A total of 62 schizophrenia patients, 67 siblings of schizophrenia patients, and 41 healthy comparison subjects were included in this study. Study participants were recruited at the University Medical Center Utrecht, during a large ongoing cohort in the Netherlands (Genetic Risk and Outcome of Psychosis; GROUP). The study was approved by the affiliated ethical committee. Study participants were between 18 and 60 years of age. All subjects provided written informed consent prior to participation. Subjects with a history of head trauma or major medical or neurological illness were excluded.

Presence or absence of current and lifetime psychopathology was established for all participants, using the Comprehensive Assessment of Symptoms and History (CASH) interview (Andreasen et al., 1992). Schizophrenia patients were eligible for the present study if they met Diagnostic and Statistical Manual of Mental Disorders fourth edition (DSM IV; American Psychiatric Association, 1994) criteria of schizophrenia or related spectrum disorders. Both siblings of schizophrenia patients and healthy comparison subjects could have no history of any psychiatric illness, including substance dependence of abuse; and healthy comparison subjects had no first- or second-degree family members with a lifetime psychotic disorder.

Study participants originated from a total number of 132 families. Within the healthy comparison group, there were two sibling pairs (both healthy control subjects, i.e., without a first- or second-degree relative with a lifetime psychotic disorder), and within the 67 siblings of patients, there were seven pairs of siblings of a schizophrenia patient and one set of three siblings of a schizophrenia patient. There were no family relationships within the patient group. Between the sibling and patient groups, there were a total number of 27 family-ties [i.e., a schizophrenia patient and (one of) their sibling(s) participated in the study].

For schizophrenia patients, symptom severity was assessed using the positive and negative syndrome scale (PANSS; Kay et al., 1987); Furthermore, the type and daily dose of antipsychotic medication at the time of scanning was recorded, and a haloperidol equivalent dose was calculated using conversion rates (risperidone 0.5:1; olanzapine 1.66:1; quetiapine 25:1; clozapine 33.33:1; aripiprazole 2.5:1; flupenthixol 0.66:1; perphenazine 2.66:1; Kroken et al., 2009). In healthy comparison subjects and siblings of patients, a shortened version of the structured interview of schizotypy-revised (SIS-R; Kendler et al., 1989; Vollema and Ormel, 2000) was employed [excluding four items: social isolation (last 3 years); antisocial behavior; dysfunction (obligatory activities); and dysfunction (personal caretaking)], to measure the overall severity of schizotypal signs and symptoms on a four-point scale (0 = absent, 1 = mild, 2 = moderate, 3 = severe). For all participants, global cognitive functioning, as measured by total IQ, was estimated using four subtests (i.e., Information, Arithmetic, Block design, and Digit symbol coding) of the Dutch version of the Wechsler adult intelligence scale (WAIS; Stinissen et al., 1970).

Finally, statistical testing of group-differences in demographic characteristics was performed using analysis of variance (ANOVA) for continuous and Chi-Square tests for categorical variables. All demographic and clinical characteristics are provided in Table 1.

Resting-state fMRI data were acquired on two 1.5 T Magnetic Resonance Imaging scanners (Philips Medical Systems, Best, The Netherlands) at the University Medical Center Utrecht, The Netherlands. BOLD time-series were recorded during 9 minutes using a 3D-PRESTO sequence (Van Gelderen et al., 1995; Ramsey et al., 1996; acquisition parameters: TR/TE 21.1/31.1 ms; voxel size 4 mm × 4 mm × 4 mm). Subsequently, a T1 weighted image was acquired for anatomical reference (3D FFE pulse sequence, TR/TE = 30/4.6 ms, flip-angle 30°, FOV 256 mm × 256 mm, voxel size 1 mm × 1 mm × 1.2 mm, 160–180 contiguous slices; Ramsey et al., 2006). Preprocessing was performed using SPM5 software. In short, all resting-state functional images were registered to the last functional scan to correct for head movements and co-registered with the T1 scan to ensure overlap between the anatomical reference scan and resting-state time-series. The registered functional images were spatially smoothed, using an 8-mm full width half–max smoothing kernel. Next, the T1 scan and resting-state time-series were normalized to standard space, matching the MNI-152 template. To create individual anatomical label maps, the spatially normalized T1 image was overlaid with the automated anatomical labeling (AAL) template (Tzourio-Mazoyer et al., 2002), distinguishing 116 anatomical brain regions. Time-series were corrected for white matter, ventricle, and motion parameters (using regression) and band pass filtered (bandwidth 0.01–0.1 Hz) to eliminate low frequency noise and influences of frequencies reflecting possible cardiac or respiratory oscillations.

