Abinaya Chandrasekaran, Pia Jensen, Fadumo A. Mohamed, Madeline Lancaster, Michael E. Benros, Martin R. Larsen, Kristine K. Freude

First published: 25 August 2021

https://doi.org/10.1002/stem.3447

Abinaya Chandrasekaran, Pia Jensen, and Fadumo A. Mohamed contributed equally to this study.

[Correction added on 6 October 2021, after first online publication: Copyright statement has been updated.]

Funding information: Innovationsfonden, Grant/Award Numbers: BrainStem, NeuroStem; Lundbeckfonden, Grant/Award Number: Developnoid

Abstract

Schizophrenia (SCZ) is a severe brain disorder, characterized by psychotic, negative, and cognitive symptoms, affecting 1% of the population worldwide. The precise etiology of SCZ is still unknown; however, SCZ has a high heritability and is associated with genetic, environmental, and social risk factors. Even though the genetic contribution is indisputable, the discrepancies between transcriptomics and proteomics in brain tissues are consistently challenging the field to decipher the disease pathology. Here we provide an overview of the state of the art of neuronal two-dimensional and three-dimensional model systems that can be combined with proteomics analyses to decipher specific brain pathology and detection of alternative entry points for drug development.

Significance statement

The present study stresses the importance to implement complex three-dimensional brain organoids (BOs) derived from human induced pluripotent stem cells. Those BOs should ideally have contributions from different brain regions as seen in so called assembloids where, for example, cortical and thalamic organoids are fused. Moreover, it points out that protein and posttranslational modifications of proteins could be more relevant for disease development and progression. This is illustrated using the example of schizophrenia, a neurodevelopmental disease with highly diverse risk gene contribution. It is envisioned that such proposed approaches are highly relevant for complex neurological diseases.

1 INTRODUCTION

Schizophrenia (SCZ) was first described by Emil Kraepelin as dementia praecox in 1883, as it was thought to be an early form of dementia, a generative disease, and then termed SCZ by Eugen Bleuler in 1908. The precise etiology of SCZ is still unknown but SCZ is by many described as a neurodevelopmental disorder with typical onset in early to mid-twenties with most having had prior psychiatric symptoms. SCZ has a high heritability, and is associated with a broad range of particularly genetic and environmental risk factors. However, research into the biological underpinnings of SCZ is hampered by the lack of direct investigations of the brains in living patients with SCZ. Prior studies have mainly been performed on peripheral biological samples, cell cultures, and postmortem brain samples by measuring mRNA using transcriptomics technologies, or with indirect brain imaging. Selected studies have investigated the proteome in postmortem brains and patient tissues.¹ Transcriptomics rarely reflect the protein content in a cell and the current disease models are either challenged by immaturity of neurons (cell cultures) or reflect end-point stages of disease (postmortem brains). Thus, proteomic analyses of cerebral organoids have the potential to advance the current knowledge into brain mechanisms and disease progression involved in SCZ. Those investigations allow assessment of neural connectivity, cell signaling, posttranslational modifications (PTMs), and protein-protein interactions in cerebral organoid models of SCZ.

We will first discuss the current evidence on brain pathology and genetic findings in SCZ, followed by critical evaluations of current two-dimensional (2D) induced pluripotent stem cells (iPSCs) models, leading to the potential of using brain organoids for SCZ research with a specific focus on proteomics and PTMomics.

2 BRAIN PATHOLOGY IN SCZ

SCZ has been associated with morphological alterations in several major brain regions. Those structural abnormalities include: ventricular enlargement, whole brain volume reduction in gray and white matter in the network of the frontal cortex, temporal, thalamic, ventral striatum, and other subcortical regions.^2, 3 Among those changes ventricular enlargements are the most consistent changes even though it has been reported that this is secondary to other underlying changes within the cortex.⁴ Additionally, changes in the cortical gray and white matter have been observed, which has been suggested to even precede disease onset.⁵ Brain imaging studies of SCZ patients indicate that SCZ in part is a progressive neurodevelopmental disease affecting both gray and white matter within the cortex, thalamus, and striatum with considerable heterogeneity.^2, 5

