Brain Heterogeneity in Schizophrenia and Its Association With Polygenic Risk | Genetics and Genomics | JAMA Psychiatry | JAMA Network
[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Mean and Dispersion of Cortical Thickness
Mean and Dispersion of Cortical Thickness

All maps were thresholded using permutation testing, threshold-free cluster enhancement, and fitting the tail of the permutation distribution to a generalized Pareto distribution (500 permutations; P < .05, familywise error). A, In the t map for the schizophrenia mean model, blue shades represent areas with decreased mean thickness in schizophrenia compared with healthy controls. Schizophrenia was associated with decreased thickness globally, with the exception of the visual cortex, and with strongest effects in frontal and temporal regions, compared with healthy controls. B, In the t map for the schizophrenia dispersion model, orange and yellow shades represent areas with increased heterogeneity in schizophrenia compared with healthy controls. Interindividual variability in cortical thickness showed a spatially global increase for the schizophrenia group compared with healthy controls. C, In an independent sample of healthy adults, the mean model showed that higher polygenic risk for schizophrenia was associated with lower cortical thickness, represented by blue shades, in frontal and temporal cortices. D, Polygenic risk was not associated with cortical thickness heterogeneity in any region.

Figure 2.  Shift Function Plots
Shift Function Plots

Top graphs, Marginal distributions for patients with schizophrenia and healthy controls. Lines show the amount of shift between the 2 distributions. Orange lines and boxes indicate that corresponding deciles are lower in schizophrenia compared with healthy control groups (purple shows the reverse). Bottom graphs, The magnitude of the group difference is plotted as a function of the distribution among healthy controls. A sloped line indicates a difference in the distributions between the groups. Error bars represent bootstrapped 95% CIs. A, Vertex values were extracted by masking the images by the schizophrenia dispersion significance map, and the mean was calculated across vertices and hemispheres and residualized for scanner, sex, and age. Schizophrenia was associated with reduced thickness, with larger differences between groups in the lower deciles. B, Values were residualized for scanner, sex, age, and estimated intracranial volume (eTIV). Schizophrenia was associated with larger volumes compared with controls, with the largest difference between groups in the upper deciles. C and D, Values were residualized for scanner, sex, age, and eTIV. Schizophrenia was associated with smaller volumes compared with controls, with the largest difference between groups in the upper deciles.

Figure 3.  Mean and Dispersion Values of Cortical and Subcortical Volumes
Mean and Dispersion Values of Cortical and Subcortical Volumes

The t statistics for mean and dispersion models are shown. Filled dots mark significant effects after correction for multiple comparisons across regions (5000 permutations; permuted P < .05, familywise error, adjusted for estimated intracranial volume). A, The schizophrenia group had decreased cortical and subcortical volumes and increased ventricle, putamen, and pallidum volumes. Cortical, hippocampal, and ventricle volumes were more heterogeneous in the schizophrenia group compared with healthy controls. B, Polygenic risk for schizophrenia was not associated with mean changes or dispersion in any of the regions. CC indicates corpus callosum; GM, gray matter; Lh, left hemisphere; Rh, right hemisphere; and WM, white matter.

Figure 4.  Mean and Dispersion of Hippocampus Subfield Volumes
Mean and Dispersion of Hippocampus Subfield Volumes

The t statistics for mean and dispersion models are shown. Filled dots mark significant effects after correction for multiple comparisons across regions (5000 permutations; permuted P < .05, familywise error, adjusted for intracranial volume). A, The schizophrenia group had decreased hippocampal volumes. This decrease was also evident in all subfields and accompanied by an increase of the hippocampal fissures. Whole hippocampal volumes were also more heterogeneous in the schizophrenia group, and among the subfields this effect was present in the left molecular layer, left CA1 (cornu ammonis 1), left granule cell layer of the dentate gyrus (GC-DG), left CA4, and left presubiculum. B, Polygenic risk for schizophrenia was associated with mean reductions of left dentate gyrus, left CA4, and bilateral CA2/3. Total hippocampal volumes and subfields showed no significant association between polygenic risk and volume heterogeneity. HATA indicates hippocampus-amygdala transition area.

