Abstract
Background The 3q29 deletion (3q29Del) is a copy number variant (CNV) with the highest known effect size for psychosis-risk (>40-fold increased risk). Systematic research on this CNV offers promising avenues for identifying mechanisms underlying psychosis and related disorders. Relative to other high-impact CNVs like 22q11.2Del, far less is known about the phenotypic presentation and pathophysiology of 3q29Del. Emerging findings indicate that posterior fossa abnormalities are common among 3q29Del carriers; however, their clinical relevance is unknown.
Methods Here, we report the first in-depth evaluation of psychotic symptoms in study participants with 3q29Del (N=23), using the Structured Interview for Psychosis-Risk Syndromes (SIPS), and compare to SIPS data from 22q11.2Del participants (N=31) and healthy controls (N=279). We also investigate the relationship between psychotic symptoms, cerebellar morphology, and cystic/cyst-like malformations of the posterior fossa in 3q29Del by structural brain imaging.
Results Cumulatively, 48% of the 3q29Del sample exhibited a psychotic disorder (N=4) or attenuated positive symptoms (N=7), with three individuals with attenuated symptoms meeting the frequency and timing criteria for clinical high risk for psychosis. Males with 3q29Del scored higher in negative symptoms than females. 3q29Del participants had more severe ratings than controls on all domains and exhibited less severe negative symptoms than 22q11.2Del participants. An inverse relationship was identified between positive symptom severity and cerebellar cortex volume in 3q29Del, while cystic/cyst-like malformations yielded no clinical link with psychosis.
Conclusions Overall, our findings establish the unique and shared profiles of psychotic symptoms across two CNVs and highlight cerebellar involvement in elevated psychosis-risk in 3q29Del.
- 3q29 deletion syndrome
- 22q11.2 deletion syndrome
- psychosis
- SIPS
- structural magnetic resonance imaging
- cerebellum
Introduction
The identification of copy number variants (CNVs) that confer high risk for serious mental illnesses has provided new substrates to investigate the etiology of psychotic disorders. The 3q29 and 22q11.2 deletions (Del) are two CNVs that significantly increase risk for a range of neuropsychiatric disorders, especially schizophrenia and other psychotic disorders (1-12). Thus, systematic research on these CNVs holds promise for shedding light on the neuropathology underlying psychosis.
22q11.2Del is caused by a hemizygous 1.5- to 3.0-Mb deletion involving 45 protein-coding genes. The deletion affects 1 in 2,148-4,000 individuals (13-17) and is estimated to convey a >20-fold increased risk for developing schizophrenia (18, 19). Standardized measures have been used to examine the clinical high risk (CHR) or the “prodromal” symptoms that may precede the onset of psychosis in 22q11.2Del syndrome and the general population. Before the first psychotic episode, typically during late adolescence and early adulthood, individuals often exhibit a period of functional decline coinciding with the emergence of attenuated psychotic symptoms (20-23). Those identified at CHR present with these “warning” signs and are at substantial risk for future psychosis (24). Published reports in 22q11.2Del have shown that about 30-60% of study participants manifest attenuated psychotic symptoms which are known to be linked with elevated risk for subsequent conversion to psychosis (25, 26).
Individuals with 3q29Del are hemizygous for a 1.6-Mb interval containing 21 protein-coding genes. Compared to 22q11.2Del, 3q29Del was only recently identified (27) and is less common, with a prevalence of approximately 1 in 30,000 individuals (10); consequently the associated phenotypes are just now being documented. Like 22q11.2Del, 3q29Del is pleiotropic; it elevates risk for a range of physical and psychiatric disabilities, including low birth weight, failure to thrive, cardiac defects, cognitive deficits, and autism spectrum disorder (9, 28). Current evidence suggests that 3q29Del confers an equal or greater risk for psychosis than 22q11.2Del. Recent reports revealed that the 3q29 deletion confers a >40-fold increased risk for schizophrenia (1, 8, 29). It is not known whether early clinical signs of psychosis risk are elevated among individuals with 3q29Del.
Our team recently documented the neurodevelopmental and psychiatric manifestations associated with 3q29Del (10, 30). 3q29Del individuals participated in in-depth evaluation of psychiatric symptoms and magnetic resonance imaging (MRI). Here, we report detailed results on psychotic symptoms in study participants with 3q29Del, as assessed by the Structured Interview for Psychosis-Risk Syndromes (SIPS) (31, 32). Using this gold-standard instrument, we determined the proportion of 3q29Del participants with attenuated or florid psychotic symptoms and contrasted this with the symptom profile in 22q11.2Del participants and healthy controls (HC).
Next, we asked whether psychosis susceptibility in 3q29Del is associated with morphometric features of the brain measured by structural MRI. Recent radiological findings by our team have revealed that cerebellar hypoplasia and cystic/cyst-like malformations of the posterior fossa, are significantly elevated in 3q29Del (10). Our group has also shown quantitively that study participants with 3q29Del exhibit abnormal cerebellar volumes compared with HCs (33). In the current study, we demonstrate that these neuroanatomical phenotypes may have clinical relevance. This is the first known investigation of the relationship between psychotic symptoms and neuroimaging findings in 3q29Del syndrome and suggests the cerebellum may be a primary site of pathology.