3 GENETIC CONTRIBUTIONS

In line with the trajectory of morphological changes affecting different brain regions, the cumulative effect of susceptibility genes and environmental insults during early neurodevelopment seems to initiate neurophysiological changes and thereby the disease course in SCZ. The latest example of such efforts is a large-scale genome-wide association study (GWAS) by the “Schizophrenia Working Group of the Psychiatric Genomics Consortium,” which has identified 128 significant single nucleotide polymorphisms (SNPs) for SCZ.⁶ This work has further identified that in addition to common genetic associations, rare variants associated with SCZ are further linked to rare copy-number variations (CNVs), since deletions at 22q11.2 substantially increase the risk of SCZ.⁶ Researchers have placed a large effort into identifying susceptibility genes. However, many more genes besides those identified in this study are known to be involved in SCZ. This is reflected in the polygenic risk score for SCZ capturing the vast majority of SCZ common risk alleles with small effect sizes, underlining the polygenic nature of SCZ. The most researched susceptibility genes include Dystrobrevin Binding Protein 1 (DTNBP1), Neuregulin-1 (NRG1), Disrupted in schizophrenia-1 (DISC1), and Dopamine D2 receptor (DRD2).⁷

DTNBP1 encodes the protein, Dysbindin-1, which executes essential functions at synaptic terminals. Linkage analyses have revealed a strong association between SCZ and the chromosomal locus 6p22.3, which includes DTNBP1. SCZ patients with genetic variations in this locus exhibit impaired synaptic connectivity and decreased spine density as evident from analysis of postmortem brains.⁸ More evidence stems from several SNPs located in different parts of DTNBP1, associated with reduced hippocampal and prefrontal gray matter volumes and cognitive impairment.⁸ Furthermore, a SCZ postmortem study has revealed that expression of DTNBP1 mRNA is decreased in the dorsolateral prefrontal cortex, midbrain, and hippocampus.⁹ Additionally, changes in Dysbindin-1 protein levels have been correlated to dysfunction of glutamatergic signaling potentially contributing to the cognitive deficits seen in SCZ.¹⁰ Reduced expression of DTNBP1 has furthermore been linked to increased expression of DRD2 at the synapses in mice models and cell cultures.¹¹ DRD2 regulates synthesis, storage, and release of dopamine. Increased levels of DRD2 have been linked to the psychotic symptoms observed in SCZ¹² and several SCZ risk SNPs are associated with DRD2.¹³

NRG1, an important mediator of neurodevelopment, has first been identified as a SCZ candidate gene via linkage and haplotype analysis.¹⁴ NGR1 plays an essential role in early fate determination, differentiation, migration, and survival of neuronal cells, promotion of synaptic plasticity, and selective increase in the expression of other neuronal transmitter receptors. Moreover, elevated levels of NRG1 signaling impairs synaptic plasticity in SCZ patients¹⁵ and postmortem studies of SCZ brains have shown that NRG1 mRNA and NRG1 protein are increased in the prefrontal cortex.¹⁶ Lastly, it has been proposed that the NRG1 signaling effect on neuronal development and plasticity may compromise cortical and hippocampal function.¹⁵

DISC1 was first identified in a large Scottish family with balanced chromosomal translocation, which was subsequently substantiated with SNPs in affected individuals.¹⁷ DISC1 is involved in multiple cellular processes during fetal and postnatal brain development as a multifunctional scaffold protein. Some of the interaction partners of DISC1 implicated in those developmental processes regulate synaptic function, neuronal signaling, neuronal migration, and neuronal progenitor proliferation.¹⁸

While GWAS studies can identify correlation between SNPs and SCZ, transcriptome-wide association (TWAS) studies are more powerful to determine actual expression changes in genes correlated to SCZ. Differentially expressed genes identified by TWAS include gamma-aminobutyric acid type a receptor subunit alpha2 (GABRA2), chloride voltage-gated channel 3 (CLCN3), and double C2-like domain-containing protein alpha (DOC2A).^19, 20 GABRA2 and CLCN3 are both involved in inhibitory GABA mediated neuronal signaling and their expression changes might contribute to imbalances in excitatory glutamate and inhibitory GABA signaling. While reduced DOC2A expression has been linked to hypofunction at glutamatergic synapses.