1.
World Health Organization. The Global Burden of Disease: 2004 Update. Geneva, Switzerland: WHO Press; 2008.
2.
Lakhan  SE, Vieira  KF.  Schizophrenia pathophysiology: are we any closer to a complete model?  Ann Gen Psychiatry. 2009;8(1):12. doi:10.1186/1744-859X-8-12PubMedGoogle ScholarCrossref
3.
Owen  MJ, Sawa  A, Mortensen  PB.  Schizophrenia.  Lancet. 2016;388(10039):86-97. doi:10.1016/S0140-6736(15)01121-6PubMedGoogle ScholarCrossref
4.
Van Rheenen  TE, Lewandowski  KE, Tan  EJ,  et al.  Characterizing cognitive heterogeneity on the schizophrenia-bipolar disorder spectrum.  Psychol Med. 2017;47(10):1848-1864. doi:10.1017/S0033291717000307PubMedGoogle ScholarCrossref
5.
Malhotra  AK.  Dissecting the heterogeneity of treatment response in first-episode schizophrenia.  Schizophr Bull. 2015;41(6):1224-1226. doi:10.1093/schbul/sbv117PubMedGoogle ScholarCrossref
6.
Huber  G.  The heterogeneous course of schizophrenia.  Schizophr Res. 1997;28(2-3):177-185. doi:10.1016/S0920-9964(97)00113-8PubMedGoogle ScholarCrossref
7.
Weinberg  D, Lenroot  R, Jacomb  I,  et al.  Cognitive subtypes of schizophrenia characterized by differential brain volumetric reductions and cognitive decline.  JAMA Psychiatry. 2016;73(12):1251-1259. doi:10.1001/jamapsychiatry.2016.2925PubMedGoogle ScholarCrossref
8.
Zhang  T, Koutsouleris  N, Meisenzahl  E, Davatzikos  C.  Heterogeneity of structural brain changes in subtypes of schizophrenia revealed using magnetic resonance imaging pattern analysis.  Schizophr Bull. 2015;41(1):74-84. doi:10.1093/schbul/sbu136PubMedGoogle ScholarCrossref
9.
Sugihara  G, Oishi  N, Son  S, Kubota  M, Takahashi  H, Murai  T.  Distinct patterns of cerebral cortical thinning in schizophrenia: a neuroimaging data-driven approach.  Schizophr Bull. 2017;43(4):900-906.PubMedGoogle Scholar
10.
Seaton  BE, Goldstein  G, Allen  DN.  Sources of heterogeneity in schizophrenia: the role of neuropsychological functioning.  Neuropsychol Rev. 2001;11(1):45-67. doi:10.1023/A:1009013718684PubMedGoogle ScholarCrossref
11.
Koutsouleris  N, Gaser  C, Jäger  M,  et al.  Structural correlates of psychopathological symptom dimensions in schizophrenia: a voxel-based morphometric study.  Neuroimage. 2008;39(4):1600-1612. doi:10.1016/j.neuroimage.2007.10.029PubMedGoogle ScholarCrossref
12.
van Erp  TGM, Walton  E, Hibar  DP,  et al.  Cortical brain abnormalities in 4474 individuals with schizophrenia and 5098 controls via the ENIGMA consortium.  Biol Psychiatry. 2018;84(9):644-654. doi:10.1016/j.biopsych.2018.04.023Google Scholar
13.
Moberget  T, Doan  NT, Alnæs  D,  et al.  Cerebellar volume and cerebellocerebral structural covariance in schizophrenia: a multisite mega-analysis of 983 patients and 1349 healthy controls.  Mol Psychiatry. 2018;23(6):1512-1520.PubMedGoogle Scholar
14.
van Erp  TGM, Hibar  DP, Rasmussen  JM,  et al.  Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium [published correction appears in Mol Psychiatry. 2016;21(4):585].  Mol Psychiatry. 2016;21(4):547-553. doi:10.1038/mp.2015.63PubMedGoogle ScholarCrossref
15.
Brugger  SP, Howes  OD.  Heterogeneity and homogeneity of regional brain structure in schizophrenia: a meta-analysis.  JAMA Psychiatry. 2017;74(11):1104-1111. doi:10.1001/jamapsychiatry.2017.2663PubMedGoogle ScholarCrossref
16.
Wolfers  T, Buitelaar  JK, Beckmann  CF, Franke  B, Marquand  AF.  From estimating activation locality to predicting disorder: a review of pattern recognition for neuroimaging-based psychiatric diagnostics.  Neurosci Biobehav Rev. 2015;57:328-349. doi:10.1016/j.neubiorev.2015.08.001PubMedGoogle ScholarCrossref
17.
Doan  NT, Kaufmann  T, Bettella  F,  et al.  Distinct multivariate brain morphological patterns and their added predictive value with cognitive and polygenic risk scores in mental disorders.  Neuroimage Clin. 2017;15:719-731. doi:10.1016/j.nicl.2017.06.014PubMedGoogle ScholarCrossref
18.
Wolfers  T, Doan  NT, Kaufmann  T,  et al.  Mapping the heterogeneous phenotype of schizophrenia and bipolar disorder using normative models.  JAMA Psychiatry. 2018;75(11):1146-1155. doi:10.1001/jamapsychiatry.2018.2467PubMedGoogle ScholarCrossref
19.
Gopal  S, Miller  RL, Michael  A,  et al.  Spatial variance in resting fMRI networks of schizophrenia patients: an independent vector analysis.  Schizophr Bull. 2016;42(1):152-160.PubMedGoogle Scholar
20.
Gopal  S, Miller  RL, Baum  SA, Calhoun  VD.  Approaches to capture variance differences in rest fMRI networks in the spatial geometric features: application to schizophrenia.  Front Neurosci. 2016;10:85. doi:10.3389/fnins.2016.00085PubMedGoogle ScholarCrossref
21.
Sullivan  PF, Kendler  KS, Neale  MC.  Schizophrenia as a complex trait: evidence from a meta-analysis of twin studies.  Arch Gen Psychiatry. 2003;60(12):1187-1192. doi:10.1001/archpsyc.60.12.1187PubMedGoogle ScholarCrossref
22.
Schizophrenia Working Group of the Psychiatric Genomics Consortium.  Biological insights from 108 schizophrenia-associated genetic loci.  Nature. 2014;511(7510):421-427. doi:10.1038/nature13595PubMedGoogle ScholarCrossref
23.
Purcell  SM, Wray  NR, Stone  JL,  et al; International Schizophrenia Consortium.  Common polygenic variation contributes to risk of schizophrenia and bipolar disorder.  Nature. 2009;460(7256):748-752.PubMedGoogle ScholarCrossref
24.
Jones  HJ, Stergiakouli  E, Tansey  KE,  et al.  Phenotypic manifestation of genetic risk for schizophrenia during adolescence in the general population.  JAMA Psychiatry. 2016;73(3):221-228. doi:10.1001/jamapsychiatry.2015.3058PubMedGoogle ScholarCrossref
25.
Kauppi  K, Westlye  LT, Tesli  M,  et al.  Polygenic risk for schizophrenia associated with working memory-related prefrontal brain activation in patients with schizophrenia and healthy controls.  Schizophr Bull. 2015;41(3):736-743. doi:10.1093/schbul/sbu152PubMedGoogle ScholarCrossref
26.
Walton  E, Geisler  D, Lee  PH,  et al.  Prefrontal inefficiency is associated with polygenic risk for schizophrenia.  Schizophr Bull. 2014;40(6):1263-1271. doi:10.1093/schbul/sbt174PubMedGoogle ScholarCrossref
27.
Neilson  E, Bois  C, Gibson  J,  et al.  Effects of environmental risks and polygenic loading for schizophrenia on cortical thickness.  Schizophr Res. 2017;184:128-136. doi:10.1016/j.schres.2016.12.011PubMedGoogle ScholarCrossref
28.
Chen  Q, Ursini  G, Romer  AL,  et al.  Schizophrenia polygenic risk score predicts mnemonic hippocampal activity.  Brain. 2018;141(4):1218-1228. doi:10.1093/brain/awy004PubMedGoogle ScholarCrossref
29.
Lieberman  JA, Girgis  RR, Brucato  G,  et al.  Hippocampal dysfunction in the pathophysiology of schizophrenia: a selective review and hypothesis for early detection and intervention.  Mol Psychiatry. 2018;23(8):1764-1772. doi:10.1038/mp.2017.249PubMedGoogle ScholarCrossref
30.
Reddaway  JT, Doherty  JL, Lancaster  T, Linden  D, Walters  JT, Hall  J. Genomic and imaging biomarkers in schizophrenia. In: Pratt J, Hall J, eds. Biomarkers in Psychiatry. Berlin, Germany: Springer Berlin Heidelberg; 2018:325-352. doi:10.1007/7854_2018_52
