Etkin Et Al 2009 Arch Gen Psychiatry

8/6/2019 Etkin Et Al 2009 Arch Gen Psychiatry.

http://slidepdf.com/reader/full/etkin-et-al-2009-arch-gen-psychiatry 1/12

ORIGINAL ARTICLE

Disrupted Amygdalar Subregion FunctionalConnectivity and Evidence of a CompensatoryNetwork in Generalized Anxiety Disorder Amit Etkin, MD, PhD; Katherine E. Prater, BA; Alan F. Schatzberg, MD; Vinod Menon, PhD; Michael D. Greicius, MD

Context : Little is known about the neural abnormali-tiesunderlying generalized anxiety disorder (GAD). Stud-ies in other anxiety disorders have implicated the amyg-dala, but work in GAD has yielded conflicting results.Theamygdala is composed of distinctsubregions that in-teract with dissociable brain networks,which have beenstudied only in experimental animals. A functional con-nectivity approach at the subregional level may there-fore yield novel insights into GAD.

Objectives : To determinewhether distinct connectivitypatternscan bereliablyidentifiedforthebasolateral (BLA)andcentromedial (CMA) subregions of thehuman amyg-dala,andto examinesubregionalconnectivitypatterns andpotential compensatory amygdalar connectivity in GAD.

Design : Cross-sectional study.

Setting : Academic medical center.

Participants : Two cohorts of healthy control subjects(consisting of17and31 subjects) and16patientswithGAD.

Main Outcome Measures : Functional connectivitywith

cytoarchitectonically determined BLA and CMA regionsof interest, measured during functional magnetic reso-

nance imaging performedwhilesubjects were resting qui-etly in the scanner. Amygdalar gray matter volume wasalso investigated with voxel-based morphometry.Results : Reproducible subregional differences in large-scaleconnectivitywereidentifiedinboth cohortsof healthycontrols. The BLA was differentially connected with pri-mary andhigher-ordersensory andmedial prefrontal cor-tices. The CMA was connected with the midbrain, thala-mus, and cerebellum. In GAD patients, BLA and CMAconnectivity patterns were significantly less distinct, andincreased gray matter volume was noted primarily in theCMA. Across the subregions, GADpatients hadincreasedconnectivity with a previously characterized frontopari-etal executive control network and decreased connectiv-ity with an insula- and cingulate-based salience network.Conclusions : Ourfindings providenew insights into thefunctionalneuroanatomyof thehumanamygdala andcon-verge with connectivity studies in experimental ani-mals. In GAD, we find evidence of an intra-amygdalarabnormality and engagement of a compensatory fronto-parietal executive controlnetwork, consistent with cog-nitive theories of GAD.

Arch Gen Psychiatry. 2009;66(12):1361-1372

G ENERALIZED ANXIETY DIS-order (GAD) is a com-mon anxiety disorder inadults, with estimatedlifetime prevalencerates

of around 5%.1,2 Generalized anxiety dis-order has a longaverage duration ofsymp-toms2,3 and is associated with significantquality-of-life impairment or disability.4

Despite its clinical importance, GAD has

receivedconsiderablylessstudy thanotheranxiety disorders. Predictions of whichbrain circuits are altered in GAD musttherefore be extrapolatedfromfindingsinother anxiety disorders. A recent meta-analysis from ourgroup5 showed thatpost-traumatic stress disorder, social anxietydisorder, and specific phobia—3 of themost frequentlystudied disorders—areallcharacterized by hyperactivityof theamyg-dala andinsula of patientsduring thepro-

cessing of negative emotion. Engage-ment of the amygdala and insula was alsoseen in this meta-analysis when healthysubjects experiencedfear,5 suggesting thatamygdalar and insular hyperactivation inpatients reflects the neural correlates of acurrent fear or anxiety state or a trait vul-nerability for excessive fear responses.

Two recent studies examined amyg-dala activity in adult patients with GADin

response to viewing fearful faces, a com-monly used probe of amygdalar engage-ment.6,7 One study found no difference inthe amygdala between patients and con-trol subjects,7 whereas the other studyfound less activity in the amygdala in pa-tients.6 By contrast, similar approaches inadolescents with GAD have found hyper-activityintheamygdala.8,9Manyfactorsmayexplain thesedifferences, including differ-ences inbaselineamygdalaractivity, theca-

Author Affiliations:Department of Psychiatry andBehavioral Sciences (Drs Etkin,Schatzberg, and Menon andMs Prater), Program inNeuroscience (Drs Menon andGreicius), and Department of Neurology and NeurologicalSciences (Dr Greicius), StanfordUniversity School of Medicine,Stanford, California.

(REPRINTED) ARCH GEN PSYCHIATRY/VOL 66 (NO. 12), DEC 2009 WWW.ARCHGENPSYCHIATRY.COM1361

©2009 American Medical Association. All rights reserved. on May 5, 2011www.archgenpsychiatry.comDownloaded from

http://www.archgenpsychiatry.com/







pacity for task-relatedactivation,overallresponses to emo-tionalexpressions, andthe natureand relativelevels ofGADpsychopathology ineach cohort.Nonetheless,these stud-ies leave undefined the role the amygdala plays in GAD.Blair et al6 used their own data and those of related stud-

iesand suggestedthat anunderstanding ofamygdalardys-function in GAD,along withdiscrepancies betweenindi-vidual studies, will require investigation of the brainnetwork contexts in which the amygdala is involved.

Analyses of functional connectivity using functionalmagnetic resonance imaging (fMRI) data acquired whilethe subjects are at rest, unbiased by task demands, havebeen previously used to identify multiple, simultaneous-ly operating, broadly connected networks of brain re-gions.10 Abnormalities in resting functional connectivityhave also been identified in major depression.11 Use of aresting-state approach maythereforeprovide a usefultoolfor examining dysfunctional amygdalar-based neural cir-cuitry in GAD.

Our understanding of the organization of the amyg-dala isderived inlargepart from investigationsof thefunc-tionsof subnuclei within the amygdala ofanimals and thedistinctbrainnetworks inwhich they participate.This ex-tensive body of research hasestablished a model whereinsensoryinformationacrossmultiplemodalitiesenters theamygdala through the nuclei of the basolateral complex(BLA) (consisting of lateral, basal, and accessory basalnuclei).12-16Neuronsin the BLA encodefearmemories re-lated to these sensory stimuli, signal the threat value of astimulus, and can modulate memory encoding and sen-sory processing inotherbrainregions.13,14TheBLAinturnactivates thecentralnucleus, which isessentialfor theba-sic species-specific defensive responses associated with

fear.12-16

The central nucleus achieves these functionsthroughprojectionstobrainstem, hypothalamic, andbasalforebrain targets.12-16

Central to this outlined circuit areanatomicalconnec-tivity findings in rodentsandnonhuman primates,whichdifferentiatethe largely corticalconnectivity pattern of theBLA from the largely subcortical connectivity pattern of the central nucleus.12,15,16 The anatomical connectivity of human amygdalar nuclei, however, is currently un-known. In this study, we therefore examined the differ-ential connectivity patterns of these amygdalar subre-

gions in healthy control subjects and GAD patients toinvestigate the functional brain networks in which theamygdala is involved and that underlie the distinct func-tions of these amygdalar subregions.

Wefirstsoughttoestablish whether thefunctionalcon-

nectivity of amygdalar subregions can be distinguishedin healthy subjects by using resting-state fMRI. Cytoar-chitectural, myeloarchitectural, and chemoarchitec-tural studies of the human amygdala support a differen-tiation between the BLA and a centromedial subregion(CMA) composed of the central andmedialnuclei.17 Wetherefore derived regions of interest (ROIs) from cyto-architectonicallydetermined probabilisticmapsof the hu-man BLA and CMA.18,19 To demonstrate the reliability of this approach, we derived functional connectivity mapsfrom 2 independent control cohorts. We then com-pared these functional connectivity patterns with thosein a matched GAD patient cohort. Finally, we comple-mented our functional connectivityanalysesby perform-

ing voxel-based morphometry (VBM), an independentand commonlyusedstructural imaging approach, focus-ing on gray matter volumes of the BLA and CMA.