The cerebellum has three anterior–posterior divisions (Kandel et al., 2000). The primary fissure separates the anterior lobe from the posterior lobe and the posterolateral fissure separates the posterior and flocculo-nodular lobe. Three mediolateral regions that are important functionally are distinguished by two longitudinal grooves that define an elevated ridge in the midline known as the vermis and the cerebellar hemispheres on either side of the vermis. Each of the 26 cerebellar regions distinguished by the AAL template was assigned to one of seven cerebellar anatomical regions of interest (ROI; Table 2), including the anterior and posterior vermis, the bilateral anterior and posterior cerebellar hemispheres and the flocculo-nodular lobe (see Figure 1, no. 3; 6; 1 and 2; 4 and 5; 7 respectively).

Figure 2 illustrates the consecutive steps of the performed functional connectivity analysis. For each subject, 116 regional mean time-series were computed by averaging the voxel-based time-series within each of the anatomically defined regions. Next, interregional correlation in resting-state time-series between each possible pair of the 26 cerebellar regions and the 90 cerebral regions in the AAL template was computed (Figure 2A). For each subject, functional connectivity per cerebellar ROI was then calculated by averaging the correlation coefficients of the AAL regions within each of the seven cerebellar regions (Figure 2B), rendering weighted correlation coefficients r_i,j for the connections {i, j} of each cerebellar ROI i (N = 7) with each cerebral region j (N = 90). Interregional correlations were first computed and subsequently averaged for each ROI (rather than averaging at the level of regional time-series) as the ROIs may comprise more than one functional region and averaging distinct regional time-series may lead signals to phase out when averaged. Fisher r-Z transforms were not applied as the correlations were normally distributed. Next, weighted correlation maps per subject group were computed for each cerebellar ROI (Figure 2C), rendering three matrices (for control, sibling, and patient groups) per ROI, with all 90 connections {i, j} of the particular ROI i on the x-axis and the number of subjects in the group (N = 41; 67; 62 respectively) on the y-axis.

In order to interpret possible differences in cerebellar functional connectivity between subject groups, the general pattern of functional connections per cerebellar ROI was investigated in the healthy control subjects. As functionally connected regions have been shown to exhibit a high degree of temporal coherence (Biswal et al., 1995; Van den Heuvel et al., 2008), higher interregional correlation in time-series was interpreted as a higher degree of functional connectivity between two regions.

Differences between subject groups in cerebellar functional connectivity patterns were examined using two functional connectivity measures. First, overall connectivity strength per cerebellar ROI was computed, providing information on the global level of communication between each cerebellar ROI and the rest of the brain.

Using linear regression analyses, the association between any possible changes in cerebellar connectivity measures and clinical variables (i.e., duration of illness; severity of PANSS symptoms and IQ) was investigated, and the dose of antipsychotic medication (in haloperidol equivalent) at the time of scanning was examined as a potential confounder.

Connectivity strength S of each cerebellar ROI and functional connectivity Fc of each individual connection of schizophrenia patients and siblings of schizophrenia patients were compared to those of the healthy comparison subjects. To examine the statistical significance of group-differences, permutation testing was used (5000 permutations; Bassett et al., 2008; Lynall et al., 2010; Van den Heuvel et al., 2010). To this end, random permutation of group assignment was performed, maintaining the original number of subjects per group, rendering three groups of 41, 62, and 67 randomly assigned subjects. Consequently, weighted correlation maps per randomly assigned subject group were computed for each cerebellar ROI. Using the resulting correlation maps, the connectivity measures were recalculated. This process was repeated 5000 times, resulting in a between-group difference null distribution of the connectivity measures. Finally, the observed differences between the original subject groups were compared to the normal distribution of differences after random permutation, to explore the null hypothesis that the observed differences were not determined by subject group membership. Finally, p-values were assigned to the group effects by computing (after random permutation) the percentage of findings that was more extreme than the observed difference between the original subject groups. In view of the number of tests performed in the connectivity strength analysis (Analysis 1, Figure 2D), a p-value of <0.01 was considered to indicate statistical significance, and a p-value of >0.01 and <0.05 to indicate a statistical trend. To account for the multiple comparisons of the second analysis (Analysis 2, Figure 2E), the effects on functional connectivity of discrete connections were subjected to a false discovery rate (FDR) threshold of q = 0.05 (Benjami and Hochberg, 1995; Storey, 2002) over all functional connections (N = 90) of each cerebellar ROI, providing control over false positive findings.

In the healthy comparison subjects, mean functional connectivity of the posterior hemispheres [mean Fc (SD) = 0.27 (0.07)] > anterior hemispheres [mean Fc (SD) = 0.21 (0.06)] and the right cerebellum [mean Fc (SD) = 0.27 (0.08)] > left cerebellum [mean Fc (SD) = 0.22 (0.06)] (both p < 0.0001). Furthermore, mean functional connectivity of the lateral cerebellum (i.e., hemispheres) [mean Fc (SD) = 0.24 (0.06)] was significantly higher than that of midline structures (i.e., vermis and flocculo-nodular lobe) [mean Fc (SD) = 0.14 (0.05)] (p < 0.0001).

