Molecular mapping of a core transcriptional signature of microglia-specific genes in schizophrenia

Besides playing a central role in neuroinflammation, microglia regulate synaptic development and is involved in plasticity. Converging lines of evidence suggest that these different processes play a critical role in schizophrenia. Furthermore, previous studies reported altered transcription of microglia genes in schizophrenia, while microglia itself seems to be involved in the etiopathology of the disease. However, the regional specificity of these brain transcriptional abnormalities remains unclear. Moreover, it is unknown whether brain and peripheral expression of microglia genes are related. Thus, we investigated the expression of a pre-registered list of 10 genes from a core signature of human microglia both at brain and peripheral levels. We included 9 independent Gene Expression Omnibus datasets (764 samples obtained from 266 individuals with schizophrenia and 237 healthy controls) from 8 different brain regions and 3 peripheral tissues. We report evidence of a widespread transcriptional alteration of microglia genes both in brain tissues (we observed a decreased expression in the cerebellum, associative striatum, hippocampus, and parietal cortex of individuals with schizophrenia compared with healthy controls) and whole blood (characterized by a mixed altered expression pattern). Our results suggest that brain underexpression of microglia genes may represent a candidate transcriptional signature for schizophrenia. Moreover, the dual brain-whole blood transcriptional alterations of microglia/macrophage genes identified support the model of schizophrenia as a whole-body disorder and lend weight to the use of blood samples as a potential source of biological peripheral biomarkers.


Microglia genes
The present study exclusively included genes that are part of a core transcriptional signature of human microglia [1].This signature was established by Patir and colleagues through the identification of co-expressed genes associated with microglia, ensuring their presence in at least three out of nine distinct human datasets of microglia.
HLA-DRB4 was not included due to the lack of specificity of the microarray probes, and TSPO was omitted as its relevance as a microglia marker has been questioned [5] with evidence suggesting a closer association with astrocytes [6].
Eventually, it should be noted that the genes included in this study exceeded the minimum requirement of being present in three datasets, as established by Patir et al. when defining the core transcriptional signature of human microglia [1].Indeed, our candidate genes were present in a larger number of datasets, specifically in at least six out of the nine co-expression derived datasets, thus reinforcing their strong association with microglia.Additionally, it is noteworthy that all the genes included in this study were also identified in a recently published list of microglia signature genes that are highly expressed in bulk brain tissues [7].

Inclusion of datasets
Since our primary aim was to map transcriptional alterations of microglia genes in brain and peripheral tissues samples, one dataset per brain region (or peripheral tissue) was selected.When more than one dataset from the same brain region (or peripheral tissue) was available, we selected the one that would maximize the following 2 criteria in this particular order of relevance: 1) the dataset is capable to explore the largest number of genes from our list of candidate genes; 2) the dataset has the largest number of subjects.Based on these criteria, 2 datasets (GSE21138 from Narayen et al., 2008 [8]; GSE38481 from de Jong et al., 2012 [9]) from the list of 12 eligible datasets were not included in the main analyses.

Datasets included
It should be noted that the datasets included in this study are slightly different from those that were pre-registered on AsPredicted.org(#67610, https://aspredicted.org/285rn.pdf).This is due to the following reasons: 1) we initially planned to include datasets using three types of Affymetrix arrays (HG-U133_Plus_2, Human Gene 1.0 ST or Human Gene 1.1 ST) which could technically interrogate our list of candidate genes.However, this criterion precluded the inclusion of other array platforms capable of exploring these candidate genes (such as Agilent or Illumina arrays).Therefore, this criterion was removed, and only custom-designed microarrays were excluded.Consequently, an additional non pre-registered dataset was included in this study (GSE62191 from de Baumont et al., 2015 [10]); 2) we excluded one preregistered dataset (GSE93987 from Arion et al., 2015 [11] since it was captured through a laser microdissection of pyramidal cells, and thus did not contain microglia cells violating our inclusion criteria; 3) we also excluded one pre-registered dataset (GSE73129 from Horiuchi et al., 2016 [12]) due to lack of probe accuracy of at least half of the candidate genes (i.e., 5 genes present a lack of variability in gene expression values for all participants, probably reflecting transcriptional noise).
Following this selection, we report results from 9 different datasets.Finally, it should be noted that for two included datasets, we could not reliably measure the expression of a few genes (CD68 and ITGAX in the superior temporal cortex; NCF4 and TMEM119 in the frontal cortex) due to a lack of probe accuracy for those genes.The presently reported non-significant differences between individuals with schizophrenia and healthy controls for those latter genes in the relevant datasets should therefore be considered with caution.