Methods and Materials
Participants
The 3q29Del sample
Twenty-three individuals with 3q29Del, ages 8.08-39.12 (mean ± SD = 16.94 ± 8.24 years) were evaluated using the SIPS (31, 32). Study participants were ascertained from the 3q29 registry (28) and between August 2017-February 2020 traveled to Atlanta, GA for deep phenotyping with gold-standard instruments and MRI. Deletion status was confirmed from the clinical genetics report and/or medical records. The project protocol and a summary of findings have been previously described (10, 30).
Comparison samples
The HC comparison sample included 279 healthy individuals, ages 12.07-34.41 (mean ± SD = 20.08 ± 4.61 years), from the second cohort of the North American Prodromal Longitudinal Study (NAPLS2), who were evaluated with the SIPS between June 2009-October 2012. NAPLS2 is a multi-site consortium investigating the trajectories of individuals at CHR (34). These HCs chosen for the present analysis neither met criteria for any psychotic disorder nor reported any attenuated psychotic symptoms and had no history of a central nervous system disorder, intellectual disability, substance dependence in the previous six months, or a first-degree relative with a psychotic disorder.
The 22q11.2Del comparison sample included SIPS data from 31 individuals with 22q11.2Del, ages 13.85-29.77 (mean ± SD = 19.75 ± 4.12 years), from the 22q11.2 Clinic in Atlanta, GA (25, 26, 35). Participants were ascertained from a case registry, with the deletion confirmed via fluorescence in situ hybridization (FISH). Participants were often referred for FISH as minors due to cardiovascular defects, language difficulties or immunological problems. Individuals identified as adults were referred as part of clinical care within a genetics or heart clinic, as described in (25). Data were collected between February 2006-June 2007.
Assessment of attenuated and florid psychotic symptoms
All study participants were administered the SIPS semi-structured clinical interview by trained personnel. The 19 items in the SIPS are grouped within four domains: positive, negative, disorganization, and general. Each item is rated on a scale from 0 (Absent) to 6 (Severe and Psychotic for positive symptoms, Extreme for others), with a rating of 3 (Moderate) indicating clinical significance. Item ratings were summed to produce a total score for each domain. For adolescents and adults with 3q29Del meeting criteria for a psychotic disorder, the Structured Clinical Interview for DSM-5 (SCID-5) (36) specified the diagnosis. All study participants provided informed consent to participate. These studies were approved by Emory University’s Institutional Review Board.
Structural MRI acquisition and processing in 3q29Del
T1- and T2-weighted structural MRI was performed in 17 3q29Del participants with available SIPS data, ages 8.08-39.12 (mean ± SD = 18.13 ± 9.03 years). Of the six 3q29Del participants who had SIPS data but did not complete MRI, five were medically ineligible and one declined to participate. Images were acquired on a Siemens Magnetom Prisma 3T scanner in the sagittal plane using a 32-channel Prisma head coil and an 80mT/m gradient. T1-weighted 3D images were acquired in the sagittal plane using a single-echo MPRAGE sequence (37) with the following parameters: TE=2.24ms, TR=2400ms, TI=1000ms, bandwidth=210Hz/pixel, FOV=256×256mm, resolution=0.8mm isotropic. T2-weighted 3D images were acquired in the sagittal plane using a SPACE sequence (38) with the following parameters: TE=563ms, TR=3200ms, bandwidth=745Hz/pixel, FOV=256×256mm, resolution=0.8mm isotropic. Further details on neuroimaging parameters have been described (10, 33).
All MR images were processed using the “minimal structural pre-processing” pipeline established by the Human Connectome Project (HCP, v.4.1.3) with FreeSurfer (v.6.0) (http://surfer.nmr.mgh.harvard.edu/). We performed automatic segmentation as described in (39), using a previously described probabilistic atlas and Bayesian classification rule that assigns anatomical labels to each voxel based on estimates derived from a manually labeled training set (40). Since it is currently unfeasible to achieve a topologically correct reconstruction and automatic segmentation of the cerebellar cortical surface from in vivo MRI data, we treated the cerebellum as a volumetric structure and extracted cerebellar cortex and cerebellar white matter volumes for each 3q29Del participant, using a “coarse parcellation” scheme (41). Tissue-specific volumes were then summed to derive a global metric of total cerebellar volume; these three measures constitute the regions of interest (ROI) in the present study. A similar approach has been successfully used in previous literature (42-50), allowing for methodological consistency.
Finally, we used estimated total intracranial volume (eICV) as a proxy for premorbid brain size (51). eICV was calculated by FreeSurfer as described in (52) and has been reported to show high correlations with manually delineated intracranial volume (53-55). We report results from both eICV-adjusted and unadjusted analyses (56), given previous reports of microcephaly (27, 57, 58), and eICV reduction in 3q29Del (33). See Supplemental Materials for extended methods.
