Hippocampal Metabolite Concentrations in A Schizophrenia Case Series Vary in Association With Rare Gene Variants For Schizophrenia Versus Autism

Background Rare variants in the TRIO gene are associated with schizophrenia and autism spectrum disorders (ASD), which are commonly comorbid. ASD may dene a specic schizophrenia subtype. This study examined person-specic hippocampal metabolite concentrations for 4 schizophrenia cases harboring rare variants in TRIO or its interaction partner ARMS/KIDINS220 and 5 cases with other rare variants. Methods Nine of 19 cases from a prior imaging study underwent targeted exome sequencing. Multi-voxel 1 H-MRS imaging of the entire 3-dimensional hippocampus found only increased Creatine [Cr] (cellular energy use) concentration, distinguished cases at the group level. However, concentrations of N-acetyl-aspartate [NAA] (neuronal integrity) and choline [Cho] (membrane turnover/myelination) concentrations had a greater variance in cases than controls. Four with rare, missense-coding mutations or non-frameshift deletions in TRIO or ARMS/KIDINS220 had signicantly lower [Cho] than the other 5 (1.78 ± 0.18 mM versus 2.67 ± 0.8 mM: t = p but similar [NAA]. Two cases harboring rare variants in the SLC39A1 zinc transporter had the lowest [NAA]. (8.41 ± 0 80 mM versus 10.35 ± 2.03 mM, t = 4.52, p = 0.001). The highest [Cho] accompanied rare variants in SORCS2 and SORT, associated with and Disease.


Background
Schizophrenia is highly comorbid with autism spectrum disorder (ASD) with a recent meta-analysis nding the prevalence of ASD in individuals with schizophrenia ranged from 3.4 to 52% (1). Both conditions are associated with rare gene variants in the Rho guanine nucleotide exchange factor (RhoGEF) TRIO gene, which is considered to be a hotspot for mutations associated with ASD (2,3). A phenotype for this comorbidity could de ne an important biological subtype of schizophrenia for research and treatment.
This study examined rare variants in TRIO and other genes with respect to hippocampal metabolite concentrations, obtained by multivoxel proton MR spectroscopic imaging ( 1 H-MRSI) of the entire structure. The different metabolites it quantitates are closely associated with cellular processes, including NAA (N-Acetyl-Aspartate and N-acetyl-aspartyl-glutamate) for neuronal integrity, Cr (creatine and phosphocreatine) for cellular energy metabolism; Cho (choline, phosphocholine, and glycerophosphocholine) for membrane turnover and myelin, and myo-inositol (mIns) to measure glial cells (4,5). Accumulating evidence suggests these metabolites are altered in patients with schizophrenia (6), but the results are highly inconsistent in distinguishing groups of schizophrenia cases versus comparisons groups (7)(8)(9)(10).
Using a novel multi-voxel, whole hippocampal 1 H-MRSI to assess metabolite concentrations in the entire bilateral 3-dimensional structure (as opposed to single voxel approaches), our prior study demonstrated only elevated Cr concentrations [Cr] signi cantly distinguish the group of cases from the control groups (11). Notably, concentrations of Cho and NAA in this study were signi cantly more variable within the group of cases than controls, consistent with heterogeneous underpinnings among schizophrenia cases.
The current study examined if variability in [Cho] and [NAA] was related to the harboring of rare variants for TRIO or its interaction partner, ARMS/KIDINS220 compared to those only harboring rare variants in other genes for psychosis. Our panel included the zinc transporter SLC39A13, which was identi ed as harboring a de novo mutation in a sporadic case from an Israeli Cohort and replicated in a New York study, as well as other genes associated with schizophrenia, some of which are also associated with Alzheimer's Disease.

Methods
The imaging and subsequent genetic study were approved by the Institutional Review Boards of NYU and Bellevue Hospital Center and all subjects provided written informed consent. Participant assessments and 1 H-MRSI procedures were previously described in 19 schizophrenia cases (11), with nine of them further agreeing to participate in genetic studies. Targeted exome sequencing of 38 CNS signaling genes was done by targeted exome capture, also previously described (3,12). Rare or novel variants of these genes were identi ed according to Genome Aggregation Database (gnomAD v 2.1.1) and 1000Genomes Project databases (13,14).  (11). All schizophrenia cases were stable and maintained on steady medication regimens for at least a month prior to the imaging study.

Page 4/16
The nine participants' ages (mean = 41.2 years; SD = 10.3, 5 women, 4 men), disease duration (mean = 19.9 years; SD = 11.9), genetic information, hippocampal [Cho] and [NAA], are compiled in Table 1, identi ed by #1 -#9, and sorted in descending order of their [Cho]. Alongside each gene are annotations indicating if it is associated through a genome-wide association study (GWAS) with (a) schizophrenia or bipolar disorder, (b) Alzheimer's disease (AD) or other dementias, and (c) autism or intellectual disability. Strikingly, each of the nine subjects harbored at least one rare (minor allele frequencies (MAF) ≤ 1%) and/or ultra-rare (MAF ≤ 0.1%) missense coding variant or had a deletion in a coding region of a gene.

