Advertisement

Static and dynamic network properties of the repetitive transcranial magnetic stimulation target predict changes in emotion regulation in obsessive-compulsive disorder

  • Linda Douw
    Correspondence
    Corresponding author. Department of Anatomy & Neurosciences, O|2 building room 13W09, Amsterdam UMC, location VUmc, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands.
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, 149 13th Street, Boston, MA 02129, USA
    Search for articles by this author
  • Mirjam Quaak
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Psychiatry, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands
    Search for articles by this author
  • Sophie M.D.D. Fitzsimmons
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Psychiatry, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands
    Search for articles by this author
  • Stella J. de Wit
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Psychiatry, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands
    Search for articles by this author
  • Ysbrand D. van der Werf
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands
    Search for articles by this author
  • Odile A. van den Heuvel
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Psychiatry, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    OCD team, Haukeland University Hospital, Postboks 1400, 5021, Bergen, Bergen, Norway
    Search for articles by this author
  • Chris Vriend
    Affiliations
    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Anatomy & Neurosciences, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands

    Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Psychiatry, Amsterdam Neuroscience, De Boelelaan 1108, 1081 HZ, Amsterdam, the Netherlands
    Search for articles by this author
Open AccessPublished:October 24, 2019DOI:https://doi.org/10.1016/j.brs.2019.10.017

      Highlights

      • Pre-treatment network topology of the target region predicts rTMS-response.
      • Greater baseline local connectivity implied greater stimulation effects.
      • Less temporal integration also predicted greater stimulation effects.

      Abstract

      Background

      Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive neuromodulation technique to treat psychiatric disorders, such as obsessive-compulsive disorder (OCD). However, the rTMS response varies across subjects.

      Objective/hypothesis

      We hypothesize that baseline network properties of the rTMS target may help understand this variation and predict response.

      Methods

      Excitatory rTMS to the dorsolateral prefrontal cortex (dlPFC) was applied in 19 unmedicated OCD patients, while inhibitory dlPFC-rTMS was applied in 17 healthy controls. The vertex was used as an active control target (19 patients, 18 controls). The rTMS response was operationalized as the individual change in state distress rating during an emotion regulation task. At baseline, subjects underwent resting-state functional MRI. The brain network was constructed by calculating wavelet coherence between regional activity of regions in the Brainnetome atlas. Local and integrative static connectivity and the dynamic network role of the target were calculated. Baseline target region network features were non-parametrically correlated to rTMS response.

      Results

      In the dlPFC-stimulated patients, greater local connectivity (Kendall’s Tau = −0.415, p = 0.013) and less promiscuous role of the target (Kendall’s Tau = 0.389, p = 0.025) at baseline were related to greater distress reduction after excitatory rTMS. There were no significant associations in healthy subjects nor in the active control stimulated patients.

      Conclusions

      Pre-treatment network topological indices predict rTMS-induced emotional response changes in OCD, such that greater baseline resting-state local connectivity and less temporal integration of the target region imply greater stimulation effects. These results may lead the way towards personalized neuromodulation in OCD.

