Gray Matter Alterations in Schizophrenia High-Risk Youth and Early-Onset Schizophrenia: A Review of Structural MRI Findings

Synopsis

The purpose of this article is to provide a review of the literature on structural MRI findings in pediatric and young adult populations at clinical or genetic high-risk for schizophrenia, as well as in early-onset schizophrenia. The authors discuss the implications of this research for understanding the pathophysiology of schizophrenia and for early intervention strategies for prevention of the illness. The evidence linking brain structural changes in pre-psychosis development and early-onset schizophrenia with disruptions of normal neurodevelopmental processes during childhood and/or adolescence are described. In addition, the authors outline future directions for research to address current knowledge gaps regarding the neurobiological basis of brain structural abnormalities in schizophrenia and to help improve the utility of these abnormalities for preventative interventions.

Introduction

Neuroimaging studies over the last four decades have provided overwhelming evidence that schizophrenia is a disorder involving widespread abnormalities of brain structure¹. It is thought that the neurobiological processes underlying these structural abnormalities are central to the pathophysiology of schizophrenia². However, the specific mechanisms involved in producing the structural deficits of schizophrenia remain incompletely understood. While no focal brain abnormality has been identified unequivocally, structural abnormalities including enlargement of the lateral and third ventricles, and reduced lateral temporal cortical, medial temporal, and prefrontal lobe volumes are consistently reported in persons with schizophrenia¹. Further, alterations of brain structure are linked with key psychotic symptoms (e.g., auditory hallucinations³, delusions⁴), neurocognitive deficits⁵, and social dysfunction⁶ in schizophrenia.

Neurodevelopmental models hypothesize that pathological processes occurring during “early” (i.e., perinatally) and/or “late” (i.e., adolescent/young adult) brain development (e.g., aberrant migration of neuronal precursor cells during gestation and/or excessive dendritic pruning during adolescence) may be key to the emergence of the brain structural alterations occurring in schizophrenia^{7, 8}. Growing evidence from studies of typically developing children shows that brain maturational processes continue well into adolescence⁹. Neuroimaging studies, for example, reveal changes in the rates of gray to white matter (WM) during the second decade of life, with increases in WM (largely comprised of myelinated axon bundles) accompanied by reductions in gray matter ([GM]; an index of cellular and unmyelinated fiber density)^{10, 11}. It is thought that, at least in part, these findings are indicative of changes in cellular-level processes, such as myelinization by oligodendrocytes (increased WM) and neuronal apoptosis resulting in dendritic pruning (GM reduction), which together contribute to improved regional communication and more efficient neuronal coding over the course of adolescence¹². Additionally, several studies suggest a characteristic temporal pattern of GM reduction, with structural decrements proceeding from posterior cortical areas (e.g., parietal cortex) during childhood to anterior brain regions (e.g., prefrontal cortex (PFC)) during late adolescence and early adulthood^{13, 14}. It has been hypothesized that this lag in PFC GM maturation may result in an “imbalance” during adolescence between earlier developing mesolimbic structures mediating responses to pleasurable stimuli (i.e., the brain’s reward circuitry) and less fully-developed prefrontal brain areas involved in response inhibition and cognitive control¹². Thus, the trajectories of structural brain changes in normal development are increasingly believed to provide a neurobiological basis for the increase in impulsivity and risk-taking behavior that contribute to making adolescence a period of heightened risk for the emergence of a broad range of psychopathology, including schizophrenia¹⁵.

Several lines of evidence provide substantial support for the neurodevelopmental hypothesis that alterations of normal brain maturational processes are implicated in the characteristic structural abnormalities of schizophrenia. This evidence, reviewed extensively elsewhere^{2, 16, 17}, includes:

1. Neuropathological findings in schizophrenia consistent with microneuroanatomical alterations (e.g., abnormal laminal organization and orientation of neurons) associated with gestational development in the prefrontal, cingulate, and lateral temporal cortices, as well as the hippocampus^18–20.

2. Structural MRI observations of abnormal prefrontal and/or temporal cortical surface morphology in adult²¹ and early-onset schizophrenia (EOS)²², as well as in adolescents at genetic high risk (GHR) for schizophrenia ²³, that are thought to reflect perturbations of gyrification during early development.

3. Reduced cortical neuropil and somal size found in post mortem studies of schizophrenia^{24, 25}, suggestive of dysregulated apoptosis and/or synaptic pruning in adolescence

4. Immunohistochemical and genetic linkage and association studies²⁶ implicating gene mutations in persons with schizophrenia that are thought to disrupt normal cortical developmental processes – e.g., early synaptogenesis (e.g., RELN²⁷) and changes in dendritic spines during adolescence (e.g., DISC 1²⁸).

5. Animal models showing that early alterations of GM development produce later abnormalities of adolescent cortical function that are analogous to those observed in schizophrenia²⁹.

Despite the accumulating evidence for brain dysmaturation during childhood and adolescent development in schizophrenia, many questions regarding the pathophysiology of the structural brain abnormalities of schizophrenia remain unanswered. For example, it is not clear when GM loss first begins during development¹ – e.g., whether less GM in schizophrenia is the product of primarily early (intrauterine/perinatal), or late (peri-adolescent/early adult) dysmaturational processes, or some combination of the two. Further, how environmental, or epigenetic risk and/or protective factors might influence the course of neurodevelopment, or how alterations of brain structure are specifically linked to the emergence of psychotic symptoms have yet to be determined. Additionally, it is acknowledged that although there has been significant growth in neuroscientific understanding of normal brain development, relatively few studies have focused directly on youth (age less than 18 years) who develop schizophrenia (i.e., EOS)³⁰, or adolescents or young adults (age less than 30 years) at risk for the illness (whether by virtue of family relatedness to a person with schizophrenia (GHR), or as a result of symptoms or functional decline thought to be indicative of clinical high-risk (CHR) for full blown psychosis). Characterizing the structural brain changes observed in studies of youth/young adults at GHR or CHR for schizophrenia, as well as in EOS, may be critical to refining current understanding of the timing and pathophysiology of these alterations. It may also help us to identify individuals who could benefit from early treatment interventions. Toward this end, we provide a selective review of structural MRI findings in pediatric at-risk populations and EOS. Our goal is to identify the common findings and gaps in current knowledge regarding structural brain changes associated with the trajectory of schizophrenia risk in youth and young adult development in order to provide future directions for research.

