Transcriptomic Data Analysis Reveals a Down-Expression of Galectin-8 in Schizophrenia Hippocampus

Schizophrenia (SCZ) is a severe psychiatric disorder with several clinical manifestations that include cognitive dysfunction, decline in motivation, and psychosis. Current standards of care treatment with antipsychotic agents are often ineffective in controlling the disease, as only one-third of SCZ patients respond to medications. The mechanisms underlying the pathogenesis of SCZ remain elusive. It is believed that inflammatory processes may play a role as contributing factors to the etiology of SCZ. Galectins are a family of β-galactoside-binding lectins that contribute to the regulation of immune and inflammatory responses, and previous reports have shown their role in the maintenance of central nervous system (CNS) homeostasis and neuroinflammation. In the current study, we evaluated the expression levels of the galectin gene family in post-mortem samples of the hippocampus, associative striatum, and dorsolateral prefrontal cortex from SCZ patients. We found a significant downregulation of LGALS8 (Galectin-8) in the hippocampus of SCZ patients as compared to otherwise healthy donors. Interestingly, the reduction of LGALS8 was disease-specific, as no modulation was observed in the hippocampus from bipolar nor major depressive disorder (MDD) patients. Prediction analysis identified TBL1XR1, BRF2, and TAF7 as potential transcription factors controlling LGALS8 expression. In addition, MIR3681HG and MIR4296 were negatively correlated with LGALS8 expression, suggesting a role for epigenetics in the regulation of LGALS8 levels. On the other hand, no differences in the methylation levels of LGALS8 were observed between SCZ and matched control hippocampus. Finally, ontology analysis of the genes negatively correlated with LGALS8 expression identified an enrichment of the NGF-stimulated transcription pathway and of the oligodendrocyte differentiation pathway. Our study identified LGALS8 as a disease-specific gene, characterizing SCZ patients, that may in the future be exploited as a potential therapeutic target.


Introduction
Schizophrenia (SCZ) is a severe psychiatric disorder characterized by a diverse set of clinical manifestations that include cognitive dysfunction, disorganized thinking, decline in motivation, and psychosis [1]. The global prevalence of SCZ is~1%, and the incidence is approximately 1.5/10,000 people [1]. Usually, the onset of SCZ is observed earlier in men, developing SCZ at age 18-25, while the onset in women is between 25 and 35 years [2]. To date, the etiology of SCZ is still elusive, and it is believed that both genetic and environmental factors may play a role, by inducing alterations in the dopaminergic, glutamatergic, serotonergic, and gamma-aminobutyric acid neurotransmitter systems [3].

Dataset Selection and Analysis
The NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/ geo/, accessed on 2 March 2021) was used to identify microarray datasets comparing the transcriptomic profiles of brain from healthy donors and SCZ patients. The GEO database was manually searched using the MeSH term (Medical Subject Headings) "schizophrenia". The datasets were selected if they met the following inclusion criteria: (a) whole-genome transcriptomic profiling; (b) tissues collected from defined brain regions; (c) consisted of both one cohort of SCZ patients and another cohort of healthy people; and (d) species of origin "Homo sapiens". Finally, the GSE53987 was selected for the analysis, as it included whole genome transcriptomic data from post-mortem samples of hippocampus, associative striatum, and dorsolateral prefrontal cortex from SCZ patients, as well as control subjects and patients suffering from bipolar disorder and major depressive disorder (MDD) [15]. The clinical data available for the patients included in the GSE53987 dataset are summarized in Table 1. Briefly, the GSE53987 dataset included 19 samples from healthy people, 15 samples from SCZ patients, 18 samples from bipolar patients, and 17 from MDD patients [15]. Data were subjected to quantile normalization, and low variance genes were excluded from the analysis. For this study, the expression levels of LGALS1, LGALS2, LGALS3, LGALS3BP, LGALS4, LGALS8, LGALS9, LGALS12, LGALS13, and LGALS14 were evaluated.

Gene Ontology Analysis
Pathway analysis, Gene Ontology, and interactome analysis were performed using the online software Metascape [16]. The hypergeometric test and Benjamini-Hochberg p-value correction are used to identify significant enriched ontology terms. Metascape analysis relies on several databases, including KEGG, MSigDB, Gene Ontology, and Reactome, among the others. Enriched terms are clustered into non-redundant groups, by calculating the pairwise similarity between the enriched terms, using a Kappa-test score [16].

