Salience–Default Mode Functional Network Connectivity Linked to Positive and Negative Symptoms of Schizophrenia

Abstract

Schizophrenia is a complex, debilitating mental disorder characterized by wide-ranging symptoms including delusions, hallucinations (so-called positive symptoms), and impaired motor and speech/language production (so-called negative symptoms). Salience-monitoring theorists propose that abnormal functional communication between the salience network (SN) and default mode network (DMN) begets positive and negative symptoms of schizophrenia, yet prior studies have predominately reported links between disrupted SN/DMN functional communication and positive symptoms. It remains unclear whether disrupted SN/DMN functional communication explains (1) solely positive symptoms or (2) both positive and negative symptoms of schizophrenia. To address this question, we incorporate time-lag-shifted functional network connectivity (FNC) analyses that explored coherence of the resting-state functional magnetic resonance imaging signal of 3 networks (anterior DMN, posterior DMN, and SN) with fixed time lags introduced between network time series (1 TR = 2 s; 2 TR = 4 s). Multivariate linear regression analysis revealed that severity of disordered thought and attentional deficits were negatively associated with 2 TR-shifted FNC between anterior DMN and posterior DMN. Meanwhile, severity of flat affect and bizarre behavior were positively associated with 1 TR-shifted FNC between anterior DMN and SN. These results provide support favoring the hypothesis that lagged SN/DMN functional communication is associated with both positive and negative symptoms of schizophrenia.

Introduction

The abnormal salience monitoring theory of schizophrenia (Sz) proposes that abnormal functional communication between the salience network (SN) and default mode network (DMN) begets wide-ranging symptoms including hallucinations, disorganized thought, and psychomotor poverty.^1,2 When healthy subjects perform cognitive tasks requiring externally focused attention, the DMN deactivates and regions essential for executive functioning (eg, lateral prefrontal and parietal cortex) become active; DMN hubs include medial prefrontal cortex (MPFC)/anterior cingulate cortex (ACC; anterior midline), posterior cingulate/precuneus (posterior midline), and angular gyrus (posterior lateral).^3,4 Both anterior and posterior midline hubs have strong structural connections to limbic regions involved in emotion and memory.⁵ But studies exploring DMN function during rest and across different tasks suggest that anterior DMN (aDMN) and posterior DMN (pDMN) hubs may play specialized functional roles. Tasks requiring explicit self-reference preferentially activate MPFC,⁶ whereas posterior midline hubs are thought to integrate self-referential judgments and play an important role in autobiographical memory.^6–8 Finally, 2 studies exploring effective (directional) connectivity within the DMN reported that the anterior prefrontal cortex acts as a sink of propagated activity (eg, anterior prefrontal activity lags behind activity of pDMN hubs).^9,10

The SN plays a critical role in monitoring the proximal salience of cues—from startling noises to changes in homeostatic state. The anterior insular hub receives convergent input from visual and auditory cortex,^11–14 whereas the dorsal ACC hub projects to the spinal cord.¹⁵ These connections allow the SN to integrate incoming perceptual information and respond quickly when confronted with salient changes to internal states of the body and external states of the environment.¹⁵ Diminished white matter integrity of anterior insula–dorsal anterior cingulate tracts in individuals with traumatic brain injury disrupts normal patterns of DMN activation/deactivation,^16,17 suggesting that the SN is required for regulating DMN activation.

Patients with Sz have an attenuated ability to deactivate DMN during task performance^18–20 and elevated DMN resting-state functional connectivity (rs-FC).^21–23 These abnormalities are associated with global assessments of positive symptoms (eg, delusions, hallucinations, and disorganized speech),¹⁸ working memory deficits,²⁴ social deficits,²⁵ and hallucinations.^26,27 Depressed rs-FC with SN hubs in Sz is linked to hallucinations,^26,28 general assessments of reality distortion (hallucinations + delusions),²⁹ and defective error monitoring.³⁰

An innovative study by Manoliu et al²⁶ first examined resting-state functional magnetic resonance imaging (rs-fMRI) signal coherence of DMN and SN in Sz, and reported that the patients with Sz had seemingly normal rs-FC between the SN and the DMN relative to healthy controls. Next, a series of time-lag-shifted rs-FC analyses²¹ explored rs-fMRI signal coherence of SN and DMN, but introduced fixed time lags between network time series. When time lags of 1 TR (2 s) and 2 TRs (4 s) were introduced between network time series, Sz had abnormal rs-FC between DMN and SN relative to healthy controls. However, the researchers did not explore potential associations between symptom severity and time-lag-shifted rs-FC between DMN and SN.