Structural MRI Findings in Genetic High-Risk for Schizophrenia Individuals

Based on the evidence regarding the strong heritability of schizophrenia – approximately 60–80% of the liability to schizophrenia is due to genes³¹ -- GHR structural magnetic resonance imaging (MRI) research has focused on the identification of neural abnormalities during the adolescent and young adult development of non-psychotic, first-degree relatives of persons with schizophrenia. This work is motivated by the drive to understand brain development prior to psychosis and to observe it largely without medication and illness state-related confounds that commonly complicate schizophrenia research. According to the GHR model³², it is hypothesized that schizophrenia results from the cumulative vulnerability of multiple genetic and environmental factors, each associated with relatively small effects. Prior to the onset of psychosis, subclinical neuroanatomical, or other abnormalities (e.g., reduced hippocampal GM volume, or neurocognitive deficits) are thought to be reliably detectable and expressed in non-psychotic, first-degree relatives of patients, who on average share 50% of genes with their affected family member³³. A particular strength of the GHR approach is that it allows for the identification of neural markers of schizophrenia risk preceding psychosis-like symptoms. Findings from GHR research, therefore, could contribute to greater understanding of pathophysiological processes associated with early development, as well as to the identification of vulnerability markers that may be particularly useful for early detection strategies for psychosis prevention^{7, 34}. Additionally, longitudinal GHR research in children and adolescents may be able to distinguish temporally distal neuroanatomical abnormalities associated with schizophrenia risk from those more closely linked to the timing of psychosis onset⁷.

Recently, our group has carried out a comprehensive review³⁵ of all GHR MRI studies involving individuals 30 years of age or younger, bringing together the results from 14 independent research groups, 12 of whom have contributed structural MRI data (Table 1). It should be noted that studies included in this review employed a variety of MRI morphometric techniques (e.g., voxel-based morphometry and manual parcellation), as well as differing MRI software packages (e.g., Statistical Parametric Mapping (SPM), FreeSurfer) and methods to correct for multiple comparisons (e.g., whole brain vs. region-of-interest (ROI)). Below, we summarize the main findings.

Table 1.

Structural MRI findings from studies of genetic high-risk for schizophrenia individuals less than age 30

Group	Author (Year)	Study type	Results
Pittsburgh	Keshavan et al. (1997)	Cross-sectional	-Smaller L amygdala and enlarged 3rd ventricle
	Keshavan et al. (2002)	Cross-sectional	-Smaller bilateral amygdala-hippocampal complex and intracranial volume
	Rajarethinam et al. (2004)	Cross-sectional	-Smaller superior temporal gyrus bilaterally
	Jou et al. (2005)	Cross-sectional	-Altered gyrification of L anterior cortical surface
	Diwadkar et al. (2006)	Cross-sectional	-GHR showed smaller PFC GM
			-GHR with symptoms showed smaller PFC, thalamus, and cuneus GM vs. GHR without symptoms
	Bhojraj et al. (2009)	Cross-sectional	-Smaller L PT, R Heschl’s gyrus, L supramarginal and R angular gyri
			-Reversed PT asymmetry, exaggerated Heschl’s gyrus asymmetry, attenuated supramarginal/angular gyri
	Prasad et al. (2010)	Longitudinal	-Reduced gyral surface area in fronto-parietal lobes
			-Increased gyral cortical thinning
			-Shrinkage of total surface area (bilateral frontal/occipital cortices) at 1-yr follow-up.
	Bhojraj et al. (2010)	Cross-sectional	-Smaller bilateral lateral temporal, R inferior parietal, and L posterior cingulate cortices
			-Smaller bilateral precuneus and R DLPFC
	Bhojraj et al., (2011a)	Longitudinal	-Reduced bilateral lateral orbitofrontal, L rostral anterior cingulate, L medial PFC, R inferior frontal gyrus, and L frontal pole over time in GHR
			-Smaller volumes predicted greater severity of symptoms at baseline and at follow-up
			-Smaller baseline volumes and longitudinal decrease in volumes predicted greater severity of prodromal symptoms over time
	Bhojraj et al. (2011b)	Longitudinal	-L surface area in auditory association cortex and laterality index showed decline over time in GHR
Edinburgh	Lawrie et al. (1999)	Cross-sectional	-Smaller L amygdala-hippocampal complex volume and bilateral thalamus in GHR
	Harris et al. (2004)	Cross-sectional	-Increased R PFC gyrification index in GHR who developed schizophrenia
	Job et al. (2005)	Longitudinal	-GHR with symptoms showed reduction in L superior lateral hippocampal surface, L fusiform gyrus, L uncus, L inferior temporal gyrus, and L STG
	Job et al. (2006)	Longitudinal	-Changes over time in inferior temporal gyrus were significantly predictive of developing schizophrenia in GHR
	Lymer et al. (2006)	Cross-sectional	-L superior temporal gyrus GM density associated with schizotypal symptoms in GHR
	McIntosh et al (2011)	Longitudinal	-Smaller bilateral PFC volume in GHR vs. controls at baseline
			-GHR who transition to psychosis show significant reduction in bilateral PFC volume
			-GHR as a whole show significant reductions in whole-brain volume, L and R temporal lobes, and L frontal lobe volume
NIMH	Gogtay et al. (2003)	Cross-sectional	-Smaller total cerebral, frontal and parietal lobe GM volume in GHR
	Gogtay et al. (2007)	Cross-sectional	-Smaller GM in L PFC and bilateral temporal cortices in GHR
	Mattai et al. (2011)	Longitudinal	-Smaller baseline GM in bilateral PFC, L temporal cortex, and parietal cortex that normalized by age 17
NYU	Li et al. (2012)	Cross-sectional	-Thinning of inferior frontal gyrus GM volume in GHR
			-Increased cortical thickness in PFC, temporal and parietal cortices in GHR
Iowa	Ho et al. (2010)	Cross-sectional	-Smaller bilateral hippocampal volume in GHR
WU	Harms et al. (2010)	Cross-sectional	-Smaller inferior frontal gyrus GM in GHR
	Karnik-Henry et al. (2012)	Cross-sectional	-Thinner parahippocampal volume in GHR
Harvard AHRS	Rosso et al. (2010)	Cross-sectional	-Smaller bilateral vmPFC and frontal pole GM volume in GHR
			-vmPFC volume negatively correlated with schizotypal symptoms in GHR
UNC	Dougherty et al. (2012)	Cross-sectional	-Greater positive association between age and hippocampal and basal gangliar volumes in GHR
Turkey	Sismanlar et al. (2010)	Cross-sectional	-Smaller bilateral hippocampal volume in GHR
Korea	Byun et al. (2012)	Cross-sectional	-Cortical thinning in R anterior cingulate, L paracingulate, PCC, bilateral frontal pole, vmPFC, and occipital cortex
Harvard LHRS	Francis et al. (2012)	Cross-sectional	-Smaller L pars triangularis and R pars orbitalis volumes in GHR with reversal of L>R pars orbitalis lateralization