METHODS

PARTICIPANTS

A total of 64 subjectsparticipated in this study (seeTable 1 fordemographics) after giving their informedconsent accordingtoinstitutional guidelines for the protection of human subjects atStanfordUniversity.Thefirstcontrol cohort(GAD controls)andtheGADpatientswere recruited throughlocal online advertise-ments. Psychiatric diagnoses based on theDSM-IV 20 weredetermined through an informal clinical interview with a psy-

chiatrist and the structured diagnostic Mini International Neu-ropsychiatricInterview.21,22 Exclusioncriteria werebipolar, psy-chotic, or posttraumatic stress disorders or substance abusediagnoses. Generalized anxiety disorder was the primary diag-nosis for all patients. Patients with comorbid major depressionwere includedif theonsetof GAD wasclearly primary to that of depression. Other exclusioncriteriaincludeda historyof a neu-rological disorder, head trauma or loss of consciousness, claus-trophobia, and regular use of benzodiazepines, opiates, or thy-roid medications. Of the 16 patients, 4 were taking regularantidepressants at a stable dose. No subject used an as-neededdose ofa benzodiazepinewithin48hoursof the scan.Comorbid

Table 1. Subject Demographics a

Study Cohorts

ComparisonsC1

(n=17)P

(n=16)C2

(n=31)

Age, y 32.5 (2.0) 30.6 (1.7) 20.5 (0.2) C1 C2b; C1 vs P, P =.15Education, y 17.5 (0.5) 17.1 (0.6) 14.3 (0.2) C1 C2b; C1 vs P, P =.34Female, % 15 (88) 14 (88) 13 (42) C1 C2c; C1 vs P, P .99Right-handed, % 17 (100) 16 (100) 31 (100) P .99

STAI-T 30.5 (1.3) 56.1 (2.6) NA C1 PbPSWQ 37.1 (2.3) 62.7 (2.3) NA C1 Pb

BAI 3.8 (1.1) 23.9 (3.2) NA C1 Pb

BDI-II 2 (0.6) 23.4 (3.3) NA C1 Pb

Abbreviations: BAI, Beck Anxiety Inventory; BDI-II, Beck Depression Inventory II; C1, generalized anxiety disorder (GAD) control cohort; C2, second controlcohort; NA, not acquired; P, GAD patients; PSWQ, Penn State Worry Questionnaire; STAI-T, Spielberger State-Trait Anxiety Inventory.

a Unless otherwise indicated, data are expressed as mean (SEM).bP .001.c P =.001.










AxisI conditions includedmajor depression (n= 4), social anxi-etydisorder (n=5), panic disorder (n=2), dysthymia (n=2), andobsessive-compulsivedisorder (n=1). Of the patients, 6 hadnocomorbidities, 6 had1 comorbidity (depressionin 3, dysthymiain 2, and social anxiety in 1), and 4 had 2 comorbidities (socialanxiety and panic disorder in 2, social anxiety and obsessive-compulsive disorder in 1, and depression and social anxiety in1).Nonehad more than 2 comorbidities.All GAD controlswerefree ofanycurrentor pastAxis I condition andpsychiatric medi-cation. No GAD patient had undergone structured psycho-

therapy.AllGADcontrolsandGADpatientscompletedtheSpiel-berger State-Trait Anxiety Inventory,23 the Penn State WorryQuestionnaire,24 theBeckAnxietyInventory,25 theBeckDepres-sionInventory II,26 andthe Mood andAnxietySymptoms Ques-tionnaire.27,28 The 31 healthy subjects in the second control co-hort were recruited as part of an unrelated study and denied ahistory of psychiatric or neurological disorders but did not un-dergo formal evaluation with a structured interview. Data from22 ofthesesubjects were used previouslyin anunrelatedstudy.29

fMRI DATA ACQUISITION

All participants underwent an 8-minute resting-state fMRI scanin which they were told to keep their eyes closed, hold still, trynot to fall asleep, and allow theirminds to wander. Images wereacquiredona 3-Tscanner(GESignascanner; GEHealthcare,Mil-waukee, Wisconsin) using a custom-built headcoil. A total of 29axial slices (4.0-mm thickness) covering the whole brain wereacquired using a T2-weighted gradient-echo spiral-pulse se-quence (repetition time, 2000 milliseconds; echotime, 30 milli-seconds; flip angle,80°; and1 interleave).30 The fieldof view was22 cmfor the GAD controlsand GAD patientsand20 cmfor thesecond control cohort, and the matrix size was 6464. To re-duce blurringandsignal lossarising from field inhomogeneities,an automated high-order shimming method based on spiral ac-quisitions was used before acquiring fMRI scans.31 A high-resolution T1-weighted spoiled grass gradient-recalled invertedrecovery 3-dimensional MRIsequence wasused with thefollow-ingparameters: time after inversionpulse,300milliseconds;rep-etitiontime, 8 milliseconds;echotime,3.6milliseconds; flipangle,15°;fieldofview,22cm;124slicesinthecoronalplane;256192matrix; number of excitations, 2; and acquired resolution,1.5 0.9 1.1 mm. The images were reconstructed as a124 256 256 matrix with a 1.5 0.9 0.9-mm spatial reso-lution.Structuralandfunctionalimageswere acquiredin thesamescan session.

fMRI DATA ANALYSIS

Preprocessing

The first 8 volumes were not analyzed to allow for signal equili-bration effects. A linear shim correction was applied separatelyfor each slice during reconstruction using a magnetic field mapacquired automaticallyby thepulsesequenceat thebeginningof the scan.30 The fMRI data were then preprocessed using SPM5

software (available at: http://www.fil.ion.ucl.ac.uk/spm) imple-mentedin a MATLABsuite (Mathworks, Inc,Natick, Massachu-setts). Images were realigned to correct formotion,correctedforerrors in slice timing, spatially transformed to standard stereo-taxic space (based on the Montreal Neurologic Institute coordi-natesystem),32resampledevery2 mm,andsmoothedwith a 6-mmfull-width half-maximum gaussian kernel. There were no par-ticipants with movement greater than 3 mm of translation or 3°of rotation. There were also no significant differences betweenthe total range of movement across any axis of translation or ro-tation between groups. Data were then bandpass filtered usingFMRIBSoftware Library tools(available at: http://www.fmrib.ox

.ac.uk/fsl/index.html)to retainfrequenciesfrom 0.008 to 0.1Hz,to remove high- and low-frequency noise sources, and to allowfor better discrimination of low-frequency resting-state net-works. Most of the power in resting-state networks is found inthis low-frequency band.33

Functional Connectivity Analyses

Seeds for theconnectivity analysis correspondedto theBLA andCMA subregions34 and were constructed from maximum prob-

abilitymaps of these 2 subregions, defined through theAnatomyToolbox in SPM5.19 Maximum probability maps allow the defi-nition ofnonoverlappingregionsfromunderlyingprobabilitymapsthat are inherently overlapping.18,19 Voxels were included in themaximum probability maps if the probability of their being as-signedto theBLA orCMAwas higherthanforother nearbystruc-tures,suchas other amygdalarsubregionsormedial temporallobestructures,withno lessthan40%likelihood.19 TheBLAandCMAmaximum probability maps were then converted to ROIs usinga toolboxforSPM (MarsBaR[MARSeilleBoıˆte À Region d’Interet];available at http://marsbar.sourceforge.net/), and the individualtime series within each ROI was obtained from bandpass-filtered images. Each time series was then put into a first-levelfixed-effectsgeneral linear model in SPM5,35 and 4 separate con-nectivity analyses were then performed for eachsubject (2 ROIsper side). A global signal regressor and the 6 motion parametersforeach subjectwere included as covariates ofno interest in eachmodel. For direct correlations of ROI time courses, variance as-sociated with the global mean or motion parameters was re-moved beforeperformingthe correlations. A similarapproachwastaken in a control connectivity analysis in which theprimary au-ditorycortexwas theseedregion, usinga previouslydefined ROI.36

Group-Level Analyses

We performed group-level analyses using the contrast imagesfrom the individual functionalconnectivityanalyses in random-effectst tests or analyses of variance (ANOVAs). The ANOVAswere performed with a flexible factorial model in which a subject factor was additionally used to model the sphericity of thedataappropriately. All group-level analysesweremasked withamap from 1-samplet tests of each relevant group, thresholdedleniently atP=.05, uncorrected, andthen combined acrosssideand region to restrict connectivity analyses to positively corre-lated voxels only. This procedure avoided potential interpreta-tional confounds related to apparently negativeconnectivityre-sulting from correction for global signal changes.37 Statisticalthresholds were setat q 0.05,whole-braincorrectedfor thefalsediscovery rate. Connectivityfociwere labeledby comparisonwithneuroanatomical atlases.38 Reported voxel coordinates corre-spond to standardizedMontreal Neurologic Institute space. Be-cause the BLA cortical targets formed 1 very large cluster atq 0.05, whole-brain corrected, we report representative differ-ential connectivity peaks separately for each lobe inTable 2and limit the number of peaks to no more than 3 per region forclarity of presentation. For the displayed sections, the right side

of the image corresponds to the right side of the brain.To determine theconnectivity specificity of amygdalarsub-regions with their respective targets, average connectivity( weights) in volumes of interest were extracted from CMA-or BLA-seeded connectivity analyses (separatedby side) usingmasks generatedfrom theANOVA effectof region forthe GADcontrol cohort (separately for BLA or CMA targets). The vol-ume of interest data then underwent ANOVA using commer-cially available statistical software (SPSS; SPSS Inc, Chicago,Illinois). Identical masks were used for both control cohortsand the GAD patient group. All error bars or standard errorranges reported refer to standard errors of the mean.