Bayesian analyses
In additional Bayesian analyses, we quantify evidence in favor of the null (H0) and alternative (H1) hypotheses using the Bayes Factor (BF).Indeed, BFs are the ratio of the likelihood of the data under the alternative hypothesis and under the null hypothesis.BF10 quantifies the evidence in favor of H1 compared with H0, while BF01 (= 1/BF10) quantifies the evidence in favor of H0 compared with H1.Conventionally a BF10 (or BF01) that exceeds the threshold of 3 represents moderate evidence in favor of H1 (or H0), while when it exceeds the threshold of 100 the evidence can be considered as decisive.Finally, it is not possible to conclude regarding the presence or absence of group differences when BF10 (or BF01) is between 1 and 3 (i.e., anecdotal evidence) [13].Erasmus University Medical Center (EMC), Rotterdam, The Netherlands All subjects provided written informed consent after complete description of the study.For those patients who were too disturbed to provide consent, consent was initially given by a first-degree relative and final written consent was sought within six weeks from the patients themselves.This study was approved by the Erasmus University Medical Center Institutional Review Board and was conducted according to the declaration of Helsinki Eligible for inclusion were male, stabilized or acutely psychotic, patients diagnosed with SZ or schizophreniform disorder according to DSM IV criteria after a Comprehensive Assessment of Symptoms and History interview (CASH) and by consensus between two senior psychiatrists.Additional criteria were recent onset (defined as duration of illness <5 years) and age (>15 and <36 years)

Supplementary
Age-matched controls were recruited from the students and staff of the EMC medical school and hospital For SZ and HC, exclusion criteria were defined as follows: presence of any somatic or neurological disorders and abuse of heroin, cocaine, or alcohol.Cannabis abuse was not an exclusion criterion.Concomitant use of mood-stabilizers and/or antidepressants was an exclusion criterion.For HC, the presence of psychiatric disorders in first-degree relatives was also an The project was approved by the local ethics committee.Written informed consent was obtained from the patients and controls.
In the case of patients with a compromised ability to provide authorization, informed consent was signed by the legally authorized representative All SZ satisfied the DSM-IV criteria for SZ.Diagnoses were confirmed using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) diagnostic scale Unrelated healthy volunteers were screened for DSM-IV Axis I disorders by expert psychologists using the Mini-International Neuropsychiatric Interview (M.I.N.I.).Only healthy volunteers without a history of drug or alcohol abuse or dependence and without a personal or first-degree family history of psychiatric disorders were enrolled in the study For SZ and HC, exclusion criteria were defined as follows: mental retardation or cognitive disorder; serious somatic illnesses; uncorrected hypothyroidism or hyperthyroidism; age <18 and >70 years; metabolic disorders (diabetes); specific dermal diseases (e.g., dermal cancer or psoriasis).For HC, the presence of a history of drug, alcohol abuse or dependence, the presence of psychiatric disorders in first-degree relatives were additional exclusion criteria

