Patients with first-episode, drug-naive schizophrenia and subjects at ultra-high risk of psychosis shared increased cerebellar-default mode network connectivity at rest

Increased cerebellar-default mode network (DMN) connectivity has been observed in first-episode, drug-naive patients with schizophrenia. However, it remains unclear whether increased cerebellar-DMN connectivity starts earlier than disease onset. Thirty-four ultra-high risk (UHR) subjects, 31 first-episode, drug-naive patients with schizophrenia and 37 healthy controls were enrolled for a resting-state scan. The imaging data were analyzed using the seed-based functional connectivity (FC) method. Compared with the controls, UHR subjects and patients with schizophrenia shared increased connectivity between the right Crus I and bilateral posterior cingulate cortex/precuneus and between Lobule IX and the left superior medial prefrontal cortex. There are positive correlations between the right Crus I-bilateral precuneus connectivity and clinical variables (Structured Interview for Prodromal Syndromes/Positive and Negative Symptom Scale negative symptoms/total scores) in the UHR subjects. Increased cerebellar-DMN connectivity shared by the UHR subjects and the patients not only highlights the importance of the DMN in the pathophysiology of psychosis but also may be a trait alteration for psychosis.

. These seeds showed intrinsic connectivity with the DMN in healthy subjects 23,24 and patients with schizophrenia 34,35 . The ROIs were defined as 6mm radius spheres with the software REST 41 for FC analysis. Pearson correlation coefficients between each seed and other voxels of the entire brain were computed to create correlation maps that were z-transformed with Fisher's r-to-z transformation. Details of FC analysis with the software REST can be found in the supplementary files.
For each seed and each group, one-sample t-tests were used to detect voxels that showed significant correlations with the seeds. The significance level was set at p < 0.005 corrected for multiple comparisons using the Gaussian Random Field (GRF) theory (min z > 2.807, cluster significance: p < 0.005). GRF correction is a cluster-extent based thresholding, which identifies statistically significant clusters according to the number of contiguous voxels, and controls the estimated false positive probability of the brain region as a whole 42,43 . Analyses of covariance (ANCOVA), followed by post hoc t-tests, were conducted to compare group differences within the union mask of one-sample t-test results of three groups. Age was used as a covariate to minimize the possible effect of this variable. Framewise displacement (FD) for each participant was calculated 44 and also used as a covariate in the group comparisons. The significance level was set at p < 0.005 (GRF corrected).
Correlation analysis. The mean z values of brain regions with abnormal FCs were extracted to examine the correlations between abnormal FCs and clinical variables in the patients and UHR subjects. Multiple linear regressions were performed with abnormal FCs as dependent variables and clinical variables as independent variables in the patients and the UHR subjects. The correlation results were Bonferroni corrected at p < 0.05.

Results
Sample characteristics. Three UHR subjects, 4 patients with schizophrenia and 3 healthy controls were excluded due to excessive head motion. As shown in Table 1, the three groups have no significant differences in age, sex ratio, education level, and FD values. As expected, the patients got higher PANSS scores than the UHR subjects and healthy controls.
FC patterns of the seed. For each group, each cerebellar seed ( Figure S1) showed distributed FCs with the DMN using one-sample t-tests ( Figure S2). A union mask for each seed was created based on the results of one-sample t-tests for the following seed-based FC analyses.
Seed-based FCs: Group differences. Compared with the controls, the UHR subjects exhibited increased FCs between the left Crus I and the left middle temporal gyrus (MTG), between the right Crus I and the left cuneus, bilateral PCC/precuneus and bilateral precuneus, and between the Lobule IX and the right middle frontal cortex and bilateral medial prefrontal cortex (MPFC)/anterior cingulate cortex (ACC) ( Fig. 1 and Table 2). By contrast, the patients showed increased FCs between the right Crus I and the left MTG, bilateral rectus, bilateral PCC/precuneus/cuneus and right angular gyrus (AG), and between the Lobule IX and the left superior MPFC relative to the controls ( Fig. 1 and Table 2). Compared with the UHR subjects, the patients showed increased FCs between the right Crus I and the right superior MPFC, and decreased FCs between the Lobule IX and the right middle frontal cortex and bilateral superior MPFC/ACC ( Fig. 1 and Table 2). Hence, the UHR subjects and the patients shared increased FCs between the right Crus I and bilateral PCC/precuneus, and between the Lobule IX and the left superior MPFC.

