Multiple Network Dysconnectivity in Adolescents with Psychotic Experiences: A Longitudinal Population-Based Study

Abstract

Abnormal functional connectivity (FC, the temporal synchronization of activation across distinct brain regions) of the default mode (DMN), salience (SN), central executive (CEN), and motor (MN) networks is well established in psychosis. However, little is known about FC in individuals, particularly adolescents, reporting subthreshold psychotic experiences (PE) and their trajectory over time. Thus, the aim of this study was to investigate the FC of these networks in adolescents with PE. In this population-based case-control study, 24 adolescents (mean age = 13.58) meeting the criteria for PE were drawn from a sample of 211 young people recruited and scanned for a neuroimaging study, with a follow-up scan 2 years later (n = 18, mean age = 15.78) and compared to matched controls drawn from the same sample. We compared FC of DMN, SN, CEN, and MN regions between PE and controls using whole-brain FC analyses. At both timepoints, the PE group displayed significant hypoconnectivity compared to controls. At baseline, FC in the PE group was decreased between MN and DMN regions. At follow-up, dysconnectivity in the PE group was more widespread. Over time, controls displayed greater FC changes than the PE group, with FC generally increasing between MN, DMN, and SN regions. Adolescents with PE exhibit hypoconnectivity across several functional networks also found to be hypoconnected in established psychosis. Our findings highlight the potential for studies of adolescents reporting PE to reveal early neural correlates of psychosis and support further investigation of the role of the MN in PE and psychotic disorders.

Introduction

Abnormal functional connectivity (FC, which measures the temporal synchronization of activation across distinct brain regions) is well established across the psychosis spectrum.^1–7 The typical age of onset of psychosis is between late adolescence and early adulthood.⁸ Thus, understanding when during development FC becomes disrupted may advance our understanding of the brain systems involved, help identify those at risk for psychosis, and provide opportunities for early intervention. Otherwise healthy young people who report subthreshold psychotic experiences (PE) are of specific interest in psychosis research, given the greater prevalence of these symptoms amongst the general population than the prevalence of psychotic disorders (5.8% and 3.0%, respectively).^9,10 The prevalence of PE is higher again amongst adolescents aged 13–18 at 7.5%.¹¹ Most importantly, PE have been associated with a relatively increased risk of psychiatric disorders, including psychosis,^12,13 suicidality,^14,15 and neurocognitive impairment.¹⁶ Investigating the integrity of large-scale functional networks in adolescents who report PE, therefore, has the potential to reveal network precursors of psychotic disorders and psychosis psychopathology.

FC studies in psychosis have demonstrated dysconnectivity within 3 primary large-scale networks: the default mode (DMN),¹⁷ the central executive (CEN),¹⁸ and the salience networks (SN).¹⁹ Meta-analyses of resting-state studies consistently describe network hypoconnectivity in established psychosis, primarily across the DMN and SN.^2,3 Similar abnormalities have been demonstrated in independent cross-sectional studies of childhood, young adulthood, and adult PE.^7,20–24 We previously reported findings from a preliminary cross-sectional analysis of a subsample of the current cohort (the “Adolescent Brain Development Study,” ABD²⁵), which also demonstrated significant hypoconnectivity of the DMN, CEN, and SN in the PE group, with the DMN displaying the most abnormalities.⁴ The DMN is involved in internally directed thought processes^26,27, whilst CEN engagement is related to external, goal-driven processes.²⁸ The SN is thought to modulate the engagement of the DMN and CEN as salient stimuli are detected,¹⁹ resulting in the allocation of attentional resources to the stimuli for further processing.²⁹ Aberrant salience attribution, and the subsequent dysfunctional engagement of the DMN and CEN, has been proposed as an explanation for the psychotic symptoms and cognitive impairments that characterize the disorder.^19,30,31