In conclusion brain imaging, postmortem studies have been as well as GWAS and TWAS studies have been instrumental to identify risk genes, but their precise contribution to disease development still needs further evaluation. Hence, it is crucial to develop human-specific and disease-relevant models of SCZ, such as human induced pluripotent stem cells (hiPSCs), to clarify the neurodevelopmental process leading to SCZ.

4 2D IPSCs MODELS OF SCZ

Since the revolutionary development of hiPSC with the capacity to differentiate into any given cell type, several studies have been implementing 2D hiPSC derived neuronal cultures to study SCZ. These 2D hiPSC models include genetic abnormalities, such as 22q11.2 deletion, mutations in DISC1, patients with a family history of SCZ or childhood onset SCZ and some researchers implemented CRISPR/Cas9 gene editing to validate the genetic contributions of their hiPSCs models.^21-23 Those studies differentiated patient SCZ hiPSCs into various neuronal subtypes, such as hippocampal and cortical glutamatergic- and GABAergic neurons, as well as dopaminergic neurons.^21, 22, 24 Neural progenitor cells and neurons can be differentiated from SCZ hiPSCs with no apparent differentiation defects, compared with their respective controls. However, mature neurons derived from SCZ hiPSC display decreased connectivity, fewer neurites' and reduced expression of synaptic markers. Intriguingly, gene expression profiles of SCZ hiPSC neurons revealed increased levels of NRG1.²² These in vitro findings support that increased expression of NRG1 is associated with dysregulation of synaptic plasticity in brains derived from patients with SCZ. Additional extensive gene expression analyses revealed further causative connections between SNPs identified through GWAS and expression levels of the affected genes. One of such genes with altered expression in SCZ hiPSC neurons is the pivotal gene DISC1.²² Besides altered expression, DISC1 mutations have been correlated to synaptic dysregulation and altered proliferation of forebrain neurons.²³ Taken together it is obvious that SCZ is complex and multifactorial, with different combinations of susceptibility genes playing a key role in disease development. The 2D hiPSCs studies recapitulate the molecular and cellular processes related to SCZ and have been instrumental to provide further neuron type specific information for SCZ in a human-specific context.

4.1 Advantages and limitations of using 2D iPSCs models

One advantage of the 2D system is the simplicity of downstream analyses allowing easier drug and compound screens.²⁵ However, a clear limitation of the 2D hiPSCs models is their poor resemblance, complexity, and functional relationship with endogenous neurons in the brain. 2D cell cultures are monolayers of specific neurons on artificial matrices. Consequently, this environment decreases the impact of neuronal activities and functionalities due to the lack of natural biophysical and biochemical factors.²⁶ Furthermore, 2D models do not mimic the complex nature of brain tissues as it lacks interaction between the diverse cells in the brain, long distance migration, oxygen diffusion, and waste removal dynamics²⁷ (Figure 1). Additional limitations of the system can be seen in cell morphology, myelination, survival, proliferation, and differentiation.²⁸ Importantly, 2D models have been shown to have reduced gene and protein expression when compared with three-dimensional (3D) models.²⁷ This is reflected in transcriptomic analysis demonstrating significant upregulation of genes in 3D models compared with the 2D models.²⁹ Additionally, mature neurons with longer neurites' and higher synaptic density are more abundant and present at earlier time points in the BOs compared with models²⁸ (Figure 1). Even though 2D models have been instrumental in deciphering aspects of cellular disease pathology they are somewhat limited in elucidating the complex interplay of different neuronal subtypes. More complex network analyses with more mature neurons can be better addressed in 3D neuronal models.