Discussion
Schizophrenia cases with rare missense coding variants in Trio Rho Guanine Nucleotide Exchange Factor (TRIO) or its interaction partner ARMS/Kidns220, had signi cantly lower [Cho] than other schizophrenia cases in this series and compared to the healthy control group in the parent study (11). TRIO is associated with autism and intellectual impairment in addition to schizophrenia (15)(16)(17), and this analysis suggests that these cases have signi cantly reduced hippocampal myelin. If con rmed, the ndings indicate a distinct pathophysiology for some schizophrenia cases, presumably some with comorbid ASD, indicating the need for distinct approaches for prevention and treatment.
TRIO codes for a rho-guanine nucleotide exchange factor, a major regulator of neuronal development downstream of neurotrophin signaling and also plays roles in cell adhesion, neuronal spine formation and other signaling pathways (18). Its protein is involved in the formation of a ternary complex with the neurotrophin p75 and SORCS2. The dissociation of TRIO from this ternary complex occurs upon binding of the proneurotrophin pro-NGF to the p75-SORCS2 complex, which leads to growth cone retraction in hippocampal neurons (19). Lower [Cho] in these cases may re ect reduced dendritic spine assembly and neurite outgrowth in the hippocampus, leading to less membrane turnover. The ARMS/KIDINS220 rare variants impacted the region between transmembrane domains 2 and 3. These transmembrane regions facilitate the functional interactions of the gene with tropomyosin receptor kinase (Trk) receptors. It promotes prolonged mitogen-activated protein kinase (MAPK) signaling, enhanced neuronal maintenance, and modulates spine formation.
Differences in [NAA], which is speci c for neuronal integrity, did not distinguish subjects with and without rare TRIO or ARMS/KIDINS220 variants. Lower [NAA] were only previously demonstrated in a metaanalysis for children with ASD, but not in adults (20). A meta-analysis of antipsychotic naive schizophrenia cases also found no signi cant group level differences in hippocampal [NAA] compared with controls (21). Nonetheless, there were two persons in this series with signi cantly reduced [NAA] than the other cases. Each of them harbored rare variants in SLC39a13 (zinc transporter 13) and DISC1 ("disrupted in schizophrenia 1"), which regulates adult neurogenesis, controlling dendritic development and synapse formation. It was one of the earliest identi ed mental-illness genes using linkage analysis and is now well-replicated for association with schizophrenia (22). It codes for a scaffold protein that is highly expressed in the hippocampus and is responsible for multiple aspects of neurogenesis, perhaps through Wnt/β-Catenin signaling (23). Both DISC and SLC39A13 are essential for neurodevelopment and have behavioral phenotypes. SLC39A13 transports vesicular Zinc to the NMDA receptor where Zinc has an inhibitory action. The gene is associated with the BMP/TGFβ pathway that is differentially expressed in pyramidal neurons in schizophrenia cases from post-mortem studies. SLC39a13 is signi cantly associated with alcohol use disorders in a GWAS study of European Americans (24).
Two cases in this series had very elevated [Cho] which is essential for the synthesis and turnover of the membrane phospholipid bilayer [and predicted more severe psychotic and manic symptoms in this cohort (25)], suggesting suggests increased membrane turnover and demyelination/remyelination. The gene variants in these cases included rare/novel mutations in genes associated with dopamine regulation, including the endocytosis related genes (SORT1, SORCS2), which are also associated with AD (26), protein tyrosine phosphatase (PTPRG) and its related synaptic scaffolding molecule (MAGI2), and DISC1, the latter three of which are identi ed in association with schizophrenia (27)(28)(29). SORT1 promotes apoptosis by endocytosis of the pro-neurotrophic factors pro-BDNF and pro-Nerve Growth Factor (pro-NGF). SORCS2 interacts with pro-BDNF and is involved in growth cone collapse, Schwann cell apoptosis and dopaminergic hyperinnervation (19). PTPRG has trans-synaptic actions, is the mRNA target of FMR1 and regulates neurotransmission. MAGI2 is a synaptic scaffolding molecule which maintains the glutamate receptor subunits of AMPA receptors. The genetic variants in these genes support our nding that elevated [Cho] is signi cantly associated with the severity of psychosis, a required symptom domain for the diagnosis (25).
Indeed, case #1, with the highest [Cho], also harbored a rare SORCS2 variant. In addition to its association with AD (above), SORCS2 is a converging point of linkage, genetic association, and GWAS ndings on chromosome 4p for bipolar disorder and schizophrenia (30). Its de ciency in a mouse model is accompanied by reduced NMDA-dependent hippocampal plasticity, impaired pre-pulse inhibition, as well as stress sensitivity, risk taking and stimuli seeking (31). Another high [Cho] case harbored a SORT1 variant, a member of the VPS10-related sortilin family which in uences dendritic spine formation and apoptosis through neurotrophin signaling, acting as a receptor for pro-BDNF and pro-NGF. It is involved in a host of biological processes-including glucose and lipid metabolism, cardiovascular conditions and AD-contributing to the overlap of diabetes and AD risk. The case with the rare SORT1 variant also harbored a rare DISC1 variant. Notably, a DISC1 knockout in rats shows disrupted morphogenesis of adult-born hippocampal neurons in the dentate gyrus. Another variant among the three cases with increased [Cho] were OXTR, which modulates the frequency of interneuron spiking in the hippocampus in addition to social bonding (32), and the neuronal pentraxin gene (NPTX2). The latter is associated with excitatory synaptogenesis, AMPA receptor aggregation, and the onset of the critical period. It is a direct transcriptional target of BDNF and is key to the BDNF-mediated modulation of glutamatergic synapses and mossy ber plasticity (33). NPTX2 has reduced expression in mouse models of AD and is reduced in the cerebrospinal uid (CSF) of persons with AD in association with cognitive de cits and hippocampal volume reduction (34).
The three cases with intermediate [Cho] harbored several potentially in uential variants that could act through other pathways. One had two rare missense-coding variants in the FAAH gene, which encodes the degradation enzyme for the endogenous endocannabinoid anandamide, which is highly expressed in the hippocampus where it modulates short-term synaptic depression (35). Reduced activity is associated with working memory de cits in a rat model (33). The mouse knockout displayed increased numbers of activated microglia and cytokines (IL-6 and IL-1β) predicting an in ammatory state (35,36). Another case carried the greatest number of rare variants and a unique novel mutation in DISC1 (I798T) that localized to the region interacting with Lis1, which is related to neurodevelopment (37). Also in this group were rare variants of SORCS2, a proneurotrophin receptor essential for BDNF and neurogenesis, and of IL1A, which is linked to schizophrenia in numerous genetic studies (38). IL1A is a cytokine that interacts with numerous pathways relevant to psychosis, including transforming growth factor beta (TGF-β), MAPK and Protein Kinase B (Akt) signaling. These ndings support the contention that de cient myelination of hippocampal parvalbumin GABAergic interneurons could underlie the pathophysiology of schizophrenia (39). Reduced numbers and function of GABAergic interneurons are well supported in models that bridge hippocampal pathology with schizophrenia (40). Their reduced function increases the activity of the far more numerous NMDA glutamatergic neurons, proposed to produce psychosis by deregulating dopamine-induced ring (41,42).
These ndings support the use of 1 H-MRSI concentrations of [Cho] for the treatment selection and monitoring of potentially distinct pathologies for schizophrenia. But notably, the genetics of schizophrenia are complex. As only a fraction of the population attributed risk is explained by any of the hundreds of common gene variants identi ed through GWAS, we have focused on the novel and rare gene variants. We propose that rare missense coding variants could plausibly be more important in explaining risk and particular illness features, e.g., symptom severity, cognitive involvement, and medical and psychiatric comorbidity (12,43).