31.
Fraser  HB, Schadt  EE.  The quantitative genetics of phenotypic robustness.  PLoS One. 2010;5(1):e8635. doi:10.1371/journal.pone.0008635PubMedGoogle ScholarCrossref
32.
Dwyer  DB, Cabral  C, Kambeitz-Ilankovic  L,  et al.  Brain subtyping enhances the neuroanatomical discrimination of schizophrenia.  Schizophr Bull. 2018;44(5):1060-1069. doi:10.1093/schbul/sby008PubMedGoogle ScholarCrossref
33.
Conley  D, Johnson  R, Domingue  B, Dawes  C, Boardman  J, Siegal  ML.  A sibling method for identifying vQTLs.  PLoS One. 2018;13(4):e0194541. doi:10.1371/journal.pone.0194541PubMedGoogle ScholarCrossref
34.
Fischl  B.  FreeSurfer.  Neuroimage. 2012;62(2):774-781. doi:10.1016/j.neuroimage.2012.01.021PubMedGoogle ScholarCrossref
35.
Fischl  B, Salat  DH, Busa  E,  et al.  Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain.  Neuron. 2002;33(3):341-355. doi:10.1016/S0896-6273(02)00569-XPubMedGoogle ScholarCrossref
36.
Dale  AM, Fischl  B, Sereno  MI.  Cortical surface-based analysis, I: segmentation and surface reconstruction.  Neuroimage. 1999;9(2):179-194. doi:10.1006/nimg.1998.0395PubMedGoogle ScholarCrossref
37.
Fischl  B, Sereno  MI, Dale  AM.  Cortical surface-based analysis, II: inflation, flattening, and a surface-based coordinate system.  Neuroimage. 1999;9(2):195-207. doi:10.1006/nimg.1998.0396PubMedGoogle ScholarCrossref
38.
Iglesias  JE, Augustinack  JC, Nguyen  K,  et al; Alzheimer’s Disease Neuroimaging Initiative.  A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI.  Neuroimage. 2015;115:117-137. doi:10.1016/j.neuroimage.2015.04.042PubMedGoogle ScholarCrossref
39.
Euesden  J, Lewis  CM, O’Reilly  PF.  PRSice: polygenic risk score software.  Bioinformatics. 2015;31(9):1466-1468. doi:10.1093/bioinformatics/btu848PubMedGoogle ScholarCrossref
40.
Winkler  AM, Ridgway  GR, Webster  MA, Smith  SM, Nichols  TE.  Permutation inference for the general linear model.  Neuroimage. 2014;92:381-397. doi:10.1016/j.neuroimage.2014.01.060PubMedGoogle ScholarCrossref
41.
Winkler  AM, Ridgway  GR, Douaud  G, Nichols  TE, Smith  SM.  Faster permutation inference in brain imaging.  Neuroimage. 2016;141:502-516. doi:10.1016/j.neuroimage.2016.05.068PubMedGoogle ScholarCrossref
42.
Nichols  TE, Holmes  AP.  Nonparametric permutation tests for functional neuroimaging: a primer with examples.  Hum Brain Mapp. 2002;15(1):1-25. doi:10.1002/hbm.1058PubMedGoogle ScholarCrossref
43.
Rosen  AFG, Roalf  DR, Ruparel  K,  et al.  Quantitative assessment of structural image quality.  Neuroimage. 2018;169:407-418. doi:10.1016/j.neuroimage.2017.12.059PubMedGoogle ScholarCrossref
44.
Clementz  BA, Sweeney  JA, Hamm  JP,  et al.  Identification of distinct psychosis biotypes using brain-based biomarkers.  Am J Psychiatry. 2016;173(4):373-384. doi:10.1176/appi.ajp.2015.14091200PubMedGoogle ScholarCrossref
45.
Mistry  S, Harrison  JR, Smith  DJ, Escott-Price  V, Zammit  S.  The use of polygenic risk scores to identify phenotypes associated with genetic risk of schizophrenia: systematic review.  Schizophr Res. 2017;197:2-8. doi:10.1016/j.schres.2017.10.037PubMedGoogle ScholarCrossref
46.
Tesli  M, Espeseth  T, Bettella  F,  et al.  Polygenic risk score and the psychosis continuum model.  Acta Psychiatr Scand. 2014;130(4):311-317. doi:10.1111/acps.12307PubMedGoogle ScholarCrossref
47.
Liu  B, Zhang  X, Cui  Y,  et al.  Polygenic risk for schizophrenia influences cortical gyrification in 2 independent general populations.  Schizophr Bull. 2017;43(3):673-680.PubMedGoogle Scholar
48.
Cordova-Palomera  A, Meer D, Kaufmann T, et al.  Genetic control of variability in subcortical and intracranial volumes.  bioRxiv. October 15, 2018. https://www.biorxiv.org/content/10.1101/443549v1. Google Scholar
49.
Agerbo  E, Sullivan  PF, Vilhjálmsson  BJ,  et al.  Polygenic risk score, parental socioeconomic status, family history of psychiatric disorders, and the risk for schizophrenia: a Danish population-based study and meta-analysis.  JAMA Psychiatry. 2015;72(7):635-641. doi:10.1001/jamapsychiatry.2015.0346PubMedGoogle ScholarCrossref
50.
Smeland  OB, Wang  Y, Frei  O,  et al.  Genetic overlap between schizophrenia and volumes of hippocampus, putamen, and intracranial volume indicates shared molecular genetic mechanisms.  Schizophr Bull. 2018;44(4):854-864. doi:10.1093/schbul/sbx148PubMedGoogle ScholarCrossref
51.
van der Meer  D, Rokicki  J, Kaufmann  T,  et al; Alzheimer’s Disease Neuroimaging Initiative; Pediatric Imaging, Neurocognition and Genetics Study.  Brain scans from 21,297 individuals reveal the genetic architecture of hippocampal subfield volumes [printed online October 2, 2018].  Mol Psychiatry. PubMedGoogle Scholar
52.
McIntosh  AR, Lobaugh  NJ.  Partial least squares analysis of neuroimaging data: applications and advances.  Neuroimage. 2004;23(suppl 1):S250-S263. doi:10.1016/j.neuroimage.2004.07.020PubMedGoogle ScholarCrossref
53.
Reuter  M, Tisdall  MD, Qureshi  A, Buckner  RL, van der Kouwe  AJW, Fischl  B.  Head motion during MRI acquisition reduces gray matter volume and thickness estimates.  Neuroimage. 2015;107:107-115. doi:10.1016/j.neuroimage.2014.12.006PubMedGoogle ScholarCrossref
54.
Schwartz  S, Susser  E.  The use of well controls: an unhealthy practice in psychiatric research.  Psychol Med. 2011;41(6):1127-1131. doi:10.1017/S0033291710001595PubMedGoogle ScholarCrossref
55.
Bernal-Rusiel  JL, Atienza  M, Cantero  JL.  Determining the optimal level of smoothing in cortical thickness analysis: a hierarchical approach based on sequential statistical thresholding.  Neuroimage. 2010;52(1):158-171. doi:10.1016/j.neuroimage.2010.03.074PubMedGoogle ScholarCrossref
56.
Insel  TR.  Rethinking schizophrenia.  Nature. 2010;468(7321):187-193. doi:10.1038/nature09552PubMedGoogle ScholarCrossref
57.
Alnæs  D, Kaufmann  T, Doan  NT,  et al.  Association of heritable cognitive ability and psychopathology with white matter properties in children and adolescents.  JAMA Psychiatry. 2018;75(3):287-295. doi:10.1001/jamapsychiatry.2017.4277PubMedGoogle ScholarCrossref
58.
Kaufmann  T, Alnæs  D, Doan  NT, Brandt  CL, Andreassen  OA, Westlye  LT.  Delayed stabilization and individualization in connectome development are related to psychiatric disorders.  Nat Neurosci. 2017;20(4):513-515. doi:10.1038/nn.4511PubMedGoogle ScholarCrossref
59.
Honnorat  N, Dong  A, Meisenzahl-Lechner  E, Koutsouleris  N, Davatzikos  C.  Neuroanatomical heterogeneity of schizophrenia revealed by semi-supervised machine learning methods [published online December 20, 2017].  Schizophr Res.PubMedGoogle Scholar
60.
Viher  PV, Stegmayer  K, Giezendanner  S,  et al.  Cerebral white matter structure is associated with DSM-5 schizophrenia symptom dimensions.  Neuroimage Clin. 2016;12:93-99. doi:10.1016/j.nicl.2016.06.013PubMedGoogle ScholarCrossref
Original Investigation
April 10, 2019

