Evidence From Structural and Diffusion Tensor Imaging for Frontotemporal Deficits in Psychometric Schizotypy

Introduction

Large-scale genome-wide association studies^1,2 have provided strong evidence that the etiology of schizophrenia (SZ) is complex and multifactorial. Moreover, such studies have demonstrated that hundreds to thousands of common genetic variants with small effects contribute to the behavioral expression of psychotic-like phenomena across traditional diagnostic boundaries.³ Such findings provide strong support for a dimensional model in which phenomenological, genetic, and cognitive factors interact to affect the behavioral expression and severity of psychotic symptoms.⁴ At the phenotypic level, evidence suggests the existence of a continuous distribution of psychotic symptoms in the general population ranging from mild, or subclinical, to severe and clinically significant, with additional evidence indicating etiological continuity between subclinical and clinically significant psychosis phenotypes.⁵ The continuity between subclinical psychosis at the population level and the clinically significant levels of psychosis observed in SZ spectrum disorders may provide a unique opportunity to elucidate the pathophysiology of psychotic disorders free of the potentially confounding effects of treatment-related factors.

Although several approaches have been employed to measure subclinical psychosis in nontreatment seeking populations, the measurement of schizotypal traits is among the most common.⁶ Schizotypy was initially conceptualized by Rado⁷, and later elaborated by Meehl^8,9, to denote the genetically determined predisposition to SZ and may be measured in nonclinical samples using psychometric self-report questionnaires such as the Schizotypal Personality Questionnaire (SPQ)¹⁰ or the Chapman Scales.¹¹ Studies have generally demonstrated schizotypy to have dimensional factor structures analogous to those observed in SZ.^12–14 Moreover, several independent studies^15–18 have shown an increased incidence of schizotypy in relatives of SZ patients that are likely related to shared genetic variation.^19,20 Thus, the examination of psychometric schizotypy within nonpsychiatric populations is ideally suited for studies seeking to better characterize the neurobiology of psychosis.²¹

Several studies have examined the neurobiological basis of schizotypal personality disorder (SPD), a disorder characterized by severe schizotypal traits including asocial tendencies, difficulties with language, paranoia, odd behavior, and magical thinking. Recent reviews^22–24 suggest that at the structural level, SPD is associated with temporal lobe abnormalities comparable to those observed in SZ while frontal lobe regions may be more spared. Moreover, the aforementioned reviews also suggest that low fractional anisotropy (FA), a measure broadly associated with white matter integrity, in the uncinate fasciculus (UF) and temporal lobe is present in SPD; a finding generally consistent with those observed in SZ.²⁵ To date, however, relatively few studies have examined the neurobiological basis of psychometric schizotypy, or subclinical psychosis, in otherwise healthy adults.

Although several studies have reported structural differences between healthy participants with high levels of subclinical positive symptoms vs low levels, overall findings have been mixed.^26–28 However, several studies have reported cortical thickness abnormalities among healthy individuals with high levels of schizotypy compared with low levels^29,30 that are broadly consistent with findings in SZ patient samples.³¹ Additionally, diffusion tensor imaging (DTI) studies have provided complementary evidence for white matter alterations in psychometrically defined schizotypy similar to those observed in SZ, but these findings have also been inconsistent. While some studies have found higher FA in some regions but lower FA in others,³² others have found consistent reductions in FA in several regions^33,34 and yet others have found increases in FA³⁵ related to high levels of subclinical psychotic symptoms. Taken together these studies implicate aberrant neurodevelopmental processes in subclinical psychosis that are similar to those believed to underlie the neurobiology of SZ without the associated confound of antipsychotic medications and suggest that the examination of psychometric schizotypy in nonpsychiatric populations may provide insight into the neurobiology of psychosis. Samples examined in prior work, however, have been relatively small and have not comprehensively assessed the brain using multimodal imaging. Moreover, several prior studies examined adolescent and young adult samples that may still be at risk for developing psychosis. In the present study, we thus investigated both gray and white matter structural variation in relation to psychometric schizotypy using structural and DTI in a large sample of healthy adults. Consistent with observations of less gray matter volume in SZ³⁶ and with models of frontotemporal lobe dysfunction in SZ,^37–39 we hypothesized that individuals characterized by high schizotypy would demonstrate less gray matter volume, lower cortical thickness, and lower FA in frontal and temporal lobe regions, compared with those characterized by low schizotypy.