FIGURE 1

Open in figure viewerPowerPoint

Comparison of two-dimensional (2D) and three-dimensional (3D) hiPSCs neuronal cultures. 2D cultures are a valuable method for brain cell-type specific studies. The main limitation of the 2D system is the lack of a natural 3D environment. In vivo cells are surrounded by a variety of brain cells and extracellular matrices in a 3D fashion. The monolayer model of 2D neuronal culture does not allow oxygen diffusion and waste removal.²⁷ In a natural 3D, environment neurons can be myelinated and create high synapse density while 2D cultures differentiate into unmyelinated neurons with reduced synapse density. Another advantage is the increased gene and protein expression profiles in 3D cultures, probably linked to advanced neuronal maturity²⁸

Another advantage, besides advanced connectivity and maturity of neurons, is the possibility to maintain 3D BOs for extended time periods to accomplish maturation milestones. Examples of such long-term BOs reach 300 days in culture to recapitulate human neural development³⁰ comparable to postnatal primary human astrocytes differentiation.³¹ These examples highlight the importance for not only long-term maturation of BOs, but also the need to introduce brain cell complexity to fully represent human cellular models for neurodegeneration.

5 CEREBRAL ORGANOIDS AND THE POTENTIAL FOR SCZ RESEARCH

Recent progress in hiPSC derived 3D organoid models hold great potential for producing brain-like structures, including cellular distribution, physiological structure, electrophysiological activities, and neuronal networks.³² So far, various methods to generate organoids have been reported with different structural complexity³¹ and cell diversity.³³ For this reason, brain organoids (BOs), also referred to as cerebral organoids, have become a unique model to explore the mechanisms of neurological disorders (Figure 2). BOs have been implemented in many studies investigating abnormal human brain development, such as congenital brain malformation and neurological disorders, including age-related neurodegenerative diseases (reviewed in Reference 37). Some studies have used single-cell RNA sequencing (scRNAseq) to profile large numbers of cells from both whole-brain organoids and patterned organoids, thereby proposing an opportunity to study species-specific cell types.³⁴ Recently, such studies have revealed preferential differentiation of NPCs into GABAergic neurons and reduced WNT signaling, providing insights into alteration of inhibitory and excitatory brain functions in SCZ.³⁸ BOs have shown varying degrees of neuronal activity and synaptic connectivity. For example, oscillatory Ca²⁺ waves synchronously activating neurons over long distances have been recorded in mouse BOs, similar to early postnatal in vivo mouse cortex activity.³⁴ More recently, a new strategy allowed the formation of circuits with functional neuronal output by culturing BOs in an air-liquid interface system.³⁰ Collectively, this validates that BOs have the ability to establish spontaneously active neuronal networks.

FIGURE 2

Open in figure viewerPowerPoint

Advantages of human brain organoids (BOs). Human BOs can be investigated with diverse approaches. BOs can be exploited to model organ development by enhancing neurodevelopment and by enriching coculture interactions. BOs can also be used for transcriptomic studies³⁴ and increasing interest are placed on vascularization^35, 36

One central aspect in optimization of BOs is whether cell fate should be aided through the addition of exogenous morphogens and signaling molecules, or not guided at all. Unaided BOs, derived from hiPSC, spontaneously aggregate and rely fully on intrinsic morphogenesis and differentiation capabilities.³⁴ One main limitation of the unguided approach is that hiPSC can randomly differentiate into non-ectodermal cell types, for example representing mesodermal cells.³⁴ Furthermore, this study also highlights high variations in the different numbers of glial and neuron subtypes in different batches of differentiated organoids. Although this approach offers a unique opportunity to model the interactions between different brain cell types, heterogeneity between brain organoids remains a major challenge. In contrast, guided BO differentiation methods require external supplementation of patterning factors to induce hiPSCs to differentiate toward desired lineages. Guided BO differentiation approaches contribute to specific brain region identities with minimal heterogeneity.³¹ The tradeoff between complexity and heterogeneity is an important feature to consider. As mentioned by Lancaster,³⁹ a more homogeneous system would be needed for drug testing while a heterogenous system would remarkably display complexity and highly expanded regions of different identities.