Gene Expression Regulation
For the identification of the transcription factors (TFs) that could regulate the expression of LGALS8, a list of putative TFs was first retrieved from John and Mishra (2016) [17]. The TFs mostly correlated (FDR < 0.05) to LGALS8 expression were then clustered using the Spearman rank correlation, as similarity metrics. For the identification of miR-NAs putatively regulating LGALS8 expression, the miRWalk database was interrogated ( http://mirwalk.umm.uni-heidelberg.de/, accessed on 23 July 2021). Among the predicted miRNAs, those negatively correlated to LGALS8 expression were selected. For the analysis of the methylation pattern of LGALS8, the GSE89703 dataset was interrogated. The dataset included post-mortem hippocampus samples: 14 schizophrenia and 13 controls. Bisulfite converted DNA from these samples were hybridized to the Illumina Infinium 450k Human Methylation Beadchip v1.0 [18].

Statistical Analysis
Statistical analysis for the microarray data was performed using the LIMMA (Linear Model for MicroArray Analysis) test. An adjusted (Benjamini-Hochberg corrected) pvalue (adj. p-value, FDR: False Discovery Rate) < 0.05 was considered as threshold for statistical significance. Hierarchical clustering and gene similarity matrix were calculated using the Spearman rank correlation, as similarity metrics. For the identification of the genes significantly correlated to LGALS8, the Pearson correlation was calculated, and the statistical significance was computed using a permutation test, with 1000 random permutations. Correlated genes were selected based on p-value < 0.01 and |r| > 0.7.