Table 2. Conjunction Analysis Connectivity Peaks

Region a Peak Side

Voxel Coordinates b

z Scorex y z

BLA CMAc

Frontal lobe

Precentral gyrus R 38 −28 62 6.34Precentral gyrus R 46 −18 50 5.11Precentral gyrus L −58 −18 38 4.53

Orbitofrontal cortex L −36 38 −18 4.18Orbitofrontal cortex R 36 28 −20 3.83Ventromedial PFC L −4 54 −16 4.13Inferior/middle frontal gyrus L −44 16 20 3.43

Parietal lobe

Postcentral gyrus R 38 −30 64 6.39Postcentral gyrus R 46 −22 52 5.28Postcentral gyrus L −54 −22 48 4.51Superior parietal lobule R 22 −84 34 4.63Superior parietal lobule L −20 −80 34 4.58Intraparietal sulcus R 28 −60 58 3.76Intraparietal sulcus L −24 −56 58 3.48

Occipital lobe

Inferior/middle occipital gyrus R 42 −80 6 5.89Inferior/middle occipital gyrus L −48 −74 4 4.91Fusiform gyrus R 26 −66 −8 5.36Fusiform gyrus L −16 −82 −8 4.58Fusiform gyrus R 30 −86 −8 4.41

Lingual gyrus L −22 −52 −4 5.18Middle/superior occipital gyrus R 22 −88 30 4.96Middle/superior occipital gyrus L −16 −88 24 4.88Occipital pole R 8 −90 18 4.56

Temporal lobe

Superior temporal sulcus/gyrus R 54 −2 −12 5.56Superior temporal sulcus/gyrus L −62 −16 2 4.83Parahippocampal gyrus L −22 −50 −6 5.46Parahippocampal gyrus R 24 −50 −2 5.06Temporal pole L −40 12 −32 5.23Temporal pole R 36 10 −32 4.53Middle temporal gyrus R 44 −60 2 5.13Middle temporal gyrus L −54 −6 −26 4.58Fusiform gyrus R 34 −34 −20 4.26Inferior temporal gyrus R 46 −58 −18 4.48Inferior temporal gyrus L −44 −54 −18 4.08Superior temporal gyrus/planum temporale R 62 −40 10 4.46

Superior temporal gyrus/planum temporale L −44 −36 12 4.21CMA BLAd

Thalamus/caudate/amygdala

Pulvinar/caudate (body) L −18 −22 18 6.63Pulvinar/caudate (body) R 22 −26 16 6.16Septal nuclei/fornix L/R −2 4 18 5.51Centromedial amygdala L −26 −14 −4 5.38Caudate (body) R 18 −4 26 5.38Midline thalamus L/R −2 −14 10 5.25Midline thalamus L/R 2 −6 8 5.12Lateral thalamus R 20 −12 10 5.12Septal nuclei/fornix L/R −4 −18 24 4.96Centromedial amygdala R 26 −16 −4 4.65Anterior thalamus (reticular nucleus, BNST) L/R 10 −4 6 4.42Pulvinar L −8 −28 12 4.39

VTA/SN/PAG VTA/SN L/R 6 −20 −16 4.26PAG L/R −4 −28 −12 3.51

Cerebellum Vermis L/R −2 −56 −32 4.11Cerebellum Posterior L/R −4 −80 −26 4.08Cerebellum Lateral L −26 −62 −28 3.98Insula . . . R 38 10 12 3.48Cerebellum Lateral R 34 −54 −32 3.46Cerebellum Lateral R 30 −70 −26 3.38

Abbreviations: BLA, basolateral amygdalar subregion; BNST, bed nucleus of the stria terminalis; CMA, centromedial amygdalar subregion; ellipses, notapplicable; L, left side; PAG, periaqueductal gray; PFC, prefrontal cortex; R, right side; SN, substantia nigra; VTA, ventral tegmental area.

a Described as large, contiguous cortical cluster divided by lobe.bBased on the Montreal Neurologic Institute coordinate system.c Indicates BLA connectivity is increased compared with CMA connectivity.d Indicates CMA connectivity is increased compared with BLA connectivity.










STRUCTURAL DATA ANALYSIS

Voxel based morphometry wasperformedusing theVBM5tool-box in SPM5 (Wellcome Department of Cognitive Neurology,London, England; available at: http://dbm.neuro.uni-jena.de /vbm/vbm5-for-spm5/).Theindividual T1-weightedimages fromthe GAD controls and the matched GAD patient group weresegmented using a unified segmentation routine through theVBM5 toolbox.39 The segmented, modulated gray matter im-ages were then smoothed at 6 mm full width half maximum.40

The volumes of interest corresponding to gray matter volume(modulated images) were extracted for the BLA and CMA onboth sides for statistical analyses in SPSS.

RESULTS

DISSOCIABLE AND REPRODUCIBLECONNECTIVITY NETWORKS FOR BLA AND CMA

We initially examined a cohort of 17 well-characterizedhealthy subjects (the GADcontrols) (Table1) by conduct-ing resting-state fMRIconnectivity analyses separately for

each of the 2 amygdalar subregion ROIs across both sidesof the brain while controlling for variations in global sig-nal (described in the “Functional Connectivity Analyses”subsection of the “Methods” section). In this cohort, theBLA and CMA ROIs explained 21% of the variance in theother ROIduringour resting-state fMRI scans, suggestingthat these ROIs have enough nonsharedvariance to allowfor identification of differential connectivity patterns.

Next, weperformed a setof connectivityanalyses,usingthe BLA and CMA ROIs as seeds, which we entered into

a voxelwise ANOVA.Directly contrasting BLA andCMAconnectivity(collapsing acrosssides) yielded robust,spa-tially coherent patterns of differential connectivity(Figure 1 A). The connectivity patterns were bilateraland similar for ROIs from the right and left sides (datanot shown), and there were no significant connectivityclusters in the region side ANOVA interaction analy-sis (q 0.05; data not shown).

The BLA was associated with exclusively cortical dif-ferential connectivity, encompassing bilaterally the en-tire occipitallobe, largeextents of theventral temporal lobe,

CMA > BLA

– 5.2 0.0 5.1

BLA > CMA

A

B

C

CMA > BLA– 8.5 0.0 7.8

BLA > CMA

CMA > BLA– 4.3 0.0 5.4

BLA > CMA

M1/S1OFC M1/S1

vmPFCVTA/SN

PAG

Occ

STGFG

6420

– 2– 4– 6

BLA > CMA

CMA > BLAMidbrain Thalamus

10

5

0

– 5

– 10

BLA > CMA


6

24

0– 2– 4– 6

BLA > CMA


Figure 1. Differential connectivity of the basolateral (BLA) and centromedial (CMA) amygdalar subregions during resting-state functional magnetic resonanceimaging. Findings are shown in 2 separate cohorts of healthy subjects (A and B) and a formal conjunction between these cohorts (C). The BLA connectivity wasprimarily cortical, whereas the CMA connectivity was primarily subcortical. Color scales representt scores for the main effect of region in a voxelwise analysis ofvariance. Red indicates that BLA connectivity is increased compared with CMA connectivity; blue, CMA connectivity is increased compared with BLA connectivity.FG indicates fusiform gyrus; M1/S1, primary somatosensory and motor cortices; Occ, occipital cortex; OFC, orbitofrontal cortex; PAG, periaqueductal gray; STG,superior temporal gyrus; vmPFC, ventromedial prefrontal cortex; and VTA/SN, ventral tegmental area/substantia nigra.










superior temporal gyrus and sulcus, precentral and post-central gyri, intraparietal sulcus, lateral posterior orbito-frontal cortex, dorsal and ventral medial prefrontal cor-tex (PFC), parahippocampal gyri, and a small region intheinferior lateralPFC(Figure1A). This broadneuralnet-work includes primary and higher-order association cor-tices for thevisual, somatosensory,andauditory systems,as well as motor regions, areas important in memory for-mation, and higher-order areas in the PFC, most notably

along its medial and inferior aspects. The CMA wasasso-ciated almost exclusivelywithsubcorticaldifferentialcon-nectivity, including the thalamus (in particular along themidline), midbrain (in the region of the substantia nigra,ventral tegmental area, and periaqueductal gray), me-dulla, cerebellum (in particular in the vermis), and cau-date (Figure 1A).

To determine the robustness and replicability of theseconnectivity patterns, we examined a second, unrelatedgroup of31healthy adults(second controlcohort) andrep-licated thedifferentialconnectivitypatternsfortheBLAandCMA that had been identified in the GAD controls(Figure 1B). Theonly difference inconnectivityresultsbe-tweenthesecohorts consisted ofinsular,dorsalanteriorcin-

gulate, andmidcingulateCMAconnectivity in thesecondcontrolcohort.Todirectlyandquantitatively identify thoseregions with the most robust and reproducible pattern of differential BLA and CMA connectivity, we conducted aformal conjunctionanalysisacross the2 independentcon-trol cohorts (Figure 1C and Table 2).41 This analysis con-firmed the exclusively cortical pattern of differential BLAconnectivityandthesubcorticalpattern ofdifferentialCMAconnectivity, as detailed in the preceding paragraphs.