Prior studies have predominately reported links between disrupted SN/DMN functional communication and positive symptoms of Sz.^18,26–28 Yet, we know that the SN and the DMN play indispensable roles in monitoring internal and environmental states, and orienting attention. At present, it remains unclear whether altered SN/DMN functional communication explains exclusively positive symptoms, or, alternatively, both positive and negative symptoms. This study explores the relationship between positive and negative symptom expression in Sz and DMN/SN functional communication. Specifically, we explore the relationship between rs-FC (zero-lag) and time-lag-shifted (1 TR = 2 s; 2 TR = 4 s) rs-FC between resting-state networks (RSNs: aDMN, pDMN, and SN) and reported severity of 9 Sz symptom dimensions: hallucinations, delusions, bizarre behavior, positive formal thought disorder, affective flattening/blunting, alogia, avolition/apathy, anhedonia/asociality, and attention.

Methods

Subjects

This study draws from the Functional Biomedical Informatics Research Network (FBIRN) Phase III study (see Hare et al,³¹ Ford et al,³² and Damaraju et al³³). For a detailed description of the multiphase FBIRN project including subject characteristics and imaging/behavior assessments, see Keator et al.³⁴ For this study, we analyzed rs-fMRI scans from a large, clinically diverse sample of 100 subjects with Sz (table 1).

Table 1.

Demographic Information

	Descriptive Statistics (For Continuous Variables, Means and SDs Are Reported)	Range
Gender	78 (male), 22 (female)	N/A
Handedness	93 (right), 5 (left), 2 (both)	N/A
Smoking status	43 (current smoker), 26 (ex-smoker), 31 (never)	N/A
Age in years	39.3 (12.0)	18–60
Duration illness in years	17.7 (11.4)	1–41
Chlorpromazine equivalents46	414.3 (407.9)a	2–1800
Scale for the Assessment of Positive Symptoms (SAPS) total score	18.9 (15.0)	0–63
SAPS hallucinations subscale score (SAPS items 1–6 total score)	3.7 (5.0)	0–22
SAPS delusions subscale score (SAPS items 8–19 total score)	6.0 (6.2)	0–33
SAPS bizarre behavior subscale score (SAPS items 21–24 total score)	1.0 (1.6)	0–8
SAPS thought disorder subscale score (SAPS Items 26–33 total score)	3.1 (5.1)	0–27
Scale for the Assessment of Negative Symptoms (SANS) total score	28.3 (17.0)	0–80
SANS affective flattening subscale score (SANS items 1–7 total score)	5.3 (6.2)	0–24
SANS alogia subscale score (SANS items 9–12 total score)	2.0 (2.4)	0–13
SANS avolition/apathy subscale score (SANS items 14–16 total score)	4.6 (3.4)	0–14
SANS anhedonia/asociality subscale score (SANS items 18–21 total score)	6.7 (5.3)	0–19
SANS attention subscale score (SANS items 23–24 total score)	2.4 (2.2)	0–8

^aWe lacked data to derive chlorpromazine equivalents for 11/100 (11%) subjects. Mean and SD calculations are based on the sample of 89 subjects without missing data.

Raw imaging data were collected from 7 sites; written informed consent was obtained from all participants. The consent process was approved by institutional review boards of the University of California, Irvine; the University of California, Los Angeles; the University of California, San Francisco; Duke University/the University of North Carolina at Chapel Hill; the University of New Mexico; the University of Iowa; and the University of Minnesota.