Brain Heterogeneity in Schizophrenia and Its Association With Polygenic Risk

Author Affiliations
  • 1Norwegian Centre for Mental Disorders Research, KG Jebsen Centre for Psychosis Research, Division of Mental Health and Addiction, Oslo University Hospital and Institute of Clinical Medicine, University of Oslo, Norway
  • 2Centre for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden
  • 3Department of Psychological and Brain Sciences, Washington University in Saint Louis, St Louis, Missouri
  • 4Psychiatric Neuroscience Group, Department of Basic Medical Sciences, Neuroscience and Sense Organs, University of Bari Aldo Moro, Bari, Italy
  • 5Department of Medical Genetics, Oslo University Hospital, Oslo, Norway
  • 6Central Institute of Mental Health, University of Heidelberg, Mannheim, Germany
  • 7Fondazione Istituto di Ricovero e Cura a Carattere Scientifico Casa Sollievo della Sofferenza, San Giovanni Rotondo, Italy
  • 8Medical Faculty Mannheim, University of Heidelberg, Mannheim, Germany
  • 9Department of Psychology, University of Oslo, Oslo, Norway
JAMA Psychiatry. 2019;76(7):739-748. doi:10.1001/jamapsychiatry.2019.0257
Key Points

Question  Are schizophrenia and its polygenic risk associated with brain structural heterogeneity in addition to mean changes?

Findings  In this case-control analysis of 1151 patients and 2010 controls, schizophrenia was associated with increased heterogeneity in frontotemporal thickness and area and cortical, ventricle, and hippocampal volumes, as well as robust reductions in mean estimates. In an independent sample of 12 490 healthy controls, polygenic risk for schizophrenia was associated with thinner frontotemporal cortices and smaller CA2/3 volume of the left hippocampus, but not with heterogeneity.

Meaning  Schizophrenia appears to be associated with increased interindividual differences in brain structure, possibly reflecting clinical heterogeneity, gene-environment interactions, or secondary disease factors.

Abstract

Importance  Between-individual variability in brain structure is determined by gene-environment interactions, possibly reflecting differential sensitivity to environmental and genetic perturbations. Magnetic resonance imaging (MRI) studies have revealed thinner cortices and smaller subcortical volumes in patients with schizophrenia. However, group-level comparisons may mask considerable within-group heterogeneity, which has largely remained unnoticed in the literature.

Objectives  To compare brain structural variability between individuals with schizophrenia and healthy controls and to test whether respective variability reflects the polygenic risk score (PRS) for schizophrenia in an independent sample of healthy controls.

Design, Setting, and Participants  This case-control and polygenic risk analysis compared MRI-derived cortical thickness and subcortical volumes between healthy controls and patients with schizophrenia across 16 cohorts and tested for associations between PRS and MRI features in a control cohort from the UK Biobank. Data were collected from October 27, 2004, through April 12, 2018, and analyzed from December 3, 2017, through August 1, 2018.

Main Outcomes and Measures  Mean and dispersion parameters were estimated using double generalized linear models. Vertex-wise analysis was used to assess cortical thickness, and regions-of-interest analyses were used to assess total cortical volume, total surface area, and white matter, subcortical, and hippocampal subfield volumes. Follow-up analyses included within-sample analysis, test of robustness of the PRS threshold, population covariates, outlier removal, and control for image quality.

Results  A comparison of 1151 patients with schizophrenia (mean [SD] age, 33.8 [10.6] years; 68.6% male [n = 790] and 31.4% female [n = 361]) with 2010 healthy controls (mean [SD] age, 32.6 [10.4] years; 56.0% male [n = 1126] and 44.0% female [n = 884]) revealed higher heterogeneity in schizophrenia for cortical thickness and area (t = 3.34), cortical (t = 3.24) and ventricle (t range, 3.15-5.78) volumes, and hippocampal subfields (t range, 2.32-3.55). In the UK Biobank sample of 12 490 participants (mean [SD] age, 55.9 [7.5] years; 48.2% male [n = 6025] and 51.8% female [n = 6465]), higher PRS was associated with thinner frontal and temporal cortices and smaller left CA2/3 (t = −3.00) but was not significantly associated with dispersion.

