The Nuclear Proteome of White and Gray Matter from Schizophrenia Postmortem Brains

Schizophrenia (SCZ) is a serious, debilitating, and incurable mental disorder that affects approximately 1% of the world's population [1]. The disease normally manifests between the end of adolescence and the beginning of adulthood [2] and is characterized by a range of cognitive, behavioral, and emotional dysfunctions. SCZ is the main cause of psychiatric incapacitation [3] and, although it is usually treated as a single disease, it is likely to be a spectrum of related disorders with distinct etiology, clinical presentation, response to treatment, and development [4]. Despite its high prevalence and severity, little is known about the biochemical mechanisms involved in its development or progression, and so there are few established molecular diagnoses or specific treatments. Thus, there is currently a great interest in obtaining new knowledge about the disease and, in this arena, proteomic analyses have already yielded promising results and opened up new avenues of research [5].

Proteomics is used to analyze the protein profile of a specific cell type, tissue or organism, or changes in specific proteins, using mass spectrometry as its main tool. Due to its capacity for profiling large numbers of proteins simultaneously, proteomics is currently one of the main techniques used to understand biochemical pathways and, consequently, multifactorial disorders such as SCZ [6]. In the last 2 decades, proteomics has contributed to a growing understanding of SCZ, and a number of such studies investigating this disorder have been published revealing alterations in several biochemical pathways of the central nervous system [5]. However, it is important to note that in proteomic analyses of whole tissues, it is often not possible to detect proteins that are present at low concentrations. This may occur due to the complexity of the sample and/or due to the presence of other proteins that are present at very high concentrations, which may obscure those of lower abundance. This could lead to an inability to detect proteins which have important roles in the disease process [7]. To circumvent this problem, the study of subproteomes appears as a satisfactory alternative.

Subproteomes are obtained by fractionating the proteins of a given sample into distinct groups, taking into account some specific criteria. Although there are several ways of fractionating the protein content of a sample, one alternative is the analysis of specific organelles. This type of fractionation is attractive for cellular proteome analyses, since the protein content of the organelles is less complex and represents a specific and directed set, which provides the opportunity of investigating entire protein networks to an extent that cannot be achieved using whole cell approaches [8].

In this study, we have used a protocol for separation of organelles in order to obtain samples enriched in cell nuclei. The nucleus is the largest organelle of most cells and occupies approximately 10% of its volume, although this varies according to the cell type [9]. The nucleus is the compartment where the genetic material of eukaryotic organisms is stored, and nuclear proteins constitute 10-20% of the total cellular proteins and exert important functions related to gene expression, transcriptional control, splicing, and generation of the final gene products [10]. Therefore, the importance of investigating the nuclear proteome for the understanding of any physiological or pathological process is clear, and few such studies have been carried out thus far in the field of psychiatric disorders such as SCZ.

As an added level of enrichment, we have analyzed nuclei obtained from the anterior temporal lobe (ATL) and corpus callosum (CC) regions of postmortem brains from patients and controls. Dysfunctions in the ATL have been implicated previously in SCZ, and it consists mostly of gray matter [11]. The CC is the largest white matter region of the brain, and there are morphological, electrophysiological, and neurophysiological studies showing significant involvement of this region in patients with SCZ [12,13,14]. Because the brain works through communication with and across the different regions, we jointly analyzed the ATL and CC nuclear proteins to provide further information on the potential role of dysfunctions of this organelle in SCZ.

CC and ATL samples were collected post-mortem from 12 patients who had suffered from chronic SCZ and from 8 healthy controls (Table 1). The patient samples were from the Nordbaden Psychiatric Center, Wiesloch, Germany, and the controls were from the Institute of Neuropathology, Heidelberg University, Heidelberg, Germany. Postmortem evaluations and procedures were approved by the Ethics Committee of the Heidelberg University Medical School, and both patients and controls gave written consent prior to death that their brains could be used for research purposes.

