Fronto-limbic, Fronto-parietal and Default-mode Involvement in Functional Dysconnectivity in Psychotic Bipolar Disorder

Abstract

Background

The temporal sequence of events or primarily anatomical antecedents of the observed abnormalities in bipolar disorder (BD) remain speculative. Understanding how independent functional subsystems integrate globally and how they relate with anatomical cortico-subcortical networks is key to understanding how the human brain’s architecture constrains functional interactions and underpins abnormalities of mood and emotion, particularly in BD.

Methods

Resting-state functional magnetic resonance time-series were averaged to obtain individual functional connectivity matrices (AFNI); individual structural connectivity matrices were derived using deterministic non-tensor-based tractography (ExploreDTI), weighted by streamline count and fractional anisotropy (FA). Structural and functional nodes were defined using a subject-specific cortico-subcortical mapping (Desikan-Killiany, Freesurfer). Whole-brain connectivity alongside a permutation-based statistical approach and structure-function coupling were employed to investigate topological variance in predominantly euthymic BD relative to psychiatrically-healthy controls.

Results

Bipolar disorder (n=41) exhibited decreased (synchronous) activity in a subnetwork encompassing fronto-limbic and posterior-occipital functional connections (T>3, p=0.039), alongside increased (anti-synchronous) activity within a fronto-temporal subnetwork (T>3, p=0.025); all relative to controls (n=56). Preserved whole-brain functional connectivity, and comparable structural-functional relationships amongst whole-brain and edge-class connections were observed in BD relative to controls.

Conclusions

We present a functional map of BD dysconnectivity that differentially involves communication within nodes belonging to functionally specialized subsystems – default-mode, fronto-parietal and fronto-limbic systems; these changes do not extend to be detected globally and may be necessary to maintain a remitted state of the illness. Preserved structure-function coupling in BD despite evidence of regional anatomical and functional deficits is suggestive of a complex, perhaps dynamic, interplay between structural and functional subnetworks.

Introduction

Bipolar disorder (BD) is a major psychiatric condition associated with widespread dysconnectivity thought to arise from changes in integration and segregation within its brain networks (1). Neuroimaging work including diffusion and functional Magnetic Resonance Imaging (fMRI) has provided evidence into altered patterns of neuroanatomical and functional connectivity; collectively suggesting that affective dysregulation associated with BD may arise from both structural and functional changes primarily involving neural circuitries responsible for emotion regulation, cognitive-control and executive functions (2,3).

Functional features of euthymic BD include preserved whole-brain functional connectivity of default-mode (DMN), fronto-parietal (FPN) and salience (SN) networks (4). However, a priori investigations reported local patterns of functional dysconnectivity within amygdala, prefrontal and cingulate cortices in euthymic BD (4) and females with BD (5). Additionally, instabilities within the DMN are present amongst BD individuals with a positive history of psychosis and these changes may persist in patients in remission (4). Furthermore, opposing spatiotemporal patterns observed within the DMN and FPN may underpin depressive and manic episodes of BD (6). Moreover, beside evidence of dysconnectivity within the limbic system there is evidence involving reward system-related structures (7–9). Additionally, abnormalities within regions anatomically connecting with the limbic system such as the prefrontal cortex (PFC) have been linked to features of emotional and cognitive control in BD (9,10). Although these a priori investigations may have been led by morphological study findings and may be task and mood-dependent, they suggest functional impairments of BD are not confined to the DMN but rather they extend to involve emotion regulatory centers. These localized functional changes may constitute a compensatory mechanism of neural activity that underlies the general stability observed across rs-fMRI networks in euthymic subjects with BD; these changes may be necessary to sustain a remitted clinical state of the illness.

The application of graph theory methods to neuroscience has allowed a network-level understanding of the cortico-subcortical organization of the brain, specifically how independent functional subsystems integrate in global processing streams. Findings from neuroimaging investigations demonstrate that BD is unlikely to arise from changes involving one brain region alone; rather, the clinical syndrome that we currently refer to as BD may originate from a disruption of the brain’s structural and functional neurocircuitries (2). Investigations of structural and functional dysconnectivity have been increasingly implemented; however, a paucity of studies to date has applied graph theoretical tools to understand the functional organization of BD in a network-like fashion (Table 1). Collectively, these studies report weak global effects (not surviving multiple comparisons) but localized changes involving fronto-temporo-parietal and limbic nodes. These effects somewhat overlap with previous structural and functional observations in BD (11), and in unaffected siblings at high-risk of BD (12–14). Therefore, a more definite interpretation of the network-level understanding of the functional organization of BD is limited by the paucity of graph-theoretical studies and clinical samples investigated, different methodological approaches and the variety of network metrics employed (Table 1). However, a trend of functional network-level changes of BD appear to involve specific functional subsystem that predominantly encompass DMN and limbic centers, as opposed to being widespread.

Table 1. Overview of connectivity network findings of today’s functional connectivity graph theory studies.