All recruited study participants were between the ages of 18 and 62 years. All subjects in this study were diagnosed with Sz by experienced clinicians using the Structural Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) Axis I Disorders.³⁵ Patients either were stable on antipsychotic medication or were not taking antipsychotic medication at the time of the study (only 4 unmedicated of 100 subjects with Sz). Exclusion criteria for all participants included history of major medical illness, insufficient eyesight to see with normal acuity with MRI compatible corrective lenses, contraindications for MRI, drug dependence in the last 5 years or a current substance abuse disorder, and an intelligence quotient less than 75.

Assessments of Symptoms

Symptom severity was assessed using the Scale for the Assessment of Positive Symptoms (SAPS)³⁶ and the Scale for the Assessment of Negative Symptoms (SANS).³⁷ Subscale scores for each symptom dimension were calculated by deriving the sum of individual items in each dimension: hallucinations (SAPS items 1–6); delusions (SAPS items 8–19); bizarre behavior (SAPS items 21–24); positive formal thought disorder (SAPS items 26–33); affective flattening/blunting (SANS items 1–7); alogia (SANS items 9–12); avolition/apathy (SANS items 14–16); anhedonia/asociality (SANS items 18–21); and attention (SANS items 23–24) (see table 1). Clinicians and research staff at each FBIRN site were designated to perform the symptom ratings. To successfully calibrate symptom ratings, they participated in mandatory training sessions, run by experienced clinicians.

Imaging

As part of the larger FBIRN Phase III study, data were acquired using six 3 T Siemens TIM Trio scanners and one 3 T GE MR750 scanner using an AC–PC aligned echo-planar imaging pulse sequence (TR/TE 2 s/30 ms, flip angle 77º, 32 slices collected sequentially from superior to inferior, 3.4 × 3.4 × 4 mm gap, 162 frames, 5:24 min) to obtain T2*-weighted images. Subjects were instructed to lie in the scanner with eyes closed.

Data Processing

Preprocessing was performed using the Data Processing Assistant for rs-fMRI toolbox that runs with the REST software.³⁸ The first 2 time frames were removed to allow for signal stabilization. Raw data underwent motion correction to the first image, slice-timing correction to the middle slice, normalization to standardized Montreal Neurological Institute (MNI) space, and spatial smoothing with an 8-mm full-width at half maximum (FWHM) Gaussian kernel. Framewise displacement (FD)—defined as the sum of the absolute values of the derivatives of the 6 realignment parameters (3 linear + 3 rotational converted from degrees to millimeters)³⁹—was calculated for each image. The FD measurement differentiates head realignment parameters across frames and generates a 6-dimensional times series that represents instantaneous head motion.³⁹ Mean FD was calculated for each subject by taking the average of the sum of the absolute values of the derivatives of the 6 realignment parameters (3 linear + 3 rotational). Although independent component analysis (ICA) has been shown to be resistant to motion artifacts,⁴⁰ we also corrected for potentially confounding effects of head motion on the fMRI signal, by including mean FD as a subject-level covariate.

Group Spatial Independent Component Analysis

Group spatial ICA and functional network connectivity (FNC) correlation analyses were performed using GIFT software.⁴¹ As part of a prior network analysis of hallucinations in Sz, we performed group spatial ICA on a large sample of FBIRN subjects and analyzed FNC between 9 RSNs: 2 auditory networks, 2 visual networks, 2 subcortical networks, aDMN, pDMN, and SN.⁴² Back-reconstruction was performed using group information guided ICA, which takes the group maps and runs a spatially constrained ICA on individual subjects, producing individual subject component maps and time courses. This approach has been shown to be robust to artifacts as well as sensitive to individual and group differences.^43,44 In this work, we performed a new analysis of spatial maps and time series of SN, aDMN, and pDMN (figure 1) to explore DMN/SN functional communication.

Fig. 1.

Anterior default mode, posterior default mode, and salience networks. The mean aggregate spatial maps of the 3 independent component networks analyzed in the functional network connectivity analysis are shown (threshold: Z > 2): anterior default mode network (red), posterior default mode network (green), and salience network (blue).