Although cognitive and perceptual disturbances are considered primary to psychosis, interest in the role of the motor network (MN) in psychosis has increased recently. This is due to emerging evidence supporting a relationship between the dysfunction of this network and the wider pathophysiology and etiology of psychosis beyond its overt role in the motor dysfunction observed in psychosis.^32–34 Across the psychosis spectrum, studies of FC have largely reported hyperconnectivity within the MN,^35,36 and more mixed connectivity findings between the MN and other regions.^21,36,37 A recent study in youth PE found MN dysconnectivity to be more pervasive than abnormalities of the DMN, SN, and CEN.²¹ Furthermore, motor dysfunction in individuals at clinical high risk for psychosis (CHR) has also been identified as a potential predictive marker for transition to psychosis.^38–40 Two studies using subsamples of the current ABD cohort observed significant fine-motor deficits amongst the PE participants at ages 11–13⁴¹ and 17–21⁴². However, the FC abnormalities possibly underlying these motor deficits have not yet been investigated.

Such findings highlight the need for further investigation of the dysconnectivity of these networks in PE specifically. Thus far, however, longitudinal approaches to the study of connectivity in PE have been neglected, and little is known about the trajectories of FC abnormalities over time in young people who report PE. Though cross-sectional studies provide a strong basis for the generation of hypotheses, longitudinal designs allow the capture of developmental trends over time—particularly important in the investigation of adolescent abnormalities in FC²¹.

With this in mind, the aim of the current study was to investigate the longitudinal changes in FC of regions representing the DMN, CEN, SN, and MN, with the rest of the brain, in a sample of adolescents who met the criteria for a PE. We hypothesized that FC would be significantly decreased primarily across regions of the DMN and SN, and increased between regions of the MN, in the PE group compared to the control group. Furthermore, these abnormalities will be apparent at both baseline and follow-up. These findings would support the dysconnectivity theory across the psychosis continuum and provide novel evidence of the trajectory of network dysconnectivity in PE over time and for the role of the MN in PE specifically.

Methods

Participants

A sample of 211 young people between the ages of 11 and 13 years old was recruited from primary schools in Dublin and Kildare, Ireland, as part of the ABD study²⁵ (recruitment details in supplementary material). All participants attended a diagnostic clinical interview with trained raters (recruitment and interview details outlined in Kelleher et al²⁵). Psychopathology was assessed by interviewing adolescents and parents separately, both answering the same questions about the adolescent, using the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS).⁴³ The psychosis section of K-SADS was supplemented by additional questions from the SOCRATES instrument, which was devised to systematically assess the presence of PE in youth populations.⁴⁴ All interviews were reviewed by a consensus committee (two psychiatrists and a psychologist) in order to confirm PE classification (further details in supplementary material).

A subsample of 100 participants with no contraindications to functional magnetic resonance imaging (fMRI) agreed to take part in a neuroimaging study that took place 1–3 years (mean 2 years) after the original interview and again 2 years later. Of the 100 participants scanned, 26 met the criteria for a PE in the original diagnostic interview. Twenty-five of this PE group were scanned at timepoint 1 (TP1) and an additional participant scanned at timepoint 2 (TP2). One PE participant was removed during the analysis due to structural abnormalities. Apart from the participant only scanned at TP2, 17 PE participants who took part in TP1 returned for TP2. From the remaining participants, 25 individuals without PE were selected to match the PE group for gender, handedness, age (at each time of scanning), socioeconomic status, and number of scans, forming the control group—one of whom was added only at TP2. Overall, at TP1: PE = 24, controls = 24; at TP2: PE = 18, controls = 18. Mean ages of both groups were 13.58 years at TP1 and 15.78 years at TP2. None of the participants in either group met the criteria for a formal psychotic disorder, and none had any history of neurological disorder (eg, epilepsy). Written parental consent and participant assent were obtained before the study began.