Abbreviations: AHRS = Adolescent High-Risk Study; DLPFC = dorsolateral prefrontal cortex GHR = genetic high-risk; GM = gray matter; L = left; PFC = prefrontal cortex; LHRS = Language High-Risk Study; NYU = New York University; PCC = posterior cingulate cortex; PT = pars triangularis; R = right; STG = superior temporal gyrus; UNC = University of North Carolina; vmPFC = ventromedial prefrontal cortex; WU = Washington University. Note: for a comprehensive listing of all structural MRI findings in GHR individuals, see Thermenos et al., 2013.

Cross-sectional findings

In cross-sectional analyses, GHR youth have most consistently shown evidence for smaller prefrontal cortical (PFC) GM, including reduced cortical thickness^36–38, volume (inferior frontal gyrus^39–42, frontal pole⁴³, medial prefrontal cortex⁴³), and/or gyral surface area⁴⁴ compared to controls. Other brain areas where GHR have reliably shown less GM in comparison to controls include: temporal cortex (decreased bilateral superior temporal gyrus volume^{45, 46} and surface area⁴⁷, bilateral temporal lobe cortical thinning^{36, 37}), parietal cortex (decreased GM volume^{37, 42, 48} and reduced cortical thickness^{38, 49}), and medial temporal/limbic regions (hippocampus^50–55, parahippocampus^{36, 56}, anterior cingulate cortex^{36, 57}). More variable findings have been reported with respect to smaller GM in GHR versus controls in occipital cortex, cerebellum, amygdala, thalamus, and basal ganglia³⁵. Significant associations between higher levels of attenuated psychotic symptoms and smaller GM in PFC^57–59, temporal cortex^60–62, parietal cortex⁵⁹, amygdala^{59, 61, 62}, and cerebellum^{61, 62} in GHR youth and/or young adults have been reported. Regarding age-related neural alterations, significantly less GM has been observed in GHR samples with children as young as age 7 years⁴⁸. However, there is insufficient data regarding neural alterations at specific ages, or developmental periods (e.g., middle childhood versus adolescence), to draw firm conclusions about the onset of GM loss in GHR youth³⁵. Only one research group has found greater GM in GHR youth compared to controls⁴⁰. This included increased cortical thickness of PFC (inferior orbital, middle frontal gyri), temporal cortex (right superior temporal gyrus), and parietal cortex (angular gyrus, inferior parietal cortex)⁴⁰. Thus, there is substantial evidence, most consistently involving PFC and hippocampus, of less GM volume in HR subjects than controls.

Longitudinal findings

Two research groups (Pittsburgh High Risk (PHR) and Edinburgh (EHR)) have carried out longitudinal studies of GHR first-degree adolescent and/or young adult relatives of patients with adult onset schizophrenia. Consistent with the cross-sectional findings, both groups showed progressive reductions of PFC volume over time (1 year follow up (PHR), and 10 year follow up (EHR)) in GHR compared to controls^{49, 58, 63}. Further, progressive decline in PFC GM has been linked with greater symptom levels in GHR individuals, including those who developed schizophrenia^{59, 63}. Similar associations between increasing levels of symptoms and significant deceases in temporal cortical GM volume over time have also been reported^{59, 63}. An association between greater symptom severity and progressive decline in parietal cortex volume is reported in one study⁵⁹.

A third research group has carried out a longitudinal GHR study focused on the development of brain structure in the “very healthy” siblings of individuals with childhood onset schizophrenia ([COS]; i.e., schizophrenia occurring in affected individuals < age 13). As with GHR studies involving first-degree relatives of people with adult-onset schizophrenia, COS relatives initially show significant GM reductions of PFC, temporal, and parietal cortex^{37, 38}. However, over time these structural alterations were found to normalize, with no significant GM cortical decrements detected among COS relatives compared to controls by the end of adolescence (ages 17 to 20)³⁸. Of note, none of the non-affected COS siblings developed psychosis during the follow-up period and, thus, may have comprised particularly resilient individuals. By contrast, on the basis of findings from studies involving families with strong evidence of genetic loading (i.e., EHR, in which relatives had at least two affected family members⁶³), or in the offspring of persons with schizophrenia (i.e., PHR⁴⁹), there is mounting evidence for accelerated reduction in PFM GM in GHR individuals, particularly those who become symptomatic or go on to develop schizophrenia (~10%).