      Keywords

      Introduction

      Obsessive-compulsive disorder (OCD) is a relatively common neuropsychiatric disorder with a lifetime prevalence of 2–3% [
      • Ruscio A.M.
      • Stein D.J.
      • Chiu W.T.
      • Kessler R.C.
      The epidemiology of obsessive-compulsive disorder in the national comorbidity survey replication.
      ], characterized by recurrent intrusive thoughts (obsessions) and/or repetitive behaviors (compulsions) that disturb daily life [
      • American Psychiatric Association
      Diagnostic and statistical manual of mental disorders.
      ]. Symptoms have been linked to decreased distress tolerance and heightened emotional reactivity [
      • Cougle J.R.
      • Timpano K.R.
      • Sarawgi S.
      • Smith C.M.
      • Fitch K.E.
      A multi-modal investigation of the roles of distress tolerance and emotional reactivity in obsessive-compulsive symptoms.
      ,
      • Mataix-Cols D.
      • van den Heuvel O.A.
      Common and distinct neural correlates of obsessive-compulsive and related disorders.
      ], which may relate to hyperactivation of ventral brain regions involved in emotion processing [
      • Whiteside S.P.
      • Port J.D.
      • Abramowitz J.S.
      A meta-analysis of functional neuroimaging in obsessive-compulsive disorder.
      ] combined with a deficit in cognitive control by dorsal brain regions [
      • van den Heuvel O.A.
      • van Wingen G.
      • Soriano-Mas C.
      • Alonso P.
      • Chamberlain S.R.
      • Nakamae T.
      • et al.
      Brain circuitry of compulsivity.
      ]. Neuromodulation influences this system imbalance: deep brain stimulation has already shown promising improvements in symptomatology [
      • Greenberg B.D.
      • Rauch S.L.
      • Haber S.N.
      Invasive circuitry-based neurotherapeutics: stereotactic ablation and deep brain stimulation for OCD.
      ], but remains an invasive and last-resort treatment.
      Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive neuromodulation technique that utilizes repetitive, rapid magnetic field pulses to suppress (low-frequency stimulation) or stimulate (high-frequency stimulation) activity of specific brain circuits [
      • Wagner T.
      • Valero-Cabre A.
      • Pascual-Leone A.
      Noninvasive human brain stimulation.
      ]. When targeted at the (dorsolateral) prefrontal cortex ((dl)PFC), activity of both locally and distantly connected brain regions [
      • van der Werf Y.D.
      • Sanz-Arigita E.J.
      • Menning S.
      • van den Heuvel O.A.
      Modulating spontaneous brain activity using repetitive transcranial magnetic stimulation.
      ] as well as behavior [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ,
      • Hamidi M.
      • Slagter H.A.
      • Tononi G.
      • Postle B.R.
      Repetitive transcranial magnetic stimulation affects behavior by biasing endogenous cortical oscillations.
      ,
      • Gerrits N.
      • van den Heuvel O.A.
      • van der Werf Y.D.
      Decreased neural activity and neural connectivity while performing a set-shifting task after inhibiting repetitive transcranial magnetic stimulation on the left dorsal prefrontal cortex.
      ] changes, likely restoring the abovementioned system imbalance [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ,
      • Cocchi L.
      • Zalesky A.
      • Nott Z.
      • Whybird G.
      • Fitzgerald P.B.
      • Breakspear M.
      Transcranial magnetic stimulation in obsessive-compulsive disorder: a focus on network mechanisms and state dependence.
      ,
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • van Balkom A.J.L.M.
      • Veltman D.J.
      • van den Heuvel O.A.
      Emotion regulation in obsessive-compulsive disorder: a functional MRI study before and after transcranial magnetic stimulation.
      ,
      • Dunlop K.
      • Woodside B.
      • Olmsted M.
      • Colton P.
      • Giacobbe P.
      • Downar J.
      Reductions in cortico-striatal hyperconnectivity accompany successful treatment of obsessive-compulsive disorder with dorsomedial prefrontal rTMS.
      ]. However, like in any treatment of OCD, there are large individual differences in rTMS treatment effects in terms of neural changes and behavioral response [
      • Hinder M.R.
      • Goss E.L.
      • Fujiyama H.
      • Canty A.J.
      • Garry M.I.
      • Rodger J.
      • et al.
      Inter- and Intra-individual variability following intermittent theta burst stimulation: implications for rehabilitation and recovery.
      ], hindering effective clinical application [
      • Nicolo P.
      • Ptak R.
      • Guggisberg A.G.
      Variability of behavioural responses to transcranial magnetic stimulation: origins and predictors.
      ]. It is essential to find predictors of individual rTMS response to achieve personalized treatment, which is the main aim of the current work. In order to investigate such prognostic factors, we employ advanced imaging in combination with network theory.
      Functional connectivity and network topological indices may offer advanced imaging-based predictors of individual response [
      • Cocchi L.
      • Zalesky A.
      • Nott Z.
      • Whybird G.
      • Fitzgerald P.B.
      • Breakspear M.
      Transcranial magnetic stimulation in obsessive-compulsive disorder: a focus on network mechanisms and state dependence.
      ,
      • Cardenas-Morales L.
      • Volz L.J.
      • Michely J.
      • Rehme A.K.
      • Pool E.M.
      • Nettekoven C.
      • et al.
      Network connectivity and individual responses to brain stimulation in the human motor system.
      ,
      • Gallen C.L.
      • D’Esposito M.
      Brain modularity: a biomarker of intervention-related plasticity.
      ]. Functional connectivity refers to the statistical interdependencies between time series representing brain activity of different locations in the brain that are presumed to represent functional communication [
      • Aertsen A.M.
      • Gerstein G.L.
      • Habib M.K.
      • Palm G.
      Dynamics of neuronal firing correlation: modulation of "effective connectivity".
      ]. Functional connectivity based on resting-state functional MRI (rsfMRI) has previously been investigated as a biomarker for rTMS success. In OCD, higher baseline connectivity between a dorsal medial prefrontal cortex (dmPFC) rTMS target and the ventral striatum predicted greater subsequent symptom reduction [
      • Dunlop K.
      • Woodside B.
      • Olmsted M.
      • Colton P.
      • Giacobbe P.
      • Downar J.
      Reductions in cortico-striatal hyperconnectivity accompany successful treatment of obsessive-compulsive disorder with dorsomedial prefrontal rTMS.
      ]. However, measures of functional connectivity between two predefined brain regions are necessarily constrained by the topology of the entire brain network, which has not been taken into account [
      • Shine J.M.
      • Poldrack R.A.
      Principles of dynamic network reconfiguration across diverse brain states.
      ].
      A more complete and sensitive prediction of rTMS outcome may be achieved by concurrently analyzing all pairwise connectivities between brain regions through network or graph analysis [
      • Greicius M.D.
      • Krasnow B.
      • Reiss A.L.
      • Menon V.
      Functional connectivity in the resting brain: a network analysis of the default mode hypothesis.
      ,
      • Watts D.J.
      • Strogatz S.H.
      Collective dynamics of ’small-world’ networks.
      ,
      • Dayan E.
      • Censor N.
      • Buch E.R.
      • Sandrini M.
      • Cohen L.G.
      Noninvasive brain stimulation: from physiology to network dynamics and back.
      ,
      • Sale M.V.
      • Mattingley J.B.
      • Zalesky A.
      • Cocchi L.
      Imaging human brain networks to improve the clinical efficacy of non-invasive brain stimulation.
      ]. In the field of functional network neuroscience, each brain region is seen as a node, and functional connectivity is represented by correlation weights between them in a connectivity matrix. Several whole-brain and nodal network properties may then be computed, such as measures of local segregation and global integration [
      • Watts D.J.
      • Strogatz S.H.
      Collective dynamics of ’small-world’ networks.
      ]. These measures are increasingly seen as fundamental characteristics of the (diseased) brain [
      • Gong Q.
      • He Y.
      Depression, neuroimaging and connectomics: a selective overview.
      ,
      • Fornito A.
      • Zalesky A.
      • Breakspear M.
      Graph analysis of the human connectome: promise, progress, and pitfalls.
      ,
      • Stam C.J.
      Modern network science of neurological disorders.
      ]. Moreover, the network profile of the rTMS target region is predictive of individual outcome in patients with Parkinson’s disease [
      • Koirala N.
      • Fleischer V.
      • Glaser M.
      • Zeuner K.E.
      • Deuschl G.
      • Volkmann J.
      • et al.
      Frontal lobe connectivity and network community characteristics are associated with the outcome of subthalamic nucleus deep brain stimulation in patients with Parkinson’s disease.
      ,
      • Horn A.
      • Reich M.
      • Vorwerk J.
      • Li N.
      • Wenzel G.
      • Fang Q.
      • et al.
      Connectivity Predicts deep brain stimulation outcome in Parkinson disease.
      ] or depression [
      • Downar J.
      • Geraci J.
      • Salomons T.V.
      • Dunlop K.
      • Wheeler S.
      • McAndrews M.P.
      • et al.
      Anhedonia and reward-circuit connectivity distinguish nonresponders from responders to dorsomedial prefrontal repetitive transcranial magnetic stimulation in major depression.
      ,
      • Wang L.
      • Xia M.
      • Li K.
      • Zeng Y.
      • Su Y.
      • Dai W.
      • et al.
      The effects of antidepressant treatment on resting-state functional brain networks in patients with major depressive disorder.
      ,
      • Leaver A.M.
      • Wade B.
      • Vasavada M.
      • Hellemann G.
      • Joshi S.H.
      • Espinoza R.
      • et al.
      Fronto-temporal connectivity predicts ECT outcome in major depression.
      ]. Greater baseline local and/or segregated connectivity of the target region, for instance quantified with node strength or clustering, generally yields larger effects as opposed to more global or integrative connectivity, e.g. high betweenness centrality, of the target region [
      • Gallen C.L.
      • D’Esposito M.
      Brain modularity: a biomarker of intervention-related plasticity.
      ]. In OCD, network topology has also been hypothesized, but not proven, to predict rTMS response [
      • Cocchi L.
      • Zalesky A.
      • Nott Z.
      • Whybird G.
      • Fitzgerald P.B.
      • Breakspear M.
      Transcranial magnetic stimulation in obsessive-compulsive disorder: a focus on network mechanisms and state dependence.
      ].
      The most recent insights in the field of network neuroscience relate to the dynamic nature of the brain network; while network topology averaged over an entire functional scan certainly relates to (alterations in) cognitive functioning, temporal fluctuations therein may even better explain cognitive variance [
      • Shine J.M.
      • Poldrack R.A.
      Principles of dynamic network reconfiguration across diverse brain states.
      ,
      • Hutchison R.M.
      • Womelsdorf T.
      • Allen E.A.
      • Bandettini P.A.
      • Calhoun V.D.
      • Corbetta M.
      • et al.
      Dynamic functional connectivity: promise, issues, and interpretations.
      ,
      • Lord L.D.
      • Stevner A.B.
      • Deco G.
      • Kringelbach M.L.
      Understanding principles of integration and segregation using whole-brain computational connectomics: implications for neuropsychiatric disorders.
      ,
      • Lurie D.J.
      • Kessler D.
      • Bassett D.S.
      • Betzel R.F.
      • Breakspear M.
      • Keilholz S.
      • et al.
      On the nature of resting fMRI and time-varying functional connectivity.
      ,
      • Jia H.
      • Hu X.
      • Deshpande G.
      Behavioral relevance of the dynamics of the functional brain connectome.
      ,
      • Calhoun V.D.
      • Miller R.
      • Pearlson G.
      • Adali T.
      The chronnectome: time-varying connectivity networks as the next frontier in fMRI data discovery.
      ,
      • Thompson G.J.
      Neural and metabolic basis of dynamic resting state fMRI.
      ], also when measured during the resting-state [
      • Jia H.
      • Hu X.
      • Deshpande G.
      Behavioral relevance of the dynamics of the functional brain connectome.
      ,
      • Betzel R.F.
      • Fukushima M.
      • He Y.
      • Zuo X.N.
      • Sporns O.
      Dynamic fluctuations coincide with periods of high and low modularity in resting-state functional brain networks.
      ,
      • Allen E.A.
      • Damaraju E.
      • Plis S.M.
      • Erhardt E.B.
      • Eichele T.
      • Calhoun V.D.
      Tracking whole-brain connectivity dynamics in the resting state.
      ]. The current study aims to expound the application of (dynamic) network analysis to predict response to excitatory dlPFC-rTMS on distress ratings during an emotional task paradigm in a relatively small sample of OCD patients that has been previously reported on [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ,
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • van Balkom A.J.L.M.
      • Veltman D.J.
      • van den Heuvel O.A.
      Emotion regulation in obsessive-compulsive disorder: a functional MRI study before and after transcranial magnetic stimulation.
      ]. Since the original experiment was designed to elucidate the neural basis of OCD, healthy controls received inhibitory rTMS to temporarily mimic OCD [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ,
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • van Balkom A.J.L.M.
      • Veltman D.J.
      • van den Heuvel O.A.
      Emotion regulation in obsessive-compulsive disorder: a functional MRI study before and after transcranial magnetic stimulation.
      ]. We hypothesize that individual response is predicted by baseline (dynamic) resting-state network topology of the target region and its associated subnetwork of connected brain regions. Participants with greater baseline local, segregated, connectivity of the target node and/or subnetwork are expected to show a greater behavioral change after rTMS compared to participants with more integratively connected target regions.

