Clinical, Brain, and Multilevel Clustering in Early Psychosis and Affective Stages

Importance

Approaches are needed to stratify individuals in early psychosis stages beyond positive symptom severity to investigate specificity related to affective and normative variation and to validate solutions with premorbid, longitudinal, and genetic risk measures.

Objective

To use machine learning techniques to cluster, compare, and combine subgroup solutions using clinical and brain structural imaging data from early psychosis and depression stages.

Design, Setting, and Participants

A multisite, naturalistic, longitudinal cohort study (10 sites in 5 European countries; including major follow-up intervals at 9 and 18 months) with a referred patient sample of those with clinical high risk for psychosis (CHR-P), recent-onset psychosis (ROP), recent-onset depression (ROD), and healthy controls were recruited between February 1, 2014, to July 1, 2019. Data were analyzed between January 2020 and January 2022.

Main Outcomes and Measures

A nonnegative matrix factorization technique separately decomposed clinical (287 variables) and parcellated brain structural volume (204 gray, white, and cerebrospinal fluid regions) data across CHR-P, ROP, ROD, and healthy controls study groups. Stability criteria determined cluster number using nested cross-validation. Validation targets were compared across subgroup solutions (premorbid, longitudinal, and schizophrenia polygenic risk scores). Multiclass supervised machine learning produced a transferable solution to the validation sample.

Results

There were a total of 749 individuals in the discovery group and 610 individuals in the validation group. Individuals included those with CHR-P (n = 287), ROP (n = 323), ROD (n = 285), and healthy controls (n = 464), The mean (SD) age was 25.1 (5.9) years, and 702 (51.7%) were female. A clinical 4-dimensional solution separated individuals based on positive symptoms, negative symptoms, depression, and functioning, demonstrating associations with all validation targets. Brain clustering revealed a subgroup with distributed brain volume reductions associated with negative symptoms, reduced performance IQ, and increased schizophrenia polygenic risk scores. Multilevel results distinguished between normative and illness-related brain differences. Subgroup results were largely validated in the external sample.

Conclusions and Relevance

The results of this longitudinal cohort study provide stratifications beyond the expression of positive symptoms that cut across illness stages and diagnoses. Clinical results suggest the importance of negative symptoms, depression, and functioning. Brain results suggest substantial overlap across illness stages and normative variation, which may highlight a vulnerability signature independent from specific presentations. Premorbid, longitudinal, and genetic risk validation suggested clinical importance of the subgroups to preventive treatments.

Multimodal Machine Learning Workflows for Prediction of Psychosis in Patients With Clinical High-Risk Syndromes and Recent-Onset Depression

IMPORTANCE

Diverse models have been developed to predict psychosis in patients with clinical high-risk (CHR) states. Whether prediction can be improved by efficiently combining clinical and biological models and by broadening the risk spectrum to young patients with depressive syndromes remains unclear.

OBJECTIVES

To evaluate whether psychosis transition can be predicted in patients with CHR or recent-onset depression (ROD) using multimodal machine learning that optimally integrates clinical and neurocognitive data, structural magnetic resonance imaging (sMRI), and polygenic risk scores (PRS) for schizophrenia; to assess models' geographic generalizability; to test and integrate clinicians' predictions; and to maximize clinical utility by building a sequential prognostic system.

DESIGN, SETTING, AND PARTICIPANTS

This multisite, longitudinal prognostic study performed in 7 academic early recognition services in 5 European countries followed up patients with CHR syndromes or ROD and healthy volunteers. The referred sample of 167 patients with CHR syndromes and 167 with ROD was recruited from February 1, 2014, to May 31, 2017, of whom 26 (23 with CHR syndromes and 3 with ROD) developed psychosis. Patients with 18-month follow-up (n = 246) were used for model training and leave-one-site-out cross-validation. The remaining 88 patients with nontransition served as the validation of model specificity. Three hundred thirty-four healthy volunteers provided a normative sample for prognostic signature evaluation. Three independent Swiss projects contributed a further 45 cases with psychosis transition and 600 with nontransition for the external validation of clinical-neurocognitive, sMRI-based, and combined models. Data were analyzed from January 1, 2019, to March 31, 2020.

MAIN OUTCOMES AND MEASURES

Accuracy and generalizability of prognostic systems.

RESULTS

A total of 668 individuals (334 patients and 334 controls) were included in the analysis (mean [SD] age, 25.1 [5.8] years; 354 [53.0%] female and 314 [47.0%] male). Clinicians attained a balanced accuracy of 73.2% by effectively ruling out (specificity, 84.9%) but ineffectively ruling in (sensitivity, 61.5%) psychosis transition. In contrast, algorithms showed high sensitivity (76.0%-88.0%) but low specificity (53.5%-66.8%). A cybernetic risk calculator combining all algorithmic and human components predicted psychosis with a balanced accuracy of 85.5% (sensitivity, 84.6%; specificity, 86.4%). In comparison, an optimal prognostic workflow produced a balanced accuracy of 85.9% (sensitivity, 84.6%; specificity, 87.3%) at a much lower diagnostic burden by sequentially integrating clinical-neurocognitive, expert-based, PRS-based, and sMRI-based risk estimates as needed for the given patient. Findings were supported by good external validation results.