Structural MRI Findings in Clinical High-Risk Individuals

Clinical high-risk (CHR) studies have provided an alternative approach to the investigation of alterations of neural structure associated with schizophrenia risk in adolescents and young adults based on the presence of clinical risk syndromes indicative of the “need-for-care”⁶⁴ – i.e., low level, attenuated positive symptoms; brief intermittent psychotic symptoms; or, genetic risk accompanied by functional decline⁶⁴. Since approximately 20 percent of persons meeting “prodromal” criteria convert to psychosis within one year of initial assessment, and 35% over about 3 years⁶⁵, CHR studies provide a method for examining brain structural alterations proximal to the emergence of frank psychosis, which could ultimately elucidate pathophysiological processes most closely associated with illness onset⁷. Structural MRI findings in CHR studies have been the subject of several recent systematic and critical reviews^{66, 67}. As with the GHR literature, structural MRI studies of CHR individuals have employed a wide-range of imaging methods, and MRI morphometric analytic techniques. Here, we summarize the structural MRI findings in CHR youth from 11 independent research groups and one multicenter study, as well as from two meta-analyses (Table 2).

Table 2.

Structural MRI findings from studies of clinical high-risk for schizophrenia individuals

Group	Author (Year)	Study type	Results
Melbourne	Phillips et al. (2002)	Cross-sectional	-Smaller hippocampal volume bilaterally in CHR
			-Larger L hippocampal volume in CHR-t vs. CHR-n, but no differences compared to controls
	Pantelis et al. (2003)	Longitudinal	-Smaller GM in R medial temporal, lateral temporal, and inferior frontal and bilateral cingulate cortices at baseline
			-CHR-t showed reduced GM in L parahippocampal, fusiform, orbitofrontal, and cerebellar cortices and cingulate gyri over time
			-CHR-n showed reduced cerebellar GM
	Yucel et al. (2003)	Cross-sectional	-Interrupted L anterior cingulate sulcus in CHR vs. controls, but no differences between CHR-t and CHR-n
	Garner et al. (2005)	Cross-sectional	-Larger baseline pituitary vol. in CHR-t vs. CHR-n
	Wood et al. (2005)	Cross-sectional	-Smaller hippocampal volume and less L anterior cingulate folding in CHR with GHR vs. CHR without GHR
	Velakoulis et al. (2006)	Cross-sectional	-Normal baseline hippocampal and amygdala volume in CHR
			-Smaller whole-brain volumes in CHR vs. controls
	Fornito et al. (2008)	Longitudinal	-Bilateral thinning of anterior cingulate in CHR-t also associated with negative symptoms
			-Baseline anterior cingulate differences in CHR-t vs. CHR-n predicted time to psychosis onset
	Takahashi et al. (2008)	Cross-sectional	-No increased prevalence of cavum septi pellucidi enlargement in CHR
	Walterfang et al. (2008)	Cross-sectional	-Smaller anterior corpus callosum in CHR-t vs. CHR-n
	Sun et al. (2009)	Longitudinal	-Greater brain contraction in R PFC in CHR-t vs. CHR-over time
	Takahaski et al., (2009)	Longitudinal	-Smaller baseline insula bilaterally in CHR-t vs. CHR-n, and in R insula vs. controls
			-Reduced GM of bilateral insula in CHR-t vs. CHR-n and controls
	Hannan et al. (2010)	Cross-sectional	-No differences in caudate volume in CHR at baseline vs. controls or in CHR-t vs. CHR-n
	Takahashi et al. (2010)	Cross-sectional	-Smaller STG bilaterally at baseline in CHR vs. controls
	Wood et al. (2010)	Cross-sectional	-Smaller L hippocampal volume in CHR vs. controls
	Dazzan et al. (2012)	Longitudinal	-Smaller frontal cortex volume in CHR-t vs. CHR-n at baseline
			-Reduced parietal cortex and temporal cortex (trend) in CHR-t vs. CHR-n
	Whitford et al. (2012)	Cross-sectional	-Smaller cuneus in CHR-HSV1+ vs. CHR-HSV1− and controls
Basel	Borgwardt et al. (2007a)	Cross-sectional	-Smaller GM at baseline in posterior cingulate and precuneus bilaterally and L superior parietal lobule in CHR-t vs. controls
	Borgwardt et al. (2007b)	Longitudinal	-Smaller L insula, STG, cingulate gyrus, and precuneus in CHR vs. controls
			-Reduced R insula, inferior frontal and STG in CHR-t vs. CHR-n
	Borgwardt et al. (2008)	Longitudinal	-Reduced orbitofrontal, superior frontal, inferior temporal, parietal cortex, and cerebellum in CHR-t vs. controls over time
	Haller et al. (2009)	Cross-sectional	-Whole brain cortical thickness asymmetry in CHR vs. controls
	Koutsouleris et al. (2009)	Cross-sectional	-Smaller GM volume in fronto-temporal and limbic structures in CHR-L vs. controls
			-Alterations of bilateral temporal and limbic structures in CHR-E vs. controls
			-Alterations of PFC in CHR-t vs. CHR-n and controls
	Buehlmann et al. (2010)	Cross-sectional	-Asymmetry between L and R hippocampus in CHR vs. controls
	Smieskova et al. (2012)	Cross-sectional	-Smaller insula GM volume bilaterally in CHR-E at baseline vs. CHR-L
			-Insular alterations associated with negative symptoms in CHR
	Walter et al (2012)	Longitudinal	-Reduced hippocampal volume in CHR over time vs. controls
			-No hippocampal volume differences in CHR-t vs. CHR-n
Berlin	Witthaus et al. (2009)	Cross-sectional	-Smaller GM volume in cingulate gyrus bilaterally, R inferior frontal, R STG, and bilateral cingulate cortex in CHR
	Witthaus et al. (2010)	Cross-sectional	-Smaller corpus and tail volume of hippocampus bilaterally in CHR vs. controls
			-Smaller R hippocampal tail volume in CHR-t vs. CHR-n
	Bohner et al. (2012)	Cross-sectional	-Smaller cingulate gyrus GM in CHR vs. controls
Seoul	Choi et al. (2008)	Cross-sectional	- Higher incidence of cavum septum pellucidum in CHR vs. controls
	Jung et al. (2011)	Cross-sectional	-Reduced cortical thickness in PFC, anterior cingulate cortex, inferior parietal cortex, STG and parahippocampal cortex vs. controls
	Han et al. (2012)	Cross-sectional	-Smaller ALIC volume in CHR vs. controls
	Soon Shin et al. (2012)	Cross-sectional	-Reduced cortical thickness in L Heschl’s gyrus in CHR vs. controls
Munich	Meisenzahl et al. (2008)	Cross-sectional	-Smaller GM volume in frontal, lateral temporal, and medial temporal areas in CHR vs. controls
London	Fusar-Poli et al. (2009)	Longitudinal	-Smaller middle and medial frontal gyrus, insula and anterior cingulate cortex volume in CHR vs. controls at baseline
			-No structural differences in CHR and controls at follow up
	Fusar-Poli et al. (2011)	Longitudinal	-Smaller GM volume in L middle and medial frontal gyri in CHR vs. controls at baseline
Utrecht	Ziermans et al. (2009)	Cross-sectional	-No structural difference in CHR vs. controls
	Ziermans et al. (2012)	Longitudinal	-Greater loss of total brain volume in CHR-t vs. CHR-n and controls
			-Cortical thinning in L anterior cingulate, precuneus, and temporo-parietal-occipital areas in CHR-t vs. CHR-n and controls
Amsterdam	Meijer et al. (2011)	Cross-sectional	-Smaller baseline GM density in R STG, MTG, R insula, and L anterior cingulate in CHR-t vs. CHR-n
			-GM reductions correlated with semantic fluency in CHR
Bonn	Hurlemann et al. (2008)	Cross-sectional	-Smaller bilateral hippocampal volume in CHR-L and CHR-E vs. controls
Tokyo	Iwashiro et al. (2012)	Cross-sectional	-Smaller bilateral PT volume in CHR vs. controls
			-Reduced PT vol. correlated with prodromal symptoms in CHR
Los Angeles	Mittal et al. (2010)	Cross-sectional	-Smaller baseline striatal GM volume in CHR-t vs. CHR-n
			-Trend association between reduced GM vol. in CHR-t and increased dyskinetic movements
Multicenter	Mechelli et al. (2011)	Cross-sectional	-Smaller frontal GM volume in CHR vs. controls at baseline
			-Smaller baseline L parahippocampal GM volume in CHR-t vs. CHR-n