Limitations
As a rst-of-its-kind study, it has two main limitations. First, it is a rather small cohort comprising patients with varying disease durations and treatments. Second, the approach of targeted exome capture on a set of pre-selected genes does not have the informational bene t of whole exome sequencing. Consequently, we could not determine the exact effect-size of these genes in relation to information that could be derived from whole exome variants on the schizophrenia-related pathologies. However, we are convinced that the co-occurrence of rare missense coding variants and deletions in the gene, in combination with the observed variability in [ΝΑΑ] and [Cho], point the way to a novel categorization method of patients and thereby a method of dissecting schizophrenia into biologically relevant subtypes.

Conclusions
Rare variants in genes associated with autism (TRIO or its interaction partner ARMS/KIDINS220) were identi ed in four of nine cases with schizophrenia (44%) undergoing multi-voxel 1 H-MRSI of their entire 3dimensional hippocampus and targeted exome sequencing. These preliminary ndings, if replicated, suggest that a person's speci c 1 H-MRSI metrics may be employed to select and monitor treatments for psychosis. Speci cally, those with 1 H-MRSI evidence of in ammation, for example, may bene t from antiin ammatory treatments. Further research may show that certain metabolic patterns suggest disruptions in certain genetic pathways that drive pathophysiologies indicating targeted treatments; for instance, patients with rare missense coding SNVs in the zinc transporter SLC39A13 may bene t from supplementation.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Hippocampus MRI and 3D 1H-MRSI of the entire structure: Position, Size and Analysis. Top: Axial (a) coronal (b) and sagittal (c) T1-weighted MRI from a 23 year old female SCZ patient (#3 in Table 1), superimposed with the 9×6×2 cm3 (LR×AP×IS) VOI and 1H-MRSI grid (thick and thin white frames) and the left hippocampus outline (transparent yellow on a). Yellow arrow on b and c indicates the level of a. Bottom, left, d: Real part of the 9×6 axial 1H spectra matrix from the VOI on a on common frequency and