Endosomal trafficking in schizophrenia

Abstract

Schizophrenia is a severe and heritable neuropsychiatric disorder, which arises due to a combination of common genetic variation, rare loss of function variation, and copy number variation. Functional genomic evidence has been used to identify candidate genes affected by this variation, which revealed biological pathways that may be disrupted in schizophrenia. Understanding the contributions of these pathways are critical next steps in understanding schizophrenia pathogenesis. A number of genes involved in endocytosis are implicated in schizophrenia. In this review, we explore the history of endosomal trafficking in schizophrenia and highlight new endosomal candidate genes. We explore the function of these candidate genes and hypothesize how their dysfunction may contribute to schizophrenia.

Introduction

Schizophrenia is a common and severe neuropsychiatric disorder characterized by both positive (delusions, hallucinations, disorganized speech, disorganized or catatonic behavior) and negative (diminished emotional expression or avolition) symptoms [1]. Schizophrenia affects about 0.25%–0.64% of the population of the United States [2–4]. There are negative consequences to unmanaged schizophrenia for both people with schizophrenia and their caregivers, but current therapeutics have side effects that lead some people to discontinue treatment. These facts highlight the need to elucidate specific mechanisms that underlie schizophrenia pathogenesis to develop more effective and targeted therapies.

Twin and other studies demonstrated that schizophrenia has an estimated heritability of ~80% [5]. These studies suggest that schizophrenia has a genetic component but not the source of genetic variation. Historically, candidate genes were identified from linkage studies with small sample sizes [6]. However, technological advancements have made sequencing the genomes of large numbers of people possible. These studies revealed that common variation, copy number variation (CNV), and rare loss-of-function (LoF) variation all contribute to schizophrenia. Rare LoF variation contributes a small but significant source of variation in schizophrenia, explaining about 0.274% of the overall liability [7]. Although identifying rare LoF variation has been challenging, recent progress has been made [8–10]. CNV explains about 0.85% of the variance in schizophrenia liability, and eight specific CNVs have been identified [11]. Finally, common variation explains about 3.4% of the variance in schizophrenia liability [12]. However, this variation only explains a small percentage of the overall heritability of schizophrenia, which underscores its highly polygenic and heterogenous nature. This missing heritability is most likely a result of rare variants that have not yet been identified and have much larger effects on risk [13]; larger samples sizes will certainly identify more rare variants.

Identifying candidate genes in schizophrenia

Genome-wide association studies (GWAS) identified common variation within 145 loci associated with schizophrenia [14]. However, many of these loci are within non-coding regions of the genome, which complicates assigning affected genes to these variants. Traditionally, non-coding variants are assigned to genes based on proximity and linkage disequilibrium (LD) using multimarker analysis of genomic annotation (MAGMA). More recently, Hi-C, a method used to detect 3D chromatin structure, has been utilized to link genome wide significant (GWS) loci to genes based on a physical interaction between the GWS loci and the gene. For example, Sey et al. linked genes to variants using Hi-C datasets from human adult dorsolateral prefrontal cortex and fetal developing cortex in a process termed H-MAGMA [15].

Gene candidates highlighted by H-MAGMA support a physical chromatin interaction with risk loci, but the effects of these interactions are not known. Quantitative trait loci (QTL) has been used to determine the effects of these genetic variants by quantifying how various traits, such as gene expression (eQTL) and splicing (sQTL) correlate with specific loci. Although eQTL provides support for single nucleotide polymorphisms (SNPs) as the cause for expression changes of a gene, there is no consensus on how to calculate effect size, or the magnitude of the change, in gene expression. Therefore, eQTL is unreliable in determining the magnitude of gene expression changes. Transcriptome-wide association studies (TWAS) can determine the magnitude of gene expression changes in schizophrenia. TWAS is similar to GWAS, but uses RNA-sequencing rather than DNA-sequencing, and, thus, can determine fold-changes in gene expression. These methods are summarized in Figure 1.

Figure 1. Functional genomic evidence can link candidate genes to GWS loci.

The majority of common variation in schizophrenia resides within noncoding regions of the genome, therefore, further support is needed to link a specific gene to GWS loci. Hi-C has been used to identify physical interactions between GWS loci and genes. eQTL and sQTL provide support that these GWS loci can explain gene expression and splicing changes, respectively. Finally, TWAS has been used to identify genes that are more or less expressed in people with schizophrenia than in neurotypical controls. We identified a number of membrane trafficking genes that are supported by Hi-C, eQTL, sQTL, and/or TWAS.