Abbreviations: ALIC = anterior limb of internal capsule; CHR = clinical high-risk; CHR-E = early course clinical high-risk; CHR-HSV1+ = clinical high-risk with herpes simplex virus 1; CHR-HSV- = clinical high-risk without herpes simplex virus 1; CHR-L clinical high-risk of long duration; CHR-n = clinical high-risk without transition to psychosis; CHR-t = clinical high-risk with transition to psychosis; GM = gray matter; L = left; MTG = middle temporal gyrus; PFC = prefrontal cortex; PT = pars triangularis; R = right; STG = superior temporal gyrus

Cross-sectional findings

Overall, studies of CHR individuals show brain structural alterations that are neuroanatomically similar to, but less severe than those commonly reported in established schizophrenia⁶⁸. For example, compared to controls, CHR groups have shown both smaller GM volume and cortical thinning in PFC^69–76, lateral temporal cortex^{69, 72, 73, 75–80} (particularly superior temporal gyrus (STG)), and, to a lesser extent, parietal cortex^{72, 81}. Further, in the largest structural MRI study of CHR to date, which involved data collected from five clinical sites, CHR individuals showed significantly less GM in the PFC bilaterally compared to controls⁸². Less PFC GM has also been associated with impaired executive function⁷⁴ and greater symptoms severity⁷¹ in CHR, while smaller STG GM has been linked with deficits involving semantic fluency⁷⁷.

Structural alterations of limbic brain areas and insula are also among the most consistently reported findings in CHR individuals compared to controls. This includes less bilateral^83–85 and ipsilateral^{86, 87} hippocampal GM volume, aberrant surface morphology^{80, 87–90} and smaller GM^{70, 76, 82, 85, 91} in anterior cingulate and paracingulate cortex, as well as asymmetry⁹² and smaller GM volume ^{69, 70, 93, 94} of the insula. In several studies, structural alterations of anterior cingulate⁸⁹ and insula^{93, 94} have been significantly associated with greater negative symptom levels in CHR. Structural abnormalities in CHR involving the cuneus⁹⁵, caudate⁹⁶, anterior limb of the internal capsule⁹⁷, and the presence of cavum septum pellucidum^{98, 99} are less frequently assessed and less consistently reported. In one study, CHR individuals showed less total whole brain volume¹⁰⁰ compared to controls. Only one study has reported no significant differences in any brain structures in CHR persons versus controls¹⁰¹.

Cross-sectional comparisons of CHR who transition to psychosis (CHR-t) to non-converters or controls have provided evidence for smaller GM volume in PFC^{102, 103} and temporal cortical (STG^{69, 88}) GM among CHR-t. CHR-t have also shown aberrant anterior cingulate morphology⁸⁹, smaller insula bilaterally⁹⁴ and on the right⁶⁹, as well as both greater⁸⁴ and smaller¹⁰⁴ hippocampal, or parahippocampal⁸² GM volume. One study reported greater pituitary volume in CHR-t, which may potentially reflect greater exposure to environmental stress in persons who transition to psychosis¹⁰⁵. An additional study of dykinesia in CHR has reported smaller striatal volume in CHR-t with a trend association between less striatal GM and greater dyskinetic symptoms¹⁰⁶. Finally, a study of CHR persons exposed to herpes simplex virus 1 (HSV1) showed smaller GM volume of the cuneus among HSV1 positive CHR-t¹⁰⁶. Overall, cross-sectional studies have shown smaller GM volume in frontal-temporal and medial temporal/limbic structures in CHR individuals compared to controls, with significantly less GM in these brain areas among individuals who transition to psychosis than in non-converters.

Longitudinal findings

In longitudinal studies, comparisons of structural brain alterations in CHR compared to controls have shown progressive GM loss in PFC (orbitofrontal cortex^{76, 102}), lateral temporal cortex (STG^{102, 107}), parietal cortex¹⁰², cingulate gyrus⁷⁶, parahippocampus⁷⁶, fusiform cortex⁷⁶, insula⁹⁴, and cerebellum^{76, 102}. Further, studies comparing structural changes in CHR-t to CHR non-converters have shown evidence for reductions over time in PFC⁸¹ and temporal cortex¹⁰⁷, as well as in the cerebellum¹⁰⁸.

