Microglia Transcriptome Changes in a Model of Depressive Behavior after Immune Challenge

Depression symptoms following immune response to a challenge have been reported after the recovery from sickness. A RNA-Seq study of the dysregulation of the microglia transcriptome in a model of inflammation-associated depressive behavior was undertaken. The transcriptome of microglia from mice at day 7 after Bacille Calmette Guérin (BCG) challenge was compared to that from unchallenged Control mice and to the transcriptome from peripheral macrophages from the same mice. Among the 562 and 3,851 genes differentially expressed between BCG-challenged and Control mice in microglia and macrophages respectively, 353 genes overlapped between these cells types. Among the most differentially expressed genes in the microglia, serum amyloid A3 (Saa3) and cell adhesion molecule 3 (Cadm3) were over-expressed and coiled-coil domain containing 162 (Ccdc162) and titin-cap (Tcap) were under-expressed in BCG-challenged relative to Control. Many of the differentially expressed genes between BCG-challenged and Control mice were associated with neurological disorders encompassing depression symptoms. Across cell types, S100 calcium binding protein A9 (S100A9), interleukin 1 beta (Il1b) and kynurenine 3-monooxygenase (Kmo) were differentially expressed between challenged and control mice. Immune response, chemotaxis, and chemokine activity were among the functional categories enriched by the differentially expressed genes. Functional categories enriched among the 9,117 genes differentially expressed between cell types included leukocyte regulation and activation, chemokine and cytokine activities, MAP kinase activity, and apoptosis. More than 200 genes exhibited alternative splicing events between cell types including WNK lysine deficient protein kinase 1 (Wnk1) and microtubule-actin crosslinking factor 1(Macf1). Network visualization revealed the capability of microglia to exhibit transcriptome dysregulation in response to immune challenge still after resolution of sickness symptoms, albeit lower than that observed in macrophages. The persistent transcriptome dysregulation in the microglia shared patterns with neurological disorders indicating that the associated persistent depressive symptoms share a common transcriptome basis.

RNA extraction from microglia followed the Tripsin method using a total RNA Kit (Omega Biotek, Norcross, GA) and a DNase step to remove DNA contamination [11]. The Agilent 2100 Bioanalyzer with RNA Pico chip (Agilent Technologies, Palo Alto, CA) was used to assess RNA Integrity Numbers. RNA Integrity Numbers were > 9 in 90% of the samples and > 7 in 100% of the 48 samples.

Identification of differential transcript isoform and gene expression
RNA libraries from individual mouse were sequenced using an Illumina HiSeq 2000 (Illumina, San Diego, CA) and 100nt long paired-end reads were obtained. Read quality control was implemented using FastQC [8,16]. Quality control analysis indicated that the Phred nucleotide quality score was > 30 across the length of the reads such that the read sequences were not trimmed.
Reads were mapped to the Genome Reference Consortium GRCm38 mouse assembly using Tophat2 (v 2.0.8) [17] with the Illumina iGenomes package (mm10; http://support.illumina. com/sequencing/sequencing_software/igenome.html). Reads were combined into transcripts and differential expression was tested using Cufflinks (v2.1.1) [18]. The specifications beyond the default settings used to obtain transcript abundance levels were: a) Upper Quartile Normalization was used to normalize the median transcript counts across libraries; b) multi-read correction for reads mapping to multiple sites; and c) fragment bias correction [19]. Four comparisons were evaluated: two pairwise contrasts between BGG-challenge groups within cell types, and two pairwise contrasts between cells within BCG-challenge groups. Results from genes with at least 10 mapped reads were considered. Multiple test adjustment used the Benjamini-Hoechberg false discovery rate (FDR) approach [20,21]. Genes exhibiting an FDRadjusted P-value < 0.05 were considered differentially expressed.

Identification of alternative splicing events
Alternative splicing events characterizing differences between groups (i.e., cell types or challenge level) were identified using a quantitative approach [22]. Quantitative characterization considered genes detected in both groups, represented by at least two transcript isoforms, and with at least one transcript isoform differentially expressed between groups (FDR-adjusted Pvalue < 0.05) between groups in one direction (over-or under-expressed) and the rest of the isoforms not differentially expressed or differentially expressed in the other direction.

Identification of functional categories over-represented among gene profiles
Functional analyses of the transcript isoforms differentially expressed between BCG-challenge groups within cell types, between cell types within BCG-challenged groups, and of transcript isoforms expressed exclusively in one cell type were performed using hypergeometric testing and Gene Set Enrichment Analysis (GSEA). These analyses allowed the identification of Gene Ontology (http://www.geneontology.org/) biological processes, molecular functions, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (http://www.genome.jp/org/). The hypergeometric test was implemented in the Database for Annotation, Visualization and Integrated Discovery (DAVID, http://david.abcc.ncifcrf.gov) [23,24]. Gene Ontology (GO) results were reported using the DAVID Functional Annotation Tool (FAT) classes to facilitate interpretation. Category enrichment was measured using Expression Analysis Systematic Explorer (EASE) scores computed based on a one-tailed jackknifed Fisher exact test [25,26]. The GO categories were clustered and the statistical significance of each cluster was assessed using an Enrichment Score (-log 10 geometric mean of the cluster members EASE scores [27,28]. Functional annotation clusters were considered significant at Enrichment Score > 2 (comparable to P-value < 0.001) using the Mus musculus genome as background. Gene Set Enrichment Analysis was implemented using the GSEA-P software package [29,30]. This approach offered functional insights complementary to DAVID based on the consideration of the expression profile of all the genes analyzed and annotations to the Molecular Signature Database (MSigDB) [31].