Expression Levels of the Galectin Gene Family Members in Different Brain Regions from Control Donors
First, we wanted to determine the expression levels of the galectin gene family members in different brain areas from control donors. As shown in Figure 1, significant differences could be observed in the expression levels of the galectins in the hippocampus, associative striatum, and the prefrontal cortex ( Figure 1A). In particular, the hippocampus was characterized by a down-expression of LGALS1, LGALS13, and LGALS14, as compared to striatum and prefrontal cortex. An over-expression of LGALS1 and LGALS8 was instead observed in the prefrontal cortex, as compared to the hippocampus and striatum. Significant higher levels of LGALS3, LGALS3BP, and LGALS14, as well as significant lower levels of LGALS2, LGALS8, LGALS9, and LGALS12 characterized the associative striatum when compared to the hippocampus and the prefrontal cortex ( Figure 1A). Along the same lines, hierarchical cluster analysis revealed that the hippocampus, prefrontal cortex, and striatum could be segregated by considering the expression levels of the galectin gene family members ( Figure 1B). Next, we determined the relative expression levels of the different members of the galectin family in SCZ patients. As shown in Figure 2, only LGALS8 was significantly downregulated in the hippocampus of SCZ patients, as compared to the control donors ( Figure 2). Interestingly, no modulation of LGALS8 was observed in either the bipolar or the MDD hippocampus samples (Figure 2). A trend of reduced levels of LGALS8 was also observed in the striatum and the prefrontal cortex from SCZ patients, but statistical significance was not reached. No modulation was observed in the expression levels of all the other members of the galectin family, in either of the brain areas considered and for either the bipolar or MDD patients ( Figure 2) Since all the SCZ patients were under antipsychotic treatment, we aimed to determine whether the modulation in LGALS8 expression was related to drug exposure rather than to the disease. To this aim, we interrogated the GSE66277, which included Next, we determined the relative expression levels of the different members of the galectin family in SCZ patients. As shown in Figure 2, only LGALS8 was significantly downregulated in the hippocampus of SCZ patients, as compared to the control donors ( Figure 2). Interestingly, no modulation of LGALS8 was observed in either the bipolar or the MDD hippocampus samples ( Figure 2). A trend of reduced levels of LGALS8 was also observed in the striatum and the prefrontal cortex from SCZ patients, but statistical significance was not reached. No modulation was observed in the expression levels of all the other members of the galectin family, in either of the brain areas considered and for either the bipolar or MDD patients ( Figure 2) whole-genome expression data from male Sprague-Dawley rats chronically treated with either haloperidol, risperidone, or vehicle. No modulation of LGALS8 was observed in the hippocampus of the treated rats (Supplementary Materials Figure S1). LGALS1 LGALS2 LGALS3 LGALS3BP LGALS4 LGALS8 LGALS9 LGALS12 LGALS13 LGALS14 LGALS1 LGALS2 LGALS3 LGALS3BP LGALS4 LGALS8 LGALS9 LGALS12 LGALS13 LGALS14 LGALS2 LGALS3 LGALS3BP LGALS4 LGALS8 LGALS9 LGALS12 LGALS13 LGALS14 Since all the SCZ patients were under antipsychotic treatment, we aimed to determine whether the modulation in LGALS8 expression was related to drug exposure rather than to the disease. To this aim, we interrogated the GSE66277, which included whole-genome expression data from male Sprague-Dawley rats chronically treated with either haloperidol, risperidone, or vehicle. No modulation of LGALS8 was observed in the hippocampus of the treated rats (Supplementary Materials Figure S1).