Meta-analyses of CHR structural MRI studies

The clinical diversity of CHR youth and the heterogeneity of MRI morphometric techniques used across studies together have posed a challenge to interpreting CHR structural findings regarding the neural alterations most closely linked to the risk for transitioning to psychosis. To shed further light on the neural correlates associated with the transition to psychosis, Smieskova and colleagues conducted a meta-analysis of structural MRI findings in both GHR and CHR, comparing HR individuals who transitioned to psychosis (HR-t) with non-coverters⁶⁷. Overall, HR-t showed significantly decreased GM volume in PFC, temporal cortex, the limbic system, and cerebellum, compared to non-converters⁶⁷. A subsequent meta-analysis by Fusar-Poli and colleagues⁶⁶ of voxel-based morphometric studies in GHR and CHR showed smaller GM volume in the PFC, temporal cortex (STG), anterior cingulate, parahippocampus, and precuneus in HR individuals⁶⁶. In the same meta-analysis, a comparison of HR-t to non-converters revealed less GM in PFC (inferior frontal gyrus) and temporal cortex (STG) in HR-t. A comparison of CHR to GHR in the same meta-analysis showed smaller GM volume in the anterior cingulate bilaterally in CHR, while GHR showed less GM in the left hippocampal gyrus, insula, and right temporal cortex (STG) compared to CHR individuals ⁶⁶. Taken together, these meta-analyses show smaller PFC, STG, and medial temporal structures across HR populations, as well as converging evidence for reduced fronto-temporal GM volume in HR individuals who develop psychosis.

Early-Onset and Childhood-Onset Schizophrenia

Schizophrenia beginning in adolescence (EOS, age 13–18) or childhood (COS, < age 13) occurs rarely (approximately 4% of cases¹⁰⁹), but is generally more clinically and neurobiologically severe than the adult-onset illness¹¹⁰. In particular, the brain structural abnormalities observed in COS have been shown to be significantly greater than in adults with schizophrenia¹¹⁰. Research over the last two decades regarding the pattern of neural alterations in COS, premorbid risk factors, and neurocognitive deficits in non-affected family members have provided strong evidence suggesting the neurobiological continuity between COS/EOS and adult-onset schizophrenia¹¹¹. Further, because of evidence for greater genetic vulnerability¹¹⁰ in COS (e.g., increased familiarity¹¹⁰, cytogenetic abnormalities¹¹², and copy number variants¹¹²), it is increasingly believed that studies of brain structural alterations in children and adolescents with schizophrenia may be particularly valuable to understanding the neurobiological basis of the GM abnormalities associated with the illness overall. Below, we summarize the findings from structural MRI studies of EOS and COS carried out by 15 independent research groups world-wide (Table 3). It should be noted that roughly half of the studies included in this review have been carried out by the NIMH research group¹¹⁰, which has focused on COS. Further, as with the HR structural MRI literature, there is considerable variability in terms of the MRI morphometric techniques and data analytic methods employed across studies.

Table 3.

Structural MRI findings from studies of early-onset and childhood-onset schizophrenia individuals