Statistical analyses
All analyses were conducted using R v.4.0.3 (59). The frequencies of florid or attenuated psychotic symptoms were compared between the 3q29Del and 22q11.2Del groups using Fisher’s exact test. To assess the variation in symptom profiles between groups, a series of analyses of covariance (ANCOVA) were performed; first, by domain and then item-wise to identify symptoms that may be elevated in 3q29Del. Partial eta-squared (η2p) was used to assess effect sizes. Pairwise comparisons for 3q29Del vs. HC and 3q29Del vs. 22q11.2Del were performed using Tukey’s test. Since we observed several violations of the statistical assumptions required for ANCOVA (partially due to a likely floor effect in the HC sample), SIPS scores were log-transformed in between-group analyses. Age and sex were considered as potential covariates.
For neuroimaging analyses in 3q29Del, we used multiple linear regression to examine the relationship between ROI volumes and SIPS ratings for positive, negative and disorganization symptom domains separately. In models where a significant relationship was identified, we subsequently added eICV as an additional covariate to test whether results reflect a link with the cerebellum beyond global variability in head size. We also employed an additional categorical approach that relies on diagnostic criteria to determine the presence of either a psychotic disorder or attenuated psychotic symptom syndrome (APSS). Binary logistic regression was used to estimate the probability that either of these diagnoses are present in the 3q29Del sample given cerebellar volumes as the predictor. Multiple comparisons correction was applied at the ROI level using the Bonferroni method, with a rounded p-value threshold of 0.02 (0.05/3 ROIs).
In all models, age and sex were considered as covariates. Standard diagnostics were performed for testing regression assumptions. Multicollinearity was assessed in models with eICV by computing variance inflation factor scores. Given observed violations of linear regression assumptions using ordinary least squares, we additionally calculated heteroscedasticity-robust estimates (estimation-type: HC1) (60); final inferences from linear regression are based on robust estimates.
Finally, we tested whether psychotic symptoms differ between 3q29Del participants with versus without cystic/cyst-like malformations of the posterior fossa, which were identified by a board-certified neuroradiologist (33). Fisher’s exact test was used for diagnostic comparisons. Student’s t-test and Wilcoxon signed-rank test were used for dimensional comparisons. Given limited power due to smaller sample size in the neuroimaging arm, we also report trend-level associations using a less conservative alpha (p ≤ 0.10) to inform future hypotheses. All analyses were two-tailed. See Supplemental Materials for extended methods.
Results
Presence of attenuated and florid psychotic symptoms in 3q29Del and 22q11.2Del
Demographics of the 3q29Del sample are presented in Table 1 and Fig. S1; information for 22q11.2Del and HCs are provided in Table S1. Four out of the 23 3q29Del participants (17%) met criteria for a psychotic disorder: schizophrenia (N = 1); schizoaffective disorder, bipolar type (N = 1); and unspecified schizophrenia spectrum and other psychotic disorder (N = 2). Two of these individuals had a prior diagnosis of a psychotic disorder; two did not. These four individuals (3 males, 1 female) tended to be older than those without a psychotic disorder (with psychosis: mean age = 26.66; without psychosis: mean age = 14.89). However, cognitive ability was similar between the groups (with psychosis: mean FSIQ = 73.00; without psychosis: mean FSIQ = 73.89). Those participants meeting criteria for a psychotic disorder reported past or current symptoms that met a rating of Severe and Psychotic for unusual thought content, suspiciousness, and/or perceptual abnormalities; these participants reported prodromal symptoms beginning in late childhood or early adolescence (on average between ages 11-12). In contrast, only one out of the 31 22q11.2Del participants (3%) met criteria for a psychotic disorder.
Descriptive statistics are reported separately for the total 3q29Del sample, and the sample stratified by the presence/absence of a psychotic disorder. Abbreviations: deletion, Del; mean, M; standard deviation, SD.
We next excluded individuals with a psychotic disorder and sought to analyze sub-threshold symptoms of psychosis in the remaining individuals with complete data (N = 19 in 3q29Del, N = 29 in 22q11.2Del). Seven 3q29Del participants (37%) exhibited clinically significant attenuated positive symptoms (i.e., at least one positive symptom ≥ 3). Three of these individuals (16%) met the frequency and timing criteria to qualify for APSS (31). Nineteen 22q11.2Del participants (66%) exhibited clinically significant attenuated positive symptoms. The information necessary for assessing APSS was unavailable for this group.
Cumulatively, 48% of the 3q29Del sample showed either florid or attenuated psychotic symptoms, as compared to 67% of the 22q11.2Del sample (Table 2); the comparison of these frequencies revealed no significant difference between the groups (p = 0.26). Too few participants with 3q29Del (N = 3) were taking antipsychotics to determine treatment effects.