BD
University of Pittsburgh brain bankBrain specimens were obtained during autopsies after consent for donation was obtained from the next-of-kin.All procedures were approved by the University of Pittsburgh Committee for the Oversight of Research and Clinical Trials Involving the Dead and the Institutional Review Board for Biomedical Research Diagnoses were made by an independent committee of experienced research clinicians, using DSM-IV criteria and based on the results of structured interviews conducted with family members and review of medical records The absence of psychiatric diagnoses was confirmed using an identical approach.The healthy controls were free of any neurological or psychiatric illness during their life courseLanz et al., Translational Psychiatry, 2019 (PMID: 31123247)    To evaluate shared transcriptional alterations across connected brain regions in SZ, bipolar disorder (BD), major depressive disorder (MDD) individuals, or HC Genome-wide expression was obtained from postmortem DLPFC, hippocampus, and associative striatum from 19 wellmatched tetrads of SZ, BD, MDD, or HC SZ showed a substantial burden of differentially expressed genes across all examined brain regions with the greatest effects in hippocampus, whereas BD and MDD showed less robust alterations.Pathway analysis of transcriptional profiles compared across diagnoses demonstrated commonly enriched pathways between all three disorders in hippocampus, significant overlap between SZ and BD in DLPFC, but no significant overlap of enriched pathways between disorders in striatum.SZ showed increased expression of transcripts associated with inflammation across all brain regions examined, which was not evident in Stanley Medical Research Institute's Neuropathology Consortium and Array Collections Specimens were collected with informed consent from next-of-kin Diagnoses were made by two senior psychiatrists, using DSM-IV criteria and based on medical records and, when necessary, telephone interviews with family members Diagnoses of unaffected controls were based on structured interviews by a senior psychiatrist with family member(s) to rule out Axis I diagnoses.Individuals over age 65 were excluded Chen et al., Molecular Psychiatry, 2013 (PMID: 23147385) To identify schizophrenia-associated gene-expression networks in the parietal cortex and cerebellum of SZ and HC.To test whether the gene modules perturbed in SZ were similarly perturbed in bipolar disorder Genome-wide expression data was used to construct gene expression networks and identify gene co-expression modules within the networks.The modules were tested for association with SZ.Identified SZ-associated modules were tested for association with bipolar disorder Two modules were differentially expressed in SZ versus HC.One, upregulated in cerebral cortex, was enriched with neuron differentiation and neuron development genes, as well as disease genome-wide association study genetic signals; the second, altered in cerebral cortex and cerebellum, was enriched with genes involved in neuron protection function.The findings were preserved in five expression datasets, including sets from three brain regions, from a different microarray platform, Tissue collection of the Charing Cross Hospital, Imperial College London, UK All patients with the agreement of their nearest relative or authorized representative, have given written informed consent for use of tissue obtained post-mortem for research.The control group were tissue donors for research from the community.Procedures have been approved by the West London Mental Health Ethical Research Committee and complies with the conditions of the Research Governance Office of the Imperial College of Science, Technology and Medicine Clinical Research Office All patients met DSM-III diagnostic criteria for chronic residual SZ with pronounced negative symptoms alongside attenuated positive symptoms and intellectual dysfunction.All patients had been treated with neuroleptic drugs except one patient who was neuroleptic naive at death Mentally normal individuals from the community Alzheimer's disease, Parkinson's disease or multiple sclerosis were excluded Maycox et al., Mol Psychiatry, 2009 (PMID: 19255580) To identify differentially expressed genes in anterior prefrontal cortex (BA 10) from SZ and HC Genome-wide expression in post-mortem brain tissue from anterior prefrontal cortex (BA 10) was compared between 28 SZ and 23 HC.Results were then compared to those from an independent prefrontal cortex dataset obtained from SZ and HC 51 gene expression changes were common between the two SZ cohorts, and 49 showed the same direction of diseaseassociated regulation.Changes were observed in gene sets Barnes et al., J Neurosci Res, 2011 (PMID: 21538462) Gene ontology pathway enrichment analysis in BA22 and BA10 from SZ and HC Genome-wide expression was determined in the post-mortem BA22 region of 23 SZ and 19 HC and compared with genomewide expression of BA10 from the same subjects.Gene ontology pathway enrichment analysis was carried out in each region In BA22 region, the highest enrichment was observed in processes mediating cell adhesion, synaptic contact, cytoskeletal remodeling, and apoptosis.In BA10 region, the strongest changes were observed in reproductive signaling, Stanley Medical Research Institute's Neuropathology Consortium The study protocol was approved by the ethics committee of A.C.Camargo Cancer Center and was performed in accordance with the Declaration of Helsinki Records of all patients were reviewed for DSM-IV psychiatric diagnosis independently by two senior psychiatrists For normal controls, a structured telephone interview with a first-degree family member was carried out in all cases Participants with samples with low RNA integrity were excluded De Baumont et al., Schizophrenia Research, 2015 (PMID: 25487697)To identify the molecular mechanisms that differentiate SZ and individuals with bipolar disorder from healthy controls Genome-wide expression data were used to identify co-expression of pairs of genes to assess differences between SZ and individuals with bipolar disorder.A Protein-Protein Interaction network was also used to identify additional proprieties potentially associated with the differentially expressed genes between SZ and individuals with bipolar disorder Co-expression analyses revealed that the pairs CCR1/SERPINA1, CCR5/HCST, C1QA/CD68, CCR5/S100A11 and SERPINA1/TLR1 present the most significant difference between SZ and individuals with BD.Moreover, network analyses showed CASP4, TYROBP, CCR1, SERPINA1, CCR5 and C1QA as having a central role in the manifestation of the disease