Nuclear proteins were obtained from the ATL and CC brain tissues according to the protocol of Cox and Emili [15]. In this protocol, each sample (20 mg tissue) was homogenized in 10 volumes of buffer containing 250 mM sucrose (Sigma-Aldrich, St. Louis, MO, USA), 50 mM Tris-HCl, 5 mM MgCl₂ (Sigma-Aldrich), 1 mM dithiothreitol (Sigma-Aldrich), 250 µg spermine, and 250 µg spermidine buffer (pH = 7.4), containing 1 tablet of protease cocktail inhibitor (Roche Diagnostics, Indianapolis, IN, USA) per 25 mL buffer. The homogenate was centrifuged at 100 g for 1 min at 4°C. The supernatant was discarded. Next, 5 volumes of the same buffer was added to the sediment. The homogenate was centrifuged at 800 g for 15 min at 4°C, and the supernatant was separated for further separation of the mitochondria and stored at -80°C as CytoI. The previous step was repeated and the supernatant was named as CytoII. The pellet was homogenized in 4 mL of the same buffer mentioned above but with a concentration of 2 M sucrose. The mixture was filtered with gauze, and the filtrate was placed on 4 mL of the last buffer. The tube was centrifuged at 80,000 g for 35 min at 4°C. The pellet contained the pure nuclei. The nuclear protein pellet was dissolved in 50 mM ammonium bicarbonate (pH 8.0) prior to protein digestion.

Protein extracts from nuclear enrichment of ATL and CC were digested by trypsin at a ratio of 1:80 (trypsin: total protein). The resulting peptides were lyophilized and frozen at -80°C before mass spectrometry analysis. Immediately prior to analysis, lyophilized peptides were dissolved in an aqueous solution of 0.1% formic acid and injected into a 2D nano high-performance liquid chromatography system (Eksigent, Dublin, CA, USA) coupled online to a LTQ XL-Orbitrap mass spectrometer (Thermo Scientific, Bremen, Germany). The specifics of data acquisition are described in detail in Maccarrone et al. [16].

The program used in the identification and quantification of proteins was MASCOT Distiller (Matrix Sciences, London, UK). For the identification and quantification, this program follows a series of statistical criteria. The main test used to indicate quantitative changes between the proteins was Student's t in log space. This analysis assigns to each protein a p value of significance with regard to differences in protein levels. In addition, samples values were applied to a data normalization process. This process is based on the hypothesis that it is reasonable to expect that only a minority of the proteins in the sample will be found to be differentially expressed, considering that the overall normalization is applied in order to make the mean or median ratios in the entire dataset equal to 1. Following this logic, the data distribution is log-normal, and the statistical test used to confirm this premise is the Shapiro-Wilk W test. If the results do not pass this test, it indicates that the values are meaningless and something has systematically gone wrong with the analysis. In these cases, the values are rejected in the normality test.

Shotgun proteomics analysis can produce high amounts of data, especially in studies of complex biological mixtures, such as postmortem brain samples. As a consequence, protein-protein interaction analysis and identification of the pathways involved are fundamental to understanding cellular phenotypes in the most complete manner possible. Due to this, we used bioinformatics tools available online in these analyses. These were: the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING, http://STRING-db.org/), Kyoto Encyclopedia Genes and Genomes (KEGG, http://www.genome.ad.jp/kegg/), and Reactome (http://reactome.org/).

In the results of the ATL nucleus enrichment, we identified a total of 4,293 unique peptides, which corresponded to 629 proteins. Of these, 224 were present at significantly different levels between the SCZ and control samples, and 76 were nuclear proteins (nuclear enrichment of 33%; Table 2). In the CC analysis, we identified 3,820 unique peptides, corresponding to 552 proteins with 119 present at different levels, and 24 of these were nuclear proteins (nuclear enrichment of 21%; Table 3). These differentially expressed proteins were analyzed using the online human protein reference database (http://www.hprd.org/) in order to find the biological processes and function/molecular classes with which they are related (Table 2, 3).

