Article Text


Structural MRI of the brain in presumed carriers of genes for schizophrenia, their affected and unaffected siblings
  1. R M Steel,
  2. H C Whalley,
  3. P Miller,
  4. J J K Best,
  5. E C Johnstone,
  6. S M Lawrie
  1. University of Edinburgh, Department of Psychiatry. Kennedy Tower, Royal Edinburgh Hospital, Morningside Park, Edinburgh EH10 5HF, UK
  1. Correspondence to: 
 Dr R M Steel, University of Edinburgh, Department of Psychiatry. Kennedy Tower, Royal Edinburgh Hospital, Morningside Park, Edinburgh EH10 5HF, UK; 


Background: Schizophrenia is a highly heritable disorder associated with structural brain abnormalities. The aim of this study was to establish if the gene(s) for schizophrenia are associated with specific abnormalities of brain structure.

Subjects: Six sibships from multiple affected families were recruited. Each sibship consisted of one patient with schizophrenia, one “obligate carrier” without the disorder but with an affected child, and one “non-affected non-carrier”. Such sibships are very rare, but present a powerful opportunity to separate the associations of genotype and phenotype. Obligates presumably have the gene(s) but not the disorder, affected siblings have both, whereas non-affected non-carrier siblings have neither.

Method: Brain MRI was conducted with a semiautomated region of interest analysis. The risk of false positive findings was reduced by collapsing brain regions and sides into five regions and comparing groups by repeated measures analysis of variance.

Results: In terms of whole brain volumes and volumes of cortical structures, obligates resembled their non-affected non-carrier siblings, both groups having significantly greater volumes than their schizophrenic siblings (p=0.01 and p=0.04). Obligates also had significantly smaller ventricles than their schizophrenic siblings (p=0.03). However, with respect to the amygdalohippocampal complex, the obligates' brains resembled those of their schizophrenic siblings, both groups showing a significant reduction in volume when compared with their non-affected non-carrier siblings (p=0.001).

Conclusions: In the families studied, reductions in volumes of cortical structures and reductions in whole brain volume seem to be associated with the phenotype of schizophrenia. By contrast, reduced volume of the amygdalohippocampal complex seems to be associated with genetic risk for the disorder even in the absence of disease.

  • schizophrenia
  • magnetic resonance imaging
  • amygdalohippocampal complex
  • PSE, present state examination;
  • SADS-L, schedule for affective disorders and schizophrenialifetime version;
  • NART, national adult reading test

Statistics from

Data from postmortem and in vivo brain imaging studies provide powerful evidence that schizophrenia is associated with abnormalities of brain structure.1–3 The most consistently identified abnormalities include reduced whole brain volume, increased lateral ventricle volume, and reduced volumes of the temporal lobes and medial temporal lobe structures. The origins and significance of these structural abnormalities are, however, not well understood. The pre-eminent contemporary theory proposes that they arise as a result of abnormal brain development.4

Many cases of schizophrenia are familial. The epidemiological evidence from twin and adoption studies strongly suggests that this familiality has a genetic origin.5 However, to date, attempts to identify specific genes that might be associated with or even responsible for the disorder have proved inconclusive.6,7 It is generally accepted that, even in highly familial cases, the development of schizophrenia is probably influenced by both genetic and environmental factors. The relative importance of genes versus environment in the development of the structural brain abnormalities associated with the illness remains unclear.

The aim of this study was to establish if the gene(s) for schizophrenia are associated with abnormalities of brain structure. We specifically hypothesised that amygdalohippocampal volumes are genetically mediated and tested this by taking a rare opportunity to compare specific sibships in multiply affected families.



The study was designed to separate out the genetic risk for schizophrenia from the illness itself (table 1). This was achieved by recruiting persons who had apparently transmitted schizophrenia from one affected parent to one or more affected children while remaining well themselves (so-called “obligate carriers”). Data from these obligates were compared with data from their schizophrenic siblings and their non-affected, presumed non-carrier siblings (as they had adult children without the disorder). An anonymised example of an ideal obligate family tree is shown in figure 1.