Conclusions and Relevance  This study suggests that schizophrenia is associated with substantial brain structural heterogeneity beyond the mean differences. These findings may reflect higher sensitivity to environmental and genetic perturbations in patients, supporting the heterogeneous nature of schizophrenia. A higher PRS was associated with thinner frontotemporal cortices and smaller hippocampal subfield volume, but not heterogeneity. This finding suggests that brain variability in schizophrenia results from interactions between environmental and genetic factors that are not captured by the PRS. Factors contributing to heterogeneity in frontotemporal cortices and hippocampus are key to furthering our understanding of how genetic and environmental factors shape brain biology in schizophrenia.

Introduction

Schizophrenia is a severe psychiatric disorder with a lifetime prevalence of about 1%, rendering it a leading cause of disability worldwide, with 26 million people affected.1 Although genetic and environmental factors contributing to disease risk have been identified, the pathophysiological process remains elusive.2,3 Patients diagnosed with schizophrenia display substantial heterogeneity in terms of their clinical characteristics and symptoms,4 treatment response,5 and long-term prognosis.6 The notion that the observed heterogeneity stems at least partially from distinct subtypes of patients with differentially affected neurobiology and clinical and cognitive profiles7-9 has not been fully confirmed to date.10 Hence, whether a single unifying pathophysiological process is shared across patients or a multitude of partly independent disease processes lead to a similar clinical syndrome remains salient.11

Schizophrenia is associated with widespread brain abnormalities, with the most robust group-level mean structural differences being ventricle enlargement, reduced thickness and area of frontotemporal cortices, and reduced hippocampal and amygdala volumes.12-14 However, substantial variability exists between patients,7,8,15 presenting a major challenge for achieving imaging-based diagnostic predictions (ie, achieving the goal of precision medicine in psychiatry, and using brain imaging as a clinical tool to guide diagnosis and treatment) with any clinical utility.16-18 Rather than simply reflecting noise, this interindividual variability in brain structure may possibly carry relevant information regarding gene-environment interactions related to the individual sensitivity to environmental and genetic perturbation. Only a few studies have investigated whether heterogeneity differs between healthy participants and patients with schizophrenia. One functional imaging study19 reported increased heterogeneity in connectivity and spatial extent of functional brain networks in schizophrenia. Regions with altered spatial variance in functional networks included areas previously implicated in schizophrenia, such as auditory and sensorimotor cortices and basal ganglia, and networks showing increased heterogeneity overlapped with those showing mean volume differences, implying that the mean and variance measures provide complementary but converging results.20 A recent meta-analysis15 reported increased interindividual volumetric variability in several cortical and subcortical structures, including the temporal lobe, thalamus, hippocampus, and amygdala in schizophrenia, and lower variability in the anterior cingulate cortex. These results point to the importance of modeling heterogeneity as well as mean changes. Detecting brain regions that are more homogenous in patients could point to a primary role in a shared underlying pathophysiological process of schizophrenia, whereas regions of increased heterogeneity might be informative of putative subtypes of disease or reflect regional differences in the sensitivity to genetic and environmental perturbations.

Schizophrenia is highly heritable,21 motivating the ongoing efforts to identify intermediate brain phenotypes associated with disease liability to elucidate the pathway from genes to illness manifestation. Several risk loci for schizophrenia have been identified,22 but the individual contribution of each identified variant is weak, and at present no common variants have been conclusively linked to the disease. The polygenic risk score (PRS) for schizophrenia, which represents a weighted sum of common genetic schizophrenia risk alleles, has been proposed to account for the polygenic nature of disease risk.23 Beyond being associated with case-control status,22 PRS has been associated with negative symptoms, anxiety, and lower cognitive ability in adolescents.24 Polygenic burden has also been linked to a thinner cortex and to prefrontal working memory– and hippocampal encoding–related activation and connectivity in patients and healthy participants.25-28 This linkage is in line with findings implicating the frontal cortex and hippocampus as core regions in the pathophysiological process of schizophrenia.29 Polygenic risk for schizophrenia, however, is only weakly associated with subcortical volumes.30 Risk alleles could also exert their effect by influencing the environmental sensitivity, which could be reflected in the phenotypic variability between individuals.31

Thus, revealing brain structures with higher or lower heterogeneity in schizophrenia could facilitate discovery of intermediate brain phenotypes that may serve to identify putative subtypes8,32 of the disease and phenotypes that are primary or common in the neurobiology of schizophrenia.15 Further, investigating how the genetic architecture of disease risk is related to brain heterogeneity could reveal regions in which the cumulative burden of common risk alleles influence the phenotypic variance.33 To this end, we directly compared within-group dispersion in several key brain structural phenotypes, including cortical thickness, as well as cortical, subcortical, and hippocampal subfield volumes between 1151 patients with schizophrenia and 2010 healthy controls. Next, to test whether between-individual variability is associated with the cumulative polygenic risk for schizophrenia in absence of a clinical syndrome, we tested for associations between dispersion in the same brain features and PRS for schizophrenia in 12 490 healthy individuals from the UK Biobank.

Methods
Samples

Data were collected from October 27, 2004, through April 12, 2018. Demographic characteristics and clinical information are presented in eTables 1 and 2 in the Supplement, respectively. The data have been used in previous publications. Details and references are presented in eMethods and eTable 3 in the Supplement, and MRI protocols appear in eTable 4 in the Supplement. Data collection was performed with each participant’s written informed consent and with approval by the respective local institutional review boards.

Image Preprocessing and Genetic Data

We processed T1-weighted MRI images using FreeSurfer (version 5.3.0) for cortical reconstruction and volumetric segmentation,34-37 and FreeSurfer (version 6.0) for hippocampus subfield segmentation38 (eMethods in the Supplement). We calculated PRS using PRSice (version 1.25),39 based on the European Caucasian subset of the 2014 Psychiatric Genomics Consortium 2 schizophrenia genome-wide association study22 (eFigure 1 and eMethods in the Supplement). The PRS based on a threshold of P < .05 was used for the main analysis because this threshold has been reported as optimal in terms of explaining case-control differences.22

Statistical Analysis

Data were analyzed from December 3, 2017, through August 1, 2018. For all included measures in the case-control sample, we used vertex- or volume-wise generalized additive models to regress out scanner effects while accounting for age, sex, and diagnosis (eMethods in the Supplement). For the case-control comparison and the PRS analysis, we modeled vertex- or volume-wise mean and dispersion using double generalized linear models, which iteratively fit a generalized linear model of the mean parameter and a second generalized linear model of the dispersion parameter on the deviances of the first model (eMethods in the Supplement). We permuted diagnostic labels for the case-control comparison and the PRS UK Biobank analysis by recalculating the mean and dispersion parameters for each iteration. For cortical thickness, the true and permuted statistical maps (t maps) were submitted to the permutation analysis of linear models tool40 to correct for multiple comparisons using threshold-free cluster enhancement and tail approximation41 (600 permutations; eMethods in the Supplement). For the region of interest–based measures (eMethods in the Supplement), we performed 5000 permutations per volume and extracted the maximum t value across regions of interest to calculate familywise error.42 Significance threshold was set at 2-tailed P < .05 for all analyses. We also performed a meta-analysis of the multiscanner thickness data. We conducted analyses with and without covarying for estimated intracranial volume (eTIV) for the volumetric measures. We performed follow-up analyses with more stringent exclusion criteria as well, with adding the Euler number as a covariate43 (eMethods in the Supplement). To assess the robustness to the PRS P value threshold, we performed analyses with PRS threshold selection of 2-sided P < .001, the first component from a principal component analysis on PRS (PRS-PC1) calculated across thresholds (eMethods in the Supplement), and exploratory analysis of the second and third components of the principal component analysis (PC2 and PC3) based on their opposite gradient pattern with regard to PRS P value threshold.

Results
Study Samples

The case-control comparison included 1151 patients with schizophrenia (mean [SD] age, 33.8 [10.6] years; 790 male [68.6%] and 361 female [31.4%]) and 2010 healthy controls (mean [SD] age, 32.6 [10.4] years; 1126 male [56.0%] and 884 female [44.0%]). The PRS analysis included 12 490 healthy participants from the UK Biobank (mean [SD] age, 55.9 [7.5] years; 6025 male [48.2%] and 6465 female [51.8%]).