      Material and methods

      Participants

      This study was approved by the VUmc Medical Ethics Committee and was executed in accordance with the Declaration of Helsinki on treatment of study participants. All participants gave written informed consent before participation.
      Patients were recruited through outpatient clinics within The Netherlands OCD Association [
      • Schuurmans J.
      • van Balkom A.J.
      • van Megen H.J.
      • Smit J.H.
      • Eikelenboom M.
      • Cath D.C.
      • et al.
      The Netherlands Obsessive Compulsive Disorder Association (NOCDA) study: design and rationale of a longitudinal naturalistic study of the course of OCD and clinical characteristics of the sample at baseline.
      ], the Academic Anxiety Center Altrecht (Utrecht, the Netherlands) and online advertisements. Healthy controls, group-level matched on age, sex and education, were recruited by local and online community advertisements. Our final sample consisted of 38 OCD patients and 35 healthy controls (HCs) randomly appointed to the verum dlPFC condition (OCD: n = 19, HC: n = 17) or active control vertex condition (OCD n = 19, HC n = 18). Since the HCs underwent inhibitory rTMS, while patients underwent excitatory rTMS, we report on this asymmetric control population only in the Supplementary materials.
      All patients were screened for the presence of psychiatric disorders with the Structural Clinical Interview for DSM-IV [
      • First M.B.
      • Spitzer R.L.
      • Gibbon M.
      • Williams J.B.
      Structured clinical Interview for DSM-IV-TR Axis I disorders. SCID-I/P.
      ]. OCD symptoms and their severity were assessed using the Yale-Brown Obsessive Compulsive Scale (YBOCS [
      • Goodman W.K.
      • Price L.H.
      • Rasmussen S.A.
      • Mazure C.
      • Delgado P.
      • Heninger G.R.
      • et al.
      The Yale-Brown obsessive compulsive scale. II. Validity.
      ]), and depression was measured with the Montgomery Åsberg Depression Rating Scale (MADRS [
      • Montgomery S.A.
      • Asberg M.
      A new depression scale designed to be sensitive to change.
      ]). The main inclusion criterion was a primary diagnosis of OCD without predominant hoarding, while psychiatric co-morbidity, such as major depressive disorder, was not an exclusion criterion. Patients were free of medication for at least four weeks. Further exclusion criteria were presence of current psychoactive medication use, current or past psychosis, a major physical or neurological illness (including current or previous severe traumatic head injuries and alcohol or drug dependence), MRI contraindications and excessive head movement during the resting-state scan session. Excessive motion was defined as a mean relative root mean squared displacement (RMS) > 0.2 mm, or >20 vol with frame-wise relative RMS displacement > 0.5 mm [
      • Ciric R.
      • Wolf D.H.
      • Power J.D.
      • Roalf D.R.
      • Baum G.L.
      • Ruparel K.
      • et al.
      Benchmarking of participant-level confound regression strategies for the control of motion artifact in studies of functional connectivity.
      ]. Behavioral outcomes and associated changes in task fMRI (tfMRI) activations of a slightly larger sample have been described before [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ]; we excluded 1 OCD patient due to excessive movement during rsfMRI.

      Experimental procedure

      The experimental procedure has been described in full before [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ,
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • van Balkom A.J.L.M.
      • Veltman D.J.
      • van den Heuvel O.A.
      Emotion regulation in obsessive-compulsive disorder: a functional MRI study before and after transcranial magnetic stimulation.
      ]. In short, participants visited Amsterdam UMC (location VUmc) for three measurements within the course of 1–4 weeks. Participants first underwent a psychiatric screening session and practiced the emotion regulation task (ERT) to be used during tfMRI. Participants then had a baseline fMRI scan session in which they first underwent rsfMRI and then tfMRI. During this ERT, participants viewed OCD-related or fearful pictures in two different instruction conditions: they were asked to either attend to the picture or regulate their emotions while viewing the pictures. During each block, three pictures were presented. After each picture the participants indicated their level of distress, using a visual analogue scale, which fed into our main behavioral outcome measure (see next section). For the current analyses we only used the fearful pictures.
      In the verum condition, baseline tfMRI activations were used to derive the coordinates for rTMS stimulation or inhibition of the dlPFC in each individual participant, the so-called rTMS hotspot [
      • Sack A.T.
      • Cohen Kadosh R.
      • Schuhmann T.
      • Moerel M.
      • Walsh V.
      • Goebel R.
      Optimizing functional accuracy of TMS in cognitive studies: a comparison of methods.
      ]. For the active control condition, the vertex coordinate (MNI coordinate [x = 0; y = −34; z = 72]) was chosen using the coactivation map tool [
      • Toro R.
      • Fox P.T.
      • Paus T.
      Functional coactivation map of the human brain.
      ].
      During the final visit, participants underwent rTMS and tfMRI directly after offline rTMS was applied using a hand-held statically cooled figure-of-eight TMS coil (Medtronic MagOption) under guided real-time neuronavigation (Visor v1.0, Eemagine GmbH, Germany). In the verum condition, patients received 10 Hz rTMS over the left dlPFC at 110% of their resting motor threshold during 20min in thirty pulse trains of 10sec with a 30sec inter-train-interval [
      • Conca A.
      • Di Pauli J.
      • Beraus W.
      • Hausmann A.
      • Peschina W.
      • Schneider H.
      • et al.
      Combining high and low frequencies in rTMS antidepressive treatment: preliminary results.
      ], which resulted in 3000 pulses in total. In the active control condition, the protocol was identical to the verum condition, apart from coil localization. Participants were naive to rTMS and blind to stimulation condition.

      Behavioral outcome measure

      We used the mean distress ratings of both attend and regulate conditions for the fearful pictures. Since we were mostly interested in the change in distress after rTMS, a relative difference score was calculated for distress using the following formula:
      posttreatmentdistressbaselinedistressbaselinedistress+cx100%=relativechangeindistress


      A constant (c) of 1 was added to the denominator to avoid dividing by zero. This relative change in distress was used as the outcome measure of rTMS effect.

      Image acquisition

      Imaging was performed on a GE Signa HDxt 3T magnet (General Electric, USA) with an 8-channel headcoil. Structural scanning entailed a sagittal 3-dimensional gradient-echo T1-weighted sequence (256 × 256 matrix; voxel size = 1 × 0.977 × 0.977 mm; 172 sections). In the current study, only baseline rsfMRI was utilized, during which participants were instructed to remain awake with their eyes closed. Whole-brain functional images were acquired with a gradient echo-planar imaging sequence (TR = 1800 ms; TE = 30 ms; 64 × 64 matrix; field of view = 24 cm; flip angle = 80°; 200 vol; total duration 6min) with 40 ascending slices per volume (3.75 × 3.75 mm in-plane resolution; slice thickness = 2.8 mm; inter-slice gap = 0.2 mm).