Next, we wanted to determine the mechanisms of regulation of LGALS8 expression in the SCZ hippocampus. As shown in Figure 3A, among the putative TFs regulating LGALS8 expression, TBL1XR1, BRF2, and TAF7 were the most positively correlated TFs (FDR < 0.05) ( Figure 3A). Moreover, among the miRNAs predicted for LGALS8, two of them, i.e., MIR3681HG and MIR4296, were negatively correlated with LGALS8 expression, suggesting a role for epigenetics in the regulation of LGALS8 levels ( Figure 3B). On the other hand, no significant differences were observed in the methylation levels of LGALS8 between SCZ and matched control donor hippocampus ( Figure 3C). Next, we wanted to determine the mechanisms of regulation of LGALS8 expression in the SCZ hippocampus. As shown in Figure 3A, among the putative TFs regulating LGALS8 expression, TBL1XR1, BRF2, and TAF7 were the most positively correlated TFs (FDR < 0.05) ( Figure 3A). Moreover, among the miRNAs predicted for LGALS8, two of them, i.e., MIR3681HG and MIR4296, were negatively correlated with LGALS8 expression, suggesting a role for epigenetics in the regulation of LGALS8 levels ( Figure 3B). On the other hand, no significant differences were observed in the methylation levels of LGALS8 between SCZ and matched control donor hippocampus ( Figure 3C). Finally, to gain insights into the biological meaning of LGALS8 modulation in the SCZ hippocampus, we first identified the genes significantly correlated, both positively and negatively, to LGALS8 expression, and Gene Ontology and pathway analysis were performed. In total, 115 genes were negatively correlated, while 177 genes were positively correlated to LGALS8 (p-value < 0.01 and |r| > 0.7). Several enriched terms were found for both the positively and negatively correlated genes (Figures 4 and 5). In particular, R-HSA-9031628 (NGF-stimulated transcription), GO:0048709 (oligodendrocyte differentiation), and GO:0070848 (response to growth factor) were found to be enriched among the genes negatively correlated to LGALS8, while GO:0043266 (regulation of potassium ion transport), GO:0032984 (protein-containing complex disassembly), and GO:0042180 (cellular ketone metabolic process) were found to be enriched among the genes positively correlated to LGALS8 (Figures 4 and 5). Finally, to gain insights into the biological meaning of LGALS8 modulation in the SCZ hippocampus, we first identified the genes significantly correlated, both positively and negatively, to LGALS8 expression, and Gene Ontology and pathway analysis were performed. In total, 115 genes were negatively correlated, while 177 genes were positively correlated to LGALS8 (p-value < 0.01 and |r| > 0.7). Several enriched terms were found for both the positively and negatively correlated genes (Figures 4 and 5). In particular, R-HSA-9031628 (NGF-stimulated transcription), GO:0048709 (oligodendrocyte differentiation), and GO:0070848 (response to growth factor) were found to be enriched among the genes negatively correlated to LGALS8, while GO:0043266 (regulation of potassium ion transport), GO:0032984 (protein-containing complex disassembly), and GO:0042180 (cellular ketone metabolic process) were found to be enriched among the genes positively correlated to LGALS8 (Figures 4 and 5).  . Gene Ontology and pathway analysis for the genes most correlating with LSGALS8 in SCZ hippocampus. Genes significantly correlated to LSGALS8 in SCZ hippocampus were identified by calculating the Pearson correlation, and the statistical significance was computed using a permutation test, with 1000 random permutations. The graph represents the hierarchical clustering of the most enriched biological processes and gene ontologies among the genes significantly correlated to LGALS8, as determined using the web-based utility, Metascape.  . Gene Ontology and pathway analysis for the genes most correlating with LSGALS8 in SCZ hippocampus. Genes significantly correlated to LSGALS8 in SCZ hippocampus were identified by calculating the Pearson correlation, and the statistical significance was computed using a permutation test, with 1000 random permutations. The graph represents the hierarchical clustering of the most enriched biological processes and gene ontologies among the genes significantly correlated to LGALS8, as determined using the web-based utility, Metascape.
Brain Sci. 2021, 11, x FOR PEER REVIEW 8 of 14 Figure 4. Gene Ontology and pathway analysis for the genes most correlating with LSGALS8 in SCZ hippocampus. Genes significantly correlated to LSGALS8 in SCZ hippocampus were identified by calculating the Pearson correlation, and the statistical significance was computed using a permutation test, with 1000 random permutations. The graph represents the hierarchical clustering of the most enriched biological processes and gene ontologies among the genes significantly correlated to LGALS8, as determined using the web-based utility, Metascape.