Currently, our insight into structural and cellular deviances that manifest in early brain development in SCZ brains is very limited. Following the pioneering work by Brennand et al,²² the number of hiPSC-based studies in SCZ has increased progressively. For example, BOs derived from SCZ hiPSCs displayed decreased proliferation activity of neurons and reduced expression of FGFR1 protein in cortical cells, accompanied by the loss of nFGFR1 signaling.⁴⁰ Blocking FGFR1 signal with the antagonist PD173074 in control organoids caused cortical growth arrest similar to what was observed in SCZ derived BOs. Overall, this study shows that activation of FGFR1 signals in cortical neurons could have the potential to prevent developmental abnormalities. Along this line, Sarkar et al²⁴ advanced a directed differentiation protocol via 3D BOs for hippocampus-patterned NPCs that closely mimics in vivo development. The hiPSC-derived DG-CA3 BOs displayed reduced spiking frequency compared with control DG-CA3 BOs. Another study examined BOs derived from SCZ with DISC1 mutation. The BOs with DISC1 mutation were morphologically distinct from control organoids and displayed over-activation of the WNT signaling pathway.⁴¹ In summary, these 3D models have the potential to facilitate long-term cell-cell contacts and could even be used to study accumulation of pathological protein aggregates.

6 PERSPECTIVES AND CHALLENGES OF ORGANOIDS TO STUDY SCZ

Cellular and molecular dysfunction in SCZ has traditionally been studied by examining either postmortem patient brain tissue or through animal models. Postmortem tissue is limited and RNA, DNA as well as epigenetic markers are rapidly degraded. Animal SCZ models rely either on targeted brain lesions or on manipulating a single gene associated with the disorder.⁴² Therefore, hiPSC-derived BOs offer a distinctive opportunity to investigate the full genetic landscape contributing to SCZ while monitoring neural development. BOs derived from SCZ hiPSCs have revealed differences in NPC proliferation affecting the ventricular zone (VZ) being reduced to one layer as opposed to controls with 2 to 3 layers of proliferating NPCs in the VZ.⁴⁰ TBR1 is a critical transcription repressor during cortical development and cortical organoids from patients with SCZ showed a reduction in TBR1-positive neurons.⁴² These studies give mechanistic insights into the cellular and structural alterations affecting subcortical neurogenesis, potentially impacted during early brain development in SCZ. Moreover, BOs allow to examine spontaneous network activity and have the ability to perform activity-dependent synaptic pruning relevant for neurodevelopmental disorders,⁴³ making them ideal pre- and postnatal in vitro models to study such deficiencies in SCZ.

In spite of having many advantages, BOs are still advancing technologies One of the current concerns in organoid technology is the variability of the methods. Furthermore, tissue maturation and vascularization are currently the main limitations associated with BOs. Recently, efforts have been made to achieve vascularization in human brain organoids.^35, 36 Those efforts are tremendously important since many brain disorders are caused or contribute to failures of the blood brain barrier (BBB). Moreover, drug screening in BOs would be much more reliable and meaningful with the possibility to mimic a blood compartment and BBB within the BOs. Other limitations of BOs are the diverse genetic backgrounds and various culture media that cause variability when comparing BOs from different research groups. Some of these limitations can now be resolved via coculturing, for example mesodermal cells with BOs in the attempt to create vascular networks.⁴⁴ Most BOs recapitulate only one specific brain region, which are useful to study the impact of genetic defects in specific disease-associated brain regions. For example, protocols to generate midbrain BOs successfully generate tyrosine hydroxylase positive dopaminergic neurons implicated in Parkinson's disease.⁴⁵ Likewise, mini bioreactors could produce forebrain BOs with a well-defined outer sub-VZ, demonstrating the presence of outer radial glial cells with appropriate molecular markers.³² Targeted inhibition of TGF-β signaling allowed for the generation of hippocampal-like BOs. Recently, BOs have been established representing distinct human brain regions, including midbrain,²⁵ hippocampus,⁴⁶ hypothalamus,³² and cerebellum.⁴⁷ Advancement of the system would be combining different BOs from diverse brain regions. To date, only few studies have generated such assembloids. One example is the fusion of cerebral and thalamic BOs, recapitulating the thalamocortical projections between thalamus and cortex.⁴⁸ The other example is the fusion of medial ganglion eminence and cortical BOs, allowing for the formation of advanced networks.⁴⁹ It remains a challenge for BOs models to combine different regions and cell types. This challenge is reflected in the usage of dorsalizing patterning cues promoting formation of an excitatory brain organoid, in which inhibitory interneurons are scarcely present.⁴²

In conclusion, organoid models represent one of the to date best platforms to investigate pathological aspects of the developing human brain and scRNA analyses have already validated specific genes relevant for disease progression and specific cell type pathologies. As a next step, it would be important to investigate the underlying molecular mechanisms for these diseases including alterations in cellular signaling, which can be efficiently studied using a protein-centric strategy employing proteomics and PTMomics for studying SCZ cerebral organoids vs controls.