Resting-State Data Acquisition

Whole-brain fMRI data were acquired for each participant using a 3T magnetic resonance imaging system (Philips Achieva, Philips Medical Systems Netherland BV) in Trinity College Institute of Neuroscience, Dublin, Ireland. During scanning, participants were presented with a fixed white cross in the center of a black screen and were asked to look at this cross for the duration of the scan while blinking normally. T2*-weighted images were acquired continuously using a gradient echo planar imaging over 5:26 mins, with the following parameters: repetition time/echo time = 2000/27 ms; flip angle = 90°; 37- × 3.2-mm slices acquired parallel to the anterior-posterior commissure plane; Field-of-view = 240 mm; matrix = 80 × 80. High-resolution T1-weighted images were acquired with a turbo gradient echo 3D sagittal sequence using the following parameters: TE/TR = 8.4/3.9 ms; flip angle = 8°; 256 × 256 matrix; 180- × 0.9-mm slices; FoV = 230. Scan duration was 5:47 min.

Preprocessing of Functional Data

All data preprocessing and analysis were performed using the CONN functional connectivity toolbox (v18a),⁴⁵ via the SPM12 platform (Wellcome Department of Imaging Neuroscience, London, UK; www.fil.ion.ucl.ac.uk/spm). Briefly, preprocessing included functional realignment and unwarping, slice timing correction, structural segmentation of the data into typical tissue classes, normalization to the Montreal Neurological Institute template, resampling at 2 mm,³ functional normalization, Artifact Detection Tools (ART) based outlier identification/scrubbing (outlier criteria in supplementary material), and smoothing using an 8-mm full width at half maximum Gaussian filter (additional details in supplementary material).

FC Analysis

Whole-brain seed-to-voxel analyses were performed using the CONN toolbox. The networks and seed regions of interest were as follows: (1) the DMN—seeds included the precuneus, ventral anterior cingulate cortex, and the parahippocampal gyrus cortex¹⁷; (2) the SN—seeds included the dorsal anterior cingulate cortex and the insula¹⁹; (3) the CEN—seeds included the dorsolateral prefrontal cortices (BA9 and BA46, separately)¹⁸; and (4) the MN—seeds included the primary motor cortex M1, premotor cortex, and the anterior cerebellum (cerebellum IV–V).⁴⁶

Seed regions were identified via inbuilt CONN atlases (based on FSL Harvard–Oxford atlas cortical and subcortical areas,^47–50 Automated anatomical labelling atlas cerebellar areas,⁵¹ and standard Brodmann areas⁴⁵ and were all bilateral. The average BOLD signal time series over each of the seed regions were then extracted from each subject's brain space. Covarying for the movement and ART outlier artifacts identified during preprocessing, the strength and significance of the bivariate Pearson correlations between the time series of each seed region and the time course from all other brain voxels were then calculated within each subject's brain space using general linear modeling (outlier details in supplementary material).

Second-level random-effects analyses were used to create group connectivity maps (ie, for the PE group and the control group), averaging the correlation coefficients resulting from the first-level analysis across participants for each seed region separately. Two-sample ANOVAs were used to identify cross-sectional differences in whole-brain seed-to-voxel FC between the PE group and the control group at TP1 and TP2. Separate repeated-measures ANOVAs were used to identify longitudinal FC differences within each group across timepoints (in these cases, only participants present for both TP1 and TP2 were included; thus, PE = 17 and controls=17). Finally, mixed-measures ANOVAs were used to assess group by time effects on FC (again, only participants present for both TP1 and TP2 were included). In each case, the individual seeds representing each network were analyzed separately (ie, individual ANOVA analyses were run for each seed). Results were initially thresholded at a false discovery rate whole-brain corrected cluster-level of P < .05. P was then adjusted to .005 to correct for the number of seeds being compared.

Results

Demographics

Participant demographic information is summarized in table 1. All participants were antipsychotic naïve at both timepoints. At TP2, 10 PE participants met the criteria for the reoccurrence of PE.

Table 1.