MTG = middle temporal gyrus; FC = functional connectivity; FCS = functional connectivity strength; DMN = default mode network; rh = right hemisphere; lh = left hemisphere; SFG = superior frontal gyrus; IFG = inferior frontal gyrus; STG = superior temporal gyrus; MFG = middle frontal gyrus; ITG = inferior temporal gyrus; PreCG = precentral gyrus; FCS = functional connectivity strength; Elocal = local efficiency; mPFC = medial prefrontal cortex; vlPFC = ventrolateral prefrontal cortex; EGlobal = Global Efficiency; CPL = characteristic path length; PI = participation Index; CC = clustering coefficient; ↓ = reduced; ↑=increased; Voxel-mirrored homotopic connectivity (VMHC) (voxel-wise whole brain analysis); HC = healthy controls; BD = bipolar disorder; MDD = major depressive disorder. BD I = bipolar disorder type I; BD II = bipolar disorder type II; FA = fractional anisotropy.

Reference	Sample	Methodology	Network measures investigated	Findings (BD vs. HC)
Zhang et al., 2019 Aberrant brain structural- functional connectivity coupling in euthymic bipolar disorder	57 young (13-28yo) euthymic BD, 42 HC	Functional network (SPM and DPABI) Structural network (AAL-90 and whole-brain DTI tractography, FSL)	Structural: global strength, network-based statistics, rich-club connectivity, functional: modularity Structure-function coupling	Structural: preserved global (FA) strength; NBS: ↓ fronto-parietal-temporal connectivity (FA); ↑ frontal cortex and subcortical regions (FA). Functional: ↓ intra-modular connectivity; (marginally) ↑inter-modular connectivity. ↓ Structure-function coupling whole-brain and involving intra-hemispheric connections.
Dvorak J. et al., 2019 Aberrant brain network topology in fronto-limbic circuitry differentiates euthymic bipolar disorder from recurrent major depressive disorder	20 euthymic BD (13 with BD I and 7 with BD II), 15 MDD, 30 HC	AAL-90, DPARSF	Network-based statistics, Global (CC, CPL, EGlobal), nodal (CC, CPL, degree, betweenness centrality)	NBS: ↑ (synchronous) FCS in rh/lh temporal regions, with ↓ (synchronous) FCS specifically between lh angular gyrus-lh temporal pole and rh parietal gyrus-rh hippocampus. Preserved global connectivity: ↑ CC (not surviving FDR correction), unchanged CPL and EGlobal. CPL: ↓ rh olfactory cortex, rh hippocampus, rh middle temporal and lh fusiform gyrus, rh/lh caudate, rh putamen. Degree: ↑ rh middle frontal gyrus.
Wang Y. et al., 2017 Shared and Specific Intrinsic Functional Connectivity Patterns in Unmedicated Bipolar Disorder and Major Depressive Disorder	(Unmedicated) 48 BD II w/depression, 48 MDD, 51 HC	DPARSF/SPM8. Grey matter probability map (SPM8). Voxel-wise whole-brain functional network analysis	FCS of short-range (<75mm) & long-range (>75mm) fibers – using	Long-fibres ↑ FCS rh MTG & cerebellum Short-fibres:↑ FCS lh/rh thalamus & lh/rh cerebellum, ↓ lh/rh precuneus
Wang Y. et al., 2017 Topologically convergent and divergent functional connectivity patterns in unmedicated unipolar depression and bipolar disorder	(Unmedicated) 31 BD II w/ depression, 32 (unmedicated) UD, 43 HC	GRETNA (SPM8). AAL-1024 random parcellation, Zalesky et al., 2012	Network-based statistics, Measures: FCS, CC, CPL, EGlobal, normalised CPL, normalised CC, small-world, Elocal, Modularity.	NBS: ≠ FCS. Nodes: DMN (69%), and FPN (20%); edges: DMN-DMN (59%), DMN-FPN (14%), FPN-FPN (10%). ↑path length, ↓EGlobal – not surviving multiple correction. ↓Elocal in DMN, Limbic system and cerebellum; ↓Strength: lh/rh precuneus and SFG, lh middle cingulum, rh temporal pole; lh posterior cerebellar lobe. Disrupted intra-modular connectivity within DMN and limbic system.
Roberts et al.,2017 Functional Dysconnection of the Inferior Frontal Gyrus in Young People With Bipolar Disorder or at Genetic High Risk	49 young (16-30yo) BD with mild depression (28 with BD I; 21 with BD II), 71 at-risk, 80 HC	SPM8. AAL-90 random parcellation into 513 ROIs, Zalesky et al., 2012	Network-based statistics on the IFG, CC, PI, CPL	NBS: ↓FCS between lh IFG and fronto-temporal regions (lh insula, lh putamen, lh/rh STG, lh/rh vlPFC, lh/rh mPFC. ↓CC in IFG (no effect of lithium/antipsych/antidepr on CC); no change in CPL.
Douchet et al., 2017 The Role of Intrinsic Brain Functional Connectivity in Vulnerability and Resilience to Bipolar Disorder	78 BD I, 64 unaffected siblings, 41 healthy controls	Crossley et al., 2010 Atlas	CPL, EGlobal, CC, small-world, nodal degree, PI, Modularity. Network-based statistics.	No ≠ in CPL, EGlobal and CC. ↑degree in supplementary motor area, MFG, supramarginal gyrus, MedialFG, ITG; ↓degree in PreCentral lobule, PostCG. Modularity: ↓ FCS in sensorimotor network ↓PI vmPFC, hippocampus. NBS: ↓sensorimotor & visual networks.
Zhao et al.,2017 Altered interhemispheric functional connectivity in remitted bipolar disorder: A Resting State fMRI Study	20 euthymic BD II, 38 HC	DPARSF/SPM8/REST. MNI template. Voxel-mirrored homotopic connectivity (VMHC).	FCS between any pair of symmetric interhemispheric voxels. Global VMHC.	FCS:↓MFG and preCG; No ↑ recorded.
Spielberg et al., 2016 Resting State Brain Network Disturbances Related to Hypomania and Depression in Medication-Free Bipolar Disorder	(Unmedicated) tot of 60 BD I and II: 30 w/ depression, 30 hypomanic	Time series extracted using a 181-ROI random parcellation, no cerebellum, Craddock et al., 2012	Network-based statistics, Global: Global efficiency, CPL, Assortativity, CC in significant NBS ROIs	NBS: ↑ connectivity mostly involving amygdala (42%). A subnetwork involving FrontalG was associated w/ YMRS; subnetwork involving OrbitofrontalG w/ HDRS. ↓EGlobal; ↓CC rh Amygdala
Wang Y. et al., 2016 Disrupted resting-state Functional connectivity in nonmedicated Bipolar Disorder	(Unmedicated) 37 BD II w/ depression, 37 HC	GRETNA (SPM8)	FCS	FCS:↓DMN; ↑ParahippocampalG, Amygdala, ACC, STG, Ling G, Cerebellum
Wang Y. et al., 2015 Interhemispheric resting state functional connectivity abnormalities in unipolar depression and bipolar depression	36 BD II w/ depression-66.7% unmedicated, 32 UD during a depressive episode-71.8% unmedicated, 40 HC	DPARSF/SPM8/REST. MNI template. Voxel-mirrored homotopic connectivity (VMHC)	FCS between any pair of symmetric interhemispheric voxels	FCS: ↓fusiform/lingual G & Cerebellum
He et al., 2015 Resting-state functional network connectivity in prefrontal regions differs between unmedicated patients with bipolar and major depressive disorders	(Unmedicated) 13 BD I and II w/depression, 40 MDD, 33 HC	SPM8. Connectivity analysis: Group Independent Component Analysis (iCa) – 48 ICNs. Graph theory analysis	FCS, nodal CC, Elocal	↑CC in DLPFC, VLPFC (↓ w/depression), SFG, ACC