CONCLUSIONS AND RELEVANCE

These findings suggest that psychosis transition can be predicted in a broader risk spectrum by sequentially integrating algorithms' and clinicians' risk estimates. For clinical translation, the proposed workflow should undergo large-scale international validation.

Altered Functional Subnetwork During Emotional Face Processing: A Potential Intermediate Phenotype for Schizophrenia

IMPORTANCE

Although deficits in emotional processing are prominent in schizophrenia, it has been difficult to identify neural mechanisms related to the genetic risk for this highly heritable illness. Prior studies have not found consistent regional activation or connectivity alterations in first-degree relatives compared with healthy controls, suggesting that a more comprehensive search for connectomic biomarkers is warranted.

OBJECTIVES

To identify a potential systems-level intermediate phenotype linked to emotion processing in schizophrenia and to examine the psychological association, task specificity, test-retest reliability, and clinical validity of the identified phenotype.

DESIGN, SETTING, AND PARTICIPATIONS

The study was performed in university research hospitals from June 1, 2008, through December 31, 2013. We examined 58 unaffected first-degree relatives of patients with schizophrenia and 94 healthy controls with an emotional face-matching functional magnetic resonance imaging paradigm. Test-retest reliability was analyzed with an independent sample of 26 healthy participants. A clinical association study was performed in 31 patients with schizophrenia and 45 healthy controls. Data analysis was performed from January 1 to September 30, 2014.

MAIN OUTCOMES AND MEASURES

Conventional amygdala activity and seeded connectivity measures, graph-based global and local network connectivity measures, Spearman rank correlation, intraclass correlation, and gray matter volumes.

RESULTS

Among the 152 volunteers included in the relative-control sample, 58 were unaffected first-degree relatives of patients with schizophrenia (mean [SD] age, 33.29 [12.56]; 38 were women), and 94 were healthy controls without a first-degree relative with mental illness (mean [SD] age, 32.69 [10.09] years; 55 were women). A graph-theoretical connectivity approach identified significantly decreased connectivity in a subnetwork that primarily included the limbic cortex, visual cortex, and subcortex during emotional face processing (cluster-level P corrected for familywise error = .006) in relatives compared with controls. The connectivity of the same subnetwork was significantly decreased in patients with schizophrenia (F = 6.29, P = .01). Furthermore, we found that this subnetwork connectivity measure was negatively correlated with trait anxiety scores (P = .04), test-retest reliable (intraclass correlation coefficient = 0.57), specific to emotional face processing (F = 17.97, P < .001), and independent of gray matter volumes of the identified brain areas (F = 1.84, P = .18). Replicating previous results, no significant group differences were found in face-related amygdala activation and amygdala-anterior cingulate cortex connectivity (P corrected for familywise error =.37 and .11, respectively).

CONCLUSIONS AND RELEVANCE

Our results indicate that altered connectivity in a visual-limbic subnetwork during emotional face processing may be a functional connectomic intermediate phenotype for schizophrenia. The phenotype is reliable, task specific, related to trait anxiety, and associated with manifest illness. These data encourage the further investigation of this phenotype in clinical and pharmacologic studies.

Genetic association signal near NTN4 in Tourette syndrome

Tourette syndrome (TS) is a neurodevelopmental disorder with a complex genetic etiology. Through an international collaboration, we genotyped 42 single nucleotide polymorphisms (p < 10(-3) ) from the recent TS genomewide association study (GWAS) in 609 independent cases and 610 ancestry-matched controls. Only rs2060546 on chromosome 12q22 (p = 3.3 × 10(-4) ) remained significant after Bonferroni correction. Meta-analysis with the original GWAS yielded the strongest association to date (p = 5.8 × 10(-7) ). Although its functional significance is unclear, rs2060546 lies closest to NTN4, an axon guidance molecule expressed in developing striatum. Risk score analysis significantly predicted case-control status (p = 0.042), suggesting that many of these variants are true TS risk alleles.

