Structural covariance in schizophrenia and first-episode psychosis: An approach based on graph analysis
Graphical abstract
Graphical representation of the measures used in this study. First, the cortex was segmented in 64 regions. A correlation matrix was generated with the volume of all regions in the first step. To capture two distinct features that can represent how the brain is affected in schizophrenia, two measures were derived from the correlations: Connectivity-closeness, and integrity closeness.
Introduction
Structural neuroimaging studies of schizophrenia have classically framed brain alterations in terms of either regional volume or thickness disruptions, whereas cerebral abnormalities in schizophrenia appear to be widely distributed (Haijma et al., 2013).
Network perspectives improve descriptions of brain function as both segregated and distributed processing are crucial to cognitive and sensorimotor functions (Bassett and Bullmore, 2006). This framework is particularly relevant to schizophrenia, as it has been suggested that the disorder could be better understood as being caused by a disruption of connections between brain regions (Crossley et al., 2009, Friston and Frith, 1995) or in the interaction of more complex circuitry such as the cortical-thalamic-cerebellar-cortical circuit (Andreasen et al., 1999). There is a growing body of evidence supporting anatomical (Canu et al., 2015, Dazzan, 2014) and functional dysconnectivity (Fitzsimmons et al., 2013) in schizophrenia. Connectivity-based disease model may link structural and functional brain abnormalities seen in the disorder (Williams, 2008). In fact, it has been recently found that connectivity patterns may be used to differentiate between schizophrenia patients and healthy controls (Kaufmann et al., 2015), and more interestingly differentiate schizophrenia group from major depressive disorder patients (Guo et al., 2013). Therefore, connectivity measures could potentially be used as disease biomarkers in future studies (Dazzan, 2014).
Structural covariance (or interregional covariance) is an approach that examines whether anatomical change in one region correlates with changes in other brain regions (Lerch et al., 2006, Sato et al., 2013). Structural covariance has been shown to be similar to structural connectivity (Lerch et al., 2006), but also to coordinated developmental changes in the brain, suggesting that it is a signature of developmental processes (Alexander-Bloch et al., 2013). As such, this approach seems suitable to explore neurodevelopmental abnormalities, which have been hypothesized to play a significant role in the pathophysiology of schizophrenia (Murray and Lewis, 1987, Weinberger, 1987). Exploring structural covariance is also suitable to study diffuse pathological brain processes in an interconnected brain (Fornito et al., 2015), such as those observed in schizophrenia (Fornito et al., 2009, Haijma et al., 2013, Wright et al., 2000). Schizophrenia does not seem to fully target specific regions, which could lead to an increase in covariance lead by common pathological factors.
A promising extension of structural covariance analysis is the utilization of graph theory descriptors. Graph analysis yields many different measures when applied to brain imaging (for review, see (Bullmore and Sporns, 2009)). It can be used to study network community structure and topology, characterize system properties (clustering coefficient, path length and network efficiency) or even provide a measure of the importance of a determined node in the system through the use of centrality measures (degree, betweenness, closeness and eigenvector centrality) (Telesford et al., 2011).
Few studies have addressed structural covariance in schizophrenia; however, there appear to be differences in structural covariance when comparing schizophrenia subjects with unaffected controls, including reduced inter-regional correlations between prefrontal, anterior cingulate and temporal regions, and between posterior cingulate and hippocampus (Woodruff et al., 1997). More recently, structural covariance graphs have been investigated in relation to mental disorders. Bassett et al. (2008) used a priori hub definitions and found that schizophrenia patients exhibit loss of frontal hubs, increased connection distance and reduced hierarchy. Another study showed that schizophrenia patients had decreased betweenness centrality in associative regions of the cortex and increased centrality in limbic and paralimbic regions (Zhang et al., 2012). Despite previous work, many questions regarding the relationship between structural covariance and complex behavioral disorders such as schizophrenia remain unresolved (Telesford et al., 2011).
Previous work shows that cortical thinning (Ziermans et al., 2012) and cognitive deficits (Bora et al., 2014) are present very early in the course of the disease. Despite differences in distribution and less consistent findings, first-episode group of patients have gray matter deficits very similar to those found in chronic schizophrenia (Williams, 2008). Additionally there is compelling evidence for a decline in gray matter volume during the course of the disease (Haijma et al., 2013, Woods et al., 2005), and this appears to be more prominent in the first years after disease onset (Takahashi et al., 2010, van Haren et al., 2007, Yoshida et al., 2009). It is now recognized that gray matter changes are related to a number of confounding factors, such as antipsychotic treatment (Ho et al., 2011), cannabis use, and disease outcome (Van Haren et al., 2013). However, there is no consensus as to when in the course of the disease volume reduction of specific regions occurs. If network measures reflect underlying biological characteristics that are central to disease onset, any illness-related abnormalities are likely to be similarly present in chronic schizophrenia and FEP patients.