For P5 and the positive symptom domain total, 22q11.2Del N = 30 due to missing data from one 22q11.2Del participant. Note that 13 3q29Del participants and 10 22q11.2Del participants had more than one P item with a ≥ 3 rating. Abbreviations: Structured Interview for Psychosis-Risk Syndromes, SIPS; deletion, Del.
Group differences in demographics and symptom profiles
SIPS symptom ratings are presented in Table 3 for each study group. There was a significant age difference between the 3q29Del sample and HCs (p ≤ 0.01), and between the 3q29Del and 22q11.2Del samples (p ≤ 0.05), with 3q29Del participants being younger on average. Given these age differences, as well as evidence from community samples indicating increases in psychotic-like experiences through adolescence followed by a decline in early adulthood (61-63), we first assessed whether the relationship between age and symptom severity differs between study samples. Results are presented in Table S2 and Fig. S2. The correlations between symptom severity and age were positive and often highest in magnitude for 3q29Del, positive and modest in 22q11.2Del, but negative in HCs, which is consistent with the group differences in the mean and range of age.
The unadjusted means and standard errors for both DS groups and HCs are shown. Pairwise comparisons reflect p-values calculated on log-transformed sex-adjusted data. Refer to Table S3 for log-transformed sex-adjusted means and standard errors for each group. ‘a’ indicates significant difference (p ≤ 0.01 for domain, p ≤ 0.003 for item) between 3q29Del and HC; ‘b’ indicates significant difference (p ≤ 0.01 for domain, p ≤ 0.003 for item) between 3q29Del and 22q11.2Del. For P5 and the positive symptom domain total, 22q11.2Del N = 30 due to missing data from one 22q11.2Del participant. For G2 and the general symptom domain total, HC N = 278 due to missing data from one HC. Abbreviations: Structured Interview for Psychosis-Risk Syndromes, SIPS; healthy controls, HC; deletion, Del; deletion syndrome, DS; analysis of covariance, ANCOVA; standard, std.
Next, ANCOVAs were performed to contrast the average symptom ratings by domain of the 3q29Del sample against HCs and 22q11.2Del, adjusting for sex. As the mean age and correlations between symptom severity and age varied between groups, age was initially not adjusted for in the comparison of symptom profiles, as covariate adjustment using the ANCOVA framework assumes homogeneity between groups. There was a significant effect of diagnostic group on all domains: positive [F(2, 328) = 112.87, p ≤ 0.001, η2p = 0.41], negative [F(2, 329) = 135.00, p ≤ 0.001, η2p = 0.45], disorganization [F(2, 329) = 166.30, p ≤ 0.001, η2p = 0.50], general [F(2, 328) = 74.56, p ≤ 0.001, η2p = 0.31]. Pairwise contrasts between 3q29Del and the two other groups were conducted correcting for multiple comparisons, with a rounded p-value threshold of 0.01 for each domain (0.05/4 domains). Comparisons revealed that individuals with 3q29Del were rated significantly higher than HCs on all SIPS domains (p’s ≤ 0.01): positive [t(328) = 8.12, p ≤ 0.001], negative [t(329) = 8.21, p ≤ 0.001], disorganization [t(329) = 11.10, p ≤ 0.001], and general [t(328) = 6.59, p ≤ 0.001]. In contrast, the average ratings for the 3q29Del group did not significantly differ from ratings in 22q11.2Del for positive [t(328) = -2.84, p = 0.01], disorganization [t(329) = -1.81, p = 0.16], or general [t(328) = -2.24, p = 0.06]. However, the 22q11.2Del group exhibited greater negative symptoms than 3q29Del [t(329) = -3.75, p ≤ 0.001].
To compare group profiles of specific items, a series of ANCOVAs were conducted with the same approach. Bonferroni correction was applied, with a rounded p-value threshold of 0.003 (0.05/19 items). For almost all items on the SIPS, there was a significant effect of diagnostic group on severity (p’s ≤ 0.003). Fig. 1 presents the symptom profile for each diagnostic group. Comparing the 3q29Del sample with HCs revealed more severe ratings on all items (p’s ≤ 0.003), except for sleep disturbance, t(329) = 2.55, p = 0.01, which was nominally significant. Differences between 3q29Del and 22q11.2Del groups were primarily in the negative symptom domain. Participants with 22q11.2Del exhibited more severe social anhedonia [t(329) = -6.09, p ≤ 0.001], avolition [t(329) = -4.19, p ≤ 0.001], experience of emotions and self [t(329) = -4.16, p ≤ 0.001], and impaired ideational richness [t(329) = -3.10, p = 0.002]. Lastly, individuals with 22q11.2Del were rated as more impaired in personal hygiene [t(329) = -4.27, p ≤ 0.001], and as experiencing greater dysphoric mood [t(328) = -4.43, p ≤ 0.001]. The only group difference in positive symptoms was more severe suspiciousness in 22q11.2Del [t(329) = -4.18, p ≤ 0.001]. When the above analysis was conducted with the inclusion of age as an additional covariate, the same pattern of results was found.