Converging Evidence for Epistasis between ANK3 and Potassium Channel Gene KCNQ2 in Bipolar Disorder

Genome-wide association studies (GWAS) have implicated ANK3 as a susceptibility gene for bipolar disorder (BP). We examined whether epistasis with ANK3 may contribute to the “missing heritability” in BP. We first identified via the STRING database 14 genes encoding proteins with prior biological evidence that they interact molecularly with ANK3. We then tested for statistical evidence of interactions between SNPs in these genes in association with BP in a discovery GWAS dataset and two replication GWAS datasets. The most significant interaction in the discovery GWAS was between SNPs in ANK3 and KCNQ2 (p = 3.18 × 10⁻⁸). A total of 31 pair-wise interactions involving combinations between two SNPs from KCNQ2 and 16 different SNPs in ANK3 were significant after permutation. Of these, 28 pair-wise interactions were significant in the first replication GWAS. None were significant in the second replication GWAS, but the two SNPs from KCNQ2 were found to significantly interact with five other SNPs in ANK3, suggesting possible allelic heterogeneity. KCNQ2 forms homo- and hetero-meric complexes with KCNQ3 that constitute voltage-gated potassium channels in neurons. ANK3 is an adaptor protein that, through its interaction with KCNQ2 and KCNQ3, directs the localization of this channel in the axon initial segment (AIS). At the AIS, the KCNQ2/3 complex gives rise to the M-current, which stabilizes the neuronal resting potential and inhibits repetitive firing of action potentials. Thus, these channels act as “dampening” components and prevent neuronal hyperactivity. The interactions between ANK3 and KCNQ2 merit further investigation, and if confirmed, may motivate a new line of research into a novel therapeutic target for BP.

FAF1, a gene that is disrupted in cleft palate and has conserved function in zebrafish

Cranial neural crest (CNC) is a multipotent migratory cell population that gives rise to most of the craniofacial bones. An intricate network mediates CNC formation, epithelial-mesenchymal transition, migration along distinct paths, and differentiation. Errors in these processes lead to craniofacial abnormalities, including cleft lip and palate. Clefts are the most common congenital craniofacial defects. Patients have complications with feeding, speech, hearing, and dental and psychological development. Affected by both genetic predisposition and environmental factors, the complex etiology of clefts remains largely unknown. Here we show that Fas-associated factor-1 (FAF1) is disrupted and that its expression is decreased in a Pierre Robin family with an inherited translocation. Furthermore, the locus is strongly associated with cleft palate and shows an increased relative risk. Expression studies show that faf1 is highly expressed in zebrafish cartilages during embryogenesis. Knockdown of zebrafish faf1 leads to pharyngeal cartilage defects and jaw abnormality as a result of a failure of CNC to differentiate into and express cartilage-specific markers, such as sox9a and col2a1. Administration of faf1 mRNA rescues this phenotype. Our findings therefore identify FAF1 as a regulator of CNC differentiation and show that it predisposes humans to cleft palate and is necessary for lower jaw development in zebrafish.

Variability of 5-HT2C receptor cys23ser polymorphism among European populations and vulnerability to affective disorder

Substantial evidence supports a role for dysfunction of brain serotonergic (5-HT) systems in the pathogenesis of major affective disorder, both unipolar (recurrent major depression) and bipolar.(1) Modification of serotonergic neurotransmission is pivotally implicated in the mechanism of action of antidepressant drugs(2) and also in the action of mood stabilizing agents, particularly lithium carbonate.(3) Accordingly, genes that code for the multiple subtypes of serotonin receptors that have been cloned and are expressed in brain,(4) are strong candidates for a role in the genetic etiology of affective illness. We examined a structural variant of the serotonin 2C (5-HT2C) receptor gene (HTR2C) that gives rise to a cysteine to serine substitution in the N terminal extracellular domain of the receptor protein (cys23ser),(5) in 513 patients with recurrent major depression (MDD-R), 649 patients with bipolar (BP) affective disorder and 901 normal controls. The subjects were drawn from nine European countries participating in the European Collaborative Project on Affective Disorders. There was significant variation in the frequency of the HT2CR ser23 allele among the 10 population groups included in the sample (from 24.6% in Greek control subjects to 9.2% in Scots, chi(2) = 20.9, df 9, P = 0.01). Logistic regression analysis demonstrated that over and above this inter-population variability, there was a significant excess of HT2CR ser23 allele carriers in patients compared to normal controls that was demonstrable for both the MDD (chi(2) = 7.34, df 1, P = 0.006) and BP (chi(2) = 5.45, df 1, P = 0.02) patients. These findings support a possible role for genetically based structural variation in 5-HT2C receptors in the pathogenesis of major affective disorder.

Gaussian beam parameters that use Coddington-based Y-NU paraprincipal ray tracing

A method is described for modeling a Gaussian beam that travels along an axial or oblique path of a principal ray in a meridional plane. The method uses paraprincipal rays, which are traced adjacent to the principal ray by using a Y-NU ray trace procedure that is based on Coddington's equations. Formulas are given that relate the parameters of the Gaussian beam to the heights and slopes of the paraprincipal rays. Two ray trace examples are provided.