Subject time courses were detrended and despiked, then filtered with a high-frequency cutoff of 0.15 Hz before computing FNC correlations (zero-lag) and time-lag-shifted FNC correlations; FNC correlations (zero-lag) are defined as the pairwise correlations between network time courses, and time-lag-shifted FNC correlations are defined as pairwise correlations between one network’s time course and another network’s time course shifted by a specified lag. In a previous analysis, Manoliu et al²⁶ performed time-lag-shifted FNC analyses with specified lags of 1 TR (2 s), 2 TR (4 s), and 3 TR (6 s), and found that Sz had abnormal 1-TR-shifted and 2-TR-shifted FNC between DMN and SN (but normal 3-TR-shifted FNC between DMN and SN) relative to healthy controls. Given these findings, we explored time-lag-shifted FNC between aDMN/pDMN and SN with specified time lags of 1 TR (2 s) and 2 TR (4 s). All FNC correlations (zero-lag and lagged) were transformed to z scores using Fisher’s transformation.

Statistical Analyses

We performed hierarchical linear regression analyses of FNC correlations (zero-lag and time-lag-shifted), controlling for confounding effects of nuisance variables in block 1 of the linear model (age, gender, scanning site, and mean FD). Symptom scores including the SANS³⁷ subscale scores (affective flattening/blunting, alogia, avolition/apathy, anhedonia/asociality, attention) and the SAPS³⁶ subscale scores (hallucinations, delusions, bizarre behavior, positive formal thought disorder) were entered in block 2 of the linear model. Subjects with residuals >3 SD from the mean were excluded (N ≥ 98 subjects for each regression analysis).

To ensure that observed associations between symptom severity and FNC were not driven by confounding effects of medication, we also performed regression analyses including total chlorpromazine equivalents⁴⁵ as an additional covariate in block 1. We lacked information to derive chlorpromazine equivalents⁴⁵ for 11 subjects with Sz, so we calculated the mean value of total chlorpromazine equivalents (based on the available data; n = 89 subjects) and interpolated the mean value for the 11 subjects with missing data. Results of these analyses are reported in supplementary table S2.

Because nicotine use is 2–3 times higher in patients with Sz than in the healthy population⁴⁶ and has been shown to significantly impact brain functional connectivity,⁴⁷ we examined Spearman correlations between FNC and smoking status (factor with 3 levels: “never smoker,” “ex-smoker,” “current smoker”). We found no significant correlations between smoking status and FNC measures, so smoking status was not included as a covariate.

Although we hypothesized that rs-FNC with DMN/SN would be linked predominately to positive symptoms,^18,26–28 our FNC analyses were largely exploratory to test whether DMN/SN connectivity might also be linked to negative symptoms and to determine whether FNC-symptom associations depend on lag direction and/or magnitude between SN/DMN time courses. For clarity of reporting the results, lag magnitude is reported parenthetically, whereas lag direction is denoted with an arrow. For instance, “lagged (1 TR) aDMN→SN connectivity” refers to the correlation between aDMN and SN rs-fMRI signal when the time series of the SN lags behind the time series of the DMN by 1 TR (2 s). For each set of time-lag-shifted FNC analyses of a specified lag (1 TR, 2 TR), confidence was initially specified as P < .05, and then Bonferroni-corrected for six tests (SN→aDMN, aDMN→SN, SN→pDMN, pDMN→SN, aDMN→pDMN, pDMN→aDMN) (P < .0083).

Results

Below, we report significant associations between time-lag-shifted FNC and symptom dimension scores of the SAPS/SANS (table 2). Results of the zero-lag FNC analyses are reported in supplementary table S1a; nominally significant (non-Bonferroni-corrected, P < .05) results of the time-lag-shifted FNC analyses are reported in supplementary table S1b. To provide estimates of effect sizes, we parenthetically report standardized regression coefficients.

Table 2.

Associations Between Symptom Dimension Scores and Network Connectivity

FNC	Lag Summary	Symptom Dimension	β	T-stat	P
aDMN→pDMN	pDMN time series lags aDMN time series by 2 TRs	Attention	−.31	−3.0	.003
aDMN→pDMN	pDMN time series lags aDMN time series by 2 TRs	Thought disorder	−.31	−2.9	.005
aDMN→SN	SN time series lags aDMN time series by 1 TR	Flat affect	.29	2.9	.005

Note: FNC, functional network connectivity; aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network.