Standard error bars are shown. In panel A) SIPS domain totals are presented for each group. In panel B) item-specific SIPS ratings are presented for each group. For P5 and the positive symptom domain total, 22q11.2Del N = 30 due to missing data from one 22q11.2Del participant. For G2 and the general symptom domain total, HC N = 278 due to missing data from one HC. Abbreviations: Structured Interview for Psychosis-Risk Syndromes, SIPS; healthy controls, HC; deletion, Del.
The pairwise comparisons are presented in Table 3, along with the untransformed values for the means and standard errors of each group for interpretability on the original scale. See Table S3 for adjusted means and standard errors after log-transformation. Of note, the 3q29Del group was found to exhibit similar estimates for general cognitive ability compared with the 22q11.2Del group (Table S1), which mitigates the potential confounding effect of IQ on between-group differences.
Sex differences within groups
We also examined sex differences for the three diagnostic groups (Fig. S3). Statistical analyses included a set of ANCOVAs adjusted for age, given the absence of violations in homogeneity of slopes within group. As these analyses were conducted to explore whether there were sex differences in symptoms similar to those observed in research on CHR and psychotic patients, multiple comparisons correction was not made. The differences in domain scores between males and females with 3q29Del were significant only for negative symptoms [F(1, 20) = 4.85, p = 0.04, η2p = 0.20], with males exhibiting more severe symptoms than females. Similarly, negative symptoms differed by sex among the 22q11.2Del group [F(1, 28) = 5.10, p = 0.03, η2p = 0.15], and the HC group [F(1, 276) = 6.94, p = 0.01, η2p = 0.02], with males exhibiting more severe symptoms than females. Lastly, only among HCs, positive symptoms varied between sexes [F(1, 276) = 3.93, p = 0.05, η2p = 0.01], with males demonstrating greater severity. Notably, relatively small effect sizes were found in the HC group for both symptom domains.
Structural cerebellar correlates of psychosis-risk in 3q29Del
Prior work by our group using structural MRI scans acquired from a larger 3q29Del sample (including the scans analyzed in the present study) has shown that cerebellar hypoplasia and posterior fossa arachnoid cysts are frequently observed among individuals with 3q29Del (10). Our team has also shown in the same sample that 3q29Del participants exhibit smaller cerebellar cortex and larger cerebellar white matter volumes than healthy controls, and the prevalence of cystic/cyst-like malformations of the posterior fossa is significantly elevated in this syndrome (33).
To determine whether the severity of psychotic symptoms in participants with 3q29Del is associated with structural cerebellar abnormalities previously documented in this population (10, 33) (see Fig. 2A for a representative MR image), we first modeled the dimensional relationship between SIPS ratings for three major symptom domains relevant to psychotic disorders (positive, negative, disorganization) and global and tissue-specific cerebellar volumes in participants with 3q29Del. Table S4 summarizes demographic, clinical, and volumetric data from the 3q29Del subsample with available neuroimaging and SIPS data. Note that there were no significant differences between left and right hemispheric cerebellar volumes (p’s > 0.05) (Fig. S4); hence, each ROI reflects bilateral volumes.
A) The sagittal view of a representative T1-weighted structural MRI scan from a 3q29Del participant with cerebellar hypoplasia and a cystic/cyst-like malformation of the posterior fossa (yellow arrow). B-D) Predictor effect plots show the relationships between SIPS ratings for three major symptom domains relevant to psychotic disorders and global and tissue-specific cerebellar volumes among individuals with 3q29Del (N = 17). Predicted values of positive, negative and disorganization symptom ratings indexed by the SIPS (y-axis) were computed from the best-fit multiple linear regression models from Table S5, while covariates (age and sex) were held fixed. The shaded area is a pointwise confidence band for the fitted values. Rug plots on both sides of each graph show the distribution of the predictor and outcome variables and data points represent partial residuals. Schematic illustrations of the three ROIs are provided next to their corresponding predictor effect plots for clarity. Parameter estimates on each plot reflect the main effect of B) cerebellar cortex volume, C) cerebellar WM volume and D) total cerebellum volume on domain-specific symptom severity ratings and reflect heteroskedasticity-robust estimates. Regression results indicate a significant inverse relationship between cerebellar cortex volume and positive symptom severity, with smaller volumes predicting more severe symptoms among 3q29Del participants (bold panel). ‘*’ indicates significant association after Bonferroni correction (p ≤ 0.02). ‘†’ indicates trend-level association (p ≤ 0.10). Abbreviations: Magnetic resonance imaging, MRI; region of interest, ROI; Structured Interview for Psychosis-Risk Syndromes, SIPS; white matter, WM; deletion, Del.