To further quantify the differential connectivity pat-terns, we constructed a separate mask for BLA and CMAtargetsderivedonlyfromtheGADcontrols,removingamyg-dalar voxels to avoid the potentially confounding effectsof correlations of ROIs with themselves. We determined

the average connectivity of the BLA or CMA target net-workswiththeBLAandCMAsourceROIsfrombothsides(Figure2 A). Asexpected, inthe GAD controlcohort, wefounda strongreciprocaldifference inBLAand CMAcon-nectivity with their respective targets (eg, stronger con-nectivity of BLA ROIs with BLA targets than CMA targets;ANOVA region target interaction,P .001), which wasconsistentacrossbothsides(ANOVAside region targetinteraction,P=.38).Strikingly,we foundhighlysimilarre-sults for the second control cohort, using the target net-work defined only from the GAD controls (ANOVAregion target interaction,P .001;region side targetinteraction,P=.59). The 2 control cohorts, moreover, didnot differ in their pattern of differential BLA-CMA con-

nectivity (ANOVA group region target interaction,P=.63).This dissociablepattern ofamygdalarsubregionalconnectivityin thesecondcontrol cohortwascapturedbytarget networks independently defined from the ANOVAon just the GAD controls. Performing the analysis in thisway would, if anything, be expected to reduce our sensi-tivity forfinding subregionalconnectivity differencesin thesecond control cohort and thus speaks to the robustnessof these findings.

Finally, we examined the relationshipbetween targetregionconnectivity and cytoarchitectonicprobability of

individual amygdalar voxels by performing seeded con-nectivity analyses for both control cohorts using the en-tire BLA or the CMA target network, defined from justtheGADcontrols,as sourceROIs. We then extractedtheconnectivity strength and cytoarchitectonic probabilityfor each voxel in the amygdala. The probability of be-longing to the BLA or CMA was found to significantlyand positively predict that voxel’s connectivity with theBLA or CMA target network, respectively, on both the

right and left sides for the GAD controls (P .001 for allcomparisons).Thisrelationshipalso held truefor thesec-ond control cohort, despite the fact that the target net-works were derived from the ANOVA data of the GADcontrols (P .001 for all comparisons).

AMYGDALAR SUBREGIONAL CONNECTIVITYAND ANATOMY IN GAD

Wenextwanted to determinewhether anamygdalar sub-regionalconnectivityanalysiscouldprovideinsightintoGAD.In 16 GAD patients who were age-, sex-, and education-matched to the GADcontrols (see Table1), we examinedtheconnectivity strengthof theBLA and CMA ROIs with

the target networks identified from the GAD controls(Figure 2A). Whereas the 2 independent control cohortsshowed similarly robust differential subregional connec-tivity,theGADpatientsshowedlessdistinctsubregion-targetconnectivitypatterns(ANOVAgroup region targetin-teraction,P=.003;partial 2=0.25),althoughthesepatientsstillshowedsomedegreeofdissociationbetweenBLAandCMAtargetconnectivity (ANOVAregion targetinterac-tionforGADpatientsonly,P=.001).Thisgroupdifferenceinspecificityofsubregionalconnectivitywas drivenby de-creased connectivity in patients betweentheBLAor CMAandtheir respective targetsandby increasedconnectivitywiththeother subregion’s targets (Figure2A). This effectcan also be clearlyseen inFigure 3 A and B, inwhich the

ANOVAeffectsofgroupareshownseparately foreachsub-region. In Figure 3A, decreased connectivity of the GADpatients’BLAwithBLAtargetsisshowninblue(eg,occipi-totemporalcortex, superior temporalgyrus,somatomotorcortices,andmedialPFC).Similarly,increasedconnectiv-ityofGADpatients’CMAwiththesameBLAtargetsisshownin red in Figure 3B. This voxelwise analysis also revealedgreaterBLAconnectivityinGADpatientswithnormalCMAtargets(the thalamus,brainstem,and cerebellum)andde-creasedCMAconnectivitywithitsnormaltargets(datanotshown). Thus, thetarget network approach in Figure 2A,in which amygdalar connectivity with all significant vox-els was averaged into a single summary value, yielded thesameresultsasthevoxelwiseanalysisandtherebyprovided

a simplegraphicrepresentationofthefindings.Moreover,thegroup region targetinteractionfindingremainedsig-nificant after controlling for theregularuseof psychiatricmedication(P=.01,partial 2=0.19)orthepresenceofco-morbidmajordepressivedisorder(P=.008,partial 2=0.21),apsychiatricconditioncloselyrelatedtoGAD.Furthermore,comparisonsof subgroups based on theuseof psychiatricmedicationorthepresenceofcomorbiddepressiondidnotyield significant differences.

We next examined whether the group differences inamygdalarconnectivity reflecteda global alteration of rest-










ing-stateconnectivity in patients, by examiningconnec-tivity within a control network previously reported byour group.36 We found that the primary auditory cortexwas strongly connected to its contralateral homologuein both control cohorts (left seed target peak voxel co-ordinates, 54, −18, 8 [z =6.0]; right seed, −46, −25, 4[z=4.87]) and patients (left seed, 46, −16, 8 [z=6.03];right seed, −52, −16, 6 [z =5.51]), with no difference be-tweengroupsatq 0.05.Theamygdalarconnectivityfind-ings are therefore not due to global changes in resting-state connectivity.

Differences insubregionalconnectivityinpatientsusingthe whole-network masks may be due to an intra-amygdalaralteration(eg, insubregionalsize,shape,or cel-lular composition) or a change in the connectivity of theamygdala with specific target brain regions. To distin-guish between these 2 possibilities, we performed a vox-elwise ANOVA group regionanalysis anddisplayedit ata lenient threshold to avoid false-negative results. Signifi-cance inthisanalysiswould meanthat there isa group dif-ference in thedissociability of BLA-CMAconnectivity, be-cause of decreasedconnectivitywith thecorrect targets or

0.06

0.03

0.04

0.05

0.02

0.01

0

TargetSource

BLA CMA

Left BLA

BLA CMA

Right BLA

BLA CMA

Left CMA

BLA CMA

Right CMA

C o n n e c t i v i t y , β

W e i g h t

A BGAD controls, n = 17

GAD patients, n = 16Second control cohort, n = 31

0.9

0.7

0.8

0.6

0.5

0.4Left Right

BLA

Left Right

CMA

G r a y

M a t t e r

V o l u m e ,

A U

C

0

5

10

15

20

25

F s c o r e

Midbrain

CMAtargets

Thalamus

BLAtargets

0.0 17.3F score

Figure 2. Specificity of amygdalar subregion connectivity with target regions in healthy subjects and patients with generalized anxiety disorder (GAD). A, Connectivityof the basolateral (BLA) or the centromedial (CMA) amygdalar regions of interest (ROIs), separately on the right and left sides, with target networks defined bydifferential BLA vs CMA connectivity maps from the analysis of variance (ANOVA) effect of region analysis from just the first control cohort (GAD controls) (shown inFigure 1A). Dissociable patterns of BLA and CMA connectivity (eg, greater connectivity of BLA ROIs with BLA targets than with CMA targets, and the reverse for CMAROIs) were found to a similar degree in both control cohorts but were significantly reduced in GAD patients, who were demographically matched with the GADcontrols. Bars represent mean values; error bars, standard error of the mean. B, Results of a voxelwise group region ANOVA comparing differential BLA-CMAconnectivity in healthy controls and GAD patients. Significance in this analysis would mean that there is a group difference in the dissociability of BLA-CMAconnectivity because of decreased connectivity with the correct targets or because of increased connectivity with the incorrect targets in 1 group. That BLA and CMAtargets (see labels) are significant suggests that there is an intra-amygdalar subregional disorganization in the GAD patients rather than alterations in the connectivityof the amygdala with some targets but not others. Results are displayed atP =.05, uncorrected, to ensure a complete view of group differences. The color scalesreflect F scores from the ANOVA. C, A voxel-based morphometry analysis of structural magnetic resonance images from the GAD controls and GAD patients, focusingon the amygdalar subregions used for the functional connectivity analyses, identified increased gray matter volume in the right CMA subregion in GAD patients.* P =.03. Bars represent mean values; error bars, standard error of the mean. AU indicates arbitrary units.










increasedconnectivity withthe incorrect targetsin1 group. We found similar ANOVA significancediffuselyacrossallBLA or CMA target regions (Figure 2B). These data argue,therefore, that there is an intra-amygdalarperturbation insubregionalorganizationin GADratherthan analterationin the connectivity of the amygdala selectively with sometargets but not others. In additional support of this idea,we found that time courses from the BLA and CMA ROIs

were more highly correlated in the GAD patients than inthe control cohorts(P=.002, Cohen’s d= 1.2), accountingfor nearly twice as much of each others’ variance as in thecontrols(37% vs21%), thus consistentwith GADpatientshaving less distinct subregional connectivity patterns.

To further test the idea that amygdalar structure is al-tered at the subregional level in GAD, we sought conver-gent evidence from a complementary neuroimaging mo-dality sensitive to changes in brain structure. To do so, weconducted a VBM analysis of gray matter volume withintheamygdala, using thesame maximum probability maps

as forthefunctional connectivityanalysisbecause this wasthe greatest level of neuroanatomical detail for which wesought to make conclusions. We found a significant in-creasein gray mattervolumein GADpatients (modulatedimages; ANOVA effect of group,P=.01, partial 2=0.55),whichwasmost notable within theright CMA (Figure 3C;2-samplet test,P=.03; Cohen’s d= 0.78).Thesedata there-fore provide convergent evidence for an intra-amygdalar

perturbation at the level of individual subregions in GADpatients,using an independent measure of brain structurefrom anatomical (nonfunctional) brain scans.