When considering these previous findings, we are challenged by a paradox. On one hand, the connectivity disruptions frequently described in schizophrenia naturally lead to a hypothesis of a reduced structural covariance in patients, when compared to controls. Alternatively, common neurodegenerative/volumetric reductions driven by the disease may result in an increase in structural covariance.
The aim of the current study was to tackle this paradox by the investigation of differences in structural covariance among healthy controls, schizophrenia and FEP groups from a graph theory perspective, based on closeness measures. We propose two different metrics: connectivity-closeness and integrity-closeness. Graph-related alterations of key hub regions in the whole-brain networks are likely to reflect the impairment of integration and contextualization of different stimuli into complex representations, as well as higher cognitive functions such as working memory or executive functioning (Rubinov and Bullmore, 2013). We hypothesize that the regions exhibiting abnormal structural covariance patterns would be related to functional brain networks that are impaired in schizophrenia, specifically alterations of the nodes related to the fronto-parietal-temporal and limbic network.
Section snippets
Subjects
One hundred and forty-three patients (68% males; mean age = 37.18 years, s.d. = 11.05) with a diagnosis of chronic schizophrenia, 82 healthy matched controls (63% males; mean age = 35.48 years, s.d. = 11.16) and 32 subjects with FEP (53.12% males; mean age = 27.09 years, s.d. = 7.98) were included in the study.
Individuals were recruited from the Schizophrenia Program (PROESQ) at the Federal University of São Paulo (UNIFESP) and from an emergency unit at the Santa Casa de Misericordia de São
Results
The study groups significantly differed in terms of age (F(2, 253): 12.95, p < 0.01), with first-episode group being younger than the chronic schizophrenia and control groups. There was no difference in sex (: 4.78, p = 0.91). The chronic schizophrenia group had lower PANSS scores and lower GAF scores (see Table 1) than the FEP group.
Integrity and connectivity closeness were negatively correlated (r = −0.302; Fig. 1) and capturing distinct features of structural covariance. There were no
Discussion
In this study, we found evidence of increased structural covariance in schizophrenia compared with healthy controls. Pars orbitalis and insula showed significantly decreased integrity-closeness values. Moreover, we conducted a complementary analysis comparing the group of individuals with FEP (and as such short-term disease history and antipsychotic exposure) to further verify whether putative differences between patients and controls were related to the stage of the disorder.
We found no
Funding
The opinions, hypothesis, conclusions and recommendations of this study are under the responsibility of the authors, which not necessary represents the opinion of the funding agencies. The authors are grateful to Sao Paulo Research Foundation – FAPESP (J.R.S grant numbers: 2013/10498-6 and 2013/00506-1; R.A.B grant number: 2011/50740-5).
Contributors
André Zugman: MRI Data acquisition and processing, writing the first draft and reviewing the manuscript.
Idaiane Assunção: Recruitment of patients, MRI data acquisition and processing, proof-reading the manuscript.
Gilson Vieira: Data processing and statistical analysis, preparation of the manuscript and proof-reading the manuscript.
Ary Gadelha: Study Design, recruitment of patients, clinical data acquisition, reviewing and editing the manuscript.
Thomas White: Manuscript preparation, reviewing
Conflicts of interest
A.Z. receives a scholarship from Brazilian Government Institution (CAPES, grant number: 10866-13-2). A.G was on the speakers' bureau and/or has acted as a consultant for Janssen-Cilag and has also received research support from Brazilian government institutions (CNPq). C.N. has received a scholarship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) and has served as a consultant or advisory board member for Janssen. Q.C. worked at Santa Casa de Sao Paulo Medical School.
Acknowledgments
We are grateful to all participants and to the staff of the Programa de Esquizofrenia (PROESQ) of the Federal University of São Paulo.
References (58)
- et al.
Defining the phenotype of schizophrenia: cognitive dysmetria and its neural mechanisms
Biol. Psychiatry
(1999) - et al.
A selective review of structural connectivity abnormalities of schizophrenic patients at different stages of the disease
Schizophr. Res.
(2015) - et al.
Cortical surface-based analysis. I. Segmentation and surface reconstruction
NeuroImage
(1999) - et al.
An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest
NeuroImage
(2006) Networks of anatomical covariance
NeuroImage
(2013)- et al.
Mapping grey matter reductions in schizophrenia: an anatomical likelihood estimation analysis of voxel-based morphometry studies
Schizophr. Res.
(2009) - et al.
Meta-analysis of gray matter anomalies in schizophrenia: application of anatomic likelihood estimation and network analysis
Biol. Psychiatry
(2008) - et al.
Convergence and divergence of thickness correlations with diffusion connections across the human cerebral cortex
NeuroImage
(2012) - et al.
Brain-wide functional inter-hemispheric disconnection is a potential biomarker for schizophrenia and distinguishes it from depression
Neuroimage Clin.
(2013) - et al.
Mapping anatomical correlations across cerebral cortex (MACACC) using cortical thickness from MRI
NeuroImage
(2006)