Methods

Results

Comparison of high vs low schizotypy groups on the total SPQ score confirmed that the groups significantly differed (M_High = 8.52 ± 6.66 vs M_Low = 0.75 ± 0.82; t = 9.49; P < .001). Moreover, the relatively low mean of the high schizotypy group suggested symptom levels were reflective of subclinical psychometric schizotypy rather than schizotypal personality disorder. Comparison of high and low schizotypy groups on additional demographic characteristics revealed no significant differences (all P’s > .05) in age (M_High = 36.17 ± 13.93 vs M_Low = 35.23 ± 12.18), estimated IQ (M_High = 101.91 ± 11.09 vs M_Low = 101.39 ± 9.80), or handedness (M_High = 0.77 ± 0.50 vs M_Low=0.80±0.38). Examination of sex distributions revealed that the proportion of females in the high schizotypy group (56.72%) was significantly greater than in the low schizotypy group (39.44%) (χ ² = 4.13, P = .04).

Mean (SD) values for the frontal and temporal lobe white matter and gray matter volumes are provided in table 1 with univariate analyses provided for descriptive purposes only. In the primary repeated measures ANCOVA, there were significant main effects of group for both frontal (F _(1,133) = 4.50, P = .036) and temporal (F _(1,133) = 4.40, P = .038) gray matter volumes such that high schizotypy individuals had less volume overall compared with low schizotypy individuals. There were also significant group × region interactions for frontal white (F _(1,133) = 2.40, P = .049) and gray (F _(1,133) = 3.89, P = .003) matter volume. Post-hoc analysis indicated lower volumes of white matter in the rostral middle frontal cortex (F _(1,133) = 3.95, P = .049), gray matter pars orbitalis (F _(1,133) = 4.06, P = .046), and gray matter rostral middle frontal cortex (F _(1,133) = 8.15, P = .005) in high schizotypy participants compared with low schizotypy participants. No main effects of group were observed for frontal or temporal white matter volume. There was a significant main effect of group for temporal gray matter thickness (F _(1,133) = 4.40, P = .038), but not for frontal gray matter thickness. Specifically, high schizotypy participants exhibited lower gray matter thickness compared with low schizotypy participants. The group-by-region interactions were not statistically significant for gray matter thickness in either the frontal or temporal lobes (P’s > .05). The frontal and temporal lobe gray matter thickness measures are provided in table 2 with univariate analyses provided for descriptive purposes only.

Analysis of FA using the tract-based measures revealed a significant main effect of group for the IFOF (F _(1,133) = 4.90, P = .029) such that high schizotypy individuals had lower FA compared with low schizotypy individuals. In addition, there was a significant group × hemisphere interaction for the UF (F _(1,133) = 6.29, 133, P = .013). Post-hoc analysis revealed that high schizotypy individuals had significantly greater asymmetry (R > L; t ₍₁₃₆₎ = −2.78, P = .006) in the UF compared with low schizotypy individuals. There were no significant group differences in FA in either the right or left UF. Neither the main effects of group nor group-by-hemisphere interactions were statistically significant for the SLF, ILF, or cingulum bundle. Mean (SD) FA values for the tracts are provided in table 3 with univariate analyses provided for descriptive purposes only.