Group	Author (Year)	Study type	Results
NIMH	Frazier et al. (1996)	Cross-sectional	-Smaller total brain and thalamic vol. and increased basal gangliar vol., as well as increased lateral ventricular vol. in COS vs. controls
	Jacobsen et al. (1996)	Cross-sectional	-Smaller total cerebral vol. in COS vs. controls
	Jacobsen et al. (1997a)	Cross-sectional	-Smaller cerebellar volume in COS vs. controls
	Jacobsen et al. (1997b)	Cross-sectional	-Larger corpus callosum vol. in COS vs. controls
	Rapoport et al. (1997)	Longitudinal	-Reduced thalamic GM volume and increased lateral ventricular volume over time in COS vs. controls
	Jacobsen et al. (1998)	Longitudinal	-Reduced R temporal lobe, bilateral STG, and L hippocampus vol. over time in COS vs. controls
			-Reduced R STS volume associated with symptom severity
	Nopoulos et al. (1998)	Cross-sectional	-Enlarged cavum septum pellucidi in COS vs. controls
	Giedd et al. (1999)	Longitudinal	-Reduced total brain vol. and hippocampus and increase lateral ventricular vol. over time in COS vs. controls
	Rapoport et al. (1999)	Cross-sectional	-Four times smaller fronto-temporal GM volume in COS vs. controls
	Kumra et al. (2000)	Cross-sectional	-Reduced total cerebral vol. in COS vs. controls
	Thompson et al. (2001)	Longitudinal	-Smaller parietal cortical GM volume with progressive GM volume loss in temporal cortex, followed by PFC over 5 years in COS vs. controls
	Keller et al. (2003a)	Longitudinal	-Reduced cerebellar GM volume over time in COS vs. controls
	Keller et al. (2003b)	Longitudinal	-Reduced volume of splenium of corpus callosum in COS over time vs. controls
	Greenstein et al. (2006)	Longitudinal	-Reduced cortical thickness in temporal cortex over time in COS vs. controls
	Nugent et al. (2007)	Cross-sectional	-Smaller bilateral total hippocampal volume in COS vs. controls
	Bakalar et al. (2009)	Longitudinal	-No baseline or follow-up asymmetry of lateral or medial cortical surface in COS vs. controls
	Greenstein et al. (2011)	Longitudinal	-Reduced cerebellar GM at baseline and over time in COS vs. controls
	Gogtay et al. (2012)	Cross-sectional	-Smaller PFC and temporal GM volume in COS vs. psychosis NOS
	Johnson et al. (2013)	Cross-sectional	-No corpus callosum vol. differences in COS vs. controls
UCLA	Sowell et al. (2000)	Cross-sectional	-Increased lateral ventricular vol. in EOS vs. controls
	Levitt et al. (2001)	Cross-sectional	-Larger L amygdala vol. in EOS vs. controls
	Marquardt et al. (2005)	Cross-sectional	-Anterior cingulate asymmetry in EOS vs. controls
	Taylor et al. (2005)	Cross-sectional	-Greater posterior STG in EOS vs. controls
Iowa	White et al. (2003)	Cross-sectional	-Reduced cortical thickness and surface morphology in fronto-temporo-parietal lobes in EOS vs. controls
	Clark et al. (2010)	Cross-sectional	-Loss of planum temporale asymmetry in EOS vs. controls
Minnesota	Kendi et al. (2008)	Cross-sectional	-Smaller fornix volume in EOS vs. controls
	Kumra et al. (2012)	Cross-sectional	-Smaller L parietal volume in EOS vs. controls
Harvard	Frazier et al. (2005)	Cross-sectional	-Smaller L amygdala vol. in males with EOS vs. controls
UNC	El-Sayed et al. (2010)	Cross-sectional	-Smaller whole brain volume and frontal-parietal GM in EOS vs. controls
Madrid	Reig et al. (2011)	Cross-sectional	-Smaller frontal/parietal cortical GM in EOS vs. controls
	Aranyo et al. (2011)	Longitudinal	-Reduced frontal/parietal GM over time in EOS vs. controls
	Janssen et al. (2012)	Cross-sectional	-Smaller thalamic volume in EOS vs. controls
Oxford	Davies et al. (2001)	Cross-sectional	-Larger fornix in EOS vs. controls
	James et al. (2002)	Cross-sectional	-No GM volume differences in EOS vs. controls
	Collinson et al. (2003)	Cross-sectional	-Smaller total brain volume in EOS vs. controls
	James et al. (2004)	Cross-sectional	-Smaller PFC and thalamic volume in EOS vs. controls
London	Matsumoto et al. (2001a)	Longitudinal	-Reduced total and R STG GM in EOS vs. controls over time
			-Severity of symptoms associated with reduced STG vol. and correlated with age of onset in EOS
	Matsumoto et al. (2001b)	Longitudinal	-Smaller GM whole brain vol. in EOS vs. controls
	Hadjul et al. (2004)	Cross-sectional	-No asymmetries of hemispheric lateralization in EOS vs. controls
Orsay	Pailière-Martinot et al. (2001)	Cross-sectional	-Smaller PFC, L insula, parahippocampal and fusiform GM volume in EOS vs. controls
	Penttila et al. (2008)	Cross-sectional	-Smaller global sulcal indices in both hemispheres in EOS vs. controls
Copenhagen	Pagsberg et al. (2007)	Cross-sectional	-No GM volume differences in EOS vs. controls
Oslo	Juuhl-Langseth et al. (2012)	Cross-sectional	-Bilateral enlargement of lateral and 4th ventricle and bilateral enlargement of caudate in EOS vs. controls
Changsha	Tang et al. (2012)	Cross-sectional	-Smaller L STG/MTG GM volume in EOS vs. controls
			-Smaller L STG/MTG negatively correlated with positive symptoms in EOS
Osaka	Hata et al. (2003)	Cross-sectional	-Enlargement of lateral ventricles in EOS vs. controls
			-Positive correlation between lateral ventricular enlargement and minor physical abnormalities in EOS
Hamamatsu	Yoshihara et al. (2008)	Cross-sectional	-Smaller parahippocampal and inferior frontal GM volume in EOS vs. controls

Abbreviations: COS = childhood-onset schizophrenia; EOS = early-onset schizophrenia; GM = gray matter; L = left; MTG = middle temporal gyrus; NIMH = National Institute of Mental Health; NOS = not otherwise specified; PFC = prefrontal cortex; R = right; STG = superior temporal gyrus; UCLA = University of California, Los Angeles; UNC = University of North Carolina

Cross-sectional findings

Similar to findings in adult onset schizophrenia, structural MRI studies of EOS and COS have consistently shown smaller GM volume in PFC^113–119 and the temporal^{114, 119–122} and parietal^{113, 117, 123} cortices in EOS and COS compared to controls. Abnormalities of PFC thickness²², cortical folding¹²⁴, and asymmetry¹²⁵ have also been reported. Additionally, less STG volume has been linked with both greater symptom severity, as well as earlier age of illness onset¹²¹. However, several cross-sectional studies have found enlargement of temporal cortical structures^126–128, raising the possibility that, alternatively, temporal GM volume reduction may occur developmentally later¹²⁸. In contrast to HR and adult-onset schizophrenia, decreased hippocampal GM is less commonly reported in COS¹²⁹, with several studies showing no significant alterations compared to controls in hippocampal volume^130–132. Other commonly reported structural findings in EOS and COS include: smaller whole brain volume^{113, 126, 130, 132, 133}, greater lateral ventricular volume^{130, 134–136}, and smaller GM in the cerebellum^137–139 and thalamus^{115, 130, 140}. Fewer, and less consistent structural alterations are reported regarding the amygdala^{127, 141}, parahippocampus¹¹⁹, insula¹¹⁶, fusiform gyrus¹¹⁶, basal ganglia¹³⁰, fornix^{142, 143}, corpus callosum^{144, 145}, and cavum septum pellucidum¹⁴⁶. In two studies, no structural brain abnormalities in any brain areas in EOS versus controls were found^{147, 148}. Thus, cross-sectional EOS and COS studies have provided consistent evidence for smaller whole brain volume, enlargement of lateral ventricles, in conjunction with smaller PFC and (somewhat less consistently) STG GM.

Longitudinal findings

Longitudinal studies comparing brain structure changes in EOS and COS to controls have shown progressive decreases in GM volume involving PFC^149–151 and the temporal^{120, 150, 151} and parietal^{149, 151} cortices in conjunction with decreases over time in cortical thickness in PFC¹⁵² and temporal cortex¹⁵². Particularly noteworthy were results from a 5 year longitudinal study conducted by the NIMH group¹⁵¹, which revealed a temporal pattern of significant GM volume loss in COS compared to controls, with the earliest deficits seen in the parietal cortex, followed by progression during adolescence to the temporal lobes, and lastly to the prefrontal cortex. Additional brain regions in which volumetric decrements have been observed over time in EOS and COS compared to controls include: the cerebellum^{137, 153}, hippocampus^{120, 154}, thalamus¹⁵⁵, and corpus callosum¹⁵⁶. Finally, progressive enlargement of the lateral ventricles has also been found in COS compared to controls^{140, 154}. Across longitudinal studies, EOS and COS individuals have most consistently shown decrements in GM volume in fronto-temporal and parietal cortices over time.