Correlations between abnormal FCs and clinical variables.
In the UHR subjects, the mean z values of the right Crus I-bilateral precuneus connectivity were positively correlated with the SIPS negative symptoms scores (r = 0.533, p = 0.001) and total scores (r = 0.444, p = 0.009), and the PANSS negative symptoms scores (r = 0.426, p = 0.012) and total scores (r = 0.350, p = 0.042) (Fig. 2). No correlations were observed between abnormal FCs and clinical variables in the patients. There were also no correlations between abnormal FCs and age, educational level in the patients and UHR subjects.

Discussion
In the present study, the cerebellar seeds linking with the DMN were used to examine the cerebellar FC patterns with the DMN in the UHR subjects and first-episode, drug-naive patients with schizophrenia. The findings revealed that the UHR subjects and the patients showed increased cerebellar-DMN FCs relative to the controls. UHR subjects and patients with schizophrenia shared increased connectivity between the right Crus I and bilateral PCC/precuneus and between Lobule IX and the left superior MPFC. In addition, positive correlations were found between the cerebellar-DMN connectivity and clinical variables (SIPS/PANSS negative symptoms/total scores) in the UHR subjects.
Previously, we observed that first-episode, drug-naive patients with schizophrenia and their unaffected siblings shared increased cerebellar-DMN connectivity that could be applied as candidate endophenotypes for schizophrenia 35 . Consistent with the previous study and our hypothesis, patients with schizophrenia and UHR subjects shared increased cerebellar-DMN connectivity in this study. Combined with our previous study 35 , the presence of increased cerebellar-DMN connectivity in schizophrenia was advanced from the first-episode stage to the prodromal stage. However, the present findings are not consistent with the "disconnection" hypothesis in schizophrenia 1 , which hypothesized that schizophrenia have overall decreased connectivity 45 . In support to the "disconnection" hypothesis, Wang et al. 34 found significantly reduced FC between cerebellar seeds and brain regions of the DMN, such as the middle frontal gyrus, ACC, thalamus and supplementary motor area in a group of chronic and medicated patients with schizophrenia. The inconsistency may come from sample heterogeneity in addition to sample size and scanners. The prevailing opinion of overall decreased connectivity in schizophrenia is based on findings with chronic and/or medicated patients. When drug-naive and early-course patients were enrolled, some studies reported increased connectivity 46,47 . Moreover, early-course patients exhibited enhanced glutamate concentrations in the frontal gyrus 48,49 , which would result in frontal hyperconnectivity. Combined with the findings from early-course and chronic patients, we can depict that the cerebellar-DMN connectivity decreases progressively with illness duration, with increased connectivity in early-course patients and decreased connectivity in chronic patients. Since UHR subjects were enrolled in the present study, the presence of increased connectivity is advanced to the prodromal stage. Compared with the UHR subjects, some cerebellar-DMN connectivities in patients with schizophrenia begin to decrease in this study. By contrast, we previously observed decreased connectivity between the left Crus I and bilateral ACC in another group of first-episode patients 35 . Therefore, we can deduce that cerebellar-DMN connectivity decreases at the early-course of the disease. However, the exact timepoint of the reduction remains unclear. The different findings between two groups of patients may come from different illness duration in addition to different scanners. The mean illness duration of the present patient group is 5.16 months, and the mean illness duration of our previous patient group is 22.45 months. The difference of illness duration between the two studies gives additional support to our speculation that the cerebellar-DMN connectivity decreases progressively with illness duration. There are two possible interpretations to increased cerebellar-DMN connectivity in UHR subjects and patients with schizophrenia. First, FC differences of the prefrontal-thalamic-cerebellar circuit were compared among healthy children, adolescents and adults with a seed-based FC analysis 50 , and the findings revealed that FC of this circuit exhibits an inverted U-curve with maximal point in adolescents. The UHR subjects and the patients in this study are aged from 14 to 30 years, the developmental period from adolescents to adults. In addition, the pathophysiological process of the disease is speculated to start before the disease onset 35,51 . Normal developmental process of the cerebellar-DMN connectivity may be halted by the disease in the UHR subjects and the patients, and remain at relatively high point of the inverted U-curve. Hence, it is not surprising that the UHR subjects and the patients exhibit increased cerebellar-DMN connectivity in this study. The shared increased cerebellar-DMN connectivity in the UHR subjects and the patients may be a trait alteration for psychosis.  