The differentially expressed proteins related to both the nuclei of the ATL region cells and the CC participated in biological processes related mainly to processes such as regulation of nucleobase, nucleoside, nucleotide, and nucleic acid metabolism (27% ATL, 40% CC) and cell communication and signaling (23% ATL, 36% CC) (Fig. 1). These processes are related to the main functions of the nucleus, such as gene expression, transcriptional control, splicing, and release of gene products [10].

The ATL is enriched in gray matter, while the cerebral region of the CC corresponds to the largest portion of white matter in the brain. This results from the preponderance of neurons in the ATL [17], whereas glial cells and neuronal axons are more predominant in the CC [18]. It is important to compare proteomic profiles of these two regions of the brain mainly because this gives a more integrated view of brain function and not only what happens only with the neurons, as is typical of most other brain proteomic studies. In the comparison of these two regions, 64 of these were specific to the ATL, 13 to the CC, and 12 were common to both regions, i.e. 13.5% of the proteins are shared for the regions, as it can be seen in the Venn diagram (Fig. 2). According to analysis carried out using the Reactome software, these latter proteins were related to cellular stress response, with a chain of reactions related to heat-shock proteins (HSPs) (p value 3.18^E-11) (Fig. 3). This type of reaction is triggered by cellular stressors such as exposure to high temperatures, hypoxia, and free radicals, and these factors can cause damage to cellular proteins and induce this type of response [19,20,21,22].

Since maintaining homeostasis is important for the proper functioning of cellular metabolism, this type of stress response must be effective and coordinated [23]. For this to occur, the main molecule responsible for transcription-mediated stress response, the heat shock transcription factor HSF1, must be present at optimal functional levels [24,25]. Under normal physiological conditions, this molecule is present in its inactive form, mediated by a series of protein-protein interactions. However, in the presence of stressors, this molecule becomes activated by a series of reactions, including its phosphorylation and interaction with DNA, promoting the cellular responses to stress (Fig. 3) [26,27,28,29,30,31].

The protein that mediates most of these reactions, the heat shock 70-kDa protein 1A/1B (HSPA1A), was found to be increased in the nuclear compartments of both the ATL and CC. This protein belongs to the family of heat shock protein 70 (HSP70), which has been implicated previously in SCZ [32]. A recent study showed that a polymorphism in the HSPA1A protein gene is associated with increased risk of developing paranoid SCZ [33], and another study found increased mRNA expression of this gene in postmortem samples from the prefrontal cortex of patients with SCZ [34].

The analysis of differentially expressed proteins in the ATL nuclei showed additional changes in 4 more heat shock proteins, with 3 of these belonging to the HSP70 family: heat shock-related 70-kDa protein 2 (HSPA2), heat shock 70-kDa protein 6 (HSPA6), heat shock cognate 71-kDa protein (HSPA8), and 10-kDa heat shock protein (HSPE1). All these proteins were found at higher levels in the patients compared to controls, indicating a large response to cellular stress in the former. There is a debate about the effects displayed in the analysis of postmortem brains in studies of psychiatric disorders. One of the main disputed points is whether or not such effects are a cause or consequence of the disease, or a consequence of prolonged treatment of these patients with different antipsychotic medications throughout their lives and through different stages of disease development. However, such data have appeared recurrently in proteomic data of postmortem brain of patients with SCZ and cannot be ignored. Nevertheless, corroborative studies of some format involving first-onset antipsychotic-naïve patients may help to resolve some of these issues.

Another family of proteins that were found differentially expressed in both the ATL (histone H3.3, histone H1.2, histone H2A type 1-B/E, histone H2A type 1-D, histone H2B type 1-B, histone H2B type 1-C/E/F/G/I, histone H4) and CC (histone H2A.V; histone H1.4; histone H3.1; histone H4), is the family of histone proteins. These proteins are linked to DNA, and they change the transcription mechanism of these molecules, thus modifying gene expression. Histones are also associated with neuronal functions such as synaptic plasticity, a function that is known to be altered in patients with SCZ [35,36]. Mass spectrometry studies have shown that posttranslational modifications in the nucleosome, which are characterized by the junction of DNA and a complex of histone proteins, regulate histone-DNA interaction, and are part of epigenetic mechanisms of genetic regulation [37]. Many recent studies associate histone dysregulation, and consequently dysregulation of epigenetic processes, with SCZ [38,39,40], and there has been much discussion on possible therapeutic targets based on targeted epigenetics in the treatment of this disease (reviewed in [41]). However, as more thorough epigenetic studies in postmortem brain tissue are still arduous and limited, little is known about this association.