Associations between symptoms and FNC between aDMN and pDMN: Lagged (2 TR) aDMN→pDMN connectivity was negatively associated with severity of attentional deficits (β = −0.31, P = .003) and disordered thought (β = −.31, P = .005) (table 2, (figure 2).

Fig. 2.

Associations between symptom severity and time-lag-shifted functional network connectivity (FNC). Partial regression plots showing negative associations between lagged (2 TR) aDMN→pDMN connectivity and reported severity of attentional deficits (top left) and thought disorder (top right), in addition to positive associations between lagged (1 TR) aDMN→SN connectivity and severity of flat affect (bottom left) and bizarre behavior (bottom right). Covariates controlled for in the linear model included age, gender, scanning site, and mean framewise displacement. aDMN, anterior default mode network; pDMN, posterior default mode network; SN, salience network.

Associations between symptoms and FNC between aDMN and SN: Lagged (1 TR) aDMN→SN connectivity was positively associated with severity of flat affect (β = .29, P = .005) (table 2) and bizarre behavior (figure 2), although the latter association did not survive Bonferroni correction for multiple tests (β = .25, P = .014) (supplemental table S1b).

Discussion

The objective of this study was to ask whether functional communication between SN and DMN explains exclusively positive symptoms, or both positive and negative symptoms. Prior research suggests that traditional (zero-lag) FNC analyses may be ill-equipped to detect time-varying communication between different brain regions as well as group differences in communication between regions. We focused on functional communication between the SN and DMN and how this communication is affected by Sz.²⁶ Specifically, we probed the roles of lag magnitude and direction to explore the relations between SN/DMN connectivity and targeted behavioral dimensions of Sz. We hypothesized that time-lag-shifted rs-FNC across 3 networks (aDMN, pDMN, and SN; see figure 1) would be linked predominately to positive symptoms.^18,26–29 Instead, we found that specific patterns of time-lag-shifted rs-FNC were associated with negative symptoms (eg, attentional deficits and flat affect) as well as positive symptoms (eg disordered thought and bizarre behavior).

First, the (2 TR) aDMN→pDMN connectivity analysis revealed that patients with more severe thought disorder had less time-lag-shifted functional communication between DMN networks (specifically, aDMN activation preceding pDMN activation by 4 s). This lag might contribute to derailment and illogicality, symptoms of thought disorder.³⁷ It is thought that the DMN supports internal mental processes (memories, thought, etc.),^48,49 but it remains unclear how exactly the DMN supports these processes. Our findings suggest that functional communication between DMN hubs may be critical for organizing thoughts into coherent, meaningful utterances. Yet, this theory remains speculative until future research provides insight into how the DMN supports complex thought processes and addresses targeted associations between disrupted DMN function and wide-ranging formal thought disturbances in Sz—from derailment (eg, where the patient’s ideas slip off topic) to blocking (eg, where the patient’s train of thoughts is interrupted).

The same pattern of lagged aDMN→pDMN connectivity was also negatively associated with severity of attentional deficits. Put another way, patients with more severe attentional deficits had less temporally coherent (4 s-lagged) functional co-activation of aDMN and pDMN. In addition, we observed numerous nominally significant associations between attentional deficits and FNC between pDMN and SN (both pDMN→SN and SN→pDMN connectivity; see supplemental table S1b). A previous study found that elevated posterior cingulate activity was observed during lapses in attention when healthy research subjects performed a demanding perceptual task.⁵⁰ In another study, increased activity in posterior midline regions predicted which words were forgotten on a memory task.⁵¹ Thus, our results are consistent with the theory that pDMN plays a critical role in regulating attention.⁵

Next, we observed flat affect was more pronounced in patients to the extent that the aDMN activation preceded SN activation by 2 s, as reflected in lagged aDMN→SN connectivity. The aDMN contains midline structures spanning the MPFC and ACC. Whitfield-Gabrieli et al⁵² reported that dorsal MPFC was preferentially engaged during performance of a task that required explicit self-reference, relative to DMN activation evoked by a rest condition. Meanwhile, ventral MPFC plays a critical role in the regulation of amygdala activity⁵³; patients with ventral MPFC damage have marked reductions in autonomic arousal to emotionally charged stimuli.⁵⁴ These findings suggest that anterior midline DMN hubs contain functional subdivisions essential for explicit self-reference (dorsal MPFC), and tracking the salience of emotional stimuli and regulating our responses to those stimuli (ventral MPFC). It is plausible that flat affect stems from elevated aDMN/SN functional communication that manifests as disturbances in emotional salience tracking/monitoring, and/or inability to disengage with self-reflective thought and engage with surroundings. Future studies should explore these functional MPFC subdivisions and their potential contributions to diminution of vocal inflection, and affective gestures, as well as inappropriately elevated displays of affect in Sz.