Discussion
Dysfunction in the prefrontal cortex, hippocampus, and associative striatum has been described for several psychiatric diseases, including SCZ [15,[19][20][21][22]. Previous reports have shown differential expression of multiple genes across brain regions in SCZ patients, as compared with control donors [15,19]. In these studies, the hippocampus has shown the highest number of transcriptional changes in SCZ [15,19]. The most striking result of these studies is the consistent upregulation of inflammatory pathways in the brain of SCZ subjects [23]. Interestingly, on the other hand, bipolar and MDD samples showed very little enrichment of the same inflammatory pathways [15,19].
The finding of the inflammatory dysregulation underlying SCZ has prompted significant interest in testing the immunomodulatory potential of psychotropic medications. Unfortunately, discordant results have been obtained among studies, which dampens the possibility of reaching conclusive results [24]. Additionally, clinical trials on the effects of anti-inflammatory drugs have shown mixed benefit in patients with SCZ [25]. This is likely due to the low passage of most current anti-inflammatory drugs across the blood-brain barrier (BBB). Hence, strategies to enhance the BBB permeation of drugs aimed at reducing neuroinflammation may represent in the future a viable therapeutic strategy for treating SCZ. It is also likely that earlier intervention may be more effective, as the persistence of inflammation in the adult SCZ brain could cut down the possibility of producing measurable improvements, when prolonged inflammatory processes cause irreversible structural alterations in the brain circuitry.
The role of inflammation in the pathogenesis of SC is supported by animal models showing that prenatal immune activation determines a series of brain alterations, such as reduction in parvalbumin interneurons, disrupted working memory, and alteration in both NMDA and dopamine signaling, which are relevant to SCZ [26,27]. Moreover, animal exposure to an acute prenatal immune activation is associated with persistent increases in cytokine production in the frontal cortex and hippocampus of young adult animals [28,29].
In accordance with the role of inflammation in SCZ (reviewed in [6]), it was shown that following treatment of the acute illness, IL-6 levels significantly decreased in SCZ patients, along with an increase in sIL-2R [30]. In another meta-analysis, brain-derived neurotrophic factor (BDNF) increased upon treatment in SCZ along with a decrease in serum IL6 and TNF-α concentrations [31]. Additionally, after risperidone treatment, IL-6, IL-10, TNF-α, and IL-4 decreased significantly [32]. No significant differences were found between the post-treatment cytokine levels in first episode psychosis patients and in healthy controls, suggesting that these patterns of altered cytokine levels may represent a marker of first episode psychosis [32]. In addition, there was a significant negative association between the risperidone-induced changes in IL-10 and the negative symptoms [32].
Galectin-8 belongs to the tandem-repeat class of galectins characterized by two CRDs, which are linked by peptides of varying length [58,59]. Unlike the other galectins, Galectin-8 specifically binds the α2,3-sialylated glycans via its N-terminal CRD [58,59]. Intracellularly, Galectin-8, by binding the cytosolic exposed luminal glycans of damaged endosomes and lysosomes to the NDP52 autophagy adaptor, promotes their autophagic removal [60]. This function represents a defense mechanism against infections and seems to protect against the aggregation of the tau protein, involved in the pathogenesis of Alzheimer's disease [60][61][62][63]. In the extracellular compartment, Gal-8 binds β1-integrins and activates the downstream ERK1/2 and PI3K/AKT signaling pathways [64][65][66]. Pardo and colleagues have shown in vitro that Galectin-8 protects hippocampal neurons against stress, including nutrient deprivation, oxidative stress, β-amyloid oligomers exposure, and glutamate-induced excitotoxicity [67]. In culture, primary hippocampal neurons secrete Galectin-8, and, in vitro, anti-Galectin-8 auto-antibodies from Systemic Lupus Erythematosus patients affect neuron survival [67]. Finally, higher levels of apoptosis are observed in the hippocampus from Galectin-8 knockout mice, upon local injection of hydrogen peroxide, suggesting the neuroprotective effect of Galectin-8 in the hippocampus [67]. Hence, it is possible that the down-expression of LGALS8 observed in the postmortem SCZ hippocampus is not a downstream result of pathological processes but could be itself a mediator of the disease.
It should be noted that the reduction of LGALS8 was disease-and region-specific, as no modulation was observed in the hippocampus from bipolar nor MDD patients, nor in the associative striatum and prefrontal cortex. This is in line with the observation that the greatest regional burden of transcriptional changes was different among SCZ, bipolar, and MDD subjects [15]. In particular, it was observed that MDD samples showed more differentially expressed genes in the striatum and hippocampus than in the prefrontal cortex, whereas a lower number of transcripts were altered in the striatum from BD patients [15]. Additionally, the most dramatic alterations were found for SCZ subjects in all of the brain regions studied, when compared to BD and MDD samples. Indeed, the hippocampus showed the highest number of transcriptomic changes in SCZ, while the lower number of differentially expressed genes was detected in the prefrontal cortex [15,68,69]. It is also interesting to note that among the altered biological pathways underlying SCZ [70], several of them were enriched by the DEGs significantly correlated to LGALS8 expression, including cell cycle proliferation, survival, and ion channel homeostasis.
The analysis of the hippocampus from rats chronically administered with antipsychotic drugs showed no modulation in LGALS8 expression. These data suggest that the modulation of LGALS8 in SCZ may not be dependent on medication and could instead independently underlie SCZ pathology. However, there are some points that need to be addressed: the analysis of the rat hippocampus was performed on animals treated with either risperidone or haloperidol. Hence, we may have missed the effect of other antipsychotic medications and also the effect of their combined administration, which is common for psychiatric patients. Moreover, the medications were dosed to healthy rats, without neurological conditions resembling SCZ pathology. Finally, differences in the metabolism between rats and humans could have biased our observation. One final limitation of this study is that bulk transcriptomic data were used for this study. Hence, future studies will need to focus on cell-specific changes, which may help to clarify the cell populations within the hippocampus undergoing the modulation of LGALS8, and the consequences to the other cellular populations. Additionally, the enrollment of larger cohorts of patients will help to better correlate the observed transcriptional change to the symptoms of SCZ and its progression.

Conclusions
Overall, with the present study, we are providing a comprehensive view of the expression levels of the different members of the galectin gene family in different brain areas of patients with SCZ. The significant reduced levels of LGALS8 observed in the hippocampus from SCZ patients appears to be a disease-specific hallmark of SCZ. However, whether this modulation of LGALS8 is a driver of the disease or a consequence of genetic/epigenetic factors remains to be determined. Upon the precise identification of the LGALS8-related pathways underlying SCZ pathogenesis, the use of drug discovery approaches will help to design tailored pharmacological strategies that may regulate these pathways.