InexaminingthegroupdifferencesinBLAorCMAcon-nectivity (Figure 3A and B), we also found that several re-gionsconsistently showed increasedor decreasedconnec-tivityinpatientsusingseedsineitherofthe2subregions—aneffectorthogonalwith the decreasedsubregional specific-ity of connectivity that we have already noted in patients.Toexplorethesefindings,weidentifiedthosevoxelsinwhichGADpatientshad increased(red) ordecreased(blue) con-

A CB

0.045

0.015

0.030

0

– 0.015BLA CMA BLA CMA BLA CMA BLA CMA

Right Left Right Left

Controls Patients

C o u p l i n g t o

D L P F C

, β W

e i g h t

D

BLAtargets

BLAseeds

CMAseeds

Bothregions

BLAtargets

Control > Patient

– 10.6 0.0 4.9

Patient > Control Control > Patient

– 5.5 0.0 5.8

Patient > Control Control > Patient Patient > Control

0.06

0.02

0.04

0.00

– 0.02

Beck Anxiety Inventory Score

A m y d a l a – R

i g h t D L P F C C o u p l i n g

E

10 20 30 40 500

R = – 0.71P = .001

Figure 3. Group connectivity differences across amygdalar subregions and evidence of a compensatory network in generalized anxiety disorder (GAD). A, The analysisof variance (ANOVA) effect of group is shown for the basolateral amygdalar subregions (BLA). B, The ANOVA effect of group is shown for the centromedial amygdalarsubregions (CMA). Red indicates connectivity was increased in GAD patients compared with control cohorts; blue, connectivity was increased in control cohortscompared with GAD patients. Thus, the majority of blue in part A indicates decreased connectivity of GAD patients’ BLA to normal BLA targets, whereas the majority ofred in part B indicates increased inappropriate connectivity of GAD patients’ CMA to normal BLA targets. C, Common increased or decreased connectivity in GAD

patients across both amygdalar subregions identifies a bilateral frontoparietal network with increased connectivity (red) and a bilateral insulocingulate network withdecreased connectivity (blue). For figure clarity, only the right side of the brain is shown. D, Subregional connectivity of the right dorsolateral prefrontal cortex(DLPFC) from part C in controls and GAD patients demonstrates that this is an adaptation in patients not normally seen in controls. Bars represent mean values; errorbars, standard error of the mean. E, Average connectivity of the right DLPFC to the amygdala in GAD patients is negatively correlated with anxiety scores, suggestingthat it reflects connectivity with a compensatory network.










nectivitywithbothamygdalarsubregions(Figure3C).Wewereparticularlyintriguedby thefindingof increased dor-solateralPFC(DLPFC)–amygdalarconnectivityintheGADpatients because we did not find amygdalar connectivitywith this regionor theposteriorparietalcortexinourcon-trol cohorts. Indeed, extraction of data for the DLPFC re-vealed no connectivity with either subregion in the con-trols but consistently increased connectivity with bothsubregions intheGADpatients(Figure3D).Moreover, av-

erageDLPFC-amygdalar connectivity in theGADpatientswas negatively correlated with all anxiety measures, mostnotablywiththeBeckAnxietyInventory(Figure3E).Thesedatathereforesuggestthatalateralprefrontalexecutivecon-trol network is abnormally coupled with the amygdala inGADpatients.That thisconnectivityis strongest intheleastanxious patients strongly suggests a compensatory neuraladaptation.

COMMENT

To study amygdalarsubregional connectivity in GAD, wefirst investigated whether the functional connectivity pat-

terns of the BLA and CMA can be robustly discriminatedusing resting-state fMRI in healthy controls and whetherthese abide by predictions of anatomical connectivity de-rived from experimental animals. As with these anatomi-calstudies, ourstudy focused on differential connectivity,thus being sensitive to subregional differences in connec-tivity even if there is some anatomical connectivity of theBLA and CMA with thesametargets.Anatomical tract trac-ingstudies in nonhuman primates have determined that awide range ofcortical regions, primarilywithsensoryfunc-tionsor inthemedialPFC, provideinputpreferentially intotheBLAandfrequently receivereciprocal projections fromthe BLA.12,16 Studies in rodents have generally supportedthis view,15 although some discrepancies exist across spe-

cies. We found robust and highly reproducible differen-tial connectivity between the BLA and the entire occipitallobe and largepartsof thetemporal lobe.These targets rep-resentprimary andhigher-orderassociationcortices forthevisual and auditory systems, with likely equal sensitivityto afferent andefferent connections.Wealso found strongdifferentialconnectivityof theBLAtotheventromedialPFC,posterior orbitofrontal cortex, and parahippocampal gy-rus. In nonhuman primates, the medial and orbital PFCandtheentire temporal lobe areprimarilyconnectedwiththeBLA, although portionsof theseregions also have lightconnectivity withthecentralor medialnuclei.42,43Our dataare therefore consistent with predictions from connectiv-ity studies in experimental animals.

By contrast with animalconnectivitystudies, we foundrobust and highly reproducible differential BLA connec-tivity bilaterally with the primary motor and primary so-matosensorycortices throughout theirmediolateralcourse.In nonhuman primates, no evidence supports amygdalarconnectivity witheither region,44,45althoughinrodents lim-ited evidence supports amygdalar connectivity with bothregions.46 Additional work will be necessary to deter-mine whether these findings reflect species-specific dif-ferences in amygdalar connectivity, greater sensitivity of our functional connectivity method compared with ana-

tomicalconnectivity methods,or connectivity of theamyg-dala and sensory/motor cortex via a third region.

Finally,anatomicalstudiesin nonhumanprimates havegenerally reported absent or very sparse connectivity of theamygdalawith lateral PFC, with connectivitymost no-table within theposterior inferiorfrontalgyrus.47,48 WithinthelateralPFC, wefounddifferentialBLAconnectivity mostrobustly also in the posterior inferior frontal gyrus. Theconnectivityof theBLA with this regionof thelateral PFC,

as well as with the ventromedial PFC, thus provides evi-dence ofa tight relationship amongthese 3 regions—a re-lationship that is thoughtto beimportant foremotionregu-lation through cognitive control.49-52

The CMA is composed of the central and medial nu-clei,whicharethought to constitute a closelyrelated func-tional subsystem within the amygdala.17 In humans, thecentralnucleus representsmost ofthevolumeof theCMA,38

and thus we focusprimarily on its connectivity and func-tions, in particular in light of a role of thecentral nucleusin the expression of fear, regulation of arousal, and pat-terning of behavior13,14 and the overall similar connectiv-ity of these 2 nuclei.12,15,53 Input into the central nucleuscomes from a range of subcortical regions, includingmid-

line thalamic and anterior gustatory nuclei, hypothala-mus, midbrain and pontine reticular formations,and vis-cerosensory andneuromodulatorybrainstemnuclei.12,15,53

The central nucleus projects to many of the regions fromwhich it receives input, as well as to the pulvinar,12,15,16,53

the mediodorsal thalamic nucleus,54 and the effector nu-clei in the brainstem.12,15,16

Consistent with this pattern of anatomical connectiv-ity in rodents andnonhuman primates,we found promi-nent differential CMA connectivity across most of thethalamus (most notably along themidline)andin thepul-vinar. Activation of the midline thalamic nuclei is seenduring endogenous and learned fear in rodents55,56 andis associated withviscerosensory functions.57 Finally,we

found differential CMAconnectivity with a midbrain re-gion containing the ventral tegmental area and substan-tia nigra and extending posteriorly to the periaqueduc-tal gray. Our findings of functional connectivity of theCMA with these regions are, therefore, consistent withfindings of anatomical studies in animals.12,15,16

Strongdifferential connectivityof theCMAwasalsoseenwith the cerebellum, most notably with the vermis—connectivitythatisnotgenerally discussed in animal tract-tracing studies. The vermishasbeen implicated in endog-enous and learned fear,58-60 and vermis stimulation leadsto amygdalar synaptic potentials with latenciesconsistentwith monosynaptic connections.61,62 Our results, there-fore, strengthen evidence of a role of the cerebellum—

and in particular the vermis—in emotional processes63

through connectivity with the CMA.Generalized anxiety disorder has received less intense

neurobiologicalstudy thanother anxietydisorders, inwhichamygdalarhyperactivity in patients during the processingof negative emotion hasbeen a consistent finding.5 Theis-sueofwhether amygdalarabnormalitiesexist inadultswithGAD hasbeen unresolved, in part because the amygdala’scontext within distributed brain networks appears to beimportant6 but has not previously been a focus of investi-gation. In this study, we found evidence of an intra-










amygdalarperturbation in GADandabnormal connectiv-ity of the amygdala with at least 2 distinct brain networks.