Participant sociodemographic information

Characteristic	PE (TP1 n = 25; TP2 n = 18)	Controls (TP1 n = 25; TP2 n = 18)	Statistics
	Mean (SD)
Age in years at TP1	13.58 (1.25)	13.58 (1.28)	P = 1
Age in years at TP2	15.78 (1.26)	15.78 (1.26)	P = 1
Gender (%male)	64	64	P = 1
Handedness (%right)	96	96	P = 1
Socioeconomic status	2.83 (1.5)	2.37 (1.013)	P = 0.55
TP2 persistent (reoccurring) PEs	10 (8 had no reoccurrence of PEs by TP2)

Note. Socioeconomic status was established via highest parental occupation level, categorized as follows: 1 = professional work, 2 = managerial and technical work, 3 = nonmanual work, 4 = skilled manual work, 5 = semiskilled work, 6 = unskilled work, 7 = unemployed.

PE, psychotic experiences group; TP1, timepoint 1; TP2, timepoint; BA, Brodmann area; MNI, Montreal Neurological Institute; FDR, false discovery rate corrected; L, left; R, right; FC, functional connectivity; DMN, default mode network; CEN, central executive network; SN, salience network; MN, motor network; Lat., lateral; dACC, dorsal anterior cingulate cortex; vACC, ventral anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; PHG, parahippocampal gyrus; SFG, superior frontal gyrus; MTG, middle temporal gyrus; ITG, inferior temporal gyrus; IFG, inferior frontal gyrus.

Differences in Whole-Brain Seed-to-Voxel FC Between and Within the PE and Control Groups

Cross-Sectional Between-Group Differences

At TP1, there were no significant group differences for the DMN, CEN, or SN regions. For the MN, PE participants exhibited significantly weaker FC between the bilateral cerebellum IV–V seed and right precuneus/superior lateral occipital cortex relative to controls (figure 1; table 2).

Fig. 1.

Baseline and follow-up cross-sectional functional connectivity (FC) abnormalities. Select peak regions are displayed, with corresponding Montreal Neurological Institute z coordinates above each brain slice. (A) Baseline—FC in psychotic experiences (PE) was weaker between the bilateral cerebellum IV–V seed (motor network [MN]) and the R. precuneus/lateral occipital cortex (z = 66, default mode network) relative to controls. (B) Follow-up—FC in PE was weaker between the L. primary motor cortex seed (M1, MN) and the L. occipital pole (z = −6, visual region). (C) Follow-up—FC in PE was weaker between the R. primary motor cortex seed (M1, MN) and the L. precentral gyrus (z = 70, MN). L. = left, R. = right.

Table 2.

Between-group differences in FC between the DMN, CEN, SN, and MN, and the rest of the whole brain at baseline and follow-up

	Seed network	Difference contrast	Seed	Effect anatomy	Cluster peak MNI coordinates			Cluster size	Cluster PFDR value
					x	y	z
Baseline	MN	PE < controls	L. cerebellum IV–V	Precuneus, lateral occipital cortex (bilateral superior), superior parietal lobule	8	−68	66	241	.0024
		PE < controls	R. cerebellum IV–V	Lat. occipital cortex (superior), precuneus	8	−68	66	254	.0013
Follow-up	CEN	PE > controls	R. dorsolateral prefrontal cortex (BA46)	Insular cortex, Heschl's gyrus, planum polare, central operular cortex	42	−14	8	225	.0031
	MN	PE < controls	L. primary motor cortex	Occipital pole	−20	−94	−6	279	.0013
		PE > controls		Frontal operculum cortex, Insular cortex	38	22	8	226	.0022
		PE < controls		Precentral gyrus, postcentral gyrus, superior frontal gyrus	42	−20	64	218	.0022
		PE > controls	R. primary motor cortex	Superior frontal gyrus, middle frontal gyrus	−24	24	50	341	.00018
		PE < controls		Precentral gyrus, postcentral gyrus	−30	−20	70	250	.0009
		PE > controls		Frontal pole	−22	60	8	193	.0029
Group × Time	DMN	PE ↓ : controls ↑	Precuneus	Insular cortex, temporal pole	36	4	−8	245	.00066
		PE ↓ : controls ↑		Supramarginal gyrus (anterior), parietal operculum cortex	−58	−36	36	158	.0049
	SN	PE ↑ : controls ↓	L. insular cortex (BA13)	Frontal pole, middle frontal gyrus, inferior frontal gyrus (pars)	−32	32	16	224	.002
		PE ↓ : controls ↑		Precuneus cortex, supracalcarine cortex	22	−54	8	203	.002
		PE ↑ : controls ↓	R. insular cortex (BA13)	Cerebellum VI, Vermis VI, Vermis VII, Crus I	−10	−68	−22	214	.0014
		PE ↓ : controls ↑		Precuneus cortex	8	−52	44	212	.0014