      Image analysis

      Nodes in the cortical brain network were defined using the Brainnetome atlas [
      • Fan L.
      • Li H.
      • Zhuo J.
      • Zhang Y.
      • Wang J.
      • Chen L.
      • et al.
      The human brainnetome atlas: a new brain atlas based on connectional architecture.
      ], which contains 210 areas spanning the entire cortical ribbon. Most imaging analysis was performed using FMRIB’s software library (FSL 5.0.10 [
      • Smith S.M.
      • Jenkinson M.
      • Woolrich M.W.
      • Beckmann C.F.
      • Behrens T.E.
      • Johansen-Berg H.
      • et al.
      Advances in functional and structural MR image analysis and implementation as FSL.
      ]). Additionally, The atlas was warped to participants’ native structural space using linear (FLIRT [
      • Jenkinson M.
      • Bannister P.
      • Brady M.
      • Smith S.
      Improved optimization for the robust and accurate linear registration and motion correction of brain images.
      ]) and subsequent nonlinear (FNIRT [
      • Jenkinson M.
      • Beckmann C.F.
      • Behrens T.E.
      • Woolrich M.W.
      • Smith S.M.
      ]) registration. FIRST was applied to obtain 14 individually segmented subcortical areas [
      • Patenaude B.
      • Smith S.M.
      • Kennedy D.N.
      • Jenkinson M.
      A Bayesian model of shape and appearance for subcortical brain segmentation.
      ]. Finally, the centroid of FSL’s cerebellar atlas was added [
      • Diedrichsen J.
      • Balsters J.H.
      • Flavell J.
      • Cussans E.
      • Ramnani N.
      A probabilistic MR atlas of the human cerebellum.
      ]. Because we focused on the baseline network role of the rTMS target region, we determined the target node by calculating the overlap of a sphere (5 mm radius) centered around the MNI coordinates used for individual TMS coil localization with all regions in the atlas. The atlas region with highest percentage overlap with the rTMS target was selected as the primary node used for network analysis.
      The rsfMRI preprocessing steps included: (a) extraction of brain tissue using BET [
      • Smith S.M.
      Fast robust automated brain extraction.
      ] and tissue type segmentation using FAST on participants’ native structural scan [
      • Zhang Y.
      • Brady M.
      • Smith S.
      Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm.
      ], (b) coregistration of rsfMRI to the structural scan using FLIRT and FNIRT in order to warp atlas regions to individual functional images, (c) application of several preprocessing steps through MELODIC [
      • Beckmann C.F.
      • DeLuca M.
      • Devlin J.T.
      • Smith S.M.
      Investigations into resting-state connectivity using independent component analysis.
      ], namely discarding the first 4 vol, motion correction using McFLIRT [
      • Jenkinson M.
      • Bannister P.
      • Brady M.
      • Smith S.
      Improved optimization for the robust and accurate linear registration and motion correction of brain images.
      ], spatial smoothing at 5 mm full width half maximum using SUSAN [
      • Smith S.M.
      • Brady J.M.
      SUSAN - a new approach to low level image processing.
      ], and a high pass filter of 100 s cut-off, (d) regression of time series from voxels in the white matter and cerebrospinal fluid, (e) exclusion of atlas regions containing <4 active voxels in any participant, which yielded a total of 200 regions common to all datasets. We then extracted time series from each of these 200 nodes in each participant. In order to mitigate the effect of movement, we scrubbed time points with >0.3 mm relative frame-wise displacement.

      Static network analysis

      Fig. 1 schematically depicts the network analysis pipeline. Statically, we focused on node strength, representing the total sum of the connection weights of a (subnetwork of) node(s), as a measure of local network connectivity of both the target node (primary predictor) and the average of the target subnetwork (secondary predictor; see Fig. 1B). As measures of more integrative connectivity, we calculated both betweenness centrality, which refers to the weighted fraction of all shortest paths that pass through a (subnetwork of) node(s), and the participation coefficient, which assesses the weighted diversity of connections a (subnetwork of) node(s) has with other subnetworks.
      Fig. 1
      Fig. 1Schematic overview of (dynamic) network analysis. (A) depicts how the target node and target subnetwork were defined on both the connectivity matrix (top left) and the brain (top right). Static and dynamic network analysis was subsequently performed on this decomposition into target node and subnetwork. The top row of (B) shows an exemplar toy network of a subject with high average node strength of the target subnetwork: the target subnetwork has an average strength of: 3 (target node in magenta) + 2 (target subnetwork node 1 in purple) + 3 (target subnetwork node 2 in purple))/3 = 2.7. Conversely, the bottom row subject has low average node strength of the target subnetwork, namely (2 + 1 + 1)/3 = 1.3. In (C), the dynamic network analysis is schematically indicated. While the line surrounding each node still indicates target node, target subnetwork or non-target node, the fill of each node indicates the module to which it belongs in each particular window of data (windows 1 to 3 displayed in rows). The subject depicted in the left column has low promiscuity of the target region: the magenta node consistently belongs to the white module. However, the subject depicted in the right column has high promiscuity of the target region, as it belongs to the white, black, and gray modules consecutively.
      In-house scripts for MATLAB R2017a (The MathWorks, Inc, Natick, MA, USA) were used. Static connectivity matrices were calculated by correlating the 200 time series using the wavelet transform in the 0.06–0.12 Hz frequency band, conforming to previous studies [
      • Bassett D.S.
      • Yang M.
      • Wymbs N.F.
      • Grafton S.T.
      Learning-induced autonomy of sensorimotor systems.
      ,
      • Zhang Z.
      • Telesford Q.K.
      • Giusti C.
      • Lim K.O.
      • Bassett D.S.
      Choosing wavelet methods, filters, and lengths for functional brain network construction.
      ]. We did not threshold connectivity matrices, as thresholding may result in the loss of potentially relevant connections [
      • Knock S.A.
      • McIntosh A.R.
      • Sporns O.
      • Kotter R.
      • Hagmann P.
      • Jirsa V.K.
      The effects of physiologically plausible connectivity structure on local and global dynamics in large scale brain models.
      ]. Since negative correlations may occur, but have no known biological substrate as of yet, we absolutized correlations. Subsequent network analysis took place on 200 × 200 weighted matrices using the Brain Connectivity Toolbox (BCT), with all formulas described in Ref. [
      • Rubinov M.
      • Sporns O.
      Complex network measures of brain connectivity: uses and interpretations.
      ].
      Although we took care to localize the target node, it is expected that the (predictive) effect of rTMS is not limited to a single node; this is why we included both the target node and its subnetwork as predictors. In order to obtain the subnetwork encompassing the target region, the global modular structure of each subject was determined using the Louvain modularity algorithm (see Fig. 1A, in which the module containing the magenta target node is depicted in purple). A gamma of 1.08 was used to create on average seven (range 5–9) subnetworks per participant in line with previous work [
      • Yeo B.T.
      • Krienen F.M.
      • Sepulcre J.
      • Sabuncu M.R.
      • Lashkari D.
      • Hollinshead M.
      • et al.
      The organization of the human cerebral cortex estimated by intrinsic functional connectivity.
      ].
      Finally, based on this target subnetwork analysis, a control node was chosen that was least frequently present in the target subnetworks across all subjects. This node was located in the right occipital lobe and therefore unlikely to be affected by dlPFC-rTMS.

      Dynamic network analysis

      We also investigated dynamic or temporal segregation versus integration through the measure of promiscuity [
      • Papadopoulos L.
      • Puckett J.G.
      • Daniels K.E.
      • Bassett D.S.
      Evolution of network architecture in a granular material under compression.
      ,
      • Sizemore A.E.
      • Bassett D.S.
      Dynamic graph metrics: tutorial, toolbox, and tale.
      ], which represents the fraction of all subnetworks in which the target node participated at least once across different windows of the entire scan (see Fig. 1C). Higher promiscuity indicates greater spatial diversity in a node’s modular allegiance over time.
      Dynamic network analysis was performed identically to the static analysis, with the exception that connectivity was established over a constant range of timepoints from the entire scan using a sliding window approach. Based on the literature, window length was 25 TRs (45sec) with 1 TR shift (1.8sec [
      • Hutchison R.M.
      • Womelsdorf T.
      • Allen E.A.
      • Bandettini P.A.
      • Calhoun V.D.
      • Corbetta M.
      • et al.
      Dynamic functional connectivity: promise, issues, and interpretations.
      ,
      • Leonardi N.
      • Van De Ville D.
      On spurious and real fluctuations of dynamic functional connectivity during rest.
      ]). Promiscuity was then calculated using two publicly available Matlab toolboxes, of which all formulas have previously been published [
      • Papadopoulos L.
      • Puckett J.G.
      • Daniels K.E.
      • Bassett D.S.
      Evolution of network architecture in a granular material under compression.
      ,
      • Sizemore A.E.
      • Bassett D.S.
      Dynamic graph metrics: tutorial, toolbox, and tale.
      ].

      Statistical analysis

      Statistical analyses were carried out in R (R Foundation for Statistical Computing, 2017, Vienna, Austria) and SPSS Statistics 22 (IBM Corp., Armonk, NY, USA). ANOVA and Mann-Whitney U tests were used to compare group demographic and behavioral characteristics. As the behavioral and network measures were not normally distributed, Wilcoxon signed rank tests were used to compare baseline measurements between the four groups of participants.
      In order to test our hypotheses, Kendall’s Tau correlations between baseline static and dynamic network properties and relative change in distress were calculated with bootstrapped 95% confidence intervals. Due to the high level of interdependence between network measures and the exploratory nature of this study, these analyses were not corrected for multiple comparisons. Significance was set at an alpha level of 0.05.