Table 1

The study design can separate the effects of gene(s) from the effects of illness

Figure 1

An ideal obligate family tree.

Exclusion criteria reflecting known risk factors for altered brain volume were age>65 years; alcohol dependence (past or present); dementia, or other “organic” brain disease; and any space occupying intracranial lesion.

Families showing the ideal pattern of inheritance described in figure 1 are obviously rare. Family trees relating to 255 multiply affected Scottish families (originally identified for the Edinburgh high risk study8) were examined. Sixty obligates were identified, 14 of whom had both an affected sibling and a non-affected, non-carrier sibling. All 14 families were contacted. Five families were unsuitable as one or more of the siblings met the exclusion criteria listed above (most exclusions were due to age). Of the remaining nine families, three did not wish or were unable to participate. Data were therefore obtained from six sibling “triples”. Five of the sibships were of the same sex (three female, two male); the remaining family included a female patient and obligate and a male non-carrier (who was a twin with the patient).

Personal data

All subjects were interviewed (by RMS) with the present state examination (PSE)9 and the schedule for affective disorders and schizophrenia—lifetime version (SADS-L).10 Premorbid IQ was tested using the national adult reading test (NART).11 Descriptive information was gathered from case notes and interviews (table 2 for details).

Table 2

Demographic and clinical details of subjects by group

Imaging data

The MRI examinations included a dual spin echo sequence to exclude any significant brain lesions and a rapid volume acquisition sequence for volumetric data analysis. Any coil inhomogeneities were corrected for by scanning a flood phantom immediately after image acquisition and normalising to this before analysis. Unfortunately, it was not possible to examine all the subjects on the same machine. The first three sibling triplets were scanned at the MRI Unit, City Hospital, Edinburgh before the unit closed, on a 1 tesla Magnetom scanner (Siemens, Erlanger, Germany) using a MPRAGE sequence, FOV 250 mm, flip angle 12°, TR=10 ms, TE=4 ms, TI=200 ms, and relaxation delay 500 ms, giving 128 partitions 1.88 mm thick. The remaining three triplets were scanned at the Royal Infirmary, Edinburgh, with a 1 tesla Signa scanner (General Electric Company, Milwaukee, USA) using a SPGR sequence, FOV 250 mm, flip angle 30°, TR=16.4 ms, TE=3.3 ms, RBW 8.93 KHz, giving 124 partitions 1.5 mm thick. Regions of interest were traced on a slice by slice basis according to well established criteria using the semiautomated computer programme “ANALYZE” (Mayo Foundation, Rochester, MN, USA). All data were analysed by the same experienced rater (HCW) who was blind to group and has previously demonstrated high interrater and intrarater reliability.8 Boundaries between brain regions were determined in accordance with criteria described in previously published studies from this department.8 Volumetric data for the following brain regions were collected: whole brain, third ventricle, fourth ventricle, right and left lateral ventricles, prefrontal lobes, temporal lobes, caudate nuclei, lentiform nuclei, thalamic nuclei, and amygdalohippocampal complexes.

Statistical analysis

To reduce the chance of false positive findings, brain regions and sides were collapsed into the following regions: whole brain; cortical structures (prefrontal and temporal lobes); subcortical structures (caudate, lentiform, and thalamus nuclei); amygdalohippocampal complexes; and intracerebral ventricular system (lateral, third, and fourth ventricles). The null hypothesis was that there would be no difference between the three groups in terms of regional brain volumes. This was tested by repeated measures analysis of variance (ANOVA). The volumes of each collapsed region in each of the three subject groups was taken as a within subjects factor, with the scanner entered as a between subjects factor.


Results from volumetric analysis of the MRI are shown in table 3. There were no statistically significant group by scanner interactions.

Table 3

Volumes of brain regions by group

Differences between schizophrenic subjects and their non-affected, non-carrier siblings were largely consistent with the existing literature. Whole brain volume was significantly smaller in the schizophrenic subjects (by about 5%), with a disproportionate reduction in the volume of the amygdalohippocampal complexes (12%). The expected increase in ventricular volume was, however, not found.