Table 2. Brain regions with abnormal cerebellar connectivity between groups. MNI = Montreal Neurological Institute; MPFC = medial prefrontal cortex; PCC = posterior cingulate cortex; AG = angular gyrus; ACC = anterior cingulate cortex.
Second, increased cerebellar-DMN connectivity is also meaningful from the physiology of FC. Increased FC is often conceived as compensatory reallocation or dedifferentiation [52][53][54][55] . Inflammation may modulate the compensatory process 46 . In the early-course of schizophrenia, proinflammatory cytokines (i.e., interleukin-6) activate the astrocytes and show hyperfunction (increased blood flow and metabolism) 56 . Regional hyperfunction can contribute to increased FC and activity of this region. Moreover, increased connectivity in the DMN has been found in early-course patients with schizophrenia 46 . As mentioned before, pathophysiological process may start before the disease onset 51 . Therefore, it is no wonder that the UHR subjects and the patients exhibit increased cerebellar-DMN connectivity in this study.
MPFC and PCC/precuneus are key nodes of the DMN, and act as key players in self-referential processing and emotional regulation [57][58][59] . Increased cerebellar-DMN connectivity may have an effect on the function of the DMN, and lead to cognitive and emotional disturbances in the UHR subjects and the patients 4 . It can also enhance the risk for the UHR subjects to transit to psychosis. Specially, positive correlations are found between the mean z values of the right Crus I-bilateral precuneus connectivity and the SIPS negative symptoms scores/ total scores, and the PANSS negative symptoms scores/total scores in the UHR subjects, indicating that the cerebellar-DMN connectivity bears clinical significance. However, these significant correlations disappear in the patients. It is speculated that alterations in cerebellar-DMN connectivity might have occurred around the disease onset in first-episode schizophrenia with predominantly positive and negative symptoms, complicating the correlations between clinical symptoms and abnormal cerebellar-DMN connectivity. Previously, significantly positive correlation was found between the thalamo-orbitofrontal cortex connectivities and Global Assessment of Functioning scores in UHR subjects, but this correlation was not observed in first-episode patients 60 . The previous study gives support to our speculation.
The present study has several limitations in addition to relatively small sample size. First, the present study is cross-sectional, and which part of the UHR subjects will convert to psychosis subsequently remains unclear. Therefore we do not know the difference of the cerebellar-DMN connectivity between the UHR subjects who will convert to psychosis later and those who will not. A follow-up study is needed to elucidate this issue. Second, we used the cerebellar seeds linking to the DMN. This method enhances the specificity of the findings from the DMN. Meanwhile, other connectivity is neglected in this study. Third, the inhomogeneous B0 field induced distortion usually happens at the prefrontal cortex and the cerebellum, where the present study mainly focuses on. Hence, the present findings should be interpreted with caution. Future study should be designed to correct he inhomogeneous B0 field induced distortion. Finally, the MRI data are acquired at resting state with a relatively long repetition time. Physiological noise such as heart and respiratory rhythm may have an effect on the data. More rigorous methods need to be developed to minimize such physiological noise.
In conclusion, our findings indicate that the UHR subjects and the patients exhibit increased cerebellar-DMN connectivity. Increased cerebellar-DMN connectivity shared by the UHR subjects and the patients may be a trait alteration for psychosis. Future studies should include longitudinal, multimodal imaging techniques to specify the possibility for cerebellar-DMN connectivity serving as predictors for transition to psychosis in the UHR subjects.