To determine whether having both volume and thickness abnormalities was associated with greater psychometric schizotypy, we conducted 3 supplementary logistic regression analyses to predict group membership for regions that differed significantly between groups including gray matter volume (rostral middle frontal and pars orbitalis), white matter volume (rostral middle frontal), and average temporal lobe gray matter thickness. Classes of regions (gray matter, white matter, and thickness) were entered into the logistic regression in blocks to determine whether the overall model improved significantly by adding the last block. Thus, as an example, in one logistic regression gray matter and white matter volumes were entered into the model as the first block followed by temporal gray matter thickness in the second block to determine whether the addition of the latter variable significantly improved the overall model. None of the regression models were significant (P > .05) suggesting that no particular abnormality predicted group membership above and beyond the others.

Supplementary ANCOVA were conducted using 3 groups (low, medium, and high psychometric schizotypy) in contrast to the median split approach to further examine the relationship between subclinical symptoms and our imaging measures; these analyses yielded results consistent with our original findings. Moreover, the most robust effects were identified between individuals with the highest SPQ total scores compared with the other groups. Specifically, patients with the highest SPQ total scores had significantly (P < .05) lower FA within the IFOF, less total gray and white matter in the rostral middle frontal gyrus, lower average temporal thickness, and less total gray matter in the pars orbitalis compared with the other 2 groups. Trend level effects (P = .08) for a group-×-hemisphere interaction (likely reflecting the lower statistical power) were evident for asymmetry within the UF such that individuals with the highest total SPQ total scores demonstrated greater asymmetry compared with the other groups.

Although we used sex as a covariate in our analyses, we conducted ancillary analyses by removing the 10 youngest males with low total schizotypy scores from the analysis to better match the groups for sex, which served to equate the sex distribution across groups (P = .22) while maintaining the age match, and reran all of our primary analyses. The group main effects for FA in the IFOF, UF FA asymmetry, rostral middle frontal white and gray matter volume, and pars orbitalis gray matter volume all remained statistically significant (Ps > .05).

Discussion

The results of this investigation indicated an association between psychometric schizotypy and measures of gray and white matter using both structural and DTI. Specifically, our study indicated that otherwise healthy adults who exhibit higher levels of psychometric schizotypy demonstrated less frontal and temporal lobe gray matter and lower temporal lobe gray matter thickness compared with participants characterized as lower in schizotypy. Regionally specific effects were also evident such that individuals characterized as higher in psychometric schizotypy had less gray and white matter volume specifically within the rostral middle frontal region compared with individuals characterized as lower in schizotypy. Investigation of cortical thickness measures indicated that individuals higher in schizotypy demonstrated lower temporal (but not frontal) cortical thickness compared with individuals lower in schizotypy. Moreover, the use of probabilistic tractography indicated that compared with individuals characterized as lower in psychometric schizotypy, those who were characterized as higher in schizotypy had lower FA in the IFOF as well as differences in UF asymmetry. Strengths of the current study include the large sample, comprised of healthy adults with no history of an Axis I disorder, and no history of psychotropic medication exposure and the use of multimodal imaging measures.

It is difficult to compare our findings to prior work given that few studies have investigated psychometric schizotypy, especially across a broad range of imaging measures. Our findings are generally consistent, however, with recent data suggesting that high levels of positive schizotypy in otherwise healthy adults is associated with significantly less gray matter volume in medial prefrontal, orbitofrontal, and temporal cortical regions.²⁸ Our findings also converge with prior work in patients with SZ⁵⁴ and SPD,²⁴ which identified less gray matter in frontal and temporal lobe regions. Moreover, our study identified less gray and white matter that was localized to the rostral middle frontal cortex among individuals higher in psychometric schizotypy compared with those lower in schizotypy. Consistent with our findings, several prior neuroimaging studies in SZ reported less gray⁵⁵ and white⁵⁶ matter in the rostral middle frontal region and thus, dysfunction involving this region may be particularly relevant for the overlap in phenotypic expression between SZ and schizotypy. It is also noteworthy that dysfunction within the rostral middle frontal region may contribute to abnormal executive functioning,⁵⁷ a cognitive domain that is impaired in both SZ⁵⁸ and SPD.⁵⁹