Finally, bizarre behavior was positively associated with the same FNC pattern (2-s-lagged aDMN→SN connectivity). However, this small effect (standardized β = .25) did not survive Bonferroni correction for multiple tests. Elevated functional communication between SN and DMN could result in awareness of mislabeled bursts of inner speech or thoughts. These experiences may, in turn, affect planning, social engagement, and engagement with the environment, resulting in bizarre behavior. Future studies in patients selected to have a broader range of bizarre behaviors may further examine this relationship.

Our findings support the hypothesis that specific patterns of lagged DMN/SN functional communication are associated with both positive and negative symptoms. We observed 2 main trends: (2 TR) lagged aDMN→pDMN connectivity was negatively associated with symptom severity, whereas (1 TR) lagged aDMN→SN connectivity was positively associated with symptom severity (figure 3). On the one hand, to the extent that lagged functional communication between aDMN and pDMN hubs is reduced, patients had more severe cognitive disturbances (disordered thought and attentional deficits). On the other hand, patients had more pronounced flat affect and engaged in more bizarre behavior to the extent that aDMN activation consistently preceded SN activation (by 2 s).

Fig. 3.

Disrupted anterior default mode network functional communication linked to positive and negative symptoms. Time-lag-shifted functional network connectivity (FNC) between anterior default mode network (red) and salience network (blue) is positively associated with bizarre behavior and severity of flat affect. Meanwhile, time-lag-shifted FNC between anterior default mode network and posterior default mode (green) was negatively associated with attentional deficits and severity of disordered thought. Arrows denote direction of lag and do not imply causal relationships.

Given Manoliu et al’s report of a significant negative correlation between strength of functional connectivity within the right anterior insula and hallucination severity in the patients with Sz,²⁶ we predicted that SN functional communication would be linked to hallucination severity. Yet, we observed no associations between hallucination severity and SN functional communication, and only a nominally significant negative association between hallucination severity and (zero-lag) aDMN–pDMN connectivity (supplemental table S1a). Notably, our analysis of 100 patients with Sz drew from a larger sample than in Manoliu et al²⁶ (n = 18 patients), and we modeled effects of symptom severity on FNC, controlling for extraneous effects of motion, age, gender, and scanning site (vs performing bivariate correlation analyses). Thus, our null findings might be treated as evidence favoring rejection of the hypothesis that abnormal SN function underlies hallucinations in Sz. However, a targeted analysis of FNC between SN and sensory networks by our group⁴² revealed that elevated FNC between SN and an auditory network was positively associated with severity of auditory hallucinations. Future analyses should continue to explore and test targeted hypotheses of hallucinations by exploring potential associations between hallucination severity and SN functional communication.

Observed associations between symptom severity and FNC were dependent on lag direction. In resting-state analyses of healthy subjects, the anterior midline DMN hub acts as a sink of propagated activity (eg, anterior midline activity lags behind posterior midline activity during rest).^9,10 In this study, we observed that symptom severity was associated with atypical aDMN→pDMN connectivity and aDMN→SN connectivity. Converging evidence from rs-FC analyses,^18–20 along with a dynamic rs-FNC analysis⁵⁵ demonstrating that Sz shows reduced dynamic switching of network states, suggests that patients may be stuck in DMN states associated with self-referential processing. As such, it makes sense that DMN activity might precede activity in networks such as the SN. Although it remains unclear why lagged FNC with aDMN (aDMN→pDMN, aDMN→SN) was associated with reported symptom severity, this is an interesting result that requires further investigation with other modalities such as electroencephalography (EEG) and/or magnetoencephalography (MEG), which provide more precise timing information.