      Results

      Participants, baseline characteristics and behavioral outcomes

      The mean age of the sample was 38.9 ± 10.5 years; about 50% of participants were men (see Table 1); the four groups were well matched in terms of age, sex and education.
      Table 1Baseline participant characteristics.
      OCD patientsHealthy controlsStatistic (df)p
      verum

      n = 19
      control

      n = 19
      verum

      n = 17
      control

      n = 18
      Number/% of males9/479/479/538/44χ(3) = 0.2630.966
      Age39.1 ± 9.5639.4 ± 11.0338.4 ± 11.1638.7 ± 11.14F(3,69) = 0.0290.993
      Education (years)12.7 ± 3.2112.9 ± 2.8613.2 ± 3.1513.1 ± 2.92F(3,69) = 0.0690.976
      YBOCS total20.0 ± 7.1922.5 ± 5.41NANAF(1,36) = 1.4980.229
      YBOCS obsession9.2 ± 3.6911.1 ± 3.46NANAF(1,36) = 2.6640.111
      YBOCS compulsion10.8 ± 4.1211.4 ± 2.50NANAF(1,36) = 0.3270.571
      Legend. Cells indicate mean ± standard deviation unless otherwise specified.
      Behavioral outcomes have been described in depth before [
      • de Wit S.J.
      • van der Werf Y.D.
      • Mataix-Cols D.
      • Trujillo J.P.
      • van Oppen P.
      • Veltman D.J.
      • et al.
      Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
      ]. In short, state distress was higher in patients compared with HCs at baseline (Mann-Whitney U = 350, p < 0.001) and post-treatment (U = 362.5, p < 0.001). In patients, distress was generally lower post-treatment as compared to baseline (Fig. 2), but paired Wilcoxon signed rank tests were not significant (verum patients Z = −1.953, p = 0.051; active control patients Z = −1.729, p = 0.084).
      Fig. 2
      Fig. 2Distress ratings within each group at baseline and after treatment. Dot and box plots are shown to visualize the distress ratings during the Emotion Regulation Task at baseline and after rTMS in OCD patients. The ‘spaghetti’ lines show the change in distress ratings from baseline to after rTMS. The distress ratings represent the subject’s appraisal of both the attend and regulate conditions combined.

      Static network topology

      We then continued to test our hypotheses whether baseline static network properties were associated with post-treatment relative change in state distress. All statistical information can be found in Supplementary Table 1. Baseline betweenness and participation of the target node or subnetwork were not associated with relative change in distress. In the verum stimulated patients, average node strength of the target subnetwork was negatively associated with relative change in distress (Kendall’s Tau = −0.415, 95% CI [-0.753, −0.006], p = 0.013; Fig. 3A), indicating that higher strength of the target subnetwork related to greater reduction in task-induced state distress after stimulatory dlPFC-rTMS. This was not the case in the active control patient group (Kendall’s Tau = −0.246 [-0.572, 0.120], p = 0.141). There was also no significant association with strength of the target node in patients (Supplementary Table 1).
      Fig. 3
      Fig. 3Significant association between change in distress and baseline network properties. In (A), the association between mean strength of the target subnetwork at baseline and the subsequent change in distress is shown for OCD patients undergoing verum rTMS to the dorsolateral prefrontal cortex (dlPFC, purple circles) or active control rTMS to the vertex (cyan triangles). In the verum condition, there was a significant association (Kendall’s Tau = −0.415, 95% CI [-0.753, −0.006], p = 0.013), which was not present in the active control condition (Kendall’s Tau = −0.246 [-0.572, 0.120], p = 0.141). In (B), the association between promiscuity of the target node at baseline and the subsequent change in distress is shown for the same groups, showing significant results in the verum condition (Kendall’s Tau = 0.389 [0.088, 0.695], p = 0.025), but not in the vertex group (Kendall’s Tau = 0.313 [-0.134, 0.663], p = 0.071). Note that a linear correlation line is drawn in these figures, but that this association was tested non-parametrically.

      Dynamic network topology

      We similarly tested our hypotheses on dynamic network properties. Promiscuity of the target node was positively associated with change in fear distress ratings in verum stimulated patients (Kendall’s Tau = 0.389 [0.088, 0.695], p = 0.025; Fig. 3B). Thus, patients with less baseline promiscuity of the target node showed greater reductions in task-induced distress. Node promiscuity of two exemplar patients is illustrated in Fig. 4.
      Fig. 4
      Fig. 4Target node promiscuity in two exemplar OCD patients. In order to illustrate the target node allegiance with different modules over all windows of data, we selected subject A with very low promiscuity and subject B with very high promiscuity. Each color indicates a module. It is evident that subject A had more stable modular allegiance and thus lower promiscuity than subject B.
      To check whether the association between promiscuity of the target node and distress change was spatially specific, we also correlated promiscuity of the control node with change in state distress. This yielded non-significant results (Supplementary Table 1).
      Finally, we tested the relationship between node strength of the target subnetwork and promiscuity of the target node, since both were related to relative change in distress in the OCD patients. However, these two network characteristics were not significantly correlated in patients (Kendall’s Tau = −0.278 [-0.566, 0.047], p = 0.110).