Vertex-Wise Thickness

Schizophrenia was associated with thinner cortex globally, with the exception of the visual cortex, as well as globally higher thickness dispersion (Figure 1A and B, Figure 2A, and eFigure 2 in the Supplement). Meta-analysis of within-sample effects with more stringent exclusion criteria also revealed and confirmed significantly higher heterogeneity in schizophrenia (left hemisphere, β = 0.31 [95% CI, 0.16-0.46; P < .001]; right hemisphere, β = 0.33 [95% CI, 0.18-0.48; P < .001]) (eFigure 3 in the Supplement), and follow-up multisite analysis (eMethods in the Supplement) did not reveal major interactions between age, sex, or diagnosis and site, indicating that dispersion effects are not simply explained by multisite variability, site by demographic interactions, or a few extreme values. The PRS was associated with lower mean thickness in the right inferior frontal gyrus, the right lateral orbitofrontal cortex, the right precentral gyrus, the right medial temporal cortex, and bilaterally in the middle and superior temporal cortex (Figure 1C and D and eFigure 4 in the Supplement). Converging results were obtained on reanalysis with the addition of the first 4 population components added as covariates (eFigure 5A in the Supplement) or with more stringent exclusion criteria (eFigure 5B in the Supplement). Follow-up analysis using the PRS-PC1 gave close to an identical pattern as the PRS model based on a threshold of P < .05 (eFigure 5C in the Supplement; vertex-wise r = 0.91), whereas a PRS threshold of P < .001 showed weaker association with mean thickness (eFigure 5D in the Supplement). We found no significant association between PRS and thickness dispersion, or between PRS-PC2 or PRS-PC3 and mean or dispersion of cortical thickness. Effects of age and sex are shown in eFigure 6 in the Supplement for case-control comparisons and eFigure 7 in the Supplement for PRS analysis.

Cortical and Subcortical Volumes

Schizophrenia was associated with lower mean cortical volume (t = −17.05), mean cortical area (t = −9.35), supratentorial volume (t = −11.43), total (t = −18.04) and subcortical (t = −4.63) gray volume, cerebellar cortical volume (left, t = −11.69; right, t = −10.69), as well as brain stem (t = −9.64), amygdala (left, t = −8.82; right, t = −6.16), thalamus (left, t = −7.53; right, t = −8.35), and nucleus accumbens (left, t = −3.26; right, t = −5.78), and several white matter volumes, as well as increased ventricle (t range, 2.29-12.24), caudate nucleus (left, t = 4.00; right, t = 2.16), pallidum (left, t = 9.57; right, t = 9.55), and putamen (left, t = 5.48; right, t = 6.44) volumes. Schizophrenia was further associated with higher dispersion in mean cortical volume (t = 3.24), mean cortical area (t = 3.34), total gray volume (t = 3.41), and ventricle volumes (t range, 3.18-5.78) (Figure 2B and Figure 3). Models without eTIV (eFigure 8A in the Supplement) revealed no significant differences in the mean volumes of caudate nucleus and left putamen and resulted in an additional significant association with dispersion in supratentorial volume (t = 2.92). Reanalysis of mean and dispersion models with more stringent exclusion criteria showed converging results (eTable 5 in the Supplement). The PRS was not associated with the mean or the dispersion in any of the subcortical volumes (Figure 3B), which was also true for models without adjustment for eTIV (eFigure 8B in the Supplement).

Hippocampal Subfields

Patients with schizophrenia had lower mean volume in the left (t = −12.68) and right (t = −13.24) whole hippocampus and in all hippocampal subfields, accompanied with larger right hippocampal fissures. We found higher dispersion in schizophrenia in the left (t = 3.54) and right (t = 2.32) whole hippocampus and in the left molecular layer (t = 3.55), left CA1 (cornu ammonis 1) (t = 2.32), left granule cell layer of the dentate gyrus (t = 3.23), left CA4 (t = 3.10), and left presubiculum (t = 2.52) (Figure 4A). Models without adjustment for eTIV gave the same results for mean volumes with the exception for the left hippocampal fissure, which did not survive correction, and for dispersion with the exception for the left presubiculum (eFigure 9A in the Supplement). When reanalyzing schizophrenia mean and dispersion models with more stringent exclusion criteria, we obtained similar results (eTable 6 in the Supplement). The PRS was associated with smaller left CA2/3 (t = −3.00). None of the subfields showed an association between dispersion and PRS (Figure 4B). Models without adjustment for eTIV revealed smaller left and right CA2/3 (t = −3.70), left granule cell layer of the dentate gyrus (t = −3.20), and left CA4 (t = −2.88) (eFigure 9B in the Supplement) in patients with schizophrenia. Reanalysis with population covariates added to the models, reanalysis with stricter exclusion criteria, and modeling PRS using PRS-PC1 did not alter conclusions (eTable 7 in the Supplement).

Discussion

In the present study we found that schizophrenia is associated with higher brain heterogeneity in cortical thickness and area and cortical, lateral and third ventricle, and hippocampal volumes. The findings, based on harmonized analysis protocols for all included data sets, were robust to strict procedures for removing outliers and quality assessment, and follow-up meta-analysis confirmed that multisite case-control differences cannot be explained by scanning site. These findings are largely in line with those of a recent meta-analysis15 showing higher volumetric heterogeneity in the temporal lobe and lateral and third ventricles. Our findings also extend this meta-analysis by showing higher heterogeneity in cortical thickness and area as well as in specific hippocampal subfields. Further, higher PRS in healthy individuals was associated with thinner frontal and temporal regions and reduced volume of the left CA2/3, but not with thickness dispersion.

We found widespread reductions in cortical thickness in patients with schizophrenia, with the characteristic pattern of stronger frontotemporal effects, as well as global reductions in cortical volume.13 In addition to these mean differences, we found that schizophrenia is also associated with higher thickness heterogeneity compared with healthy participants. No cortical region showed the opposite pattern of higher homogeneity among patients. In line with previous studies, we found mean reductions in several brain volumes, with the most robust effects for cortical volume, cerebellum, and hippocampus, as well as ventricle enlargement. These regions additionally showed higher heterogeneity in patients compared with controls, and again no region showed higher homogeneity, as might result if a particular region was similarly affected by a common pathophysiological mechanism.15 Instead, the results are in line with previous studies suggesting substantial neurobiological heterogeneity in schizophrenia18 and may reflect putative subtypes32 and symptom profiles.44 Results are mostly in line with a recent report15 of higher volumetric heterogeneity in schizophrenia but contrasts with the finding of higher homogeneity for the anterior cingulate cortex in schizophrenia. One possible explanation is differing sample inclusion criteria, because the previous study15 included only patients with first-episode psychosis. An earlier disease stage may offer a more direct window into core aspects of the pathophysiological processes, which later shift toward increased interindividual variability as patients vary across different illness stages and degrees of severity, as well as differences in treatment and medication status.

The PRS reflects cumulative risk across multiple genetic loci, and the PRS for schizophrenia has been associated with several phenotypic traits, including liability for psychiatric disease such as bipolar disorder and schizoaffective disorder, negative symptoms, IQ, working memory performance, and brain activation.25,45,46 The PRS has been associated with cortical gyrification in healthy participants47 and with global cortical thickness.27 Our results show that a higher PRS in healthy controls is associated with a mean decrease in thickness in frontotemporal cortex. These shifts in mean thickness were not associated with brain heterogeneity, as was found for patients, pointing to differential genetic effects on mean thickness and heterogeneity.48 Risk for schizophrenia is also associated with socioeconomic status and family history of psychiatric disorders, with the latter being partially mediated by the PRS.49 This finding underscores the importance of investigating environmental risk factors, as well as gene-environment interplay, and their role in explaining the observed clinical and neurobiological heterogeneity.

With regard to hippocampal volumes, we found that a higher PRS was associated with smaller volumes of the left CA2/3, in the absence of an effect on total hippocampal volume and after correcting for total intracranial volume, suggesting a specific effect of genetic risk on this region. The hippocampus has been hypothesized to play a primary role in the pathophysiological processes of schizophrenia, through progressive changes to its neural circuits as the disease evolves.29 Our results also complement recent studies reporting that polygenic risk for schizophrenia is associated with hippocampal activation during memory encoding28 and of polygenic overlap between schizophrenia and hippocampus volume,50 with possible subfield specificity.51 Also, although the patients showed higher hippocampal heterogeneity, only mean volumes were associated with the PRS, mirroring the findings on cortical thickness. Thus, the CA2/3 emerge as key regions for the manifestation of and the genetic risk for schizophrenia and are potentially informative for the classification of subtypes and degrees of severity.