In terms of whole brain volume and volumes of cortical structures, the obligates were indistinguishable from their non-carrier siblings (and significantly larger than their schizophrenic siblings). They also had significantly smaller ventricles than their schizophrenic siblings. However, with respect to the amygdalohippocampal complex, the obligates' brains did not differ significantly from their schizophrenic siblings' brains (both groups showing a significant reduction in volume compared with the non-affected, non-carriers).


In this study, presumed carriers of the gene(s) for schizophrenia (“obligates”) were found to share a significant reduction in amygdalohippocampal volume with their schizophrenic siblings. However, by contrast with their affected siblings, they were found not to have reduced frontal or temporal lobe or whole brain volumes; nor did they have large ventricles. This suggests genetic and phenotypic effects respectively.

The use of two separate MRI scanners in this study is an obvious and potentially serious limitation. However, the risk of bias was minimised by scanning all members of any given family on the same machine and by entering the scanner as a between subjects factor in the statistical analysis. Indeed, as no significant scanner interactions were found, our study provides some evidence that regional volumetric analyses are reliable across scanners (although obviously this comparison has low power).

One interesting and unexpected finding of this study was the relatively high IQ of the obligates (as shown in table 2 their average NART score of 107 was 10 points higher than either their schizophrenic or non-affected siblings). This raises the possibility that, amongst people at high genetic risk of schizophrenia, intelligence may be, or reflect, a protective factor, although various interpretations of this finding are clearly possible.

The design adopted for this study allows effects attributable to genetic risk to be separated from effects of disease without the need to identify the gene or genes responsible. The use of siblings as controls also carries the advantages of eliminating a number of potential sources of bias (reducing in particular non-specific variance in regional brain volumes), thereby increasing the power of the study. Sibling “triples” as described above are very rare and any study of this design will inevitably be small. Previous studies reporting volumetric data from brain imaging in relatives of schizophrenic patients suggest a range of brain abnormalities with relatives falling midway between schizophrenic patients and normal controls.12 Most of these studies adopt a straightforward “schizophrenic patients versus relatives versus controls” design which does not allow persons with the gene(s) to be separated from those without. The only previously published study of obligates13 did not report amygdalohippocampal volumes but found increased left lateral ventricle volume (which we did not find). That study adopted a slightly different design comparing normal unrelated controls with obligates, schizophrenic patients, and non-obligate relatives recruited from 16 multiply affected families. This design did not allow such a clear separation of genetic and environmental factors as the “sibling triples” design described above.

In any condition with apparently familial and non-familial cases, it can be argued that highly familial cases are unrepresentative of the condition as a whole. This study involved subjects from families in which schizophrenia was not only unusually prevalent but also inherited in an apparently dominant pattern. The genetic mechanisms involved in these families may therefore be different from those acting in families with a non-dominant pattern of inheritance and may be of little relevance to non-familial cases of schizophrenia. The data from this study suggest that reduced volume of the amygdalohippocampal complex in schizophrenia is genetically determined whereas cortical changes and increased ventricular volume are not (and are presumably associated with the phenotype). However, data from the Edinburgh high risk study8 does not support such a straightforward explanation. In that study genetic liability to schizophrenia in those from multiply affected families (with a range of patterns of inheritance) was not linearly associated with amygdalohippocampal volumes.12 The genetic mechanisms involved in the development of schizophrenia are, however, not well understood and many different genes may be involved. We interpret our results as suggesting that obligates in this study have inherited some of the responsible genes (those causing amygdalohippocampal change) but not others (those responsible for cortical changes and symptoms). It seems likely that, even in families with an apparently dominant pattern of inheritance, highly complex gene-environment interactions are involved not only in the development of symptoms, but also in the genesis of structural brain abnormalities such as reduced brain volume and enlarged ventricles.

Genetic research into schizophrenia is currently a highly active area with many linkage and association studies in progress. The data from this study suggest that genes may play a part in the development of some, but not all, of the structural brain changes associated with the condition.


This study was funded by the Theodore and Vada Stanley Foundation. We thank radiographers and radiologists at the two hospitals in Edinburgh for conducting the scans and thank the subjects and their extended families for their participation.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.