7 PROTEOMICS AND PTMomics OF ORGANOIDS TO DECIPHER SCZ PATHOLOGY

High-throughput omics technologies have revolutionized biological research. However, most available omics data are presently those concerning the analysis of nucleic acids (genomics, epigenomics, and transcriptomics) whereas other omics technologies such as proteomics, lipidomics, and metabolomics are more limited. While analysis of gene transcription is routine, with the ability of amplifying nucleic acids due to PCR technology, other omics must rely on the presence of the endogenous molecules, often limiting their analysis. Transcript levels are repeatedly interpreted as indicators of gene expression; however, it is well-known that protein and mRNA levels only correlate between 40% and 60% at best⁵⁰ and that multiple layers of regulation of protein generation from mRNA exist. Thus, direct investigation of protein expression by analyzing their presence using proteomics technologies is mandatory for understanding biological processes in health and disease.

To investigate proteins on a global level, the method of choice is bottom-up proteomics, where proteins are cleaved into smaller peptides using specific enzymes, employing advanced tandem mass spectrometry (MS/MS) instruments, usually coupled with liquid chromatography (LC-MS/MS) to handle highly complex samples. Over the past decade there has been a remarkable improvement in MS-based proteomics due to development of new MS instruments with higher resolution, accuracy and speed which increase throughput, enabling analysis of proteins present at large dynamic range using very low amount of sample.⁵¹ Initiatives to target single cell proteomes are presented but are still in their infancy covering a relatively low fraction of the total proteome of a cell.⁵² In addition, developments in software and sophisticated sample preparation strategies, including extensive pre-fractionation strategies have advanced the field enormously.

An additional complexity of proteomics is the presence of PTMs, that is, various chemical groups, lipids, or small proteins that are enzymatically attached to proteins after they are made from the mRNA template. The presence and abundance of such PTMs are not encoded in the genome and can therefore not be predicted or analyzed from genomics data. However, approximately 5% of the total proteome in eukaryotic cells encodes enzymes participating in PTMs of proteins. These PTMs can change the proteins activity, location, and interaction and as such are extremely important for understanding biological processes and cellular signaling. It is currently believed that the complexity of the proteome exceeds that of the genome with more than a million proteoforms⁵³ originating from the 19 000 to 21 000 available genes in the genome,⁵⁴ as a result of alternative splicing and the multitude of PTMs, which can be added at any point during the proteins lifespan.⁵⁵ Large-scale analysis of PTMs is significantly more challenging than the analysis of proteins. Reasons for this are the often low stoichiometry of the modification, immense spatial distribution of PTMs in individual proteins together with challenges associated with the MS analysis itself (reviewed in Reference 56). Thus, the study of PTMs often requires specific tailored enrichment techniques to separate the PTM of interest from the non-modified peptides prior to LC-MSMS analysis. Such enrichment techniques only exist for a small subset of PTMs for example, phosphorylation,⁵⁷ glycosylation,⁵⁸ ubiquitination,⁵⁹ and lysine-acetylation.⁶⁰ Furthermore, due to the low relative abundance of PTMs, more material is needed in PTMomics compared with conventional proteomics analysis.