By first focusing on differential amygdalar subregionalconnectivity patterns, we found that, although GAD pa-tientsstill showedevidence of differentialBLA-CMA con-nectivity, thisdifference wasconsiderably lessrobust thanin controls. Further investigation revealed that this find-ingwasdue to disorganization at an intra-amygdalar sub-regional level in theGADpatients.This conclusion issup-

ported by (1) a higher correlation between BLA and CMAtime series in the GAD patients than in the controls, (2)the lower connectivityof the BLA or the CMA ofGAD pa-tients to all of their normal targets and increased connec-tivity to all of the other subregions’ targets, and (3) ourfinding of increasedamygdalar graymattervolumeinGADpatients (inparticular withintheright CMA) using a VBMapproachwith structural scansacquiredfrom thesame subjects. These findingsthereforeprovide evidencethatGAD,like other anxiety disorders, involves important alter-ations in the amygdala, even if these have not been con-sistently detected during emotional activation para-digms.Moreover,ourdata raisenewquestionsabout howto interpret hypoactivation or the absence of differences

inamygdalaractivation inGADinlightof subregionaldis-organization within the amygdala.Another striking finding in GADpatients wasthepres-

ence of significantly increased or decreased connectivityof several regions with both amygdalar subregions. Wefound significant decreased amygdalar connectivity bilat-erally with the insula, dorsal/midcingulate, supplemen-tary motorarea, thalamus,caudate,putamen,superior tem-poral gyrus, andventrolateral PFC.Extraction of datafromthe insula and cingulate revealed that, in controls, theseregions were connected with both amygdalar subregions(data not shown). This connectivity pattern closely re-sembled a previously identified resting-state network, im-plicatedinsalienceprocessing,whichconsistsof limbicand

paralimbic regions andof which theamygdala isnormallya part.64 Activation in these regions, most notably in thefrontoinsular anddorsalanterior cingulate cortices, tendsto track with sympathetic nervous systemarousal.65 Aber-rant autonomicnervous system activity hasbeen noted inGAD, and in particular a decrease in autonomic flexibil-ity,66,67 suggesting that decreased coupling of the amyg-dala to the salience network may be related to abnormali-tiesin modulationoftheautonomicnervoussystem.Furtherwork, however, will be needed to explore this possibility,which at present remains speculative.

Significantlyincreasedamygdalar connectivitywasseenbilaterally with the dorsolateral, ventrolateral, dorsome-dial, and ventromedial PFCs; posterior parietal cortices;

andinferior occipitotemporalcortices. Thispatterncloselyresembles the canonical frontoparietal executive con-trol network identified in many studies of cognitivecon-troloveremotional49 and nonemotional68,69 material. Thiscoordinatednetworkhasalso beenobservedusing resting-state fMRI and has been reliably dissociated from the sa-lience network.64

At rest, the executive control network in healthy con-trols does not normally include the amygdala,64 and thuscouplingoftheamygdalawiththisnetwork inpatientslikelyreflectsa network-levelneural adaptation in GAD.Extrac-

tion of data from the DLPFC and posterior parietal corti-cesbilaterallyconfirmedthatnoneoftheseregionswascon-nectedwitheitheramygdalarsubregionincontrols(Figure3;other data not shown). Amygdalofrontoparietal couplingin GADpatients maythus reflect thehabitual engagementofa cognitivecontrol systemto regulate excessiveanxiety.Indeed, cognitive theories of GAD suggest that the use of compensatory cognitive strategies, such as worry, reflectsattemptsat diminishingtheimpactofemotions,withwhich

GADpatientsareotherwiseless able to deal.70,71

Thenega-tive correlation between amygdala-DLPFC coupling andanxiety furthermore suggests a relationship between thisadaptation and successful control of anxiety.

Aberrantcouplingofthe BLA and CMA withthe DLPFCisalso consistentwith severalother recentfindings.Greaterright DLPFC activation has been found in GAD patientsduringemotionalprocessing.6 A magneticresonance spec-troscopy study focusing on the DLPFC found a higherN-acetylaspartate to creatinine ratio,a markerof neuronalviability, inGADpatients.72Moreover, theright DLPFChasrecently been successfully targeted in a small open-labeltreatmentstudyofGADusing transcranialmagneticstimu-lation.73 In sum, our findings demonstrate abnormal en-

gagement of the amygdala in at least 2 distinct brain net-works and argue that one of these networks reflects acompensatory neural adaptation, consistent with cogni-tive theories of GAD.

Several limitations are also important to mention. Weused conventionalfMRImethods inthis study,and, assuch,one concern may be that the limited spatial resolution of conventional fMRI may be insufficient for discriminatingbetweenamygdalarsubregions.Weandothers74-76haveusedactivationtask-basedapproaches to demonstrate thatamyg-dalar subregions may be functionally differentiated dur-ingemotionalprocessing. In thepresent study,despite theuseof conventionalfMRI acquisition parameters, theBLAandCMAwere associated with dramatically differentcon-

nectivity patterns, which were highly replicable and con-sistent with predictions from anatomical studies in ani-mals, thus strongly suggesting that fMRI is sufficientlysensitive to discriminate between amygdalar subregions.

Second, althoughtheuse of resting-state fMRI data pre-cludes conclusions about specific ongoing cognitive pro-cessesin subjectsat rest, ourresults provide proof that thisapproachhassignificantutility inunderstanding network-level abnormalities in psychiatric disorders. Although thedegree to which resting-state connectivity reflects real-time,ongoingstateproperties vs longer-standingtraitprop-erties remains uncertain, we and others10 favor the traitinterpretation forseveralverycompelling reasons.A num-ber of studies have examined the topography of resting-

state networks during different human behavioral states,including performance of tasks,77-81 sleep,82,83 and anes-thesia,84 andinother species,84 andhavefoundthatresting-state networks are largelyunaltered despite the highly di-vergent neural and behavioral contexts being assessed ormanipulated. In addition, a large number of simulta-neouslyoperating resting-state networks havebeeniden-tified, andit is unlikely that ongoingmentalactivity couldaccount formomentto momentfluctuationsacrossso manydifferentbrainnetworks. Asa result, it islikelythat resting-state networks reflect intrinsic activity possibly relatedto










synaptic maintenance in critical networks or priming of these networks to allow for rapid activation rather thantransient task-related variations in activity.10 Therefore,group differences in resting-stateconnectivitymost likelyreflect group differences in brain structure and connec-tivity thatimportantlyrelateto symptoms, presentwhetheror not the individual is at rest. Moreover, the groups didnot differ in connectivity within a control primary audi-tory network, demonstrating that the group difference in

amygdalar connectivity isnotdue to a global alterationof resting-state connectivity in patients.Finally, given the correlational nature of fMRI, 1 pos-

sible explanation for theconnectivity data is that they re-flect theconfounding effectsof physiologicalvariationnotrelated to neural connectivity, arising from cardiac, res-piratory, or vascular sources. There are several reasons,however, why these are unlikely to account for our find-ings. First, we directly accounted for the effects of globalsignal variation in our regression models and eliminatedhigh-frequency sourcesof noise by restricting the restingfluctuation frequency band analyzed. In addition, thedif-ferentiation between BLA and CMA connectivity washeavily motivated by a large body of anatomical connec-

tivitystudies inexperimentalanimals, theresults of whichwere consistent with our functional connectivity find-ings that we replicated in 2 cohorts of healthy subjects.Moreover, we note that the highly divergent patterns of differentialBLA-CMA connectivityarose fromsourceROIsthat abuteachotherwithin the small structure of theamyg-dala. Ours was a study of differential connectivity, com-paring the BLA with the CMA, and thus any common ef-fects should be subtracted out in this comparison.

Submitted for Publication: December 5, 2008; final revi-sion received February11,2009; acceptedMarch 17,2009.Correspondence: Amit Etkin, MD, PhD, Department of Psychiatry and Behavioral Sciences, Stanford UniversitySchool of Medicine,401 Quarry Rd,Stanford, CA 94305([email protected]).Author Contributions: Dr Etkin had full access to all of the data in the study and takes responsibility for the in-tegrity of the data and the accuracy of the data analysis.Financial Disclosure: None reported.Funding/Support: This study wassupported by fundsfora residency-research program of the Veterans Affairs–Palo Alto Health Care System (Dr Etkin) and grantsNS048302 (Dr Greicius), HD047520 (Dr Menon), andNS058899 (Dr Menon) from the National Institutes of Health.Additional Contributions: Eric Kandel, MD, providedhelpful comments on themanuscript, andJennifer Keller,PhD, assisted with planning of the study.

REFERENCES

1. Kessler RC, Keller MB, Wittchen HU. The epidemiology of generalized anxietydisorder. Psychiatr Clin North Am . 2001;24(1):19-39.

2. Wittchen HU, Hoyer J. Generalized anxiety disorder: nature and course.J Clin Psychiatry . 2001;62(suppl 11):15-21.

3. Keller MB. The long-term clinical course of generalized anxiety disorder.J Clin Psychiatry . 2002;63(suppl 8):11-16.

4. Hoffman DL, Dukes EM, Wittchen HU. Human and economic burden of gener-alized anxiety disorder.Depress Anxiety . 2008;25(1):72-90.