To date, the temporal sequence of events or primarily anatomical antecedents of the observed abnormalities in BD remain speculative and will emerge subsequent to future longitudinal population-based studies. Despite the evidence presented to date, disentangling whether these abnormalities are intertwined and relate to BD affective dysregulation is unclear, and the field may benefit from the application of network-based analyses and structure-function integration approaches. This study builds from previous studies on structural and functional brain connectivity and brain network organization in BD and aims to further clarify unresolved questions related to the neurobiology of BD. The aim of this study is to investigate functional changes in euthymic BD individuals and specifically to describe features of whole-brain and subnetwork functional connectivity as they relate to the illness affective dysregulation. We anticipate preserved whole-brain functional connectivity and aberrant connectivity patterns involving nodes belonging to limbic and DMN systems as previously highlighted by the literature (11). Furthermore, we examined whether gender and diagnosis have a differentials effect on whole-brain patterns of functional connectivity on the basis of previous structural connectivity differences in females with BD (15); finally, we explored the relationship between functional interactions and previously identified white matter abnormalities in this clinical sample (15).

Methods and Materials

Participants

Patients included in this study overlap with those in a previously presented structural connectivity analysis from our research group (15) (n=36, 88%). Participants aged 18-65 were recruited from the western regions of Ireland’s Health Services via referral or public advertisement. A diagnosis of BD was confirmed using the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) Structured Clinical Interview for DSM Disorders (American Psychiatric Association, 1994) conducted by an experienced psychiatrist. Euthymia was defined during medical screening and at MRI scanning using the Hamilton Depression (HDRS-21<8), Anxiety (HARS<18), and Young Mania (YMRS<7) Rating Scales. To ensure significant findings were features of BD type I, analyses were repeated removing BD type II subjects; further, to ensure significant findings were trait-features of BD, analyses were repeated removing subjects not meeting criteria for euthymia. Healthy controls had no personal history of psychiatric illness or history among first-degree relatives (DSM-IV Non-patient edition). Ethical approval was received by the University College Hospital Galway Clinical Research Ethics Committee, and participants gave written fully informed consent before participating.