Currently, more than 450 PTMs are listed in the Uniprot database, the most common of them being phosphorylation, glycosylation, acetylation, and ubiquitination. Phosphorylation is highly dynamic and influences cell signaling by networks of kinases and phosphatases that regulate phosphorylation and dephosphorylation. Target sites for phosphorylation include serine, threonine, and tyrosine residues, with the latter being the least abundant. Glycosylation is one of the most structurally complex modifications. It occurs when glycosyltransferases add glycan chains of varying length and structure to the protein.⁶¹ Most often glycosylation takes place on proteins that are located in membranes or secreted by the cell. Glycosylation is involved in various biological processes including cell-cell interactions, cell migration, and responses to ligand stimulation and as such has a large influence on cellular signaling. Acylation is another broad class of PTMs including, among others, acetylation, malonylation, and succinylation. Protein acylation generally involves a reaction catalyzed by acyl-specific transferases that transfer an acyl group from energy metabolism intermediates such as acetyl-CoA, malonyl-CoA, and succinyl-CoA to lysines in substrate proteins, among these lysine-acetylation is the best studied.⁶² Most often proteins are modified at several locations with the possibility of cross-talking with other PTMs and thereby modulating the specific biological function of the targeted protein. For example, sialylated N-linked glycosylation have been shown to influence internal phosphorylation in glucose stimulated insulin secretion in pancreatic β-cells.⁶³ PTM enrichment workflows^64, 65 combined with isobaric peptide labeling, for relative quantitation, and fast and sensitive MS instruments provide a robust strategy for discovery-based quantitative PTMomics (Figure 3).

FIGURE 3

Open in figure viewerPowerPoint

Simplified mass spectrometry-based proteomics and PTMomics workflow using isobaric tandem mass tags (TMT) quantification. After protein extraction and enzymatic digestion peptides from each sample are labeled with TMT followed by combination of all samples for simultaneous posttranslational modifications (PTM) enrichment and subsequent pre-fractionation and liquid chromatography-tandem mass spectrometry of each enriched sample as well as the unmodified sample (all peptides not containing the PTMs). After protein and PTM-peptide identifications various bioinformatics tools are used for data interpretation

8 PROTEIN-CENTRIC INITIATIVES IN SCZ RESEARCH

To date, proteomics investigations in SCZ have focused on postmortem brain tissue, peripheral tissue, and body fluids (blood, plasma, CSF) (recently reviewed in Reference 66). As described previously, the number of iPSC-based cerebral organoid studies on SCZ are still limited and none of these include proteomics analysis. One study from iPSC-derived neurons in 2D cultures from four SCZ individuals and six controls used stable isotope labeling by amino acids in cell culture (SILAC) quantitative proteomic mass spectrometry analyses to investigate protein profiles. They found abnormal protein levels and gene expression (assessed by microarray analysis) related to cytoskeletal remodeling and oxidative stress as well as aberrant migration in SCZ hiPSC neural precursor cells.⁶⁷ Postmortem brain investigations have provided some understanding regarding altered protein levels in the orbitofrontal cortex and anterior cingulate cortex as well as subcellular domains such as synaptic structures^68, 69 in SCZ patients compared with controls. Notably, most of the proteins found altered in SCZ as compared with healthy controls in postmortem studies are located in the synapses highlighting the need for proteomics/PTMomics analysis as synaptic proteins are not locally synthesized but transported to the synapse.⁷⁰ As such, transcriptomics will not reveal information on the synaptic proteome. In addition, most of the protein functions and interactions in nerve terminals are controlled and modulated by PTMs to facilitate fast responses. Compared with postmortem studies hiPSC-derived cerebral organoids will enable dynamic investigations of changes in synaptic transmission both at the proteomic and PTMomics level.