In silico analysis of the proteins found at different levels in cell nuclei of the gray matter region showed that these proteins are mainly associated with spliceosome complex (p value 5.9^E-6) (online suppl. Table 1; see www.karger.com/doi/10.1159/000477299 for all online suppl. material). The spliceosome is a complex of 5 multi-megadalton ribonucleoproteins (snRNPs), which are abundant RNA-binding proteins that carry out processing of pre-mRNA transcripts. This complex removes introns of these molecules and rearranges exons, so that these conform and can give rise to the correctly configured mRNA transcripts [42]. Due to the exon rearrangement, a single pre-mRNA can give rise to different functioning mRNAs, and any perturbations of this process may lead to disease development or contribute to the severity of preexisting diseases [43].

The proteins related to this pathway that were found at altered levels in this study were heterogeneous nuclear ribonucleoproteins C1/C2 (HNRNPC), heterogeneous nuclear ribonucleoprotein K (HNRNPK), heterogeneous nuclear ribonucleoprotein U (HNRNPU), splicing factor, arginine/serine-rich 3 (SRSF3), and HSPA1A, HSPA2, HSPA6 and HSPA8, as described above. These proteins are the main constituents of the snRNP complex, which is the main component of the splicing operation (Fig. 4; KEGG).

Eight of the proteins belonging to the hnRNP family were found deregulated in a study of oligodendrocyte cells treated with clozapine, an antipsychotic used in the treatment of SCZ [44]. Moreover, studies of silencing and super-expression of proteins showed a crucial role of hnRNP proteins in the myelination of neurons by oligodendrocytes, independently of the indirect regulation of quaking proteins as previously proposed [45,46]. This dysfunction in myelination was related to dysregulation of the synaptic connection [47,48,49]. One of these proteins was HNRNPC, which was found at decreased levels in SCZ patients compared to controls in this study. It is known that the regulation process of myelination undergoes a precise control dependent on the alternative splicing [45], and if the hnRNP protein complex is deregulated, this may cause aberrant alternative splicing, as reported in neurodegenerative and neuropsychiatric diseases such as frontotemporal dementia with Parkinsonism, amyotrophic lateral sclerosis and SCZ [50,51,52].

The deregulation of hnRNP proteins, in addition to being related to oligodendrocyte cells, is also related to the dysfunction of the neurotransmitter system. In a study performed using polymerase chain reaction analysis of SCZ prefrontal cortex tissue, alterations in splicing of mRNA molecules related to the GABAergic system were found [51]. Another study has shown that proteins related to dopamine receptors are also regulated by alternative splicing and that perturbed splicing of these proteins can cause dysfunctions in the dopaminergic system [53]. These systems are known to be altered in patients with SCZ [54]. Thus far, there are only a few recent studies that correlate perturbations in splicing and SCZ. However, the converging evidence suggests that this could play a major role in the disease process and therefore warrants further study, particularly accounting for effects on both neuronal and oligodendrocyte cells.

It should be considered that some of the alterations may be related to heterogeneity of the postmortem tissues including potential differences in cell types and cell density. Therefore, it may not be possible to identify which specific cell types the changes are associated with. However, it is likely that to the changes are associated with glial cells when analyzing white matter and with neurons when analyzing gray matter, considering the natural abundance of such cells in these tissues. It should also be considered that data generated using postmortem tissue has other disadvantages such as potential differences in postmortem intervals or agonal states. Therefore, functional analyses, such as super-expression and knockout studies, should be done to test hypotheses resulting from such investigations using postmortem tissues. For example, microdissection of specific cell types could be used to assess significant cell-specific differences between patients and controls.