5. EtkinA, Wager TD.Functionalneuroimaging of anxiety: a meta-analysis of emo-tional processing in PTSD, social anxiety disorder, and specific phobia.Am J Psychiatry . 2007;164(10):1476-1488.

6. Blair K, ShaywitzJ, Smith BW,RhodesR, Geraci M, Jones M, McCaffreyD, Vythil-ingam M, Finger E, Mondillo K, Jacobs M, Charney DS, Blair RJ, Drevets WC,Pine DS. Response to emotional expressions in generalized social phobia andgeneralized anxiety disorder: evidence for separate disorders.Am J Psychiatry .2008;165(9):1193-1202.

7. WhalenPJ, JohnstoneT, Somerville LH,Nitschke JB,Polis S,Alexander AL,Da-vidson RJ,Kalin NH.A functional magneticresonance imaging predictorof treat-ment response to venlafaxine in generalized anxiety disorder.Biol Psychiatry .2008;63(9):858-863.

8. McClureEB, Monk CS,NelsonEE, Parrish JM,Adler A,BlairRJ, FrommS, Char-ney DS, Leibenluft E, Ernst M, Pine DS. Abnormal attention modulation of fearcircuit function in pediatric generalized anxiety disorder.Arch Gen Psychiatry .2007;64(1):97-106.

9. MonkCS, TelzerEH, MoggK, BradleyBP,Mai X,LouroHM, Chen G, McClure-ToneEB, Ernst M, Pine DS. Amygdala and ventrolateral prefrontal cortex activation tomaskedangry faces inchildren andadolescents with generalizedanxiety disorder.Arch Gen Psychiatry . 2008;65(5):568-576.

10. FoxMD,Raichle ME.Spontaneousfluctuationsin brainactivityobserved withfunc-tional magnetic resonance imaging.Nat Rev Neurosci . 2007;8(9):700-711.

11. Greicius MD, Flores BH, Menon V, Glover GH, Solvason HB, Kenna H, Reiss AL,Schatzberg AF. Resting-state functional connectivity in major depression: ab-normallyincreased contributionsfrom subgenual cingulate cortex andthalamus.Biol Psychiatry . 2007;62(5):429-437.

12. AmaralDG,PriceJL,PitkanenA, CarmichaelST.Anatomicalorganizationof thepri-mateamygdaloidcomplex. In:Aggleton JP,ed.TheAmygdala: NeurobiologicalAs- pectsof Emotion,Memoryand MentalDysfunction.NewYork,NY: Wiley-Liss;1992:1-66.

13. Davis M, Whalen PJ.The amygdala: vigilance andemotion.Mol Psychiatry . 2001;6(1):13-34.

14. LeDouxJE. Emotion circuits inthe brain.AnnuRev Neurosci . 2000;23:155-184.15. PitkanenA. Connectivityof therat amygdaloid complex.In: AggletonJP, ed.The

Amygdala: A Functional Analysis.New York, NY: Oxford UniversityPress; 2000:31-115.

16. Price JL, Fokje TR, Amaral DG. The limbic region, II: the amygdaloid complex.In: Bjorklund A, Hokfelt T, Swanson LW, eds.Handbook of Chemical Neuro- anatomy. Vol 5. New York, NY: Elsevier Science Inc; 1987:279-388.

17. Heimer L, de Olmos JS, Alheid GF, Pearson J, Sakamoto N, Shinoda K, Mark-steinerJ, SwitzerR. Thehumanbasalforebrain, II.In: Bloom FE,Bjorklund A,Hok-felt T, eds. Handbook of Chemical Neuroanatomy: The Primate Nervous System,Part III.Vol 15. New York, NY: Elsevier Science Inc; 1999:57-226.

18. Eickhoff SB, Heim S, Zilles K, Amunts K. Testing anatomically specified hypoth-eses in functional imaging using cytoarchitectonic maps.Neuroimage . 2006;32(2):570-582.

19. Eickhoff SB, Stephan KE, Mohlberg H, Grefkes C, Fink GR, Amunts K, Zilles K.A newSPM toolboxfor combining probabilistic cytoarchitectonicmaps andfunc-tional imaging data.Neuroimage . 2005;25(4):1325-1335.

20. American Psychiatric Association.Diagnostic and Statistical Manual of Mental Disorders . 4th ed. Washington, DC: American Psychiatric Association; 1994.

21. Sheehan DV,Lecrubier Y, Sheehan KH,Amorim P, JanavsJ, Weiller E, Hergueta T,Baker R, Dunbar GC. The Mini-International Neuropsychiatric Interview (M.I.N.I.):the development andvalidation of a structured diagnostic psychiatric interview forDSM-IV and ICD-10. J Clin Psychiatry . 1998;59(suppl 20):22-33, quiz 34-57.

22. Sheehan DV,LecrubierY,Harnett-SheehanK,JanavsJ, WeillerE, BonaraLI,KeskinerA, Schinka J, Sheehan MF, Dunbar GC. Reliability and validity of the MINI Interna-tionalNeuropsychiatric Interview (MINI):according to the SCID-P.Eur Psychiatry .1997;12:232-241.

23. Spielberger CD, Gorsuch RL, Lushene RE.Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press; 1970.

24. Meyer TJ, Miller ML, Metzger RL, Borkovec TD. Development and validation ofthe Penn State Worry Questionnaire.Behav Res Ther . 1990;28(6):487-495.

25. Beck AT, Steer RA.Beck Anxiety Inventory Manual.San Antonio, TX: Psycho-logical Corp; 1993.

26. Beck AT,SteerRA, BrownGK.Manual for BeckDepression Inventory II (BDI-II).San Antonio, TX: Psychological Corp; 1996.

27. Watson D, Clark LA,Weber K, Assenheimer JS,Strauss ME,McCormickRA. Test-inga tripartite model,II: exploringthe symptomstructureof anxiety anddepressionin student, adult, and patient samples. J Abnorm Psychol . 1995;104(1):15-25.

28. Watson D, Weber K, Assenheimer JS,ClarkLA, StraussME, McCormick RA.Test-inga tripartite model,I: evaluating theconvergent anddiscriminant validity of anxi-ety and depression symptom scales. J Abnorm Psychol . 1995;104(1):3-14.

29. Sridharan D, Levitin DJ, Menon V. A critical role for the right fronto-insular cor-tex in switching betweencentral-executiveand default-mode networks.ProcNatl Acad Sci U S A. 2008;105(34):12569-12574.










30. Glover GH,Lai S. Self-navigated spiral fMRI: interleavedversussingle-shot.Magn Reson Med . 1998;39(3):361-368.

31. Kim DH, Adalsteinsson E, Glover GH, Spielman DM. Regularized higher-order invivo shimming.Magn Reson Med . 2002;48(4):715-722.

32. Friston KJ, Ashburner J, Poline JB, Frith CD, Heather JD, Frackowiak RS. Spatialregistration and normalization of images.Hum Brain Mapp . 1995;3(3):165-189.

33. CordesD, Haughton VM,Arfanakis K,CarewJD, TurskiPA, MoritzCH, Quigley MA,Meyerand ME. Frequencies contributing to functional connectivity in the cerebralcortex in “resting-state” data.AJNR Am J Neuroradiol . 2001;22(7):1326-1333.

34. Amunts K,KedoO, KindlerM, Pieperhoff P,MohlbergH, ShahNJ, Habel U,SchneiderF, Zilles K. Cytoarchitectonic mapping of the human amygdala, hippocampal re-gionand entorhinal cortex: intersubjectvariability andprobabilitymaps.AnatEm- bryol (Berl) . 2005;210(5-6):343-352.

35. Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RS. Statis-tical parametric mapsin functional imaging:a generallinearapproach.HumBrain Mapp . 1995;2(4):189-210.

36. Eckert MA,KamdarNV, Chang CE,Beckmann CF,Greicius MD,Menon V. A cross-modalsystem linking primary auditory and visual cortices: evidence from intrin-sic fMRI connectivity analysis.Hum Brain Mapp . 2008;29(7):848-857.

37. Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA. The impact ofglobal signal regression on resting state correlations: are anti-correlated net-works introduced? Neuroimage . 2009;44(3):893-905.

38. MaiJK,AssheuerJ, Paxinos G.Atlasof theHuman Brain.NewYork, NY:Thieme-Stratton Inc; 1997.

39. Ashburner J, FristonKJ. Unifiedsegmentation.Neuroimage . 2005;26(3):839-851.40. Ashburner J, Friston KJ. Voxel-based morphometry: the methods.Neuroimage .

2000;11(6, pt 1):805-821.41. Nichols T, Brett M, Andersson J, Wager T, Poline JB. Valid conjunction infer-

ence with the minimum statistic.Neuroimage . 2005;25(3):653-660.

42. CarmichaelST, PriceJL. Limbic connections of the orbital and medial prefrontalcortex in macaque monkeys.J Comp Neurol . 1995;363(4):615-641.

43. Stefanacci L, Amaral DG. Some observations on cortical inputs to the macaquemonkey amygdala: an anterograde tracing study. J Comp Neurol . 2002;451(4):301-323.