Note. Cluster size in voxels; cluster peaks are indicated in bold. Abbreviations are explained in the first footnote to table 1.

At TP2, the PE group exhibited significantly weaker FC between the left M1 seed (MN) and the left occipital pole (figure 1; table 2) and the right precentral gyrus and between the right M1 seed (MN) and left precentral gyrus (figure 1; table 2) relative to controls (table 2). The PE group also exhibited significantly stronger FC between the right dorsolateral prefrontal cortex (CEN) and right insular cortex (table 2), between the left M1 seed (MN) and frontal operculum/insular cortex (table 2), and between the right M1 seed (MN) and left superior frontal gyrus and left frontal pole (table 2).

Longitudinal Within-Group Differences

The within-group analyses revealed greater changes within the controls over time than in the PE group, with the majority of the changes occurring between the DMN and SN regions and between MN regions. Briefly, in the controls, FC between MN regions, between DMN regions, and between DMN and SN regions strengthened over time (table 3). In the PE group, the majority of changes occurred between DMN regions and between SN and DMN regions, with 2 instances of FC changes between CEN and DMN regions, though the directions of these changes were more mixed than in the control group (table 4). No significant differences in MN-related FC across the timepoints were observed within the PE group.

Table 3.

Within-group differences in FC over time between the DMN, CEN, SN, and MN, and the rest of the whole brain, for the control group

Seed network	Difference contrast	Seed	Effect anatomy	Cluster peak MNI coordinates			Cluster size	Cluster pFDR value
				x	x	x
DMN	TP2 > TP1	L. vACC (BA24)	Precuneus, posterior cingulate	18	–56	28	425	< 0.0001
	TP2 > TP1		Angular gyrus, lateral occipital cortex	36	–48	20	209	0.00024
	TP1 > TP2		Precentral gyrus, postcentral gyrus	54	0	34	208	0.00024
	TP1 > TP2	R. vACC (BA24)	Precentral gyrus, postcentral	58	–12	24	277	< 0.0001
	TP2 > TP1	Precuneus	Insular cortex, central opercular cortex, precentral gyrus, frontal operculum, heschl's gyrus, temporal pole, planum polare	36	12	4	594	< 0.0001
	TP2 > TP1		Central opercular cortex, insular cortex, planum polare, heschl's gyrus	–46	–4	4	229	0.00011
	TP2 > TP1		dACC, paracingulate	–12	10	42	190	0.0003
	TP2 > TP1	L. PHG complex (BA36)	Paracingulate gyrus, dACC, SFG	–10	20	46	212	0.00039
	TP2 > TP1		Frontal pole	32	42	26	137	0.0035
	TP1 > TP2	R. PHG complex (BA36)	MTG, ITG, lat occip ortex	–46	–46	–8	314	< 0.0001
SN	TP1 > TP2	L. dACC (BA32)	Precentral gyrus, post central gyrus	–38	–18	70	180	0.0016
	TP2 > TP1	L. insular cortex (BA13)	Angular gyrus, lat occip cortex, supramarginal gyrus	60	–58	22	497	< 0.0001
	TP1 > TP2		Putamen, pallidum, caudate	18	14	0	259	< 0.0001
	TP2 > TP1		Precuneus cortex	18	–62	26	197	0.0005
	TP2 > TP1		Precuneus cortex	18	–46	38	154	0.0019
	TP2 > TP1		IFG, MTG, temporal pole	46	–2	–32	148	0.0019
	TP2 > TP1	R. insular cortex (BA13)	Precuneus cortex, posterior cingulate	–12	–46	42	481	< 0.0001
	TP2 > TP1		Angular gyrus, lat occip cortex	58	–58	30	335	< 0.0001
	TP2 > TP1		Lat occipital cortex	–44	–76	24	167	0.0011
	TP2 > TP1		Temporal pole, MTG, ITG	50	–6	–30	164	0.0011
	TP1 > TP2		Cerebellum 6, occip fusiform gyrus, lingual gyrus, crusI	16	–72	–16	145	0.0018
	TP1 > TP2		Cerebellum VI, lingual gyrus	–12	–72	–16	124	0.0036
MN	TP2 > TP1	L. primary motor cortex	Precentral gyrus, postcentral gyrus	–28	–22	70	271	<0.0001
	TP2 > TP1		Postcentral gyrus, precentral gyrus	36	–30	66	264	<0.0001
	TP2 > TP1	R. primary motor cortex	Postcentral gyrus, precentral gyrus	40	–28	66	590	<0.0001
	TP2 > TP1		Precentral gyrus, postcentral gyrus	–26	–32	54	350	<0.0001
	TP1 > TP2		Superior frontal gyrus, middle frontal gyrus	–20	22	50	188	0.00044