MRI acquisition

MRI data were obtained on a 3 Tesla Achieva scanner (Philips, The Netherlands) at the Welcome Trust Health Research Board National Centre for Advanced Medical Imaging (CAMI) at St. James’s Hospital Dublin, Ireland. High-resolution 3D T1-weighted turbo field echo magnetization-prepared rapid gradient-echo (MPRAGE) sequence was acquired using an eight-channel head coil (TR/TE = 8.5/3.046 ms, 1 mm³ voxel size). Diffusion-weighted images were acquired at b=1200 s/mm2 along with a single non-diffusion weighted image (b=0), using high angular resolution diffusion imaging (HARDI) involving 61 diffusion gradient directions, 1.8x1.8x1.9 mm voxel dimension and field of view (FOV) 198x259x125 mm. Resting-state functional MRI data were acquired using a single-shot gradient echo planar imaging (EPI) sequence and involved whole-brain acquisition of 180 volumes (TR/TE = 2000/28 ms, flip angle 90°, field of view (FOV) 240x240x133 mm, 3 mm³ resolution, 80x80 matrix size and 38 axial slices of 3.2 mm each); each subject was asked to lie still in the scanner with their eyes open and fix a crosshair on the screen for the entire duration of the scan. Structural and functional MR images were visually inspected before/after processing for accuracy of segmentation/parcellation, registration, motion and outliers.

MRI Data Analysis

MRI data processing was optimized to reduce motion and physiological noise as much as possible. Individual structural T1-weighted scans underwent cortico-subcortical segmentation/parcellation (Freesurfer v5.3, http://surfer.nmr.mgh.harvard.edu/). EPI images underwent despiking to remove spikes of activation (i.e., outliers) across time series (3dDespike; AFNI v18.1.19 http://afni.nimh.nih.gov/afni), motion correction (3dvolreg; AFNI), and registration to the FreeSurfer skull-stripped structural image (FLIRT, FSL v5.0.4, https://fsl.fmrib.ox.ac.uk/fsl/fslwiki), followed by nuisance regression of six movement parameters (3dDeconvolve; AFNI), and slice timing (3dTshift, Fourier’s transformation; AFNI). The global signal was not regressed out to reduce the likelihood of introducing spurious negative activation measures in the subsequent analyses (16). ANATICOR (17) bandpass filtering (0.01-0.1 Hz) and motion scrubbing were performed in a single step (3dTproject; AFNI). In-scanner motion-induced corruption of EPI volumes was defined as framewise displacement >0.5 mm (Euclidean normalization); subjects with >30 corrupted volumes were excluded from all subsequent analyses, ensuring at least 5 min of rs-fMRI data for all subjects. The Desikan-Killiany atlas (18) containing information about the cortico-subcortical mapping was used to define spatial regions of interest (ROIs) – 34 cortical and 9 subcortical brain regions bilaterally including cerebellum, for a total of 86 nodes. The rs-fMRI data were not normalized or smoothed, and all analyses were done at the individual level to preserve subject’s native space. Regional time series were extracted for each ROI by averaging the time series of all voxels within each node. The pairwise Pearson’s and partial correlations of neural times series between two nodes in the network were carried out to generate individual (86 x 86) weighted undirected functional connectivity matrices (Matlab r2017b, Figure 1). Functional connectivity matrices and Pearson’s/partial coefficients distribution were visually inspected, screening for widespread and inflated positive or widely distributed correlation values within and across subjects.

Figure 1. Construction of functional resting-state connectivity matrices.

The average time series were correlated between cortico-subcortical regions to obtain undirected weighted functional connectivity matrices (Pearson’s, top image, and Partial correlation, bottom image). Functional connectivity matrices were used to for: (1) whole-brain network analysis (MATLAB r2017b), (2) permutation-based analysis (NBS v1.2), and (3) were combined with structural connectivity matrices in a structure-function coupling analysis (MATLAB r2017b).

Whole-brain functional connectivity measures

Global parameters summarizing whole-brain connectivity properties of BD functional networks included characteristic path length, global efficiency, global (positive/negative) strength and clustering coefficient, were calculated as the mean of the 86 regional estimates. Furthermore, a global measure of influence and centrality, betweenness centrality and network resilience, assortativity, were investigated (Brain Connectivity Toolbox v1.52) (19). Functional connectivity matrices were thresholded (r>0) to retain only positive weights due to computational difficulties and the trivial interpretability introduced by negative edges particularly for network measures that depend on shortest paths. Negative correlations where set to 0 for all measures, with the exception of whole-brain functional strength, excluding an average of 16% of the connections (16% controls, 16% patients). Recently, a test-retest reliability of functional connectivity measures reported that weighted whole-brain network metrics are more reliable than binarized ones (20); therefore, we obtained whole-brain functional measures using threshold weighted matrices. Statistical analyses were performed with diagnosis and gender as fixed factors covarying for age (*p<0.05; two-tailed; SPSS v23).

Permutation-based analysis of the Functional Connectome

A non-parametric statistical analysis, the network-based statistics (NBSv1.2) (21) was employed to perform mass univariate hypothesis testing at every functional connection comprising the graph to identify a differently functionally connected sub-graph component meanwhile controlling for the familywise error rate (FWER). Independent t-tests and ANCOVAs, with fixed factors diagnosis and gender (co-varying for age) were computed to test for group functional connectivity strength differences (M=5000, *p<0.05; one-tailed). Connections were thresholded (T=1.5-3.5) to obtain a set of suprathreshold connections which were tested for main effects of diagnosis, gender and gender-by-diagnosis interaction. The choice of primary threshold is a user-determined parameter; however, FWER control is guaranteed regardless of the threshold choice (21). NBS was run on functional connectivity matrices including both positive and negative weights. Post-hoc investigation of functional dysconnectivity of BD was carried out retaining positive or negative weights only; underpinning synchronous and anti-synchronous functional activity, respectively.