      Discussion

      The aim of the present study was to explore whether baseline static and dynamic networks predict dlPFC-rTMS response in OCD patients. Confirming our hypothesis, OCD patients with strong local connectivity (high subnetwork strength) and low dynamic integration (low promiscuity) at the target location were more sensitive to rTMS-induced behavioral effects.
      We found an association between baseline average strength of the target subnetwork and post-treatment reduction in task-induced distress in OCD patients, indicating that greater pre-treatment resting-state strength of local connections in the subnetwork of the target region facilitates the excitatory effects of rTMS. Although we did not test whether this association in the verum group was significantly different from the correlation coefficient found in the active control group, the latter did not show a significant relationship between network topology and change in distress. These results are in line with earlier rsfMRI studies that show greater effectiveness of rTMS when stimulated regions are more strongly connected pre-treatment [
      • Cardenas-Morales L.
      • Volz L.J.
      • Michely J.
      • Rehme A.K.
      • Pool E.M.
      • Nettekoven C.
      • et al.
      Network connectivity and individual responses to brain stimulation in the human motor system.
      ,
      • Liston C.
      • Chen A.C.
      • Zebley B.D.
      • Drysdale A.T.
      • Gordon R.
      • Leuchter B.
      • et al.
      Default mode network mechanisms of transcranial magnetic stimulation in depression.
      ]. It has already been suggested that local, but not integrative, connectivity during the resting-state may be of particular relevance for subsequent intervention effects [
      • Gallen C.L.
      • D’Esposito M.
      Brain modularity: a biomarker of intervention-related plasticity.
      ]. Similarly, greater segregation into separate modules or subnetworks is predictive of cognitive improvement after an intervention [
      • Arnemann K.L.
      • Chen A.J.
      • Novakovic-Agopian T.
      • Gratton C.
      • Nomura E.M.
      • D’Esposito M.
      Functional brain network modularity predicts response to cognitive training after brain injury.
      ,
      • Baniqued P.L.
      • Gallen C.L.
      • Voss M.W.
      • Burzynska A.Z.
      • Wong C.N.
      • Cooke G.E.
      • et al.
      Brain network modularity predicts exercise-related executive function gains in older adults.
      ,
      • Gallen C.L.
      • Baniqued P.L.
      • Chapman S.B.
      • Aslan S.
      • Keebler M.
      • Didehbani N.
      • et al.
      Modular brain network organization predicts response to cognitive training in older adults.
      ]. In the current study, we did not find an association between integrative connectivity of the target region and treatment effect. Speculatively, our results may suggest that local spatial segregation may potentiate the excitatory effect of rTMS.
      We also report the value of innovative dynamic network properties in predicting rTMS outcome in the verum OCD patient group. Again, we did not test whether the verum association was greater than the correlation in the active control group, although dynamic network topology did not significantly relate to change in distress in the active control condition. Whereas static network measures may assess predominantly spatial segregation and integration, dynamic network measures are tailored to pick up temporal signs of segregation and integration. Promiscuity is a dynamic network measure that is based on the notion that the brain network operates in modules or subnetworks [
      • Sizemore A.E.
      • Bassett D.S.
      Dynamic graph metrics: tutorial, toolbox, and tale.
      ]. The composition of these subnetworks is dynamic over time, resulting in nodes that participate in several subnetworks and, assumedly, different neural processes to a lesser or greater extent [
      • Fedorenko E.
      • Thompson-Schill S.L.
      Reworking the language network.
      ]. Our current results suggest that patients with target regions that dynamically participate in fewer subnetworks (i.e. show lower promiscuity) at baseline show greater reduction in task-induced distress after excitatory rTMS. Analogous to the spatial interpretation offered for static network results, we speculate that target nodes with high promiscuity participate in many different processes, thereby diffusing rTMS-induced excitation and decreasing its behavioral impact. Furthermore, our results suggest that both spatial and temporal measures of segregation are predictive of dlPFC-rTMS effects in OCD patients, while there was no significant association between the two measures.
      Several limitations apply to the current work. First, the small sample size limits the power of statistical analyses and may inflate our results [
      • Yarkoni T.
      Big Correlations in Little Studies: inflated fMRI Correlations Reflect Low Statistical Power-Commentary on Vul et al. (2009).
      ]. Moreover, the small sample size in combination with relative change in behavioral results led to a non-normal distribution of the dependent variable, thereby limiting our statistics to non-parametric analyses. Comorbidity in the patient population may also have impacted the variation in rTMS response, but could not be controlled for in this small sample. Second, our healthy controls received a different type of rTMS than the OCD patients, limiting interpretation of the specificity of our results for OCD patients: it could be that healthy subjects show the same vulnerability to excitatory TMS depending on their prestimulation network topology. Third, due to excessive movement during tfMRI in our cohort, only resting-state scans were analyzed in the current work, even though the contrast in network organization between emotional states may explain more behavioral variance [
      • Douw L.
      • Wakeman D.G.
      • Tanaka N.
      • Liu H.
      • Stufflebeam S.M.
      State-dependent variability of dynamic functional connectivity between frontoparietal and default networks relates to cognitive flexibility.
      ,
      • van Geest Q.
      • Douw L.
      • van ’t Klooster S.
      • Leurs C.E.
      • Genova H.M.
      • Wylie G.R.
      • et al.
      Information processing speed in multiple sclerosis: relevance of default mode network dynamics.
      ]. Nevertheless, most variation in regional network characteristics remains stable across states (rest versus task) and conditions (time of day), but instead largely depends on inter-individual differences [
      • Gratton C.
      • Laumann T.O.
      • Nielsen A.N.
      • Greene D.J.
      • Gordon E.M.
      • Gilmore A.W.
      • et al.
      Functional brain networks are dominated by stable group and individual factors, not cognitive or daily variation.
      ]. Taken together, combining rest and task conditions may add predictive power of rTMS response and should be investigated in future studies. Fourth, we applied a commonly used sliding window approach to assess dynamic network properties, but the chosen window length may have influenced our results [
      • Leonardi N.
      • Van De Ville D.
      On spurious and real fluctuations of dynamic functional connectivity during rest.
      ]. Several alternatives have been proposed to overcome choosing a specific window length, including a window-less approach [
      • Yaesoubi M.
      • Adali T.
      • Calhoun V.D.
      A window-less approach for capturing time-varying connectivity in fMRI data reveals the presence of states with variable rates of change.
      ], adjustable window lengths [
      • Patel A.X.
      • Bullmore E.T.
      A wavelet-based estimator of the degrees of freedom in denoised fMRI time series for probabilistic testing of functional connectivity and brain graphs.
      ] or a frame-wise approach where only relevant time points are included [
      • Preti M.G.
      • Bolton T.A.
      • Van De Ville D.
      The dynamic functional connectome: state-of-the-art and perspectives.
      ]. Future studies may explore whether these hypothetically more sensitive approaches may yield greater predictive power in the context of rTMS. Fourth, we consider our current work preliminary evidence for the potential relevance of (dynamic) network properties as biomarkers for rTMS effects. However, since correlational studies do not allow statistical inferences at the level of individual patients, future studies are necessary to develop this potential.
      In conclusion, we show that baseline resting-state network properties of the rTMS target region relate to post-treatment change in task-induced distress, such that dlPFC targets with strong local connections and low promiscuity are most susceptible to the effects of rTMS. As emotion regulation is disturbed in OCD patients, these results may have implications for the application of rTMS in OCD treatment by improving the choice of a stimulation target or selecting the optimal individualized treatment.

      Financial disclosures

      Dr. Douw was sponsored by NWO (NWO-ZonMw Veni 016.146.086) and Society in Science (Branco Weiss Fellowship). Prof. van den Heuvel received an NIMH grant (R01MH113250-01) and an NWO-ZonMW grant (BEO-60-63600-98-326). Dr. Vriend received grants from the Brain Foundation Netherlands (HA-2017-00227), Amsterdam Neuroscience and the Dutch Parkinson Patient Association. Ms. Quaak, Ms. Fitzsimmons, Dr. de Wit and Prof. van der Werf reported no biomedical financial interests or potential conflicts of interest.

      Acknowledgements

      We thank Dr. Niels J.H.M. Gerrits and Dr. Froukje E. de Vries for their help in acquiring and analyzing these data. This study was funded by the Netherlands Organisation for Scientific Research (NWO-ZonMw Veni 916.86.036 and Vidi 917.17.306 for OAvdH), the Brain & Behavior Research Foundation (NARSAD Young Investigators Award 2009 for OAvdH), the Netherlands Brain Foundation (2010(1)-50), and the Amsterdam Brain Imaging Platform. These funding bodies did not have any involvement in the design, data collection, analysis and interpretation of this study.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:
      Multimedia component 3