While these techniques provide valuable information for linking genes with GWS loci, these methods rely upon bulk Hi-C and RNA-sequencing and thus do not provide cell-type specific information. Attempting to interpret cell-type information from these data is generally biased, relying upon assumptions about cell-types by their expression of different markers, and thus may not represent the actual distribution of cell types within the brain. Therefore, single-cell RNA-sequencing is a critical next step in confidently determining which genes are affected by different GWS loci. Nonetheless, quantitative data like QTL and TWAS can provide support for specific gene candidates and affected biological pathways that lead to the cellular and physiological phenotypes in schizophrenia [16]. For example, Wang et al. integrated direct assignment, Hi–C interaction maps, QTL, and gene regulatory network data to highlight 321 “high-confidence” candidate genes (genes supported by at least two lines of evidence) for schizophrenia risk [17]. Interestingly, a number of these genes are predicted to be involved in membrane trafficking, particularly endosomal trafficking.

Endosomal trafficking was historically linked to schizophrenia

Cells contain numerous membrane-bound compartments responsible for the sorting, recycling, and degradation of protein cargo. Endocytosis transports molecules and proteins into the cell, whereas exocytosis inserts cargo into the plasma membrane and secretes cargo out of the cell. Endocytosed cargo is sorted at early endosomes. Cargo can be recycled back to the plasma membrane through rapid recycling compartments or slower recycling endosomes. Alternatively, cargo is sent to late endosomes and lysosomes for degradation. Transport carriers, including vesicles and elongated tubules, bud from and fuse with endosomes to provide both soluble and membrane-associated components to their target compartment [18].

Endosomal trafficking was implicated in schizophrenia prior to modern genomic studies. Dystrobrevin-binding protein 1 (DTNBP1), encoding the protein dysbindin-1, was linked to schizophrenia through linkage analysis [19]. Post-mortem examination suggested DTNBP1 mRNA and dysbindin-1 protein in the hippocampus and prefrontal cortex was reduced in individuals with schizophrenia compared to neurotypical controls [20]. Dysbindin-1 is a subunit of the biogenesis of lysosome-related organelles complex 1 (BLOC-1). BLOC-1 functions in the biogenesis of lysosome-related organelles (LROs), such as melanosomes [21].

Despite years of effort to explore how dysfunction of dysbindin and BLOC-1 contribute to schizophrenia pathogenesis, common variants of BLOC-1 subunits were no more associated with schizophrenia than controls in genomic studies of schizophrenia [22]. Does this mean that mutations in BLOC-1 do not contribute to schizophrenia? Not necessarily; rare variants of BLOC-1 components may exist that have not yet been implicated in schizophrenia.

Endosomal trafficking is a pathway of interest in schizophrenia

To identify membrane trafficking genes associated with schizophrenia, we compiled a list of 921 genes with gene ontology (GO) biological processes related to membrane trafficking, including both endocytic and exocytic pathways. We found 97 of 921 membrane trafficking genes were supported by both the adult and fetal H-MAGMA schizophrenia datasets. An additional 48 were supported by the adult H-MAGMA alone, and 28 were supported by the fetal H-MAGMA alone. In total, 173 genes were associated with schizophrenia GWS loci (Figure 2a). We compared the list of 921 membrane trafficking genes with eQTL and sQTL identified from adult human brain tissue [17]. We found nine genes that have significant schizophrenia-associated eQTL and 1 that has significant schizophrenia-associated sQTL. We further explored whether any membrane trafficking genes were identified in schizophrenia TWAS of human adult frontal cortex were supported by both eQTL and TWAS (TSNARE1, PHETA2, VPS45, PACSIN3, and SNAP91). These candidates are outlined in Figure 2b. While many of these candidates are considered housekeeping genes and would therefore likely be ubiquitously expressed, we wondered if any of these genes were enriched in particular cell-types in the brain. We data-mined single-cell RNA-sequencing data from human adult brain cortex [23]. The endosomal candidate genes tended to be more enriched in neurons than in glial cells (Figure 2c).

Figure 2. Candidate genes involved in endosomal trafficking are well supported in schizophrenia.

(a) 173 membrane trafficking genes are supported by either fetal and/or adult H-MAGMA (FDR adjusted p-value < 0.05). (b) 15 membrane trafficking genes are supported by eQTL, sQTL, and/or TWAS (QTL = H4 posterior probability >0.7 or TWAS = Bonferroni adjusted p-value < 0.05) and are listed with their gene name, Ensembl gene stable ID, and chromosome location (chr). If these genes were supported by the adult H-MAGMA, the FDR adjusted p-value and the number of SNPs (n SNPs) associated with that gene are shown. Blank cells suggest it was not significant. (c) Within-row normalized values of transcripts per million (TPM) per candidate gene. Rows are black if there were no data for that gene. Abbreviation: oligodendrocyte precursor cell (OPC).