Table 2 .
Description of the original studies from which the datasets were obtained Department of Psychiatry, University Medical Center, Utrecht, The Netherlands / Parnassia PsychoMedical Center, The Netherlands / Center for Neuropsychiatric Schizophrenia Research, Psychiatric Center Glostrup, Denmark The study was approved by Medical Research Ethics Committee (METC) of the University Medical Center Utrecht, The Netherlands and the Committees on Biomedical Research Ethics for the Capital Region of Denmark.All participants gave written informed consent Psychiatric diagnoses of SZ were made according to DSM-IV-TR criteria by trained clinicians using Standardized Psychiatric interviews either The Comprehensive Assessment of Symptoms and History (CASH) or the Composite international diagnosticTo identify SZ-associated gene co-expression modules in the whole blood of SZ compared to HC Genome-wide expression profiling from whole blood of 106 SZ and 96 HC.SZ-associated gene co-expression modules Identification of 12 large gene co-expression modules associated with SZ.Two of the SZ-associated modules were replicated in an independent second dataset involving antipsychotic-free SZ and HC.One of these SZ-associated modules is significantly enriched with brain-expressed genes and with genetic risk variants for SZ, the hub gene in this module (ABCF1) is located in

Table 3 .
Cattane et al., PLoS ONE, 2015 (PMID: 25658856) Differential expression analysis in skin fibroblasts of SZ and HC Genome-wide expression study comparing skin fibroblast transcriptomic profiles from 20 SZ and 20 HC Six genes (JUN, HIST2H2BE, FOSB, FOS, EGR1, TCF4) were strongly and significantly upregulated at the genome-wide level and confirmed by RT-PCR in SZ compared to HCAbbreviations: GEO, Gene Expression Omnibus; SZ, individuals with schizophrenia; HC, healthy controls; DLPFC, dorsolateral prefrontal cortex; PBMCs, peripheral blood mononuclear cells; BA, Brodmann Area Supplementary Genes with altered expression in the postmortem brain samples of individuals with schizophrenia compared with healthy controls 1 Fold change represents the expression of the target gene in individuals with schizophrenia relative to that in healthy controls 2 Significance of adjusted p-values set at 0.05 Supplementary

Table 5 .
Genes with altered expression in the peripheral tissue samples of individuals with schizophrenia compared with healthy controlsFold change represents the expression of the target gene in individuals with schizophrenia relative to that in healthy controls 2Significance of adjusted p-values set at 0.05