients were seeking treatment for acute, severeanxiety symptoms, and none were on medica-tions. Medication-free, acutely ill subjects aredifficult to recruit. In fact, small sample sizeshave been the rule in prior research with suchsamples. Second, candidate gene studies need tobe interpreted with caution. The associationbetween a disorder and a gene is complexbecause of a variety of possible influences,including gene–gene and gene–environmentinteractions. Thus, replication in large samplesis necessary to confirm findings and to gaininsight into the contribution of a gene to psy-chopathology. A third limitation relates to theuse of two different scanning acquisitionsequences. This concern is moderated by thefacts that all groups were well matched on scansequence, and that findings did not differbetween scan sequences. Finally, we did notassess parental psychopathology, which wasbeyond the scope of the current study. However,future studies should include such measures, asthey could inform neurobiological predictors ofat-risk individuals.

In summary, the current findings suggest thatGMV reductions in the amygdala and anteriorhippocampus are already present in adolescentswith an anxiety disorder. Importantly, this is thefirst study to show in adolescents, that dACC and

www.jaacap.org 193

MUELLER et al.

insula volume are modulated by the BDNFVal66Met polymorphism, thus extending priorstructural work in adults8,13 and functional workin adolescents.15 &

19

Clinical Guidance

� Few studies exist on gray matter volume changes inadolescent anxiety, and genetic impact has notbeen examined� BDNF is critically involved in developmental

processes and has been linked to mood andanxiety disorders in adults� Gray matter volume of patients with anxiety

disorders was reduced in amygdala and anteriorhippocampus relative to comparison adolescents,whereas insula volume was slightly increased� However, BDNF genotype modulated insula

volume, which showed a smaller volume inindividuals with anxiety with the Met allele relativeto Val/Val homozygous patients

4
JOURNA
www.jaacap.org

Accepted November 21, 2012.

This article was reviewed under and accepted by ad hoc editor GuidoFrank, M.D.

Drs. Mueller, Pine, and Ernst are with the Section on Development andAffective Neuroscience at the National Institute of Mental Health(NIMH). Dr. Mueller is also with the University of Ghent, Belgium.Dr. Aouidad is with the Pierre and Marie Curie Faculty of Medicine,Sorbonne Universite, Paris. Dr. Gorodetsky is with the Mood and AnxietyDisorders Program (NIMH). Dr. Goldman is with the Laboratory ofNeurogenetics at the National Institute on Alcohol Abuse and Alcoholism(NIAAA).

The study was funded by the intramural program of the NIMH and anNIMH Richard J. Wyatt Memorial Fellowship (S.C.M.).

The authors acknowledge the assistance of Colin Hodgkinson, Pei-Hong Shen, and Longina Akhtar with NIAAA.

Disclosure: Drs. Mueller, Aouidad, Gorodetsky, Goldman, Pine, andErnst report no biomedical financial interests or potential conflicts ofinterest.

Correspondence to Sven C. Mueller, Ph.D., Department ofExperimental Clinical and Health Psychology, Ghent University,Henri Dunantlaan 2, B-9000, Ghent, Belgium; e-mail:[email protected]

0890-8567/$36.00/C 2013 American Academy of Child andAdolescent Psychiatry

http://dx.doi.org/10.1016/j.jaac.2012.11.016

REFERENCES
1. De Bellis MD, Casey BJ, Dahl RE, et al. A pilot study of amygdala
volumes in pediatric generalized anxiety disorder. Biol Psychiatry.

2000;48:51-57.2. De Bellis MD, Keshavan MS, Shifflett H, et al. Superior temporal

gyrus volumes in pediatric generalized anxiety disorder. Biol

Psychiatry. 2002;51:553-562.3. Milham MP, Nugent AC, Drevets WC, et al. Selective reduction in

amygdala volume in pediatric anxiety disorders: a voxel-basedmorphometry investigation. Biol Psychiatry. 2005;57:961-966.

4. Ferrari MC, Busatto GF, McGuire PK, Crippa JA.

Structural magnetic resonance imaging in anxiety disorders:an update of research findings. Rev Bras Psiquiatr. 2008;

3:251-264.5. Shin LM, Liberzon I. The neurocircuitry of fear, stress, and

anxiety disorders. Neuropsychopharmacology. 2010;35:

169-191.6. Paulus MP, Stein MB. An insular view of anxiety. Biol Psychiatry.