Note. Cluster size in voxels; cluster peaks are indicated in bold. Abbreviations are explained in the first footnote to table 1.

Table 4.

Within-group differences in FC over time between the DMN, CEN, SN, and MN, and the rest of the whole brain, for the PE group

Seed network	Difference contrast	Seed	Effect anatomy	Cluster peak MNI coordinates			Cluster size	Cluster pFDR value
				x	y	z
DMN	TP1 > TP2	Precuneus	Middle frontal gyrus, frontal pole, inferior frontal gyrus (pars)	–48	38	26	246	0.00015
	TP2 > TP1		Frontal medial cortex, frontal pole, paracingulate gyrus	10	54	–10	165	0.0015
	TP1 > TP2		Supramarginal gyrus (anterior)	–58	–36	36	138	0.003
SN	TP2 > TP1	L. dACC (BA32)	Precuneus cortex	6	–58	50	216	0.00036
	TP2 > TP1	L. insular cortex (BA13)	Frontal pole	38	54	12	127	0.0049
	TP2 > TP1	R. insular cortex (BA13)	Inferior temporal gyrus	–48	–54	–12	163	0.0014
CEN	TP1 > TP2	L. DLPFC (BA46)	Precuneus cortex	6	–54	68	193	0.00075
	TP2 > TP1		Middle temporal gyrus, supramarginal gyrus (posterior), middle temporal gyrus (posterior), superior temporal gyrus (posterior)	44	–38	2	171	0.00088

Note. Cluster size in voxels; cluster peaks are indicated in bold. Abbreviations are explained in the first footnote to table 1.

Group × Time

The group × time analyses revealed significant interactions for the DMN, specifically, between the precuneus seed and clusters with peaks in the right insular cortex and left anterior supramarginal gyrus (table 2). In both instances, a crossover interaction effect was observed, with FC decreasing over time in the PE group and increasing over time in the controls. Significant crossover interactions were also observed between the left insular cortex (SN) and left frontal pole, with FC increasing in the PE group and decreasing in the controls over time (table 2). Where the right insular cortex (SN) was the seed, FC with the left cerebellum VI increased in the PE group and decreased in the controls over time (table 2). For the bilateral insular seeds, FC with the right precuneus decreased in the PE group and increased in the controls over time (table 2). No significant group × time interactions were observed for the CEN or MN seed regions.

Observational Findings

Several of the changes within the control group over time overlapped with the cross-sectional and group × time findings. Specifically, the increase in FC between MN regions (primary motor cortex and precentral and postcentral gyri) over time in the controls corresponded with the follow-up findings (ie, weaker FC in the PE group between these regions). Similarly, the increase in FC between DMN and SN regions (precuneus and insula, respectively) and decrease between SN and MN regions (insula and cerebellum VI, respectively) and also between SN and CEN regions (insula and frontal gyri, respectively) in controls over time corresponded with the group × time findings (ie, decrease in FC between the precuneus and insula and increase in FC between the insula and both the cerebellum VI and frontal cortex in the PE group).