After enriching the pathways of the 24 proteins found differentially expressed in the nucleus of the CC region, only 4 proteins were present in all of the enriched pathways (online suppl. Table 2) and were present in the same interaction network (Fig. 5). These 4 proteins were calmodulin (CALM1), calcium/calmodulin-dependent protein kinase type II subunit alpha (CAMK2A), calcium/calmodulin-dependent protein kinase type II subunit beta (CAMK2B), and calcium/calmodulin-dependent protein kinase type II subunit gamma (CAMK2G), which are all calcium dependent for their activation and response, as well as being constituents of the serine/threonine protein kinase family, which has been widely associated with metabolic events such as muscle contraction, cellular metabolism/proliferation, gene expression, and neurotransmitter release (reviewed in [55]).

The correlation between deregulation of calcium and SCZ began in the 1970s [56], and several studies confirming and discussing various aspects of this correlation have been performed. Among the many hypotheses that correlate calcium signaling to SCZ, there is the deregulation of the dopaminergic system, which is one of the hypotheses of SCZ development of the disease [57]. Calcium participates in the uptake of dopamine by synaptic vesicles and, in this process, calcium is associated with CaMKII proteins [26]. In this study, we found increased levels of the alpha, beta, and gamma CAMK2 proteins, suggesting that there may be a high uptake of dopamine by synaptic vesicles and, consequently, a greater release of the neurotransmitter in the synaptic cleft [54]. The major focus in most studies in SCZ has been on the proteins of neuronal cells, as can be seen in the results presented above. However, one study quantified the calmodulin proteins in nuclei of neuronal and glial cells and showed that the abundance of these molecules is greater in the latter cell type [58]. Even so, little research has been done investigating the role of the deregulation of calmodulin proteins in glial cells.

A few recent studies have associated calcium/calmodulin-dependent protein with glial cells in functions such as actin cytoskeleton remodeling, glutamate and glycine transport, and regulation of neurotrophic factors [59,60,61,62]. This involves the activation of the PI3K protein which, according to Pérez-Garcia et al. [62], is regulated by calmodulin proteins associated with calcium. The process of activation of the PI3K protein involves its autophosphorylation, which was found to be reduced in a recent study of postmortem CC from patients with SCZ performed by our group [63]. Given that neurotrophic factors are decreased in patients [64], we can associate the increased amount of CAMK found in this study to a form of cellular compensation which aims to normalize the activation of the neurotrophic factors. Another possible association is that the high concentration of CaMK proteins may desensitize the P13K activation process, causing the observed disruption in this pathway.

Also present in the network identified above are the nucleic acid metabolism-related proteins, which are proteins of the HSP family, and proteins linked to the genetic transcription, which belong to the family of hnRNPs. These results are related to those found in the ATL region, showing a possible relationship between the deregulation of the white and gray matter, with changes in cell stress and regulation of genetic expression.

The results found in this study provide an overview of the participation of nuclear proteins in the pathophysiology of the disease. When it comes to the gray matter, the results point to the dysregulation of the spliceosome, an area which has not been investigated previously. In the case of the white matter, the role of calcium/calmodulin protein deregulation in glial cells has only been partly explored in previous studies, and may be related to altered production of neurotrophic factors. These results represent the first in-depth study comparing effects on gray and white matter in SCZ and lay the groundwork for further studies in this area to help increase our understanding of this complex disease. This could lead to identification of novel biomarkers and drug targets which, in turn, may result in development of newer and better treatment options for people suffering from this disease for improved therapeutic outcomes.

V.M.S.C. and D.M.S. are supported by FAPESP (São Paulo Research Foundation), grants 2016/07332-7, 2013/08711-3, and 2014/10068-4. D.M.S. is also supported by The Brazilian National Council for Scientific and Technological Development (CNPq), grant 460289/2014-4.

The authors declare no conflict of interest.