Structure-function coupling analysis

We conducted network structural-functional connectivity relationship analyses in line with previous research (22). In brief, structural matrices were derived within a graph-theoretical structural connectome analysis using a subject-specific cortico-subcortical non-tensor-based connectome approach (15); details in Supplemental Material. Non-zero structural connectivity weights, i.e., the number of streamlines (NOS-weights) were isolated and mapped using an inverse Gaussian transformation (22,23) in order to achieve a structural weight distribution which was practically bounded within a [+4; - 4] range (i.e., +/-4*sigma of an N(0,1) distribution). Prior to normalization, all NOS values below set thresholds were set to 0 (i.e., noise floor thresholding); group differences were carried out at the primary threshold T_NOS=1 (i.e., no noise thresholding) and confirmed across different thresholds (T_NOS=2-10) – thresholding effects on network density are depicted in Figure S1. The resulting NOS-weights were correlated (Pearson’s correlation) with the corresponding functional connectivity (Pearson’s) values, for each individual (22). Furthermore, FA-weighted matrices were also investigated by the present analysis (FA-weights were not rescaled). Group-level comparisons (MANOVA with fixed factors diagnosis and gender; Matlab r2017b) were performed at the whole connectome level and within hubs, feeder (hub-to-non-hub) and local (non-hub-to-non-hub) connections (22). Specifically, a single Pearson’s r value was extracted for each subject representing their whole-brain structure-function measure of integration; additionally, a Pearson’s r value was extracted for each subject for each connection class (hubs, feeder, local). Hubs were defined for BD and controls within a previous structural rich-club analysis performed in an overlapping clinical sample of patients (n=37, 93% overlap) and controls (n=45) at the statistically different subnetwork (k>30; Z=3.78, p<2.2^e-16) in BD relative to controls (15); furthermore, structure-function relationship was assessed for the statistically different (FA-weighted) subnetwork between BD and controls (T>1.5, p=0.031) defined in a previous subnetwork analysis that included limbic system and basal ganglia nodes (15). We used the same nodal parcellation scheme for both structural and functional networks to allow direct comparison between these two networks.

Correlations with clinical variables

Significant network measures were correlated with symptom severity as rated using the HDRS, HARS and YMRS clinical mood scales, age of onset and illness duration.

Results

Participants Clinical and Demographic Characteristics

A total of 41 participants with BD and 56 age, gender and education level-matched healthy volunteers were investigated. Patients and controls did not differ in age across diagnosis-by-gender subgroups (F(3,96)=1.25, p=0.298, Table 2). Thirty-three BD type I and 8 type II participants were included in the analyses. The vast majority of the BD participants were euthymic at MRI scanning (68%) with only n=13 displaying mild manic (YMRS>7), anxiety (HARS>18) or depressive (HDRS-21>8) signs and symptoms. Significant findings were confirmed for BD removing subjects not meeting criteria for euthymia (HDRS-21>8; YMRS>7), as well as those with HARS>18, and those with type II. Further, illness duration/age of onset had no effect on the significant network measures.

Table 2. Clinical and socio-demographic details of bipolar disorder and healthy controls.

Participants were age and gender matched across groups. *p<0.05. n=14 with HDRS >8; n=8 subjects were BD type II.

Sample	Healthy Controls	Bipolar Disorder	Statistical Comparison Diagnostic Groups, F, p-value
Number of participants	56	41	-
Age (years),    male, mean±SD    female, mean±SD	40.63±13.52 41.84±13.31 39.65±13.83	43.59±12.71 40.50±13.51 46.52±11.44	t(95)=1.09, p=0.277
Gender,    Male/Female (n)	25/31	21/20	X2(1)= 0.16, 0.686
Level of Education (SES scale)    median, range	6, 2-7	5, 2-7	X2(5)= 10.58, 0.060
Age of Onset (years),    mean±SD	-	26.6±10.0	-
Illness Duration (years),    mean±SD	-	16.4±10.7	-
Hamilton Depression Rating Scale (HDRS),    mean±SD    range    median	1.13±1.7 0-7 0	7.0±7.4 0-28 5	U=1,82, p<0.001*
Young Mania Rating Scale (YMRS),    mean±SD    range    median	0.8±1.5 0-6 0	1.9±2.6 0-10 1	U=1,45, p=0.010*
Hamilton Anxiety Rating Scale (HARS),    mean±SD    range    median	0.7±1.6 0-8 0	5.0±6.4 0-27 3	U=1,75, p<0.001*
Mood stabilizers medication naïve lithium only (0.4-1.2 g/day) sodium valproate only (0.3-1.4 g/day) lamotrigine only (0.05-0.45 g/day) combination	-	3 5 3 8 8	-
Antidepressants    SNRI/ SSRI/NaSSA	-	7/7/2	-
Antipsychotics    atypical/typical	-	29/1	-
Benzodiazepine	-	2	-
Other Psychotropic	-	9	-

Whole-brain measures of integration

The BD group whole-brain organization did not differ from that of controls for path length, global efficiency, global (positive/negative) strength, clustering coefficient, betweenness centrality and assortativity (F(7,86)=1.863, p=0.086; Table 3). Further, there was no main effect of gender (F(7,86)=0.770, p=0.614), nor gender-by-diagnosis interaction (F(7,86)=1.636, p=0.136). Analysis of whole-brain connectivity using partial correlations confirmed stability of whole-brain resting-state networks in BD relative to controls.