Despite reliable associations between schizophrenia and brain morphometry,14 the PRS was only weakly associated with subcortical volumes. The lack of associations between PRS and subcortical volumes in the present study is in line with most previous reports of PRS30; however, a recent study50 found polygenic overlap between schizophrenia and hippocampal, putamen, and intracranial volumes.

Limitations

An important source of heterogeneity in the present case-control sample could be related to different scanning sites. However, in addition to residualizing for scanner site in the main analysis, we also performed within-sample analysis and ran a meta-analysis, which ruled out the scanner as a major contributor. A multivariate approach, such as partial least squares,52 might be more sensitive to dispersion, and better capture site-related variance.

Heterogeneity could be associated with differences in medication status and duration of illness. Investigation of such effects requires carefully controlled settings and is therefore difficult to address in large-scale multisite studies. Another possible explanation is that the increased variability is caused by movement artifacts, which are typically greater in clinical populations53; however, running the analysis in a subset with stricter criteria for data set exclusion did not alter the conclusions, and the results were robust to correction for image quality using the Euler number.43

In addition to the concern that clinically defined diagnostic categories do not necessarily comply with biology,18 an important consideration for case-control studies in general is the possibility that healthy controls are higher functioning compared with the general healthy population owing to selection bias and strict exclusion criteria.54 Jointly, these concerns underscore the importance of studying the full range of phenotypic variability in the population.

Further, the validity of choosing a given P value threshold among several possible thresholds when calculating PRS is uncertain. We addressed this by performing a principal components analysis across PRS calculated across a wide range of thresholds, to derive a more general PRS. This approach yielded results converging with the main analysis using a threshold of P < .05. The lack of association between PRS and brain heterogeneity suggests that the current PRS does not strongly reflect variance-controlling variants. However, the PRS might have nonlinear and gene-environment effects that are not detectable in a healthy sample.

As a composite score, PRS likely also hides a substantial genetic heterogeneity. A PRS calculated using a variance-controlling trait loci approach would likely be more sensitive in detecting such effects. Parsing the genetic and clinical contributions to heterogeneity in schizophrenia is an important follow-up, which was not possible herein owing to lack of availability of such data for several of the samples (eTable 2 in the Supplement).

The extent of image smoothing affects sensitivity and anatomical specificity of results,55 and future studies are needed to determine the influence of analysis pipeline on dispersion effects. Last, schizophrenia is increasingly understood as a neurodevelopmental disorder,56 and disentangling the sources of heterogeneity in the adult patient population likely requires investigation of the life-span trajectories and aberrant developmental paths.57,58

Conclusions

Ongoing efforts are attempting to account for neurobiological and brain heterogeneity by means of delineating patient subtypes,8,59 as well as characterizing patients by their differential degree of affectedness along one or multiple clinical domains.11,60 Herein we report that schizophrenia appears to be associated with widespread and increased heterogeneity in cortical thickness and cortical and hippocampal volumes, beyond the known mean differences, compared with controls. The results seem to support the notion that schizophrenia is a highly heterogeneous disorder and suggest that important information may be overlooked when only assessing mean differences between cases and controls.18 In healthy adults, the PRS was associated with mean differences in brain areas implicated in schizophrenia, but not with brain heterogeneity. Together these findings warrant future longitudinal studies that can disentangle the genetic and environmental factors contributing to diverging trajectories and neurobiological heterogeneity.

Back to top
Article Information

Accepted for Publication: January 14, 2019.

Published Online: April 10, 2019. doi:10.1001/jamapsychiatry.2019.0257

Correction: This article was corrected on July 17, 2019, to fix incorrectly filled dots for "CC anterior" and "Fifth ventricle" in Figure 3A.

Corresponding Author: Dag Alnæs, PhD, Norwegian Centre for Mental Disorders Research, KG Jebsen Centre for Psychosis Research, Division of Mental Health and Addiction, Oslo University Hospital and Institute of Clinical Medicine, University of Oslo, PO Box 4956 Nydalen, 0424 Oslo, Norway (dag.alnas@psykologi.uio.no).

Author Contributions: Drs Alnæs and Westlye had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis

Concept and design: Alnæs, Cordova-Palomera, Eisenacher, Andreassen, Westlye.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Alnæs, Kaufmann, Cordova-Palomera, Bertolino, Zink, Westlye.

Critical revision of the manuscript for important intellectual content: Alnæs, Kaufmann, van der Meer, Cordova-Palomera, Rokicki, Moberget, Bettella, Agartz, Barch, Brandt, Cervenka, Djurovic, Doan, Eisenacher, Fatouros-Bergman, Flyckt, Di Giorgio, Haatveit, Jonsson, Kirsch, Lund, Meyer-Lindenberg, Pergola, Schwarz, Smeland, Quarto, Zink, Andreassen, Westlye.

Statistical analysis: Alnæs, van der Meer, Cordova-Palomera, Barch, Bertolino, Schwarz, Quarto, Westlye.

Obtained funding: Agartz, Cervenka, Kirsch, Pergola, Zink, Andreassen, Westlye.

Administrative, technical, or material support: Kaufmann, Rokicki, Bettella, Brandt, Cervenka, Djurovic, Fatouros-Bergman, Haatveit, Jonsson, Lund, Smeland, Andreassen.

Supervision: Bertolino, Cervenka, Flyckt, Meyer-Lindenberg, Pergola, Zink, Andreassen, Westlye.

Group Information: Members of the Karolinska Schizophrenia Project Consortium include the following: Lars Farde, PhD, Lena Flyckt, PhD, Karin Collste, PhD, Pauliina Victorsson, MD, Helena Fatouros-Bergman, PhD, Simon Cervenka, PhD, and Ingrid Agartz, PhD, Centre for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm County Council, Stockholm, Sweden; Göran Engberg, PhD, Sophie Erhardt, PhD, Anna Malmqvist, MD, Mikael Hedberg, MD, Funda Orhan, PhD, and Carl M. Sellgren, PhD, Department of Physiology and Pharmacology, Karolinska Institutet; Lilly Schwieler, PhD, Department of Physiology and Pharmacology, Karolinska Institutet; and Fredrik Piehl, PhD, Neuroimmunology Unit, Department of Clinical Neuroscience, Karolinska Institutet.

Conflict of Interest Disclosures: Dr Bertolino reported being a stockholder of Hoffmann-La Roche, Ltd; receiving consulting fees from Biogen; and receiving lecture fees from Otsuka, Janssen, and Lundbeck. Dr Cervenka reported receiving grant support from AstraZeneca as a coinvestigator and participating in a speaker meeting organized by Otsuka. Dr Zink reported speaker and travel grants from Otsuka, Servier, Lundbeck, Roche, Ferrer, and Trommsdorff. No other disclosures were reported.

Funding/Support: This study was supported by grants 213837, 223273, 226971, 229129, 204966/F20, and 249795 from the Research Council of Norway; grants 2014097, 2015073, 2016083, and 2017112 from the South-Eastern Norway Regional Health Authority; KG Jebsen Stiftelsen; grant 602450 (IMAGEMEND) from the European Commission Seventh Framework Programme; grants 2006-2992, 2006-986, K2007-62X-15077-04-1, 2008-2167, 2008-7573, K2010-62X-15078-07-2, K2012-61X-15078-09-3, 14266-01A,02-03, 2017-949, and 523-2014-3467 from the Swedish Research Council; and grants KI 576/14-2, ZI1253/3-1, and ZI1253/3-2 from the German Research Foundation.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Information: This research has been conducted using the UK Biobank Resource (access code 27412). This study includes data from several sources.