Associations between symptom severity and FNC were also dependent on lag magnitude. In healthy subjects, brief delays are observed between network sources of propagated activity and subsequent activation of network sinks such as the anterior frontal cortex (typically <0.5 s).¹⁰ We observed that symptom severity was associated with lagged FNC with longer, atypical delays of 2 and 4 s. However, our methodological approach in this study limits us in making the strong claim that symptoms are caused by these lags. Future investigations must explore precise timing of activation of functional network hubs, and how this relates to behavioral task performance and Sz symptomology.

Given that the physiological basis of the blood-oxygenation-level-dependent (BOLD) fMRI signal remains controversial, this entails some speculation will be required when considering the significance of multisecond lags between BOLD hemodynamic responses of RSN hub regions. Lags in BOLD fMRI signaling may be caused by vascular effects, changes in neural signaling, or a combination of factors. Prior findings suggest that changes in neural signaling contribute to observed BOLD hemodynamic lags¹⁰ and that vascular effects alone cannot account for BOLD signal lag structure.⁵⁶ Prior research also suggests that infra-slow neuronal oscillations (0.01–0.1 Hz) play a key role in generating the BOLD-fMRI response.^57,58 Although, direct (causal) links between BOLD fluctuations and infra-slow neuronal oscillations in humans remain unestablished, it is widely acknowledged that proper functional network communication depends on dynamic phase coupling of fast neural rhythms (eg, γ; >30 Hz) to slower rhythms (eg, δ, θ; <8 Hz).^59,60 It is plausible that coherent BOLD signal fluctuations in RSN hubs of healthy subjects may reflect frequency-dependent coupling of network hub activation. In Sz, cross-frequency coupling of activity across DMN hubs is disrupted.^60,61 We propose that these disruptions may manifest as measurable lags between hemodynamic responses of RSNs. At the same time, we acknowledge that additional physiological factors/interactions are associated with BOLD signal fluctuations and that exact (causal) relationships between oscillatory coupling disturbances and measurable changes in FNC using BOLD fMRI remain unknown.

Although our study was the first to examine targeted relationships between time-lagged FNC between SN and DMN and wide-ranging Sz symptoms, we must acknowledge several limitations. Although we were able to probe potential links between rs-FNC and a relatively broad set of 9 symptom dimensions, the SAPS/SANS clinical assessments limited our ability to explore links with an even broader array of symptoms and targeted behavioral outcomes such as working memory deficits. Next, the cross-sectional nature of this analysis limited our ability to explore how neural function changed in patients over time; it remains unclear whether observed FNC effects reflect chronic dispositions. Third, all but 4 of the 100 subjects with Sz were taking antipsychotic medication at the time of the FBIRN study, introducing potentially confounding effects on brain FNC. We controlled for these potentially confounding effects by introducing total chlorpromazine equivalents⁴⁵ as an additional covariate in our regression analyses of FNC. Including chlorpromazine equivalents as covariate had no significant impact on the results (see supplemental table S2). Finally, our FNC analyses explore correlations between the rs-fMRI signals of DMN and SN. In our discussion of results, we use arrows to denote direction of lag. This effort to enhance clarification should not be taken to imply causation (eg, that one network’s activity exerts causal influence over another network’s activity).

The objective of this study was to address whether functional communication between SN and DMN explains exclusively positive symptoms, or both positive and negative symptoms. To achieve this aim, we explored associations between time-lag-shifted FNC between SN and DMN and heterogeneous behavioral outcomes in Sz. We found strong associations between time-lag-shifted FNC with aDMN (specifically aDMN→SN and aDMN→pDMN) and both positive and negative symptoms of Sz (figure 3); all other reported FNC-symptom associations did not survive Bonferroni correction for multiple tests. Our results suggest that altered aDMN functional communication may play a crucial role in the pathophysiology of Sz and etiology of both positive and negative symptoms. Future studies should build upon these findings and explore time-lag-shifted FNC between SN/DMN hubs and sensory networks, motor networks, and attention networks to gain a more complete, nuanced understanding of the neural mechanisms underlying specific symptoms.