This is the first study to our knowledge to demonstrate predictors of outcome in a clinically at-risk sample for psychosis using a baseline functional neuroimaging phenotype. Here we identified a network of brain regions that was differentially involved for those at risk for developing psychosis compared to healthy matched controls during performance on a naturalistic discourse processing task. Overall, the task elicited activity in the expected network of regions typically engaged in language tasks, including the inferior frontal gyrus (IFG), bilateral medial prefrontal regions, inferior and middle temporal gyri, as well as the anterior cingulate. However, clinical-high risk (CHR) participants showed increased neural activity relative to healthy controls, in the medial prefrontal lobe bilaterally and anterior cingulate, as well as left IFG and left inferior and middle temporal gyri. Further, relative to CHR participants who did not develop psychosis, CHR subjects who subsequently developed psychosis showed a pattern of relative over-activation in language –associated brain regions, including the left inferior frontal gyrus and inferior temporal lobe, as well as bilateral striatum and thalamus, and the frontal operculum. Our findings of relative hyper-activation of language-related brain regions during discourse processing in youth at high risk for psychosis are consistent with a hypothesis of neural inefficiency in these vulnerable individuals (Karlsgodt et al., 2007
; Potkin et al., 2009
). These patterns are consistent with data from high-functioning patients with schizophrenia who show compensatory increases in brain activity during task performance (Karlsgodt et al., 2009
Our results also indicate that baseline neural activity during discourse processing was related to severity of positive formal thought disorder at follow-up, as well as social outcome. In particular, increased activity in the left ventral inferior frontal gyrus (during task performance in the Reasoning condition) was associated with increased severity of formal thought disorder at follow-up. Conversely, increased activity in the dorsal region of the inferior frontal gyrus was associated with poorer social outcome at follow-up, indicating that the observed patterns of relative hyper-activation in language-related brain regions have prognostic significance. Interestingly, cross-sectional associations between neural activity in these regions and concurrent clinical status (psychosocial functioning and severity of thought disorder) were not nearly as robust as were these predictive relationships. Specifically, there was only a significant inverse correlation (at the p<.01 level) between neural activity in the anterior cingulate and Social Attainment scores (r=−.39; p<.008). The significant associations between baseline neural activity in frontal and temporal ROIs and measures of thought disorder that were observed at follow-up did not reach this level of significance in the cross-sectional baseline comparisons. Because thought disorder scores tended to increase over the follow-up period in those who converted, the range of scores at baseline was restricted, which likely attenuated the findings. The results of our predictive analyses may help improve the accuracy of existing algorithms that attempt to predict those at greatest risk for psychosis based on clinical characteristics. Developing multivariate prediction algorithms for determining risk - that include functional and structural neuroimaging data- could provide additional information, as changes in neural activity could predate symptoms.
In this study, we found portions of both dorsal and ventral left inferior frontal gyrus were activated to a greater degree during semantic reasoning judgments in CHR individuals relative to healthy controls. Numerous fMRI studies have revealed the contributions of the left ventral aspects of IFG to language processing (Badre and Wagner, 2004
; Dapretto and Bookheimer, 1999
; Sabb et al., 2007
), and specifically the evaluation of semantic incongruencies in sentence processing. For example, Cardillo and colleagues (2004)
used a sentence processing task with semantically congruent and incongruent endings (e.g., “the boy bounced the ball/wall”). They found that activity was greatest in the IFG during incongruent trials, which may reflect detection of an irregular or unexpected semantic ending. Dorsal regions of the left IFG have been linked to other components of language processing, including sequencing of information (Gelfand and Bookheimer, 2003
), verbal working memory (Cabeza and Nyberg, 2000
), and other executive processes including response planning (Derrfuss et al., 2004
). Our data provide further support for the role of the left IFG in semantic processing through judgments of reasoning, and additionally suggest that these processes are over-active in CHR youth.
As expected, we also found group differences in temporal lobe language areas, including the inferior and middle temporal gyrus. Evidence from neuropsychological lesion studies (Warrington and Shallice 1984
) as well as fMRI studies of semantic processing (Mummery et al.,1999
; Rossell et al.,2001
) highlight the importance of the temporal lobe in language processing. There is also substantial evidence that this structure is perturbed in schizophrenia, and that the degree of temporal volume reduction is related to severity of formal thought disorder (Shenton et al.,2001
). Semantic priming studies in schizophrenia patients find that automatic activation of lexico-semantic representations is normal and, in thought-disordered patients, even increased, suggesting a wider, possibly faster, automatic spread of activation (Kuperberg et al.,2008
). In a review of this literature, Kircher (2008)
hypothesizes that dysfunction of the semantic network, may be due to lateral temporal lobe abnormalities, and that temporal pathology is predominantly involved in the generation of positive formal thought disorder in schizophrenia. Our findings may extend this model by suggesting that lateral temporal dysfunction is present prior to the onset of illness in those at clinical high-risk.
Our discourse processing task also modulated activity in the anterior cingulate (ACC). While ACC activity is associated with a range of cognitive processes, converging evidence suggests that one specific role of this region is in error monitoring, or conflict detection of competing representations (Carter et al.,2001
; Carter and van Veen 2007
) Detecting and resolving conflicting/incongruent responses is a critical component of both conditions in our paradigm that require on-line context processing. Disruption of these cognitive processes – and corresponding ACC dysfunction - has been frequently demonstrated in patients with schizophrenia (Carter et al., 2001
; Carter and van Veen, 2007
), their relatives (MacDonald et al., 2005
), as well as genetic high-risk samples (Whalley et al., 2006
). Our results provide further evidence of abnormal ACC activity in an at-risk population, although further studies are needed to address the precise nature of this dysfunction.
Certain limitations of this study should be noted. We used a naturalistic task that had previously been validated in adolescents, in order to maximize comprehension and accurate performance. While we found strong activation patterns in the Reasoning condition, results were less consistent in the Topic-Maintenance condition, with much more variability in fMRI signal between subjects. This could be due partially to the block design employed here. Although this design allowed maximum power to detect subtle differences in conditions between groups, we were unable to investigate the timecourse of processing of incongruous information. Future investigations using trial-based designs in CHR populations are clearly warranted to further characterize the contribution of the network of brain regions seen here that are putatively important for a complex phenotype such as discourse processing. In addition, as with all studies involving psychiatric populations, medication is a potential confound. In our study, several of the CHR participants were taking psychoactive medication although only a minority were taking atypical antipsychotics. However, baseline medications did not differ between those who later converted and those who did not, and therefore could not account for group differences in activation patterns. In addition, there was no relationship between antipsychotic medication use and activation in the brain regions that were differentially engaged across groups (p>.20). Nevertheless, we fully acknowledge that our study was not designed to examine differential effects of medications, and this could be better addressed in the context of a randomized clinical trial in which treatment is standardized.
Identifying predictors and mechanisms of conversion to psychosis among individuals ascertained in a clinical high risk or prodromal state are critical next steps along the pathway to prevention strategies. Our results suggest that neural activity during discourse processing has important predictive implications, as it is associated with subsequent thought disorder severity and social outcome. This is the first study, to our knowledge, to demonstrate functional differences in activity between those who convert to psychosis and those who do not, prior to illness onset. These findings may suggest strategies for the development of novel preventive treatments that can correct or compensate for the specific neurodevelopmental and psychological changes associated with the formation of psychotic symptoms during late adolescence and early adulthood.