Table 3. Whole-brain functional connectivity network measures.

The BD group global network organisation did not differ from that of controls across all whole-brain network measures. Specifically, there was no main effect of (a) diagnosis (F(7,86)=1.863, p=0.086), (b) gender (F(7,86)=0.770, p=0.614), or (c) interaction between gender and diagnosis (F(7,86)=1.636, p=0.136) in BD relative to controls. MANCOVA, *p<0.05.

	Healthy Controls (mean±SD)	Bipolar Disorder (mean±SD)	(a) Statistical Comparison Diagnosis (F, p-value)	(b) Statistical Comparison Gender (F, p-value)	(c) Interaction Gender and Diagnosis ((F, p-value)
Positive Strength Male Female	15.44±3.70 14.95±3.84 15.85±3.57	15.93±4.06 16.7±4.42 15.20±3.65	0.52, 0.47	0.12, 0.73	2.07, 0.07
Negative Strength Male Female	2.90±1.24 3.16±1.31 2.69±1.16	3.02±1.06 2.99±0.98 3.05±1.16	0.12, 0.60	0.77, 0.38	0.96, 0.33
Betweenness Centrality Male Female	244.77±19.64 244.88±23.27 244.69±16.55	251.11±22.16 252.28±27.24 250.01±16.55	2.65, 0.11	0.03, 0.86	0.000133, 0.99
Global Efficiency Male Female	0.27±0.04 0.28±0.05 0.27±0.04	0.27±0.04 0.26±0.03 0.27±0.04	0.83, 0.36	1.12, 0.29	3.5, 0.07
Characteristic Path Length Male Female	4.42±0.54 4.48±0.54 4.36±0.54	4.38±0.53 4.27±0.56 4.47±0.5	0.23, 0.63	0.13, 0.72	1.92, 0.17
Clustering Coefficient Male Female	0.17±0.05 0.17±0.05 0.18±0.04	0.18±0.05 0.19±0.05 0.17±0.04	1.04, 0.31	0.29, 0.59	2.56, 0.11
Assortativity Male Females	0.05±0.06 0.05±0.06 0.05±0.06	0.05±0.06 0.05±0.06 0.05±0.06	0.20, 0.89	0.02, 0.89	0.22, 0.64

Permutation-based subnetwork analysis

We identified a functionally dysconnected subnetwork for BD relative to controls comprising parietal, cingulate and fronto-temporal (synchronous/anti-synchronous) functional connections (T-test: Pearson’s: T>3, p=0.039; Partial: T>3.5, p=0.037; Figure 2A, Table S1). A subnetwork of comparable dysconnectivity (77% overlap) was observed using positive correlations only (post-hoc T-test Pearson’s: T>3, p=0.048, Figure 2B-C, Table S2). We did not detect increased functional connectivity in BD relative to controls. When covarying for age and gender (ANCOVA), we observed a comparable (synchronous/anti-synchronous) dysconnected subnetwork in BD compared to controls (Pearson’s: T>3, p=0.039; not using Partial). No weaker/stronger sub-network was associated with gender nor diagnosis-by-gender interaction. A small subnetwork of stronger fronto-temporal (anti-synchronous)components was observed in BD relative to controls (post-hoc T-test: Pearson’s: T>3, p=0.014, Table S2; Figure 2B-C), and a weaker subnetwork of (anti-synchronous) components involving superiorfrontal, inferiortemporal and postcentral gyri alongside caudate and hippocampus connections was recorded in females relative to males (Partial: ANCOVA, T>3, p=0.025, but not Pearson’s).

Figure 2. Permutation-based subnetwork analysis of the functional connectivity matrices.

(A) significant original subnetwork of synchronous/anti-synchronous components (Pearson’s, T-test, T>3, p=0.039, in orange) with patients showing a differential subnetwork of (synchronous/anti-synchronous) components relative to controls. (B) post-hoc significant subnetworks of synchronous (in yellow) and anti-synchronous (in red) components, as well as the original significant (synchronous/anti-synchronous) subnetwork (in orange) plotted together. Bipolar disorder, relative to controls, showed a weaker subnetwork of functional (synchronous, Pearson’s T-test, T>3, p=0.048) components and a stronger subnetwork of functional (anti-synchronous, Pearson’s T-test, T>3, p=0.025) components. Synchronous components explain most functional connections within the original significant subnetwork. (B) ring-like visualisation of the original and post-hoc subnetworks of connected components. Images were obtained using NeuroMArVL software (http://immersive.erc.monash.edu.au/neuromarvl/).