Summary

Here, we reviewed the structural neuroimaging literature in youth and young adults at high-risk for schizophrenia and in EOS and COS. The most consistent finding was that, compared to normal development, there is accelerated fronto-temporal cortical GM volume reduction across the spectrum of schizophrenia risk and in EOS/COS. Specifically, progressive GM decline in these brain regions occurs in HR youth and young adults who eventually transition to psychosis, and also occurs during adolescence in persons with EOS/COS. Progressive volumetric decline and morphological alterations of limbic structures (e.g., hippocampus, parahippocampus, anterior cingulate) are also prominent among HR individuals who later develop psychosis. Structural alterations over time in limbic areas are less common in COS, although there is some evidence to suggest that these abnormalities may emerge as COS individuals are followed through the end of adolescence.

Overall, these structural neuroimaging findings are broadly consistent with the hypothesis that schizophrenia involves, at least partly, the disruption of normal neurodevelopment occurring during childhood and/or adolescence. Structural MRI findings in HR individuals suggest the potential involvement of both early and late brain dysmaturational processes in the trajectory of GM alterations during pre-psychosis development. For example, evidence for altered surface morphology of PFC⁴⁴ and STG⁴⁷ in GHR individuals are thought to potentially reflect abnormalities of neuronal migration and mini-columnar formation during gestation^{2, 16}. At the same time, it has been proposed that GM volume loss in frontal-temporal brain regions of CHR individuals who transition to psychosis^{102, 107} could reflect dysregulation of synaptic pruning during adolescence⁶⁸. Further, the progressive reduction in GM from posterior (parietal) to anterior (prefrontal) cortical brain areas over time found in COS follows the pattern of decline in GM observed during typically developing adolescents^{13, 14}, and, thus, has been interpreted as an indication of aberrant acceleration of normal brain maturational processes¹⁵⁷. Although speculative, taken together these findings lend support to the “2-hit” model proposed by Keshavan and colleagues^{2, 7, 8}, in which neural dysmaturation occurring during early development is thought to produce a vulnerability to later abnormalities of adolescent brain development that ultimately result in the emergence of psychosis.

Nevertheless, despite the evolving evidence implicating aberrant neural developmental processes in the pathophysiology of schizophrenia, it is acknowledged that the findings from both GHR and CHR structural MRI studies are quite variable and difficult to replicate^{35, 158}. Issues pertaining to the clinical heterogeneity of CHR and GHR subjects, as well as the diversity of neuroimaging methods employed for acquisition and analysis of MRI data have been identified as central to the difficulties of comparing results between research groups^{35, 158}. As a result, in part, structural MRI findings currently lack sufficient specificity and sensitivity to be used clinically to identify biomarkers for the prospective identification of individuals at risk for developing schizophrenia.

However, we suggest that future neuroimaging studies of GHR/CHR and EOS/COS populations might take several further steps in order to address the gaps in current knowledge regarding premorbid and prodromal structural brain alterations preceding schizophrenia onset, and to address the challenges of improving the clinical applicability of structural MRI findings to early intervention and prevention strategies for persons at risk for psychosis. First, the predictive value of structural MRI findings might be enhanced if the volumetric or morphological alterations observed in CHR and GHR individuals are incorporated within a multivariate approach, in which structural changes are combined with clinical and neurocognitive measures in models to predict later psychopathology¹⁵⁹. Additionally, there is recent evidence suggesting that the use of machine learning techniques to identify patterns of structural abnormalities associated with the transition to psychosis in HR individuals could be used prospectively to improve the predictive specificity of structural MRI findings during the pre-psychosis period^{73, 160}. Second, given the extensive clinical and neurobiological overlap between schizophrenia and bipolar affective disorder¹⁶¹, studies of young first-degree relatives of probands across the psychotic spectrum may help determine which structural MRI abnormalities are most specific to schizophrenia risk¹⁶². Third, while our review is limited to structural MRI findings, how GM alterations develop in conjunction with changes in WM, impairments of brain function, cognitive deficits, as well as other potential markers of schizophrenia risk (e.g., inflammatory markers and oxidative stress), all of which appear to evolve during the early phase of schizophrenia¹⁶³, remains to be determined. Future longitudinal studies need to address these questions and control for diagnostic variability, as well as differences of age and gender. Fourth, the potential influence of early (e.g., perinatal complications) and later (e.g., substance misuse, psychosocial stress) environmental stress on neural development in the context of risk needs to be clarified in an effort to elucidate the neurobiology of schizophrenia, and to identify the risk markers that can be most useful to early intervention strategies to preempt illness onset. Finally, given the evidence reviewed here of early developmental pathology underlying schizophrenia risk, future research on younger individuals at GHR (i.e., preteen children) will be critical to the further clarification of the origins of brain structural abnormalities associated with the development of schizophrenia.

Key Points.

1. Structural MRI evidence indicates that the adolescent/young adult development of individuals at genetic and clinical high-risk for schizophrenia, as well as of persons with early-onset schizophrenia, is associated with smaller brain volumes, particularly in fronto-temporal cortical areas.

2. There is evidence implicating the disruption of both “early” (i.e., perinatal) and “late” (i.e., adolescent) normal neurodevelopmental processes, which lends support to the “two hit” neurodevelopmental model of schizophrenia.

3. Future longitudinal studies that control for diagnostic variability, age and gender effects, and which examine the evolution of structural brain changes in at-risk youth/young adults in the context of evolving changes of white matter, brain function, and neurocognition will contribute to improving the clinical applicability of structural MRI findings during the premorbid and prodromal periods to early intervention strategies for illness prevention.