References
1.
World Health Organization. The Global Burden of Disease: 2004 Update. Geneva, Switzerland: WHO Press; 2008.
2.
Lakhan  SE, Vieira  KF.  Schizophrenia pathophysiology: are we any closer to a complete model?  Ann Gen Psychiatry. 2009;8(1):12. doi:10.1186/1744-859X-8-12PubMedGoogle ScholarCrossref
3.
Owen  MJ, Sawa  A, Mortensen  PB.  Schizophrenia.  Lancet. 2016;388(10039):86-97. doi:10.1016/S0140-6736(15)01121-6PubMedGoogle ScholarCrossref
4.
Van Rheenen  TE, Lewandowski  KE, Tan  EJ,  et al.  Characterizing cognitive heterogeneity on the schizophrenia-bipolar disorder spectrum.  Psychol Med. 2017;47(10):1848-1864. doi:10.1017/S0033291717000307PubMedGoogle ScholarCrossref
5.
Malhotra  AK.  Dissecting the heterogeneity of treatment response in first-episode schizophrenia.  Schizophr Bull. 2015;41(6):1224-1226. doi:10.1093/schbul/sbv117PubMedGoogle ScholarCrossref
6.
Huber  G.  The heterogeneous course of schizophrenia.  Schizophr Res. 1997;28(2-3):177-185. doi:10.1016/S0920-9964(97)00113-8PubMedGoogle ScholarCrossref
7.
Weinberg  D, Lenroot  R, Jacomb  I,  et al.  Cognitive subtypes of schizophrenia characterized by differential brain volumetric reductions and cognitive decline.  JAMA Psychiatry. 2016;73(12):1251-1259. doi:10.1001/jamapsychiatry.2016.2925PubMedGoogle ScholarCrossref
8.
Zhang  T, Koutsouleris  N, Meisenzahl  E, Davatzikos  C.  Heterogeneity of structural brain changes in subtypes of schizophrenia revealed using magnetic resonance imaging pattern analysis.  Schizophr Bull. 2015;41(1):74-84. doi:10.1093/schbul/sbu136PubMedGoogle ScholarCrossref
9.
Sugihara  G, Oishi  N, Son  S, Kubota  M, Takahashi  H, Murai  T.  Distinct patterns of cerebral cortical thinning in schizophrenia: a neuroimaging data-driven approach.  Schizophr Bull. 2017;43(4):900-906.PubMedGoogle Scholar
10.
Seaton  BE, Goldstein  G, Allen  DN.  Sources of heterogeneity in schizophrenia: the role of neuropsychological functioning.  Neuropsychol Rev. 2001;11(1):45-67. doi:10.1023/A:1009013718684PubMedGoogle ScholarCrossref
11.
Koutsouleris  N, Gaser  C, Jäger  M,  et al.  Structural correlates of psychopathological symptom dimensions in schizophrenia: a voxel-based morphometric study.  Neuroimage. 2008;39(4):1600-1612. doi:10.1016/j.neuroimage.2007.10.029PubMedGoogle ScholarCrossref
12.
van Erp  TGM, Walton  E, Hibar  DP,  et al.  Cortical brain abnormalities in 4474 individuals with schizophrenia and 5098 controls via the ENIGMA consortium.  Biol Psychiatry. 2018;84(9):644-654. doi:10.1016/j.biopsych.2018.04.023Google Scholar
13.
Moberget  T, Doan  NT, Alnæs  D,  et al.  Cerebellar volume and cerebellocerebral structural covariance in schizophrenia: a multisite mega-analysis of 983 patients and 1349 healthy controls.  Mol Psychiatry. 2018;23(6):1512-1520.PubMedGoogle Scholar
14.
van Erp  TGM, Hibar  DP, Rasmussen  JM,  et al.  Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium [published correction appears in Mol Psychiatry. 2016;21(4):585].  Mol Psychiatry. 2016;21(4):547-553. doi:10.1038/mp.2015.63PubMedGoogle ScholarCrossref
15.
Brugger  SP, Howes  OD.  Heterogeneity and homogeneity of regional brain structure in schizophrenia: a meta-analysis.  JAMA Psychiatry. 2017;74(11):1104-1111. doi:10.1001/jamapsychiatry.2017.2663PubMedGoogle ScholarCrossref
16.
Wolfers  T, Buitelaar  JK, Beckmann  CF, Franke  B, Marquand  AF.  From estimating activation locality to predicting disorder: a review of pattern recognition for neuroimaging-based psychiatric diagnostics.  Neurosci Biobehav Rev. 2015;57:328-349. doi:10.1016/j.neubiorev.2015.08.001PubMedGoogle ScholarCrossref
17.
Doan  NT, Kaufmann  T, Bettella  F,  et al.  Distinct multivariate brain morphological patterns and their added predictive value with cognitive and polygenic risk scores in mental disorders.  Neuroimage Clin. 2017;15:719-731. doi:10.1016/j.nicl.2017.06.014PubMedGoogle ScholarCrossref
18.
Wolfers  T, Doan  NT, Kaufmann  T,  et al.  Mapping the heterogeneous phenotype of schizophrenia and bipolar disorder using normative models.  JAMA Psychiatry. 2018;75(11):1146-1155. doi:10.1001/jamapsychiatry.2018.2467PubMedGoogle ScholarCrossref
19.
Gopal  S, Miller  RL, Michael  A,  et al.  Spatial variance in resting fMRI networks of schizophrenia patients: an independent vector analysis.  Schizophr Bull. 2016;42(1):152-160.PubMedGoogle Scholar
20.
Gopal  S, Miller  RL, Baum  SA, Calhoun  VD.  Approaches to capture variance differences in rest fMRI networks in the spatial geometric features: application to schizophrenia.  Front Neurosci. 2016;10:85. doi:10.3389/fnins.2016.00085PubMedGoogle ScholarCrossref
21.
Sullivan  PF, Kendler  KS, Neale  MC.  Schizophrenia as a complex trait: evidence from a meta-analysis of twin studies.  Arch Gen Psychiatry. 2003;60(12):1187-1192. doi:10.1001/archpsyc.60.12.1187PubMedGoogle ScholarCrossref
22.
Schizophrenia Working Group of the Psychiatric Genomics Consortium.  Biological insights from 108 schizophrenia-associated genetic loci.  Nature. 2014;511(7510):421-427. doi:10.1038/nature13595PubMedGoogle ScholarCrossref
23.
Purcell  SM, Wray  NR, Stone  JL,  et al; International Schizophrenia Consortium.  Common polygenic variation contributes to risk of schizophrenia and bipolar disorder.  Nature. 2009;460(7256):748-752.PubMedGoogle ScholarCrossref
24.
Jones  HJ, Stergiakouli  E, Tansey  KE,  et al.  Phenotypic manifestation of genetic risk for schizophrenia during adolescence in the general population.  JAMA Psychiatry. 2016;73(3):221-228. doi:10.1001/jamapsychiatry.2015.3058PubMedGoogle ScholarCrossref
25.
Kauppi  K, Westlye  LT, Tesli  M,  et al.  Polygenic risk for schizophrenia associated with working memory-related prefrontal brain activation in patients with schizophrenia and healthy controls.  Schizophr Bull. 2015;41(3):736-743. doi:10.1093/schbul/sbu152PubMedGoogle ScholarCrossref
26.
Walton  E, Geisler  D, Lee  PH,  et al.  Prefrontal inefficiency is associated with polygenic risk for schizophrenia.  Schizophr Bull. 2014;40(6):1263-1271. doi:10.1093/schbul/sbt174PubMedGoogle ScholarCrossref
27.
Neilson  E, Bois  C, Gibson  J,  et al.  Effects of environmental risks and polygenic loading for schizophrenia on cortical thickness.  Schizophr Res. 2017;184:128-136. doi:10.1016/j.schres.2016.12.011PubMedGoogle ScholarCrossref
28.
Chen  Q, Ursini  G, Romer  AL,  et al.  Schizophrenia polygenic risk score predicts mnemonic hippocampal activity.  Brain. 2018;141(4):1218-1228. doi:10.1093/brain/awy004PubMedGoogle ScholarCrossref
29.
Lieberman  JA, Girgis  RR, Brucato  G,  et al.  Hippocampal dysfunction in the pathophysiology of schizophrenia: a selective review and hypothesis for early detection and intervention.  Mol Psychiatry. 2018;23(8):1764-1772. doi:10.1038/mp.2017.249PubMedGoogle ScholarCrossref
30.
Reddaway  JT, Doherty  JL, Lancaster  T, Linden  D, Walters  JT, Hall  J. Genomic and imaging biomarkers in schizophrenia. In: Pratt J, Hall J, eds. Biomarkers in Psychiatry. Berlin, Germany: Springer Berlin Heidelberg; 2018:325-352. doi:10.1007/7854_2018_52
31.
Fraser  HB, Schadt  EE.  The quantitative genetics of phenotypic robustness.  PLoS One. 2010;5(1):e8635. doi:10.1371/journal.pone.0008635PubMedGoogle ScholarCrossref
32.
Dwyer  DB, Cabral  C, Kambeitz-Ilankovic  L,  et al.  Brain subtyping enhances the neuroanatomical discrimination of schizophrenia.  Schizophr Bull. 2018;44(5):1060-1069. doi:10.1093/schbul/sby008PubMedGoogle ScholarCrossref
33.
Conley  D, Johnson  R, Domingue  B, Dawes  C, Boardman  J, Siegal  ML.  A sibling method for identifying vQTLs.  PLoS One. 2018;13(4):e0194541. doi:10.1371/journal.pone.0194541PubMedGoogle ScholarCrossref
34.
Fischl  B.  FreeSurfer.  Neuroimage. 2012;62(2):774-781. doi:10.1016/j.neuroimage.2012.01.021PubMedGoogle ScholarCrossref
35.
Fischl  B, Salat  DH, Busa  E,  et al.  Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain.  Neuron. 2002;33(3):341-355. doi:10.1016/S0896-6273(02)00569-XPubMedGoogle ScholarCrossref
36.
Dale  AM, Fischl  B, Sereno  MI.  Cortical surface-based analysis, I: segmentation and surface reconstruction.  Neuroimage. 1999;9(2):179-194. doi:10.1006/nimg.1998.0395PubMedGoogle ScholarCrossref
37.
Fischl  B, Sereno  MI, Dale  AM.  Cortical surface-based analysis, II: inflation, flattening, and a surface-based coordinate system.  Neuroimage. 1999;9(2):195-207. doi:10.1006/nimg.1998.0396PubMedGoogle ScholarCrossref
38.
Iglesias  JE, Augustinack  JC, Nguyen  K,  et al; Alzheimer’s Disease Neuroimaging Initiative.  A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI.  Neuroimage. 2015;115:117-137. doi:10.1016/j.neuroimage.2015.04.042PubMedGoogle ScholarCrossref
39.
Euesden  J, Lewis  CM, O’Reilly  PF.  PRSice: polygenic risk score software.  Bioinformatics. 2015;31(9):1466-1468. doi:10.1093/bioinformatics/btu848PubMedGoogle ScholarCrossref
40.
Winkler  AM, Ridgway  GR, Webster  MA, Smith  SM, Nichols  TE.  Permutation inference for the general linear model.  Neuroimage. 2014;92:381-397. doi:10.1016/j.neuroimage.2014.01.060PubMedGoogle ScholarCrossref
41.
Winkler  AM, Ridgway  GR, Douaud  G, Nichols  TE, Smith  SM.  Faster permutation inference in brain imaging.  Neuroimage. 2016;141:502-516. doi:10.1016/j.neuroimage.2016.05.068PubMedGoogle ScholarCrossref
42.
Nichols  TE, Holmes  AP.  Nonparametric permutation tests for functional neuroimaging: a primer with examples.  Hum Brain Mapp. 2002;15(1):1-25. doi:10.1002/hbm.1058PubMedGoogle ScholarCrossref
43.
Rosen  AFG, Roalf  DR, Ruparel  K,  et al.  Quantitative assessment of structural image quality.  Neuroimage. 2018;169:407-418. doi:10.1016/j.neuroimage.2017.12.059PubMedGoogle ScholarCrossref
44.
Clementz  BA, Sweeney  JA, Hamm  JP,  et al.  Identification of distinct psychosis biotypes using brain-based biomarkers.  Am J Psychiatry. 2016;173(4):373-384. doi:10.1176/appi.ajp.2015.14091200PubMedGoogle ScholarCrossref
45.
Mistry  S, Harrison  JR, Smith  DJ, Escott-Price  V, Zammit  S.  The use of polygenic risk scores to identify phenotypes associated with genetic risk of schizophrenia: systematic review.  Schizophr Res. 2017;197:2-8. doi:10.1016/j.schres.2017.10.037PubMedGoogle ScholarCrossref
46.
Tesli  M, Espeseth  T, Bettella  F,  et al.  Polygenic risk score and the psychosis continuum model.  Acta Psychiatr Scand. 2014;130(4):311-317. doi:10.1111/acps.12307PubMedGoogle ScholarCrossref
47.
Liu  B, Zhang  X, Cui  Y,  et al.  Polygenic risk for schizophrenia influences cortical gyrification in 2 independent general populations.  Schizophr Bull. 2017;43(3):673-680.PubMedGoogle Scholar
48.
Cordova-Palomera  A, Meer D, Kaufmann T, et al.  Genetic control of variability in subcortical and intracranial volumes.  bioRxiv. October 15, 2018. https://www.biorxiv.org/content/10.1101/443549v1. Google Scholar
49.
Agerbo  E, Sullivan  PF, Vilhjálmsson  BJ,  et al.  Polygenic risk score, parental socioeconomic status, family history of psychiatric disorders, and the risk for schizophrenia: a Danish population-based study and meta-analysis.  JAMA Psychiatry. 2015;72(7):635-641. doi:10.1001/jamapsychiatry.2015.0346PubMedGoogle ScholarCrossref
50.
Smeland  OB, Wang  Y, Frei  O,  et al.  Genetic overlap between schizophrenia and volumes of hippocampus, putamen, and intracranial volume indicates shared molecular genetic mechanisms.  Schizophr Bull. 2018;44(4):854-864. doi:10.1093/schbul/sbx148PubMedGoogle ScholarCrossref
51.
van der Meer  D, Rokicki  J, Kaufmann  T,  et al; Alzheimer’s Disease Neuroimaging Initiative; Pediatric Imaging, Neurocognition and Genetics Study.  Brain scans from 21,297 individuals reveal the genetic architecture of hippocampal subfield volumes [printed online October 2, 2018].  Mol Psychiatry. PubMedGoogle Scholar
52.
McIntosh  AR, Lobaugh  NJ.  Partial least squares analysis of neuroimaging data: applications and advances.  Neuroimage. 2004;23(suppl 1):S250-S263. doi:10.1016/j.neuroimage.2004.07.020PubMedGoogle ScholarCrossref
53.
Reuter  M, Tisdall  MD, Qureshi  A, Buckner  RL, van der Kouwe  AJW, Fischl  B.  Head motion during MRI acquisition reduces gray matter volume and thickness estimates.  Neuroimage. 2015;107:107-115. doi:10.1016/j.neuroimage.2014.12.006PubMedGoogle ScholarCrossref
54.
Schwartz  S, Susser  E.  The use of well controls: an unhealthy practice in psychiatric research.  Psychol Med. 2011;41(6):1127-1131. doi:10.1017/S0033291710001595PubMedGoogle ScholarCrossref
55.
Bernal-Rusiel  JL, Atienza  M, Cantero  JL.  Determining the optimal level of smoothing in cortical thickness analysis: a hierarchical approach based on sequential statistical thresholding.  Neuroimage. 2010;52(1):158-171. doi:10.1016/j.neuroimage.2010.03.074PubMedGoogle ScholarCrossref
56.
Insel  TR.  Rethinking schizophrenia.  Nature. 2010;468(7321):187-193. doi:10.1038/nature09552PubMedGoogle ScholarCrossref
57.
Alnæs  D, Kaufmann  T, Doan  NT,  et al.  Association of heritable cognitive ability and psychopathology with white matter properties in children and adolescents.  JAMA Psychiatry. 2018;75(3):287-295. doi:10.1001/jamapsychiatry.2017.4277PubMedGoogle ScholarCrossref
58.
Kaufmann  T, Alnæs  D, Doan  NT, Brandt  CL, Andreassen  OA, Westlye  LT.  Delayed stabilization and individualization in connectome development are related to psychiatric disorders.  Nat Neurosci. 2017;20(4):513-515. doi:10.1038/nn.4511PubMedGoogle ScholarCrossref
59.
Honnorat  N, Dong  A, Meisenzahl-Lechner  E, Koutsouleris  N, Davatzikos  C.  Neuroanatomical heterogeneity of schizophrenia revealed by semi-supervised machine learning methods [published online December 20, 2017].  Schizophr Res.PubMedGoogle Scholar
60.
Viher  PV, Stegmayer  K, Giezendanner  S,  et al.  Cerebral white matter structure is associated with DSM-5 schizophrenia symptom dimensions.  Neuroimage Clin. 2016;12:93-99. doi:10.1016/j.nicl.2016.06.013PubMedGoogle ScholarCrossref
×