Structure-function coupling analysis

Structure-function association analysis was performed on 38 BD and 45 age (U=1,042, p=0.087) and gender-matched controls (X²(1)=0.092, p=0.762). There was no main effect of diagnosis (F(8,72)=0.521, p=0.837; Figure 3), gender (F(8,72)=1.345, p=0.236) or gender-by-diagnosis interaction (F(8,72)=0.353, p=0.941) using FA (range of Pearsons’ BD r:-0.06 to -0.11; HC: r:-0.06 to - 0.12), or NOS-weights across the primary threshold (T=1; range of Pearsons’ BD r:0.20 to 0.28; HC: r:0.19 to 0.27). These were confirmed across all thresholds (T=2-10). Additionally, no difference was observed when increasing the rich-club nodal definition at 70% of connections common to participants.

Figure 3. Structural-Functional coupling analysis across whole-brain and edge-class connections.

Plots are shown for Pearson’s correlation values between functional connectivity matrices and FA-weighted (top) and NOS-weighted (bottom, T=1) structural connectivity matrices; Median±SD. Across all measures, there was no main effect of diagnosis (F(32,48)=1.337, p=0.178), no main effect of gender (F(32,48)=0.952, p=0.551), and no diagnosis-by-gender interaction (F(32,48)=0.356, p=0.999). Across all considered measures, there was no main effect of diagnosis (F(32,48)=1.337, p=0.178), no main effect of gender (F(32,48)=0.952, p=0.551), no diagnosis-by-gender interaction (F(32,48)=0.356, p=0.999). These findings were confirmed across the other considered threshold (NOS-weights: T=2-10). ‘Whole-brain’ represents representing structure-functioncoupling of all nodes in the network; ‘Hubs’ representing structure-function coupling of rich-club hubs; ‘Feeder’ representing structure-function coupling of nodes connecting to rich-club hubs; ‘NBS nodes’ representing structure-function coupling of nodes implicated in the statistically different NBS subnetwork between bipolar disorder and controls; ‘Nodes feeder’ representing structure-function coupling of connecting to the NBS nodes; ‘Local’ representing structure-function coupling of nodes composing the remaining of the network.

Discussion

We examined the functional organization of BD networks both globally and locally and reported preserved whole-brain functional connectivity with localized differences involving nodes belonging to default-mode, fronto-temporal and limbic systems, relative to controls. These findings appeared to be trait-features of BD type I. Furthermore, the present study provides preliminary evidence of preserved structure-function relationships globally and within edge-class in euthymic individuals with BD, relative to controls.

Whole-brain measures of connectivity

We did not detect changes in whole-brain functional connectivity in BD relative to controls; this is in line with previous studies conducted in euthymic BD individuals (4,13,24,25) and in young BD individuals with (mild) depression (12). Changes in whole-brain functional connectivity have been reported in actively depressed or unmedicated subjects with BD relative to controls, defined by longer paths and lower global efficiency (26,27). Therefore, functional large-scale changes may be characteristic of unmedicated and symptomatic patients with BD, and our findings of preserved whole-brain functional connectivity may be considered a trait-feature, rather than state, of this illness. This was supported by preserved whole-brain (weighted) structural connectivity in an overlapping clinical sample relative to controls (15), and collectively suggest that network-level abnormalities in BD do not extend to be detected globally, rather they may be localized to specific subnetworks.

Permutation-based analysis of the functional connectivity matrices

Although there was no difference in whole-brain functional connectivity in BD, changes were observed at the subnetwork level. Nodes thought to be associated with the DMN such as the prefrontal, caudal middle frontal and posterior cingulate gyri were implicated in the significantly reduced (synchronous) subnetwork in BD (Figure 2). This suggest reduced functional interactions in BD between structures that play key roles in self-referential thinking and emotion-regulation processing and cognitive control – features that are known to be functionally altered in BD (28). Post-hoc observations into BD functional network organization corroborated and expanded the reduced functional (synchronous) connectivity observed between fronto-limbic and parieto-occipital nodes to reveal increased (anti-synchronous) connectivity in a subnetwork encompassing fronto-temporo-parietal nodes in BD, relative to controls(Table S2). These abnormalities collectively present a functional map of BD dysconnectivity that differentially involves communication between regions spanning across multiple brain subsystems. This may underpin a compensatory mechanism of neural activity underlying whole-brain functional stability in BD, that may be necessary to sustain a remitted clinical state of the illness. Whereas Recent evidence suggests that anti-synchronous activity observed in fMRI is not an analytic artefact nor represent antagonistic relationships between brain regions but rather they are an expression of the different (dynamic) spatiotemporal reconfigurations of functional networks around the same anatomical skeleton (29).

Resting-state fMRI investigations of the brain have largely focused on the synchronous activation of regions of the DMN; these have shown to be particularly relevant to psychiatric disorders and a robust feature of BD functional dysconnectivity (6,11,30,31) (Table 1). Interestingly, increased functional connectivity within DMN nodes has been reported in unaffected siblings at high-risk compared to those with the illness (13) that may be considered a biological feature underlying resilience to BD (32). Therefore, a specific subset of regions emerges as being vulnerable to functional changes in BD that may be considered a viable target for future interventions to ameliorate symptoms.