      References

        • Ruscio A.M.
        • Stein D.J.
        • Chiu W.T.
        • Kessler R.C.
        The epidemiology of obsessive-compulsive disorder in the national comorbidity survey replication.
        Mol Psychiatry. 2010; 15: 53-63
        • American Psychiatric Association
        Diagnostic and statistical manual of mental disorders.
        fifth ed. American Psychiatric Association, Arlington, VA2013
        • Cougle J.R.
        • Timpano K.R.
        • Sarawgi S.
        • Smith C.M.
        • Fitch K.E.
        A multi-modal investigation of the roles of distress tolerance and emotional reactivity in obsessive-compulsive symptoms.
        Anxiety Stress Coping. 2013; 26: 478-492
        • Mataix-Cols D.
        • van den Heuvel O.A.
        Common and distinct neural correlates of obsessive-compulsive and related disorders.
        Psychiatr Clin N Am. 2006; 29 ([viii]): 391-410
        • Whiteside S.P.
        • Port J.D.
        • Abramowitz J.S.
        A meta-analysis of functional neuroimaging in obsessive-compulsive disorder.
        Psychiatry Res. 2004; 132: 69-79
        • van den Heuvel O.A.
        • van Wingen G.
        • Soriano-Mas C.
        • Alonso P.
        • Chamberlain S.R.
        • Nakamae T.
        • et al.
        Brain circuitry of compulsivity.
        Eur Neuropsychopharmacol. 2016; 26: 810-827
        • Greenberg B.D.
        • Rauch S.L.
        • Haber S.N.
        Invasive circuitry-based neurotherapeutics: stereotactic ablation and deep brain stimulation for OCD.
        Neuropsychopharmacology. 2010; 35: 317-336
        • Wagner T.
        • Valero-Cabre A.
        • Pascual-Leone A.
        Noninvasive human brain stimulation.
        Annu Rev Biomed Eng. 2007; 9: 527-565
        • van der Werf Y.D.
        • Sanz-Arigita E.J.
        • Menning S.
        • van den Heuvel O.A.
        Modulating spontaneous brain activity using repetitive transcranial magnetic stimulation.
        BMC Neurosci. 2010; 11: 145
        • de Wit S.J.
        • van der Werf Y.D.
        • Mataix-Cols D.
        • Trujillo J.P.
        • van Oppen P.
        • Veltman D.J.
        • et al.
        Emotion regulation before and after transcranial magnetic stimulation in obsessive compulsive disorder.
        Psychol Med. 2015; 45: 3059-3073
        • Hamidi M.
        • Slagter H.A.
        • Tononi G.
        • Postle B.R.
        Repetitive transcranial magnetic stimulation affects behavior by biasing endogenous cortical oscillations.
        Front Integr Neurosci. 2009; 3: 14
        • Gerrits N.
        • van den Heuvel O.A.
        • van der Werf Y.D.
        Decreased neural activity and neural connectivity while performing a set-shifting task after inhibiting repetitive transcranial magnetic stimulation on the left dorsal prefrontal cortex.
        BMC Neurosci. 2015; 16: 45
        • Cocchi L.
        • Zalesky A.
        • Nott Z.
        • Whybird G.
        • Fitzgerald P.B.
        • Breakspear M.
        Transcranial magnetic stimulation in obsessive-compulsive disorder: a focus on network mechanisms and state dependence.
        Neuroimage Clin. 2018; 19: 661-674
        • de Wit S.J.
        • van der Werf Y.D.
        • Mataix-Cols D.
        • van Balkom A.J.L.M.
        • Veltman D.J.
        • van den Heuvel O.A.
        Emotion regulation in obsessive-compulsive disorder: a functional MRI study before and after transcranial magnetic stimulation.
        J Obsessive-Compulsive Relat Disord. 2013; 2: 226
        • Dunlop K.
        • Woodside B.
        • Olmsted M.
        • Colton P.
        • Giacobbe P.
        • Downar J.
        Reductions in cortico-striatal hyperconnectivity accompany successful treatment of obsessive-compulsive disorder with dorsomedial prefrontal rTMS.
        Neuropsychopharmacology. 2016; 41: 1395-1403
        • Hinder M.R.
        • Goss E.L.
        • Fujiyama H.
        • Canty A.J.
        • Garry M.I.
        • Rodger J.
        • et al.
        Inter- and Intra-individual variability following intermittent theta burst stimulation: implications for rehabilitation and recovery.
        Brain Stimul. 2014; 7: 365-371
        • Nicolo P.
        • Ptak R.
        • Guggisberg A.G.
        Variability of behavioural responses to transcranial magnetic stimulation: origins and predictors.
        Neuropsychologia. 2015; 74: 137-144
        • Cardenas-Morales L.
        • Volz L.J.
        • Michely J.
        • Rehme A.K.
        • Pool E.M.
        • Nettekoven C.
        • et al.
        Network connectivity and individual responses to brain stimulation in the human motor system.
        Cerebr Cortex. 2014; 24: 1697-1707
        • Gallen C.L.
        • D’Esposito M.
        Brain modularity: a biomarker of intervention-related plasticity.
        Trends Cogn Sci. 2019; 23: 293-304
        • Aertsen A.M.
        • Gerstein G.L.
        • Habib M.K.
        • Palm G.
        Dynamics of neuronal firing correlation: modulation of "effective connectivity".
        J Neurophysiol. 1989; 61: 900-917
        • Shine J.M.
        • Poldrack R.A.
        Principles of dynamic network reconfiguration across diverse brain states.
        Neuroimage. 2018; 180: 396-405
        • Greicius M.D.
        • Krasnow B.
        • Reiss A.L.
        • Menon V.
        Functional connectivity in the resting brain: a network analysis of the default mode hypothesis.
        Proc Natl Acad Sci U S A. 2003; 100: 253-258
        • Watts D.J.
        • Strogatz S.H.
        Collective dynamics of ’small-world’ networks.
        Nature. 1998; 393: 440-442
        • Dayan E.
        • Censor N.
        • Buch E.R.
        • Sandrini M.
        • Cohen L.G.
        Noninvasive brain stimulation: from physiology to network dynamics and back.
        Nat Neurosci. 2013; 16: 838-844
        • Sale M.V.
        • Mattingley J.B.
        • Zalesky A.
        • Cocchi L.
        Imaging human brain networks to improve the clinical efficacy of non-invasive brain stimulation.
        Neurosci Biobehav Rev. 2015; 57: 187-198
        • Gong Q.
        • He Y.
        Depression, neuroimaging and connectomics: a selective overview.
        Biol Psychiatry. 2015; 77: 223-235
        • Fornito A.
        • Zalesky A.
        • Breakspear M.
        Graph analysis of the human connectome: promise, progress, and pitfalls.
        Neuroimage. 2013; 80: 426-444
        • Stam C.J.
        Modern network science of neurological disorders.
        Nat Rev Neurosci. 2014; 15: 683-695
        • Koirala N.
        • Fleischer V.
        • Glaser M.
        • Zeuner K.E.
        • Deuschl G.
        • Volkmann J.
        • et al.
        Frontal lobe connectivity and network community characteristics are associated with the outcome of subthalamic nucleus deep brain stimulation in patients with Parkinson’s disease.
        Brain Topogr. 2018; 31: 311-321
        • Horn A.
        • Reich M.
        • Vorwerk J.
        • Li N.
        • Wenzel G.
        • Fang Q.
        • et al.
        Connectivity Predicts deep brain stimulation outcome in Parkinson disease.
        Ann Neurol. 2017; 82: 67-78
        • Downar J.
        • Geraci J.
        • Salomons T.V.
        • Dunlop K.
        • Wheeler S.
        • McAndrews M.P.
        • et al.
        Anhedonia and reward-circuit connectivity distinguish nonresponders from responders to dorsomedial prefrontal repetitive transcranial magnetic stimulation in major depression.
        Biol Psychiatry. 2014; 76: 176-185
        • Wang L.
        • Xia M.
        • Li K.
        • Zeng Y.
        • Su Y.
        • Dai W.
        • et al.
        The effects of antidepressant treatment on resting-state functional brain networks in patients with major depressive disorder.
        Hum Brain Mapp. 2015; 36: 768-778
        • Leaver A.M.
        • Wade B.
        • Vasavada M.
        • Hellemann G.
        • Joshi S.H.
        • Espinoza R.
        • et al.
        Fronto-temporal connectivity predicts ECT outcome in major depression.
        Front Psychiatry. 2018; 9: 92
        • Hutchison R.M.
        • Womelsdorf T.
        • Allen E.A.
        • Bandettini P.A.
        • Calhoun V.D.
        • Corbetta M.
        • et al.
        Dynamic functional connectivity: promise, issues, and interpretations.
        Neuroimage. 2013; 80: 360-378
        • Lord L.D.
        • Stevner A.B.
        • Deco G.
        • Kringelbach M.L.
        Understanding principles of integration and segregation using whole-brain computational connectomics: implications for neuropsychiatric disorders.
        Philos Trans A Math Phys Eng Sci. 2017; 375
        • Lurie D.J.
        • Kessler D.
        • Bassett D.S.
        • Betzel R.F.
        • Breakspear M.
        • Keilholz S.
        • et al.
        On the nature of resting fMRI and time-varying functional connectivity.
        PsyArXiv. 2018;
        • Jia H.
        • Hu X.
        • Deshpande G.
        Behavioral relevance of the dynamics of the functional brain connectome.
        Brain Connect. 2014; 4: 741-759
        • Calhoun V.D.
        • Miller R.
        • Pearlson G.
        • Adali T.
        The chronnectome: time-varying connectivity networks as the next frontier in fMRI data discovery.
        Neuron. 2014; 84: 262-274
        • Thompson G.J.
        Neural and metabolic basis of dynamic resting state fMRI.
        Neuroimage. 2018; 180: 448-462