Prior work reported that compared with patients with SZ, patients with SPD may demonstrate preservation of some frontal lobe white matter regions^60,61 that could, at least in part, reflect a compensatory mechanism. In contrast, temporal gray matter abnormalities may be a feature shared by both patients with SZ and SPD. This hypothesis is consistent with volumetric findings from the current study wherein a main effect of group was not apparent for frontal lobe white matter volume in contrast to the gray matter where a main effect of group was observed. This suggests white matter volume at the gross anatomic level in the frontal (perhaps except for the rostral middle frontal region) could conceivably be protective for higher levels of psychometric schizotypy and is consistent with the hypothesis that different subregions within the frontal cortex may differentiate SZ from SPD. In addition, neither global nor regionally specific effects were evident between groups in frontal cortical thickness further implicating an additional possible protective mechanism from frank psychosis. It should be noted that while gray matter volume and cortical thickness share similar properties, they are fundamentally different measures.^62,63 For example, although both thickness and area can influence volume, volume may be more closely associated with surface area than cortical thickness,⁶⁴ which appears to be specifically influenced by cell type and/or neuronal density.⁶⁵

Our finding of lower FA within the IFOF among individuals higher in psychometric schizotypy is consistent with prior work implicating dysfunction of this tract in the neurobiology of SZ. For example, Yao et al⁶⁶ reported white matter deficits in first-episode SZ in the left IFOF using an activation likelihood estimation meta-analysis involving 8 studies that included 271 first-episode patients and 297 healthy controls. Moreover, the observation of lower FA within the left IFOF among first-episode neuroleptic-naive patients compared with healthy volunteers suggests these effects are not an artifact of antipsychotic medication exposure.⁶⁷

There are several limitations to the current study. Our use of a self-report instrument to assess schizotypy may allow for over or underreporting of positive and negative subclinical psychotic symptoms. In our larger sample comprising 658 participants, however, we found high concordance (rho > 0.80) between several measures of subclinical psychosis, including the SPQ, and symptoms assessed using a clinician administered diagnostic interview (SCID-NP) (P. DeRosse, unpublished data). While this does not rule out the possibility of over- or underreporting, it lends support to the convergent validity of the self-report measure. An additional limitation is that because SPQ scores were not normally distributed, we utilized a median split to identify high and low schizotypy groups. This approach, although maximizing our power, may have limited our ability to detect more subtle relationships between the level of subclinical psychotic symptoms and imaging measures. However, supplementary analyses comparing 3 groups (high, medium, and low SPQ) revealed results nearly identical to those obtained using 2 groups. Another potential weakness of the current study is the observed sex difference between the high and low schizotypy groups. Although we included sex as a covariate in our analyses, to ensure that the sex difference was not driving the result we conducted ancillary analyses demonstrating that this did not contribute to the observed findings. Finally, it should also be noted that tractography measures do not map directly onto the brain regions examined using FreeSurfer and thus, we are limited in our interpretation across imaging modalities.

In sum, the present study provided evidence for a link between psychometric schizotypy and a range of structural and DTI-derived measures encompassing frontotemporal regions in otherwise healthy adults without the confound of antipsychotic medications. These findings contribute to a growing literature suggesting that psychosis can be examined along a continuum of severity and suggest that this continuum may relate to subtle variation in brain structure and function. Moreover, results suggest that some frontal regions, at least at the gross anatomic level, may serve as a compensatory mechanism that may be relevant to the distinction between subclinical and clinically significant psychotic symptoms.

Funding

National Institute of Mental Health (R01 MH076995 to Dr P.R.S., R00 MH086756 to Dr P.D.); the NSLIJ Research Institute General Clinical Research Center (M01 RR018535); an Advanced Center for Intervention and Services Research (P30 MH090590); and a Center for Intervention Development and Applied Research (P50 MH080173).