Although the DMN is composed of selective regions that are thought to execute functions that are categorically different from those of other networks, there is evidence that this system may not act independently but rather be in continuum with other networks (33). This global organization was further supported by the observed high correspondence between cortical gradients of functional connectivity and myelin density across most cortical areas (34). Thus, it is important to understand the connectivity relationship between nodes belonging to this functional subsystem with other existing networks to further appreciate the DMNs functional role specifically in the context of mood regulation in BD. Furthermore, the fronto-limbic system has also been heavily implicated both structurally and functionally in the pathophysiology of BD due to its key role in emotion processing (2,3). Additionally, the fronto-parietal functional system participates in mechanisms of attention, memory and cognitive control (35) – cognitive features altered functionally in BD (36). Considering that regions belonging to the implicated functional subsystems have been shown to be anatomically in continuity with each other (33), they may depend on the synergistic functioning of each region to optimally underpin higher-order cognitive functioning specifically in BD.

We did not detect differences in regional or global functional connectivity in females with BD, despite our previous report of increased rich-club connectivity in this subgroup (15); although speculative, this ‘rewiring’ may be necessary to sustain a comparable functional state to that of controls. Future studies aimed at investigating gender differences across functional interactions, alongside increased sample size, may clarify these structural changes.

Structure-function relationship

To the best of our knowledge, this is the first study to explore structural-functional connectome coalescence in an adult sample of euthymic individuals with BD. Previous investigations include a young euthymic BD group (37) and offspring (38) reporting decreased structural-functional coupling in BD (37) and increased coupling in offspring (38) with structure-function breakdown involving intra-hemispheric and whole-brain connectivity (37) or long-range connections (38). We failed to detect any difference in structure-function associations in BD relative to controls globally and within connection classes. The discrepant findings could be explained by the different structural and functional network reconstruction methods employed, and the different clinical characteristics of these cohorts; furthermore, in BD, significant changes in structure-function coupling may be occurring and thus be detectable at the onset of the illness rather than be visible at later stages of the disease when the critical period of brain network development has passed.

Structural networks are thought to place significant physical constraints on functional connectivity both globally and locally (39) so that a change in the relationship between these two measures would be suggestive of illness expression (22,23). Crucially, while anatomical connections give rise and shape functional connections, it is likely to observe several possible spatiotemporal reconfigurations of functional connectivity expressed around the same anatomical skeleton, even within a short timescale (29,40). This implies that brain function is not static, but rather is dynamic in nature constantly switching between large-scale metastable wave patterns (41). Herein, preserved whole-brain structure-function coupling corroborates intact whole-brain structural (15) and functional connectivity in BD. However, in BD relative to controls, we observed preserved structure-function coupling within connection classes despite evidence of regional structural and functional deficits, suggestive of more complex, perhaps dynamic, interactions may be occurring between structure and function at the subnetwork level. Additionally, while anatomical connectivity may inform functional interactions, it is not per se a sufficient description of connectivity and optimal models should be identified to examine structure-function relationships (40).

We did not detect an effect of lithium on the significant functional connectivity measures; however, we may have been underpowered to investigate this outcome (BD on-lithium=14, off-lithium=27, noting that all, but 3, BD on-lithium were taking other medications). Furthermore, our findings can be considered trait-features of BD type I as these were confirmed when removing subjects with BD type II and those not meeting criteria for euthymia.

Collectively, BD is associated with localized functional dysconnectivity that does not extend to be detected globally, suggestive of reduced regional functional interactions in individuals with a history of psychotic and depressive episodes. These changes predominantly involve regions belonging to the default-mode network and limbic system both of which play key roles in several cognitive domains known to be functionally altered in BD. Although we observed structural deficits between and within basal ganglia and limbic system connections (15), basal ganglia did not appear to be involved in the functional dysconnected subnetwork. We conclude that these localized changes are suggestive of traitlike features of euthymic BD subjects which may be necessary to maintain a remitted clinical state of the illness. The applications of different conceptualizations of how information can flow around the human connectome, such as the dynamic representations of these functional systems, may be used to more comprehensively describe intermittent behaviors characteristic of neuropsychiatric disorders such as BD. Despite striking evidence of cognitive deficits in BD and its social and personal burden, there is no approved pharmacological treatment that is specific to the management of core symptoms of BD, possibly made challenging by the cyclic nature of BD illness and the wide array of symptoms and cognitive deficits individuals with BD experience. The theory that BD is a ‘dysconnection syndrome’ provides an intuitive explanation for the vast heterogeneity in symptomatology that characterizes the illness, as it is shifting away from localizing specific symptoms to specific grey matter and white matter regions or functional seeds moving towards the exam of abnormal interaction between brain regions considering the brain’s network as a whole. The observed decoupling of structural and functional connectivity in BD highlights the need to examine network abnormalities at both anatomical and physiological levels, as well as to incorporate multimodal imaging for a more meaningful understanding of dysconnectivity in psychotic illness such as BD.