A number of prioritized genes function at the late endosome and lysosome (Figure 3a). For example, the gene that is the third most significant of the membrane trafficking genes in the adult H-MAGMA is TSNARE1. TSNARE1 is additionally supported by eQTL and TWAS, which suggests that tSNARE1 is overexpressed in schizophrenia. TSNARE1 encodes a syntaxin-like SNARE protein called t-SNARE domaining containing 1 (tSNARE1). SNARE proteins are involved in membrane fusion. We recently characterized tSNARE1 isoforms that all contain a Qa-SNARE domain. However, the majority of tSNARE1 in the brain lacks a transmembrane domain, which suggests it likely functions as an inhibitory SNARE. tSNARE1 isoforms localized primarily to the late endosome. tSNARE1 expression delayed trafficking of the somatodendritic transmembrane protein Nsg1 into late endosomes and lysosomes. Nsg1 is endocytosed from the plasma membrane and trafficked primarily from the early endosome to the late endosome and lysosome [24,25]. tSNARE1 therefore likely functions as a negative regulator to early endosome to late endosome trafficking [26] (Figure 3b). Interestingly, TWAS suggests NSG1 is downregulated in schizophrenia, although there is no evidence for a common variant being responsible for this change. In fact, NSG1 is the only gene from our list of membrane trafficking genes that was supported by TWAS but not by Hi-C or QTL.

Figure 3. Schizophrenia-associated genes function in endosomal trafficking.

(a) Overview of endocytosis and exocytosis. Schizophrenia candidate genes that are supported by at least two lines of evidence are listed according to their known localization and function. If a candidate is supported by TWAS, the magenta arrow displays whether the gene is more or less expressed in schizophrenia than in neurotypical controls. Abbreviations: early endosome (EE), late endosome (LE), lysosome (Lyso), rapid recycling (RR), recycling endosome (RE), trans-Golgi network (TGN). (b) A model by which tSNARE1 likely negatively regulates early endosome to late endosome trafficking, where the absence of its transmembrane domain blocks membrane fusion. (c) SNAP91 regulates clathrin-mediated endocytosis in conjunction with AP2 and clathrin. (d) Model of how ESCRT complexes mediate late endosome intralumenal invagination.

Adaptor protein (AP) complexes are also implicated by the inclusion of AP complex subunits AP3B2 and AP5B1. AP complexes are heterotetrameric and have roles in the regulation of membrane trafficking. AP3B2 encodes a subunit of AP-3 that is specific to neurons [27]. AP-3 recognizes sorting signals on transmembrane proteins that dictate cargo be sent to lysosomes or synaptic vesicles [28]. For example, AP-3 recognizes LAMP proteins on early endosomes and directs them to late endosomes and lysosomes, thus playing critical roles in lysosomal cargo sorting [29]. Interestingly, there are significant sQTL for AP3B2 in schizophrenia, indicating that common schizophrenia variants alter the splicing of AP3B2. However, how splicing affects AP-3 neuronal expression and function is not known. AP5B1 encodes a subunit of AP-5. AP-5 localizes to late endosomes and lysosomes and is involved in a sorting step out of the late endosome [30].

SNAP91 is a gene of interest that encodes the protein SNAP91/AP180. SNAP91 is thought to function with the adaptor protein complex AP-2 in directly recruiting the v-SNARE, VAMP2, for clathrin-dependent endocytosis and recycling from the plasma membrane following exocytosis [31] (Figure 3c). At the presynapse, SNAP91 regulates the number of VAMP2 synaptic vesicles—presumably via endocytic retrieval [32]. Another prioritized candidate, PACSIN3, encodes PACSIN3 or syndaptin 3 in rodents. PACSIN3 is less studied than either PACSIN1 or PACSIN2, which have well described roles in synaptic vesicle and trans-Golgi vesicle transport. PACSIN3, however, localizes to the cell periphery and is involved in transferrin endocytosis [33], TRPV4 endocytosis [34], and glucose uptake [35].

We identified several vacuolar protein sorting (VPS) genes as candidates for schizophrenia risk: VPS4A, VPS37B, and VPS45. VPS37 and VPS4 are involved in the biogenesis of late endosomes in conjunction with endosomal sorting complex required for transport (ESCRT) complexes by promoting intraluminal vesicle biogenesis (Figure 3d). Specifically, ESCRT-0 binds to ubiquitinated cargo on endosomal membranes and recruits and transfers ubiquitinated cargo to ESCRT-I. VPS37 is a subunit of the ESCRT–I complex. ESCRT-I subsequently recruits ESCRT-II, which invaginates the membrane. ESCRT-II activates ESCRT-III. ESCRT-III promotes scission of the invaginated membrane in conjunction with the ATPase VPS4 [36]. VPS4 mediates ESCRT-III subunit exchange that drives its assembly and disassembly [37]. Besides biogenesis of late endosomes, VPS4 and ESCRT-III additionally function in a number of other membrane remodeling events [36]. VPS45 is a Sec1p/Munc18-like (SM) protein that has roles in trafficking to the early endosome and recycling back to the plasma membrane [38]. More recently, VPS45 was shown to function in early to late endosome maturation [39].