Discussion

In keeping with the hypotheses, we observed significantly altered resting-state FC across all networks in the PE group compared to controls. Specifically, the PE group largely displayed hypoconnectivity at both timepoints, while the controls demonstrated more instances of significant within-group change across all the networks over time (primarily increases in connectivity) corresponding to the cross-sectional and group × time findings. At both the baseline and 2-year follow-up, MN abnormalities were the most prevalent.

Cross-Sectional Findings

No significant within- or between-network group differences were observed for the DMN, SN, and CEN regions at baseline. However, we demonstrated extensive FC abnormalities of the MN at both baseline and follow-up. Specifically, connectivity in the PE group at baseline was weaker between bilateral motor regions and regions of the posterior DMN compared to the controls. Little has been reported on resting-state dysconnectivity between the MN and the DMN, SN, and CEN across the psychosis spectrum. This is particularly true for the MN and the SN and CEN. It is worth noting that the few instances of hyperconnectivity occurred between these 3 networks. One recent study in a CHR youth sample identified weaker dynamic FC between the posterior cingulate cortex (DMN) and the precentral gyrus (MN),⁶ though these alterations were not associated with any symptomology. Studies of Parkinson's and Huntington's diseases, both neurodegenerative disorders characterized in part by dopamine-related motor dysfunction, have described abnormal connectivity between the DMN and the primary motor cortex.^52,53 Such dysconnectivity may lead to deficits in motor-related mental imagery, integration of visuomotor information, and executive control of movement.^52,53 This subsample of the ABD cohort did not perform the fine-motor skills task; thus, it was not possible to investigate the relationship between DMN–MN connectivity and motor performance. Nonetheless, MN dysconnectivity could underlie the impairments in motor function and, possibly, those in cognitive function, observed in PE groups and groups at high risk for psychosis.^34,41,42,54 Furthermore, as abnormal FC of the DMN has been linked to primary aspects of psychosis psychopathology,^55–58 it is possible that the proposed relationship between motor circuit dysfunction and wider psychosis psychopathology may be mediated via functional connections with the integrative hubs of the large-scale DMN. To explore this further, future studies should investigate any effects of DMN–MN connectivity on cognitive-motor functioning across the psychosis spectrum, and how this may vary over time.

Hypoconnectivity between MN regions (bilateral primary motor cortices and precentral gyri) became more prominent at follow-up in keeping with the findings of a recent meta-analysis of resting-state FC in schizophrenia exploring a range of networks.²⁴ Within-MN hypoconnectivity may be indicative of a neural inefficiency related to psychosis vulnerability as previous longitudinal studies in individuals who are at risk for psychosis have demonstrated associations between dysconnectivity of motor regions and transition to psychosis.^54,59 In contrast, hyperconnectivity is commonly reported between the thalamus (the primary relay center for motor and sensory signals to the cerebral cortex) and sensorimotor regions in both schizophrenia and CHR samples.^{35,36,60–63} Hypoconnectivity and hyperconnectivity of motor regions with divergent functions may underlie separate deficits; ie, planning and execution of movements (related to the primary motor cortex and precentral gyrus); and the sensory integration and feedback of information between the thalamus and other motor areas. Given these findings, and the extent of the pathways connecting the thalamus to the rest of the brain, a detailed investigation of longitudinal sensorimotor and thalamic subregion FC in PE is warranted.