44. Avendano C, PriceJL, Amaral DG. Evidence for an amygdaloid projectionto pre-motor cortex but not to motor cortex in the monkey.Brain Res . 1983;264(1):111-117.

45. ShiCJ, Cassell MD.Cascade projections fromsomatosensorycortexto therat ba-solateral amygdala via the parietal insular cortex.J Comp Neurol . 1998;399(4):469-491.

46. Sripanidkulchai K, Sripanidkulchai B, WyssJM. The cortical projection of the ba-solateralamygdaloid nucleus inthe rat:a retrogradefluorescent dyestudy. J Comp Neurol . 1984;229(3):419-431.

47. Amaral DG, Price JL. Amygdalo-cortical projections in the monkey (Macaca fascicularis ). J Comp Neurol . 1984;230(4):465-496.

48. Ghashghaei HT, Hilgetag CC, Barbas H. Sequence of information processing for

emotions basedon theanatomic dialoguebetween prefrontalcortexand amygdala.Neuroimage . 2007;34(3):905-923.49. Ochsner KN,Gross JJ. The cognitivecontrol of emotion.Trends CognSci . 2005;

9(5):242-249.50. Johnstone T, van Reekum CM, Urry HL, Kalin NH, Davidson RJ. Failure to regu-

late:counterproductive recruitment of top-down prefrontal-subcortical circuitryin major depression. J Neurosci . 2007;27(33):8877-8884.

51. UrryHL, vanReekumCM, Johnstone T, Kalin NH,ThurowME, SchaeferHS, Jack-son CA, Frye CJ, Greischar LL, Alexander AL, Davidson RJ. Amygdala and ven-tromedial prefrontal cortex are inversely coupled during regulation of negativeaffect and predict the diurnal pattern of cortisol secretion among older adults.J Neurosci . 2006;26(16):4415-4425.

52. Wager TD, Davidson ML, Hughes BL, Lindquist MA, Ochsner KN. Prefrontal-subcortical pathways mediating successful emotion regulation. Neuron . 2008;59(6):1037-1050.

53. Price JL, Amaral DG. An autoradiographic study of the projections of the centralnucleus of the monkey amygdala.J Neurosci . 1981;1(11):1242-1259.

54. Reardon F, MitrofanisJ. Organisationof the amygdalo-thalamicpathways in rats.Anat Embryol (Berl) . 2000;201(1):75-84.

55. Holschneider DP, Yang J, Sadler TR, Nguyen PT, Givrad TK, Maarek JM. Map-ping cerebral blood flow changes during auditory-cued conditioned fear in thenontethered, nonrestrained rat. Neuroimage . 2006;29(4):1344-1358.

56. YasoshimaY, ScottTR, Yamamoto T. Differential activationof anterior and mid-line thalamic nuclei following retrieval of aversively motivated learning tasks.Neuroscience . 2007;146(3):922-930.

57. Vander Werf YD,WitterMP, Groenewegen HJ.The intralaminarand midline nucleiof the thalamus: anatomical and functional evidence for participation in processesof arousal and awareness. Brain Res Brain Res Rev . 2002;39(2-3):107-140.

58. Supple WF Jr, Leaton RN, Fanselow MS. Effects of cerebellar vermal lesions onspecies-specific fear responses, neophobia, and taste-aversion learning in rats.Physiol Behav . 1987;39(5):579-586.

59. Sacchetti B, Baldi E, Lorenzini CA,Bucherelli C. Cerebellar role in fear-conditioningconsolidation. Proc Natl Acad Sci U S A. 2002;99(12):8406-8411.

60. Sacchetti B, Sacco T, Strata P. Reversible inactivation of amygdala and cerebel-lum but not perirhinalcortex impairs reactivatedfear memories.EurJ Neurosci .2007;25(9):2875-2884.

61. Heath RG, Dempesy CW, Fontana CJ, Myers WA. Cerebellar stimulation: effectson septal region, hippocampus, and amygdala of cats and rats. Biol Psychiatry .1978;13(5):501-529.

62. Snider RS, MaitiA. Cerebellarcontributionsto the Papezcircuit.J NeurosciRes .1976;2(2):133-146.

63. Schmahmann JD,ShermanJC. Thecerebellar cognitive affectivesyndrome.Brain .1998;121(pt 4):561-579.

64. Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, Reiss AL,Greicius MD. Dissociable intrinsic connectivity networks for salience process-ing and executive control.J Neurosci . 2007;27(9):2349-2356.

65. CritchleyHD. Neural mechanismsof autonomic,affective,and cognitiveintegration.J Comp Neurol . 2005;493(1):154-166.

66. Hoehn-Saric R, McLeod DR, Funderburk F, Kowalski P. Somatic symptoms andphysiologicresponses in generalized anxiety disorderand panic disorder: an am-bulatory monitor study. Arch Gen Psychiatry . 2004;61(9):913-921.

67. Thayer JF, Friedman BH,Borkovec TD. Autonomiccharacteristicsof generalizedanxiety disorder and worry.Biol Psychiatry . 1996;39(4):255-266.

68. Duncan J, Owen AM. Common regions of the human frontal lobe recruited bydiverse cognitive demands. Trends Neurosci . 2000;23(10):475-483.

69. Miller EK, Cohen JD. An integrative theory of prefrontal cortex function.Annu Rev Neurosci . 2001;24:167-202.

70. Aikins DE, Craske MG. Cognitive theories of generalized anxiety disorder.Psy- chiatr Clin North Am . 2001;24(1):57-74, vi.

71. Wells A. A cognitive model of generalized anxiety disorder.Behav Modif . 1999;23(4):526-555.

72. Mathew SJ, Mao X, Coplan JD, Smith EL, Sackeim HA, Gorman JM, Shungu DC.Dorsolateral prefrontal cortical pathology in generalized anxiety disorder: a protonmagnetic resonance spectroscopic imaging study. Am J Psychiatry . 2004;161(6):1119-1121.

73. Bystritsky A, Kaplan JT, Feusner JD, Kerwin LE, Wadekar M, Burock M, Wu AD,Iacoboni M. A preliminary study of fMRI-guided rTMS in the treatment of gen-eralized anxiety disorder.J Clin Psychiatry . 2008;69(7):1092-1098.

74. Etkin A, Klemenhagen KC, Dudman JT, Rogan MT, Hen R, Kandel ER, Hirsch J.Individualdifferencesin trait anxietypredictthe responseof thebasolateral amyg-dala to unconsciously processed fearful faces.Neuron . 2004;44(6):1043-1055.

75. Morris JS, Buchel C, Dolan RJ. Parallel neural responses in amygdala subregionsandsensory cortexduring implicit fearconditioning.Neuroimage . 2001;13(6,pt 1):1044-1052.

76. Whalen PJ, Shin LM, McInerney SC, Fischer H, Wright CI, Rauch SL. A func-tionalMRI studyof humanamygdalaresponsesto facialexpressions offear ver-sus anger. Emotion . 2001;1(1):70-83.

77. Arfanakis K, Cordes D, Haughton VM, Moritz CH, Quigley MA, Meyerand ME.Combining independent component analysis and correlation analysis to probeinterregionalconnectivityin fMRItask activation datasets.MagnReson Imaging .2000;18(8):921-930.

78. FoxMD, Corbetta M, SnyderAZ, Vincent JL,RaichleME. Spontaneous neuronalactivitydistinguisheshumandorsaland ventral attentionsystems. ProcNatl Acad Sci U S A. 2006;103(26):10046-10051.

79. FoxMD, SnyderAZ, Vincent JL,CorbettaM, VanEssenDC, Raichle ME.The hu-manbrain is intrinsicallyorganized intodynamic, anticorrelatedfunctionalnetworks.Proc Natl Acad Sci U S A. 2005;102(27):9673-9678.

80. FranssonP. Howdefaultis thedefaultmodeof brainfunction? further evidence fromintrinsic BOLD signal fluctuations.Neuropsychologia . 2006;44(14):2836-2845.

81. Greicius MD, Menon V. Default-mode activity during a passive sensory task: un-coupled from deactivation but impacting activation.J Cogn Neurosci . 2004;16(9):1484-1492.

82. Fukunaga M, Horovitz SG, van Gelderen P, de Zwart JA, Jansma JM, Ikonomi-dou VN, Chu R, Deckers RH, Leopold DA, Duyn JH. Large-amplitude, spatiallycorrelatedfluctuations in BOLDfMRI signalsduringextendedrest andearly sleepstages. Magn Reson Imaging . 2006;24(8):979-992.

83. Horovitz SG,FukunagaM, de Zwart JA,van Gelderen P, FultonSC, BalkinTJ, DuynJH.Low frequency BOLD fluctuations duringresting wakefulness andlightsleep: asimultaneous EEG-fMRI study.Hum Brain Mapp . 2008;29(6):671-682.

84. Vincent JL, Patel GH, Fox MD, Snyder AZ, Baker JT, Van Essen DC, Zempel JM,Snyder LH, Corbetta M, Raichle ME. Intrinsic functional architecture in the an-aesthetized monkey brain.Nature . 2007;447(7140):83-86.


M 5 2011h hiD l d d f






Documents

Etkin Et Al 2009 Arch Gen Psychiatry