Our results (Fig. 2B-D, Tables S5A-J) indicate a significant inverse relationship between positive symptom severity and cerebellar cortex volume in 3q29Del, while correcting for age and sex. Smaller cerebellar cortical volumes were associated with greater positive symptoms (b = -0.43, p = 0.02), and this effect remained significant in a secondary model that includes eICV for head size correction (b = -0.29, p = 0.03) (Fig. 2B, Tables S5B-C). Results also indicate a trend-level inverse association between positive symptom severity and total cerebellum volume (b = -0.27, p = 0.10) (Fig. 2D, Table S5A), and a trend-level positive association between disorganization symptom severity and cerebellar white matter volume (b = -0.28, p = 0.09) (Fig. 2C, Table S5J), while correcting for age and sex. There was no relationship between negative symptom severity and ROI volumes (p’s > 0.10).
Given ongoing discussions regarding the merits and limitations of dimensional versus categorical assessment of psychosis (64-66), we next took a categorical approach and asked whether the probability of meeting diagnostic criteria for either a psychotic disorder or APSS would recapitulate the results from our dimensional approach. Our findings from logistic regression indicate no significant relationship between interrogated volumes and the likelihood of these outcomes in 3q29Del, while correcting for age and sex (p’s > 0.10) (Tables S6A-C). Detailed regression results can be found in Tables S5-6. See Fig. S5 for sensitivity analysis.
Finally, we found no significant association between diagnostic and dimensional indices of psychotic symptoms and the presence of cystic/cyst-like malformations of the posterior fossa in 3q29Del (p’s > 0.05) (Table S7). There was a trend-level relationship between these radiological findings and the odds of APSS (p = 0.08), which may warrant future consideration. See Supplemental Materials for extended results.
Discussion
Rare pathogenic CNVs that arise from recurrent chromosomal rearrangements are now robustly implicated in psychosis-risk (2-4, 67, 68), with eight CNV loci surpassing genome-wide significance (1). Among these loci, the 3q29 deletion has the largest estimated effect size (1, 8), offering a promising opportunity to link a specific genetic mechanism to brain and behavioral phenotypes underlying at least one form of psychosis. Recently, we documented the broader phenotypic spectrum of 3q29Del by direct evaluation of the largest sample of 3q29Del study participants reported to date (10, 30). In the present study, we substantially extended these findings by systematically characterizing the comprehensive profile of psychotic symptoms in 3q29Del and investigating the potential effects of sex and age on severity. Furthermore, we compared the profiles of 3q29Del to HCs and 22q11.2Del. This is particularly important since cross-CNV similarities or differences in symptomology may suggest overlapping or divergent pathogenic mechanisms. Finally, we investigated the neuroanatomical correlates of psychosis-risk in 3q29Del by employing a hypothesis-driven approach, focusing on macrostructural properties of the cerebellum. Overall, our findings target important gaps in our understanding of the link between 3q29Del and exceptionally increased genetic risk for psychotic illnesses.
First, the present study confirmed that a significant proportion of 3q29Del participants meet diagnostic criteria for psychosis (4/23, 17%) or manifest one or more clinically significant attenuated positive symptoms (7/23, 30%), with 43% (3/7) meeting the frequency and timing criteria for APSS. Cumulatively, 48% of the 3q29Del sample showed either florid or attenuated psychotic symptoms, as compared to 67% of the 22q11.2Del sample. The rates of psychosis-related phenotypes observed here are especially striking considering the young ages of the 3q29Del participants. Based on published demographic data on CHR groups, the mean age at ascertainment is between 17 and 18 (69-71). Among those 3q29Del individuals not already meeting criteria for a psychotic disorder, the mean age was 15. Thus, most of the 3q29Del sample is below the typical age of onset for psychosis and the prodrome. Notably, the prevalence of psychosis or clinically significant attenuated positive symptoms is 67% (6/9) among 3q29Del individuals aged 17 and older (Fig. S1). Given this finding and previously published rates of transition to psychosis in CHR groups (72), it is likely that more transitions to florid psychosis are to be expected in the present 3q29Del sample.
In this context, we note that the 22q11.2Del sample was ascertained through cardiology or immunology clinics, whereas 3q29Del participants were ascertained through a registry. Contrary to the expectation that a clinically ascertained sample may be more severely affected, there were no significant differences in the rates of florid or attenuated psychotic symptoms between the 3q29Del and 22q11.2Del samples. However, since we were unable to determine the prevalence of APSS among 22q11.2Del participants, the true frequency of formal psychosis-risk syndromes could not be compared between these samples. 3q29Del participants with a psychotic disorder presented with prodromal symptoms around late childhood to early adolescence, consistent with several reports in 22q11.2Del (11, 73-75), although findings vary (76-79). The profile documented in the current sample indicates that monitoring and identification of early signs among individuals with 3q29Del are warranted. Proper assessment and treatment is necessary to reduce distress and improve functioning; among CHR, interventions may delay or prevent the onset of florid psychosis (80); thus, early intervention among 3q29Del may also attenuate psychosis severity and improve outcomes.