2006;60:383-387.7. Chen ZY, Jing D, Bath KG, et al. Genetic variant BDNF (Val66Met)

polymorphism alters anxiety-related behavior. Science. 2006;

314:140-143.8. Bueller JA, Aftab M, Sen S, Gomez-Hassan D, Burmeister M,

Zubieta JK. BDNF Val66Met allele is associated with reduced

hippocampal volume in healthy subjects. Biol Psychiatry. 2006;59:812-815.

9. Egan M, Kojima M, Callicott J, et al. The BDNF val66met

polymorphism affects activity-dependent secretion of BDNFand human memory and hippocampal function. Cell. 2003;

112:257-269.10. Huang EJ, Reichardt LF. Neurotrophins: roles in neuronal devel-

opment and function. Annu Rev Neurosci. 2001;24:677-736.11. Gallinat J, Schubert F, Bruhl R, et al. Met carriers of BDNF Val66Met

genotype show increased N-acetylaspartate concentration in the

anterior cingulate cortex. Neuroimage. 2010;49:767-771.12. Takahashi M, Shirakawa O, Toyooka K, et al. Abnormal expression

of brain-derived neurotrophic factor and its receptor in the

corticolimbic system of schizophrenic patients. Mol Psychiatry.2000;5:293-300.

13. Montag C, Weber B, Fliessbach K, Elger C, Reuter M. The BDNF

Val66Met polymorphism impacts parahippocampal and amyg-

dala volume in healthy humans: incremental support for a genetic

risk factor for depression. Psychol Med. 2009;39:1831-1839.14. Montag C, Weber B, Jentgens E, Elger C, Reuter M. An epistasis

effect of functional variants on the BDNF and DRD2 genes

modulates gray matter volume of the anterior cingulate cortex

in healthy humans. Neuropsychologia. 2010;48:1016-1021.15. Lau JY, Goldman D, Buzas B, et al. BDNF gene polymorphism

(Val66Met) predicts amygdala and anterior hippocampus

responses to emotional faces in anxious and depressed adoles-

cents. Neuroimage. 2010;53:952-961.16. Pine DS, Ernst M, Leibenluft E. Imaging-genetics applications in

child psychiatry. J Am Acad Child Adolesc Psychiatry. 2010;

49:772-782.17. Katoh-Semba R, Takeuchi IK, Semba R, Kato K. Distribution of

brain-derived neurotrophic factor in rats and its changes with

development in the brain. J Neurochem. 1997;69:34-42.18. Toro R, Chupin M, Garnero L, et al. Brain volumes and Val66Met

polymorphism of the BDNF gene: local or global effects? Brain

Struct Funct. 2009;213:501-509.19. Kaufman J, Birmaher B, Brent D, et al. Schedule for Affective

Disorders and Schizophrenia for School-Age Children—Present

and Lifetime Version (K-SADS-PL): initial reliability and validity

data. J Am Acad Child Adolesc Psychiatry. 1997;36:980-988.20. Wechsler D. Wechsler Abbreviated Scale of Intelligence (WASI).

San Antonio, TX: Harcourt Assessment; 1999.21. Faul. F, Erdfelder E, Lang A-G, Buchner A. G*Power 3: A flexible

statistical power analysis program for the social, behavioral, and

biomedical sciences. Behav Res Methods. 2007;39:175-191.22. Hodgkinson CA, Yuan Q, Xu K, Shen PH, et al. Addictions biology:

haplotype-based analysis for 130 candidate genes on a single array.

Alcohol Alcohol. 2008;43:505-515.23. Ashburner J. A fast diffeomorphic image registration algorithm.

Neuroimage. 2007;38:95-113.24. Ashburner J, Friston KJ. Voxel-based morphometry—the methods.

Neuroimage. 2000;11:805-821.

L OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


mailto:[email protected]

dx.doi.org/10.1016/j.jaac.2012.11.016

dx.doi.org/10.1016/j.jaac.2012.11.016

dx.doi.org/10.1016/j.jaac.2012.11.016


25. Szeszko PR, Robinson D, Alvir JM, et al. Orbital frontal andamygdala volume reductions in obsessive-compulsive disorder.Arch Gen Psychiatry. 1999;56:913-919.

26. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Auto-mated anatomical labeling of activations in SPM using a macro-scopic anatomical parcellation of the MNI MRI single-subjectbrain. Neuroimage. 2002;15:273-289.

27. Massana G, Serra-Grabulosa JM, Salgado-Pineda P, et al. Para-hippocampal gray matter density in panic disorder: a voxel-basedmorphometric study. Am J Psychiatry. 2003;160:566-568.

28. Massana G, Serra-Grabulosa JM, Salgado-Pineda P, et al. Amyg-dalar atrophy in panic disorder patients detected by volumetricmagnetic resonance imaging. Neuroimage. 2003;19:80-90.

29. Wignall EL, Dickson JM, Vaughan P, et al. Smaller hippocampalvolume in patients with recent-onset posttraumatic stress disorder.Biol Psychiatry. 2004;56:832-836.

30. De Bellis MD, Keshavan MS, Frustaci K, et al. Superior temporalgyrus volumes in maltreated children and adolescents with PTSD.Biol Psychiatry. 2002;51:544-552.

31. Beesdo K, Lau JY, Guyer AE, et al. Common and distinct amygdala-function perturbations in depressed vs anxious adolescents. ArchGen Psychiatry. 2009;66:275-285.

32. McClure EB, Monk CS, Nelson EE, et al. Abnormal attentionmodulation of fear circuit function in pediatric generalized anxietydisorder. Arch Gen Psychiatry. 2007;64:97-106.

33. Kalmar JH, Wang F, Chepenik LG, et al. Relation betweenamygdala structure and function in adolescents with bipolardisorder. J Am Acad Child Adolesc Psychiatry. 2009;48:636-642.

34. Siegle GJ, Konecky RO, Thase ME, Carter CS. Relationshipsbetween amygdala volume and activity during emotional infor-mation processing tasks in depressed and never-depressed indi-viduals: an fMRI investigation. Ann N Y Acad Sci. 2003;985:481-484.



35. Kuhn S, Schubert F, Gallinat J. Structural correlates of trait anxiety:

reduced thickness in medial orbitofrontal cortex accompanied by

volume increase in nucleus accumbens. J Affect Disord. 2011;

134:315-319.36. Craig A. How do you feel? Interoception: the sense of the

physiological condition of the body Nat Rev Neurosci. 2002;

3:655-666.37. Davidson RJ. Cerebral asymmetry, emotion, and affective style. In:

Davidson RJ, Hughdahl K, eds. Brain Asymmetry. Cambridge,

MA: MIT Press; 1995:361-387.38. Blair K, Geraci M, Devido J, et al. Neural response to self- and other

referential praise and criticism in generalized social phobia. Arch

Gen Psychiatry. 2008;65:1176-1184.39. Lorberbaum JP, Kose S, Johnson MR, et al. Neural correlates of

speech anticipatory anxiety in generalized social phobia. Neurore-

port. 2004;15:2701-2705.40. Mukherjee P, Whalley HC, McKirdy JW, et al. Effects of the BDNF

Val66Met polymorphism on neural responses to facial emotion.

Psychiatry Res. 2011;19:182-188.41. Mesulam M-M, Mufson EJ. Insula of the Old World monkey. III:

Efferent cortical output and comments on function. J Compar

Neurol. 1982;212:38-52.42. Shi C-J, Cassell MD. Cascade projections from somatosensory

cortex to the rat basolateral amygdala via the parietal insular

cortex. J Compar Neurol. 1998;399:469-491.43. Real E, Gratacos M, Soria V, et al. A brain-derived neurotrophic

factor haplotype is associated with therapeutic response in

obsessive-compulsive disorder. Biol Psychiatry. 2009;66:674-680.44. Casey BJ, Glatt CE, Tottenham N, et al. Brain-derived neuro-

trophic factor as a model system for examining gene by

environment interactions across development. Neuroscience.

2009;164:108-120.

www.jaacap.org 195

gray matter volume in adolescent anxiety: an impact of the brain-derived neurotrophic factor...

Documents