PHETA2 is a candidate that encodes the protein PHETA2 (also known as FAM109B/Ses2/IPIP27B). PHETA2 interacts with OCRL, an inositol phosphatase, and localizes to early endosomes, recycling endosomes, and the trans Golgi [40]. COG8 encodes a subunit of the conserved oligomeric Golgi (COG) complex, which regulates Golgi trafficking [41]. C17orf75 encodes a largely unstudied protein C17orf75, which forms a complex with WDR11 and ICP0 and localizes to the trans-Golgi network, where it is involved in tethering of vesicles [42]. FES encodes the protein Fes, a tyrosine-protein kinase whose role in membrane trafficking is not well described but colocalizes with Golgi and endosomal markers [43]. The rest of the prioritized candidates are predicted to be involved in membrane trafficking due to sequence similarity but are not well-studied, including TMED4 and ARL17.

How might dysfunction of endocytosis and membrane trafficking contribute to schizophrenia?

Several hypotheses may explain the etiology of schizophrenia, including alterations in synaptogenesis, synaptic plasticity, synaptic pruning, and synaptic function [44–49]. The brain undergoes a period of rapid synaptogenesis after birth that peaks around 2–3 years of age, after which a period of synaptic pruning, which removes unneeded synapses, extends through late adolescence. Synapses are strengthened or weakened based on how often they are used, a process termed synaptic plasticity. Ultimately, dysregulation of the dopamine system is thought to be central to schizophrenia pathogenesis, and recent evidence suggests that alterations in the circuits that control midbrain dopamine neurons may cause this dysregulation [50].

Several lines of evidence support the idea that schizophrenia is caused by altered synaptic connectivity, particularly at the post-synapse. First, neurons are associated with schizophrenia. Schizophrenia-associated genes are enriched in neurons [15,17,51,52]. Single cell RNA-sequencing data from human samples suggest that medium spiny neurons, pyramidal neurons in the hippocampus, pyramidal cells in the somatosensory cortex, and cortical interneurons are particularly correlated with schizophrenia [53]. Second, alterations in neuronal morphology are associated with schizophrenia; for example, there a decreased density of dendritic spines is observed [54]. Third, schizophrenia associated genes are associated with GO terms related to vesicle trafficking and the post-synapse [15,17,51]. However, the exact cause of the altered synaptic connectivity in schizophrenia is not well understood.

Membrane trafficking is essential for proper synaptic plasticity, pruning, and function (Figure 4). At the presynapse, the endosomal network regulates the readily releasable pool of synaptic vesicles, both by generating new synaptic vesicles or degrading vesicles that are damaged or unneeded. The endosomal network regulates receptor endocytosis and recycling at the postsynaptic membrane. Endosomal trafficking transports, sorts, and degrades a number of receptors at the synapse, such as AMPA receptors (AMPARs) [55,56]. The number of receptors on the post synaptic membrane dictates the strength of the excitatory or inhibitory signal that is propagated, therefore alterations in endocytic trafficking can have dramatic effects on postsynaptic function and plasticity. In line with this observation, Kim et al. optogenetically disrupted early to late endosomal trafficking and found that transient changes in the balance of endocytosis and exocytosis of AMPARs led to long-lived alterations in synaptic plasticity, presumably by sustained alterations to the ratio of receptors at the cell surface [57]. Whether degradation of these receptors occurs locally at spines or if receptors are transported to the soma for degradation is not understood. A number of studies describe polarized endosomal transport, where mature late endosomes and degradative lysosomes are localized closer to the cell body [25]. In support of local degradation at synapses, Goo et al. observed lysosomes that trafficked to the base of dendritic spines in response to activity [58].

Figure 4. Endosomal trafficking at the synapse.

A cartoon of a pyramidal neuron (top), which is a cell type associated with schizophrenia. Zoomed in region of a synapse (below) displays how endosomal trafficking plays multiple roles both at the pre- (synaptic vesicle trafficking) and post-synapse (receptor trafficking). Abbreviations: early endosome (EE), late endosome (LE), lysosome (Lyso), recycling compartments (R), trans-Golgi network (TGN), and synaptic vesicles (SV).

Endosomal trafficking is also critical for a number of developmental processes, which is another possible mechanism for schizophrenia pathogenesis. For example, proper endocytic control mediated by Rab5 and ESCRT contributes to proper axonal and dendritic thinning and pruning in Drosophila [59–61]. Dysregulation of axonal and dendritic thinning and pruning may contribute to the altered synaptic connectivity in schizophrenia. In conclusion, we have highlighted 15 schizophrenia gene candidates involved in endosomal trafficking. As endosomal trafficking is critical for processes thought to be disrupted in schizophrenia, understanding how these candidates contribute to schizophrenia pathogenesis should be further investigated.