Osimo et al performed a meta-analysis on synaptic loss in SCZ based on case-control postmortem studies on synaptic protein or mRNA levels in brain tissue. They found a significant reduction in synaptophysin in SCZ in the hippocampus, frontal, and cingulate cortices and reductions in SNAP-25, PSD-95, synapsin, and Rab3A protein levels in the hippocampus but inconsistency in other regions.⁷¹ However, only one study of the 36 included quantitative studies used mass spectrometry based proteomics in their analysis while the rest of the protein data came from Western blotting or ELISA. This study from Föcking et al used label free shotgun proteomics to identify differentially expressed proteins in the postsynaptic density enriched samples of anterior cingulate cortex in 20 SCZ and 20 control samples. Their findings provide insight into the contribution of the PSD to SCZ and suggested mechanisms, involving endocytosis, LTP, and NMDA receptor function.⁶⁹ In another study, Velasquez et al performed proteomic characterization of the synaptosome fraction of postmortem orbitofrontal cortex from eight SCZ patients compared with a pool of eight healthy controls free of mental diseases. They found imbalance in the calcium signaling pathway and proteins such as reticulon-1 and cytochrome c, related to endoplasmic reticulum stress and programmed cell death.⁶⁸ However, postmortem brain tissue suffers from extensive protein degradation, oxygen deprivation, free radical damage and has no biological activity. In addition, postmortem samples resemble end-point disease levels and cannot be used to assess changes during disease development.

Besides studies on glycosylation and lipidation very little is known about PTM patterns in SCZ related samples. Several studies have shown abnormal glycosylation patterns in postmortem brains from individuals with SCZ. Proteins involved in both excitatory and inhibitory neurotransmission exhibit altered glycans in the disease state, including AMPA and kainate receptor subunits, glutamate transporters EAAT1 and EAAT2, and the GABA_A receptor. In addition, enzymes involved in several pathways of glycan synthesis show differential expression in postmortem brains from individuals with SCZ (reviewed by Mueller and Meador-Woodruff⁷² and Williams et al⁷³). Studying phosphorylation is most suitable in a dynamic model system where samples can be collected repeatedly and protected from rapid dephosphorylation. As an example, Jaros et al investigated phosphorylation patterns in plasma samples (using IMAC enrichment and label free LC-MS/MS) and found around 60 phosphoprotein with altered phosphorylation patterns in SCZ patients compared with healthy controls.⁷⁴ However, the high level of phosphatase activity in blood is blocking for in-depth analysis of the plasma phosphoproteome.

To date very few studies on cerebral organoids have used proteomics.^75, 76 Nascimento et al analyzed human embryonic stem cell-derived cerebral organoids using proteomics and identified around 3000 proteins associated with different developmental stages being neural progenitors, neurons, astrocytes, and oligodendrocyte precursors.⁷⁵ They only investigated one differentiation timepoint at the proteome levels; however, future studies with higher throughput over an expanded time course would be important. In another study, Pellegrini et al used mass spectrometry to analyze fluid extracted from choroid plexus organoids and found a high degree of similarity in the protein pattern to real in vivo CSF samples.⁷⁷ Furthermore, they found that the proteins exclusively found in the in vivo CSF samples were most likely coming from the blood highlighting the usefulness of this organoid model as an isolated system to study CSF components.

9 FUTURE DIRECTIONS

As outlined above BOs are very powerful tools to study developmental brain disorders, such as SCZ. These organoids are superior in terms of producing tissue architecture that can capture snapshots during disease progression and have advanced mature states of neurons, highly desirable to understand synaptic dysfunction and neural network deficits observed in SCZ. In terms of gaining better insights into disease pathology and progression, it is of utmost importance to focus on a protein-centric approach in the subsequent characterization. Of specific interest are here, PTMs that are presently not included in most strategies for disease characterization. Identification of changes at the PTM level would lead to novel discoveries regarding understanding underlying disease mechanisms, discovery of new biomarkers for SCZ diagnosis and patient stratification and open up for novel pharmaceutical intervention studies. In summary, we believe that a combination of cerebral organoid modeling with protein and PTM analyses at divergent timepoints will be key to better understand SCZ progression and identify new angles for pharmaceutical intervention strategies.

ACKNOWLEDGMENTS

This work was supported by awards from: Innovation Fund Denmark (BrainStem: 4108-00008B & NeuroStem) (K.K.F.) and Lundbeckfonden (Developnoid: R336-2020-1113) (M.R.L.). All figures were designed using Biorender.

CONFLICT OF INTEREST

The authors declared no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

A.C., P.J., F.A.M.: writing of the manuscript, generated the figures, read and approved the final version; M.L., M.E.B., M.R.L.: contributed to the writing and editing of the manuscript, read and approved the final version; K.K.F.: writing of the manuscript, read and approved the final version.