        • Betzel R.F.
        • Fukushima M.
        • He Y.
        • Zuo X.N.
        • Sporns O.
        Dynamic fluctuations coincide with periods of high and low modularity in resting-state functional brain networks.
        Neuroimage. 2016; 127: 287-297
        • Allen E.A.
        • Damaraju E.
        • Plis S.M.
        • Erhardt E.B.
        • Eichele T.
        • Calhoun V.D.
        Tracking whole-brain connectivity dynamics in the resting state.
        Cerebr Cortex. 2014; 24: 663-676
        • Schuurmans J.
        • van Balkom A.J.
        • van Megen H.J.
        • Smit J.H.
        • Eikelenboom M.
        • Cath D.C.
        • et al.
        The Netherlands Obsessive Compulsive Disorder Association (NOCDA) study: design and rationale of a longitudinal naturalistic study of the course of OCD and clinical characteristics of the sample at baseline.
        Int J Methods Psychiatr Res. 2012; 21: 273-285
        • First M.B.
        • Spitzer R.L.
        • Gibbon M.
        • Williams J.B.
        Structured clinical Interview for DSM-IV-TR Axis I disorders. SCID-I/P.
        Biometrics Research, New York State Psychiatric Institute, New York2002
        • Goodman W.K.
        • Price L.H.
        • Rasmussen S.A.
        • Mazure C.
        • Delgado P.
        • Heninger G.R.
        • et al.
        The Yale-Brown obsessive compulsive scale. II. Validity.
        Arch Gen Psychiatr. 1989; 46: 1012-1016
        • Montgomery S.A.
        • Asberg M.
        A new depression scale designed to be sensitive to change.
        Br J Psychiatry. 1979; 134: 382-389
        • Ciric R.
        • Wolf D.H.
        • Power J.D.
        • Roalf D.R.
        • Baum G.L.
        • Ruparel K.
        • et al.
        Benchmarking of participant-level confound regression strategies for the control of motion artifact in studies of functional connectivity.
        Neuroimage. 2017; 154: 174-187
        • Sack A.T.
        • Cohen Kadosh R.
        • Schuhmann T.
        • Moerel M.
        • Walsh V.
        • Goebel R.
        Optimizing functional accuracy of TMS in cognitive studies: a comparison of methods.
        J Cogn Neurosci. 2009; 21: 207-221
        • Toro R.
        • Fox P.T.
        • Paus T.
        Functional coactivation map of the human brain.
        Cerebr Cortex. 2008; 18: 2553-2559
        • Conca A.
        • Di Pauli J.
        • Beraus W.
        • Hausmann A.
        • Peschina W.
        • Schneider H.
        • et al.
        Combining high and low frequencies in rTMS antidepressive treatment: preliminary results.
        Hum Psychopharmacol. 2002; 17: 353-356
        • Fan L.
        • Li H.
        • Zhuo J.
        • Zhang Y.
        • Wang J.
        • Chen L.
        • et al.
        The human brainnetome atlas: a new brain atlas based on connectional architecture.
        Cerebr Cortex. 2016; 26: 3508-3526
        • Smith S.M.
        • Jenkinson M.
        • Woolrich M.W.
        • Beckmann C.F.
        • Behrens T.E.
        • Johansen-Berg H.
        • et al.
        Advances in functional and structural MR image analysis and implementation as FSL.
        Neuroimage. 2004; 23: S208-S219
        • Jenkinson M.
        • Bannister P.
        • Brady M.
        • Smith S.
        Improved optimization for the robust and accurate linear registration and motion correction of brain images.
        Neuroimage. 2002; 17: 825-841
        • Jenkinson M.
        • Beckmann C.F.
        • Behrens T.E.
        • Woolrich M.W.
        • Smith S.M.
        Fsl. Neuroimage. 2012; 62: 782-790
        • Patenaude B.
        • Smith S.M.
        • Kennedy D.N.
        • Jenkinson M.
        A Bayesian model of shape and appearance for subcortical brain segmentation.
        Neuroimage. 2011; 56: 907-922
        • Diedrichsen J.
        • Balsters J.H.
        • Flavell J.
        • Cussans E.
        • Ramnani N.
        A probabilistic MR atlas of the human cerebellum.
        Neuroimage. 2009; 46: 39-46
        • Smith S.M.
        Fast robust automated brain extraction.
        Hum Brain Mapp. 2002; 17: 143-155
        • Zhang Y.
        • Brady M.
        • Smith S.
        Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm.
        IEEE Trans Med Imaging. 2001; 20: 45-57
        • Beckmann C.F.
        • DeLuca M.
        • Devlin J.T.
        • Smith S.M.
        Investigations into resting-state connectivity using independent component analysis.
        Philos Trans R Soc Lond B Biol Sci. 2005; 360: 1001-1013
        • Smith S.M.
        • Brady J.M.
        SUSAN - a new approach to low level image processing.
        Int J Comput Vis. 1997; 23: 45-78
        • Bassett D.S.
        • Yang M.
        • Wymbs N.F.
        • Grafton S.T.
        Learning-induced autonomy of sensorimotor systems.
        Nat Neurosci. 2015; 18: 744-751
        • Zhang Z.
        • Telesford Q.K.
        • Giusti C.
        • Lim K.O.
        • Bassett D.S.
        Choosing wavelet methods, filters, and lengths for functional brain network construction.
        PLoS One. 2016; 11e0157243
        • Knock S.A.
        • McIntosh A.R.
        • Sporns O.
        • Kotter R.
        • Hagmann P.
        • Jirsa V.K.
        The effects of physiologically plausible connectivity structure on local and global dynamics in large scale brain models.
        J Neurosci Methods. 2009; 183: 86-94
        • Rubinov M.
        • Sporns O.
        Complex network measures of brain connectivity: uses and interpretations.
        Neuroimage. 2010; 52: 1059-1069
        • Yeo B.T.
        • Krienen F.M.
        • Sepulcre J.
        • Sabuncu M.R.
        • Lashkari D.
        • Hollinshead M.
        • et al.
        The organization of the human cerebral cortex estimated by intrinsic functional connectivity.
        J Neurophysiol. 2011; 106: 1125-1165
        • Papadopoulos L.
        • Puckett J.G.
        • Daniels K.E.
        • Bassett D.S.
        Evolution of network architecture in a granular material under compression.
        Phys Rev E. 2016; 94032908
        • Sizemore A.E.
        • Bassett D.S.
        Dynamic graph metrics: tutorial, toolbox, and tale.
        Neuroimage. 2018; 180: 417-427
        • Leonardi N.
        • Van De Ville D.
        On spurious and real fluctuations of dynamic functional connectivity during rest.
        Neuroimage. 2015; 104: 430-436
        • Liston C.
        • Chen A.C.
        • Zebley B.D.
        • Drysdale A.T.
        • Gordon R.
        • Leuchter B.
        • et al.
        Default mode network mechanisms of transcranial magnetic stimulation in depression.
        Biol Psychiatry. 2014; 76: 517-526
        • Arnemann K.L.
        • Chen A.J.
        • Novakovic-Agopian T.
        • Gratton C.
        • Nomura E.M.
        • D’Esposito M.
        Functional brain network modularity predicts response to cognitive training after brain injury.
        Neurology. 2015; 84: 1568-1574
        • Baniqued P.L.
        • Gallen C.L.
        • Voss M.W.
        • Burzynska A.Z.
        • Wong C.N.
        • Cooke G.E.
        • et al.
        Brain network modularity predicts exercise-related executive function gains in older adults.
        Front Aging Neurosci. 2017; 9: 426
        • Gallen C.L.
        • Baniqued P.L.
        • Chapman S.B.
        • Aslan S.
        • Keebler M.
        • Didehbani N.
        • et al.
        Modular brain network organization predicts response to cognitive training in older adults.
        PLoS One. 2016; 11e0169015
        • Fedorenko E.
        • Thompson-Schill S.L.
        Reworking the language network.
        Trends Cogn Sci. 2014; 18: 120-126
        • Yarkoni T.
        Big Correlations in Little Studies: inflated fMRI Correlations Reflect Low Statistical Power-Commentary on Vul et al. (2009).
        Perspect Psychol Sci. 2009; 4: 294-298
        • Douw L.
        • Wakeman D.G.
        • Tanaka N.
        • Liu H.
        • Stufflebeam S.M.
        State-dependent variability of dynamic functional connectivity between frontoparietal and default networks relates to cognitive flexibility.
        Neuroscience. 2016; 339: 12-21
        • van Geest Q.
        • Douw L.
        • van ’t Klooster S.
        • Leurs C.E.
        • Genova H.M.
        • Wylie G.R.
        • et al.
        Information processing speed in multiple sclerosis: relevance of default mode network dynamics.
        Neuroimage Clin. 2018; 19: 507-515
        • Gratton C.
        • Laumann T.O.
        • Nielsen A.N.
        • Greene D.J.
        • Gordon E.M.
        • Gilmore A.W.
        • et al.
        Functional brain networks are dominated by stable group and individual factors, not cognitive or daily variation.
        Neuron. 2018; 98: 439-452 e5
        • Yaesoubi M.
        • Adali T.
        • Calhoun V.D.
        A window-less approach for capturing time-varying connectivity in fMRI data reveals the presence of states with variable rates of change.
        Hum Brain Mapp. 2018; 39: 1626-1636
        • Patel A.X.
        • Bullmore E.T.
        A wavelet-based estimator of the degrees of freedom in denoised fMRI time series for probabilistic testing of functional connectivity and brain graphs.
        Neuroimage. 2016; 142: 14-26
        • Preti M.G.
        • Bolton T.A.
        • Van De Ville D.
        The dynamic functional connectome: state-of-the-art and perspectives.
        Neuroimage. 2017; 160: 41-54