Further, the relation of sex and age with symptoms in the 3q29Del group is similar to that observed in CHR and psychotic samples (61-63). Males tend to score higher in negative symptoms than females, and both positive and negative symptoms tend to increase with age, although due to the small sample of 3q29Del participants this correlation was not statistically significant. It should be noted that studies of both community and clinical samples reveal increases in psychotic-like experiences during adolescence that extends to about 19 years of age, then begins to decline for those who do not transition to psychosis. The mean age and age-range of the HC group in the present study is in the period when experiences are normatively in decline, hence the inverse relation between age and severity in that group. Given this finding, we performed both age-adjusted and unadjusted analyses to assess the variation in psychotic symptom profiles between groups; both approaches produced the same pattern of results.
Our findings from these group comparisons indicate that individuals with 3q29Del are rated significantly higher than HCs on all SIPS domains, while they show remarkable similarity to individuals with 22q11.2Del in average ratings for the positive, disorganization, and general symptom domains. However, the 22q11.2Del group exhibited a more severe profile than 3q29Del in the negative symptom domain. This finding parallels an earlier study that compared individuals at CHR with 22q11.2Del to individuals at CHR without 22q11.2Del; the 22q11.2Del group had greater negative symptoms, although the two groups showed comparable positive and global symptoms (81). Interestingly, individuals with 22q11.2Del also present with greater negative symptoms when compared to Williams Syndrome or idiopathic developmental disabilities (82). Several lines of evidence also point to higher rates of major depressive disorder in 22q11.2Del (83, 84), which is a potential secondary source of negative symptoms. In contrast, most individuals with 3q29Del had relatively well-preserved hedonic experiences and social motivation (9). It is conceivable that the pronounced profile of negative symptoms observed in 22q11.2Del compared to 3q29Del reflects some level of divergence in etiopathogenetic mechanisms. Disturbances of the brain reward system have been suggested to be central in the pathogenesis of negative symptoms (85); hence, reward processing may be an important future direction for cross-comparison studies in these CNVs. We also note that the less severe negative symptoms seen in 3q29Del may suggest a more favorable profile for functional outcomes and response to intervention in psychotic symptoms, considering the previously reported inverse relationship between negative symptom severity and global functioning (86, 87), as well as prior reports of negative symptoms being more resistant to treatment (88).
Moreover, our results revealed a significant relationship between cerebellar morphology and severity of psychosis-risk symptoms in 3q29Del, driven by both a tissue-specific and symptom-specific association between cerebellar cortex volume and positive symptoms. Cystic/cyst-like malformations of the posterior fossa yielded no link with psychotic symptoms, suggesting the relationship is specific to cerebellar cortex. Besides its critical role in motor coordination, the cerebellum has been suggested to perform a parallel role in the coordination of thoughts (89-92) through internal models that enable predictive function. An internal model is a system that, when given its current state, can predict the outcome of an action without performing the action (93). The cerebellum is proposed to be a key node in the neural systems that enable the formation and implementation of such models (94). Accordingly, cerebellar dysfunction has been hypothesized to lead to perceptual abnormalities, such as auditory hallucinations (95, 96) and a loss of coherence in speech output (97, 98), which are among the core symptoms of psychosis. Notably, perceptual abnormalities and disorganized communication showed the most severe positive symptom ratings in the present 3q29Del sample.
The potential involvement of cerebellar dysfunction in schizophrenia has been speculated since the 1990s (99-101). Congruently, emerging findings indicate an increased prevalence of psychosis among patients with cerebellar pathology (102, 103). The cerebellar cognitive-affective syndrome, which follows from cerebellar lesions, has substantial overlap with the phenomenology of psychosis (104-107). In addition, there are reports of cerebellar abnormalities in idiopathic psychosis (108-119) and pathology in Purkinje cells, the output neurons of the cerebellar cortex, has been documented in schizophrenia (120-122). Further, altered functional or structural connectivity and morphology of the cerebellum has been identified in CHR groups (50, 123-131). Notably, a recent study of the Philadelphia Neurodevelopmental Cohort found cerebellar gray matter volume to be a robust predictor of psychotic-like experiences in a large community sample of youths (132), which is consistent with our findings.
Thus, various lines of evidence point to cerebellar involvement in psychotic symptoms. However, how exactly cerebellar abnormalities relate to psychosis remains unknown and the considerable heterogeneity reported in extant literature constitutes a major challenge in addressing this gap. The association that we identified between cerebellar cortex volume and positive symptom severity in 3q29Del may help elucidate one part of this question, as this finding suggests that one or more genes affected by the hemizygous deletion of the 3q29 interval modulate the link between cerebellar development and psychosis-risk. In fact, many genes located in this interval, including DLG1 and BDH1, which have been proposed as drivers of neuropsychiatric phenotypes (133, 134), show medium-high expression in the cerebellum (https://www.proteinatlas.org/). Its protracted development may increase the cerebellum’s susceptibility to perturbations of these genes (135, 136).