Reduced connectivity between motor and visual regions has been demonstrated previously in an independent youth PE sample,²¹ and in individuals with schizophrenia.³⁷ Notably, the occipital cortex encompasses the majority of the extrastriate body area involved in both the visual perception of the human body and body parts and in the planning, execution, and imagining of limb movements.⁶⁴ The “internal forward model system” suggests that the ability to differentiate between self and others is underpinned by motor control processes.⁶⁵ This ability is impaired in psychosis, resulting in passivity symptoms (ie, the belief that one's thoughts or actions are influenced or controlled by an external force).⁶⁶ It is possible that reduced FC between visual and sensorimotor areas, as observed in the current PE group, could underlie impairments in self/other action processing.³⁷

Longitudinal Findings

The typically developing control group demonstrated greater changes in FC over time than the PE group. This reflects previous findings in childhood and early adolescent development, describing delays in neurocognitive growth for individuals with psychotic symptoms,⁶⁷ and in white matter growth for siblings of patients with childhood-onset schizophrenia,⁶⁸ compared to controls. This suggests that the dysconnectivity observed here is underpinned by a delay in FC development in groups that are vulnerable to psychosis, adding a dimension to the neurodevelopmental hypothesis of psychosis. Furthermore, in the study involving siblings of patients, the lag in white matter development was found to normalize with age (ie, >14 years). Age-based normalization in groups who meet the criteria for PE should be explored in future studies.

It is notable that abnormal connectivity related to MN seed regions was the only finding at baseline, and the most prevalent finding for the PE group at the follow-up, in keeping with the findings of Mennigen et al²¹. This could establish MN dysfunction as an early biomarker for PE prior to any overt dysfunction of the traditional resting-state networks. Given the trend for the baseline PE dysconnectivity to extend between MN and DMN regions, and spread more globally by follow-up, this restructuring could also underlie a compensatory mechanism resulting from inefficient activity in broader cortical networks.

Group × Time Findings

For the majority of the crossover findings, FC increased in the controls over time, and decreased in the PE group, in keeping with a recent activation study in adolescent PE⁶⁹. Together with the cross-sectional and longitudinal findings, this further implicates aberrant neurodevelopment as a potential neural underpinning of PE⁶⁹. Both the mixed measures and follow-up between groups findings are also consistent with literature relating to established psychosis, generally describing hypoconnectivity within the DMN, and between the DMN and SN,^2,3,70 and supports the key roles for the SN and DMN proposed by the large-scale network model of psychopathology.¹⁹

These findings should be considered in view of certain limitations. Primarily, due to the sample size, it was not possible in the current analyses to investigate differences between transient (PE reported in childhood but not in adolescence) and persistent PE. It is possible that FC abnormalities may occur in a continuum, reflecting the full spectrum of PE and psychotic disorders. Nevertheless, though roughly 80% of cases are transient⁷¹; neurocognitive deficits,⁴² poorer global functioning,^72,73 and greater risk for later psychopathology⁷⁴ are still observed in those with transient PE compared to healthy controls. Notably, 8 of the returning participants who met the criteria for a PE at baseline had no reoccurrence of PEs by follow-up. Thus, the current findings demonstrate an ongoing effect on FC related to childhood PE, regardless of later reoccurrence. Additionally, due to the varied times between clinical interviews and scans, analysis of any associations between current FC findings and previous symptom data were avoided as these may be misleading. Such investigations should be considered in future studies to achieve a greater understanding of the implications of specific patterns of dysconnectivity. Strengths of the current study include all participants being antipsychotic naïve at both timepoints and having no diagnosis of psychotic disorder. Thus, it was possible to avoid the potentially confounding effects of antipsychotic treatment and illness chronicity. Furthermore, though differences are observed across a range of network regions, the overlap between the within- and between-group findings support their reliability. Additional methodological considerations are discussed in the supplementary material.

In conclusion, largely in keeping with our hypotheses, we identified hypoconnectivity across the DMN, SN, AN, and MN, across 2 timepoints, reflecting findings in established psychosis. These findings may provide age-based markers for PE during adolescence and provide insight into the impaired neural development in adolescents with PE.

Funding

This research was funded by the European Research Council (ERC 724809 to M.C.) and the Irish Health Research Board (HRA‐PHR‐2015‐1323 to M.C.).

Conflicts of Interest

The authors have declared that there are no conflicts of interest in relation to the subject of this study.