The highest levels of functional connectivity were found between the right anterior cerebellar hemisphere and right lingual gyrus (mean Fc = 0.54, SD 0.14), hippocampus [mean Fc (SD) = 0.50 (0.18)] and parahippocampal gyrus [mean Fc (SD) = 0.50 (0.15)]. Overall, cerebellar connections to posterior medial regions of the cerebral cortex [i.e., (pre)cuneus, calcarine sulcus, lingual gyrus] showed high levels of functional connectivity, irrespective of cerebellar ROI [mean Fc (SD) of all ROIs 0.26 (0.08), 0.26 (0.10); 0.34 (0.08), 0.32 (0.08); 0.32 (0.09), 0.29 (0.10); 0.29 (0.09), 0.33 (0.12); for the left and right cuneus; precuneus; calcarine sulcus; and lingual gyrus respectively]. The lowest levels of functional connectivity were observed for connections with the bilateral lentiform nucleus [mean Fc (SD) = 0.11 (0.15)], in particular the globus pallidus [left; right mean Fc (SD) = 0.05 (0.15); 0.09 (0.15)], and with the gyrus rectus [left; right mean FC (SD) = 0.11 (0.17); 0.09 (0.17)] (Figure 3).

Overall connectivity strength of the posterior and flocculo-nodular lobes of the cerebellum with the rest of the brain did not differ between groups, suggesting intact functional connectivity of posterior and flocculo-nodular cerebellar regions in schizophrenia. Connectivity strength of the right anterior hemisphere of the cerebellum however, was found to be 23% less (p = 0.01) in schizophrenia patients compared to healthy comparison subjects. Furthermore, on trend level, connectivity strength of the anterior vermis was less in both schizophrenia patients (−27%, p = 0.04) and siblings (−22%, p = 0.05), compared to healthy comparison subjects (Figure 4).

Impaired functional connectivity was most pronounced in the connections between the right anterior cerebellar hemisphere and vermis, and left-cerebral regions. Particularly between the right anterior vermis and the left-sided triangular part of the inferior frontal gyrus (Fc = −0.13, p = 0.0024), supplementary motor area (Fc = −0.12, p = 0.0014), middle cingulate gyrus (Fc = −12, p = 0.0004), hippocampus (Fc = −0.17, p = 0.0004), and thalamus (Fc = −0.16, p < 0.0001). Functional connectivity between the latter two regions (i.e., left hippocampus and thalamus) and the anterior cerebellar vermis was also decreased in patients (Fc = −0.14, p = < 0.0001; Fc = −0.13, p = 0.0008 respectively). The only significant increase in functional connectivity in schizophrenia patients relative to controls was found between the posterior vermis and left lingual gyrus (Fc = 0.15, p < 0.0001). All these findings survived FDR-correction (Figure 5).

Compared to schizophrenia patients, the siblings demonstrated both similarities and disparities in functional connectivity of the cerebellum. Siblings showed decreased functional connectivity (compared to control subjects) between the anterior cerebellar vermis and the triangular part of the left inferior frontal gyrus (Fc = −0.08, p = 0.009), left insula (Fc = −0.08, p = 0.007), and left hippocampus (Fc = −0.11, p = 0.002), as well as between the right anterior cerebellar hemisphere and left hippocampus (Fc = −0.12, p = 0.003), resembling the findings in patients. However, these findings were not as strong as in patients, as they failed to reach whole-brain significance (note that impaired functional connectivity between the anterior cerebellar vermis and left insula did not reach whole-brain significance in both siblings and patients), and should be considered exploratory. Conversely, functional connectivity to other brain regions that were found to be dysconnected from the cerebellum in schizophrenia patients (e.g., left thalamus, middle cingulate gyrus, and supplementary motor area) was not decreased in siblings of schizophrenia patients, compared to the healthy control subjects. Table 3 depicts all findings at p < 0.01 in schizophrenia patients and siblings of patients.

Importantly, as siblings were psychiatrically healthy (see inclusion criteria and SIS-R scores, Table 1) and, on average, around 30 years of age, it is unlikely that any would still convert to illness. Lastly, repeating the functional connectivity analyses in a reduced number of study participants that were completely unrelated did not alter the findings.

Cerebellar functional connectivity measures that were significantly different in schizophrenia patients, compared to healthy comparison subjects, were investigated for associations with medication dose, duration of illness, symptom severity, and global cognitive performance, using linear regression analysis. No association between the dose of antipsychotic medication at the time of scanning and cerebellar functional connectivity was found (all p > 0.420). After multiple comparison correction, there were no significant associations between aberrant cerebellar functional connectivity measures and clinical variables. On trend level, a counter-intuitive association between more negative symptoms and increased Fc between the cerebellar right anterior hemisphere and left hippocampus was observed (Table 4).