Our results indicate that diagnostic dichotomization does not recapitulate the dimensional brain-behavior relationship identified in 3q29Del. There may be several reasons for this finding. One explanation is the statistical drawback (reduced power) associated with dichotomizing continuous variables (137). Another explanation may be misclassification bias (138); young participants have not yet moved through the highest risk period for psychosis and may transition to APSS or psychosis at a later timepoint; this may have biased the categorical results toward the null. Additionally, a binary split presupposes a true cutpoint between opposite sides of diagnostic criteria. For example, an APSS diagnosis requires attenuated psychotic symptoms to have begun within the past year or to have been rated one or more points higher compared to 12 months ago (31), which could lead to the diagnostic exclusion of individuals who have a longer and more gradual prodrome (139). Hence, such criteria, although useful in other applications, may lead to arbitrary divisions of a complex continuum. To conclude, our findings support cerebellar cortex morphology as a likely trait marker of psychosis-risk in 3q29Del, which we operationalize as a multidimensional continuum with varying degrees of severity, distress, and functional impairment.
Our study has several limitations. First, our sample size was limited due to the rare prevalence of 3q29Del, which prevented analyses of interactions. Demands of study participation (e.g., travel to GA) may have barred individuals with more severe presentations from participating; thus, symptom severity could be underestimated. Additionally, the complex architecture of the cerebellum presents challenges for MRI; future work with improved segmentation techniques will explore finer associations with cerebellar subregions. Furthermore, cerebellar connections with cerebral association cortices (140), as well as the basal ganglia (141), warrant future investigation to address questions related to circuitry. Finally, given the cross-sectional design, our findings do not necessarily reflect causation. Longitudinal follow-up as well as targeted investigations of the cerebellum in the mouse model of 3q29Del (142) will be conducted for causal inference and mechanistic insights.
Altogether, our findings indicate that clinical signs of psychosis risk are elevated in 3q29Del participants. Furthermore, our results establish the unique and shared profiles of psychotic symptoms across 3q29Del and 22q11.2 DS and highlight cerebellar involvement in elevated psychosis-risk in 3q29Del.
Data Availability
The 3q29DS data collected in this study are deposited in the NIMH Data Archive (nda.nih.gov) (behavioral data: collection 2614, embargoed until September 2022; neuroimaging data: collection 3126, embargoed until November 2023). Prior to these dates, the 3q29DS data are available from the corresponding author upon reasonable request. The NAPLS2 data used in this article are available from Elaine F. Walker on behalf of the NAPLS Consortium upon reasonable request. The 22q11.2DS data used in this article are available from Opal Ousley and Joseph F. Cubells upon reasonable request.
Data availability
The 3q29Del data collected in this study are deposited in the NIMH Data Archive (nda.nih.gov) (behavioral data: collection 2614, embargoed until September 2022; neuroimaging data: collection 3126, embargoed until November 2023). Prior to these dates, the 3q29Del data are available from the corresponding author upon reasonable request. The NAPLS2 data used in this article are available from Elaine F. Walker on behalf of the NAPLS Consortium upon reasonable request. The 22q11.2Del data used in this article are available from Opal Ousley and Joseph F. Cubells upon reasonable request.
Acknowledgements and disclosures
We acknowledge the 3q29Del study population, their families and the 3q29 Project members. We thank the Marcus Autism Center for providing clinical assessment resources. Funding for this work was provided by National Institutes of Health (NIH) grants R01 MH110701, R01 MH118534, and National Institutes of Mental Health (NIMH) grant U01MH081988. REDCap is supported by grant UL1 TR000424. Funding for work in 22q11.2Del was provided by grants awarded to Joseph Cubells and Opal Ousley from the Robert W. Woodruff Fund and the Predictive Health Initiative of Emory University. We would like to thank the following NAPLS2 PIs for their contributions to the healthy control dataset: Jean Addington, Carrie E. Bearden, Kristin Cadenhead, Tyrone D. Cannon, Barbara A. Cornblatt, Daniel H. Mathalon, Thomas H. McGlashan, Diana O. Perkins, Larry J. Seidman, Ming T. Tsuang, and Scott W. Woods.
The authors have no competing financial interests related to this study to declare.
Footnotes
References
- 1.↵
- 2.↵
- 3.
- 4.↵
- 5.
- 6.
- 7.
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.
- 15.
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.
- 44.
- 45.
- 46.
- 47.
- 48.
- 49.
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.
- 63.↵
- 64.↵
- 65.
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.
- 71.↵
- 72.↵
- 73.↵
- 74.
- 75.↵
- 76.↵
- 77.
- 78.
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.
- 91.
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.
- 106.
- 107.↵
- 108.↵
- 109.
- 110.
- 111.
- 112.
- 113.
- 114.
- 115.
- 116.
- 117.
- 118.
- 119.↵
- 120.↵
- 121.
- 122.↵
- 123.↵
- 124.
- 125.
- 126.
- 127.
- 128.
- 129.
- 130.
- 131.↵
- 132.↵
- 133.↵
- 134.↵
- 135.↵
- 136.↵
- 137.↵
- 138.↵
- 139.↵
- 140.↵
- 141.↵
- 142.↵