Gene network visualization
Interpretation of findings and discovery of differences in gene co-expression associated with BCG challenge or cell type were enhanced using network visualization. Networks were visualized using the BisoGenet plug-in [27,28,32] within the Cytoscape environment [24,33]. Biso-Genet enabled the visualization of associations between genes detected in the present study using information from the SysBiomics repository. Networks depicting genes as nodes and gene associations as edges and including at most one non-measured gene connecting observed genes were considered. The resulting gene networks were augmented with information on differential gene expression results from the comparison between BCG-challenged and cell type groups obtained in this study. The node size represented the differential expression P-value and the node color denoted over-or under-expression between challenge levels or cell types.

Results and Discussion
Global transcriptome profiles across immune challenge groups and cell types Among more than 15,000 genes tested, 562 and 3,851 were differentially expressed (FDRadjusted P-value < 0.05) between the BCG-challenged and Control groups in microglia and macrophages, respectively (Table 1). Differential abundance of transcript isoforms between BCG-challenged groups followed similar patterns albeit at lower absolute count number. Significant gene expression differences between cell types were more prevalent and extreme than between BCG-challenged groups. Approximately 9,117 genes (9,780 transcript isoforms) were differentially expressed between microglia and macrophages in the BCG-challenged group. Among these, 6,590 genes (8,144 transcripts) overlapped between cell types across BCG-challenged groups. The prevalent differential expression between cell types is consistent with prior transcriptomic comparisons and has been linked to differences in cell type origin and local environment [7,34,35].

Differences in the microglia transcriptome between BCG-challenged and Control mice
Differential gene and transcript isoform expression in microglia between BCG-challenged and Control mice are presented in Table 1. Among the differentially expressed genes in the microglia, 518 genes were over-expressed whereas only 44 genes were under-expressed in the BCGchallenged relative to Control mice. The predominance of over-expressed genes (92%) in response to BCG-challenge is consistent with previous reports of transcriptome changes in the microglia in response to lipopolysaccharide (LPS) challenge [36]. Differentially expressed genes between BCG-challenged and Control in microglia ((|log 2 (fold change)| > 3, FDR-adjusted P-value < 2.0 x 10 −3 ) are listed in Table 2 together with supporting literature. An extended list of differentially expressed genes in microglia is provided in S1 Table. Studies of behavioral and molecular changes in response to BCG challenge have exposed the relationship between brain inflammation and incidence of depression-like symptoms [1,2]. These studies have reported increased levels of inflammatory cytokines expression in the microglia. Similarly, the present study detected over-expression of genes related to inflammatory response 7 days after challenge ( Table 2, S1 Table) including Serum amyloid A3 (Saa3), S100 calcium binding protein A8 (S100a8), S100 calcium binding protein A9 (S100a9), Fcgamma receptors (Fcgr4 and Fcgr2b), prostaglandin-endoperoxide synthase 2 (Ptgs2), interleukin 1beta (Il1b), and interleukins receptors (Il1r1 and Il4ra). Consistent with our findings, an immune challenge event can elicit over-expression of Saa3 in the brain. Also, the levels of Saa3 in microglia increase in response to mouse hepatitis virus-JHM [62] and a strain of Creutzfeldt-Jakob disease [63]. The expression of Saa3 could be induced by S100A8 and S100A9 and these were over-expressed as well. Altered expression of S100A9 has been reported in neurological diseases associated with inflammation and depression such as cerebral ischemia, traumatic brain injury, and Alzheimer's disease [64,65].
Fc receptors are expressed in immune cells and connect humoral and cell-mediated response to pathogen infection [66]. Polymorphism in these receptors were associated with susceptibility to Guillain-Barré syndrome and multiple sclerosis, disorders that exhibit depression symptoms [67]. Increased levels of Fcgr4 and Fcgr2b expression, similar to that observed in this study, induce vascular damage and exacerbate neurodegenerative conditions in humans [68]. Activation of microglia through Fc gamma receptors results in phagocytosis and polarization to an M2b phenotype [68] and similarly, Fc receptor overexpression was correlated with over-expressed pro-inflammatory cytokines in the present and previous studies [69,70]. Similarly, Ptgs2 was over-expressed in the microglia in our study and has been related to inflammatory response in the brain and to neurodegeneration processes [71]. This gene has been proposed as a therapeutic target for neurodegenerative diseases that encompass depressive behaviors such as Parkinson, Alzheimer, and Huntington's disease [72]. Over-expression of pro-inflammatory cytokines such as interleukin Il1b has been linked to neurodegenerative disorders such as Alzheimer's disease [65,73]. Il1b has been linked to depressive behaviors [74,75] and therapies blocking Il1b expression in a mouse model of Alzheimer's disease decreased the synthesis of S100 proteins, decreased fibrillar deposition, and protected mice from cognitive deficits [76]. Our transcriptome findings in the microglia are consistent with reports that mice express depression-like behaviors 7 days after of inoculation with BCG [5,77]. Table 2. Most extreme differentially expressed genes between BCG-challenged and Control mice within microglia (|log 2 (fold change)| > 3, FDRadjusted P-value < 2.0 x 10 −3 ) and macrophages (|log 2 (fold change)| > 6, FDR-adjusted P-value < 3.0 x 10 −4 ) and supporting literature review. Ccl8 chemokine (C-C motif) ligand 8 7.45 [41] 2010002M12Rik interferon induced protein with tetratricopeptide repeats 1B like 2 7.32 Spon1 spondin 1, (f-spondin) extracellular matrix protein 7.23 [59] S100a8 S100 calcium binding protein A8 (calgranulin A) 7.15 [60] Cxcr2 chemokine (C-X-C motif) receptor 2 7.02 Ifng interferon gamma 6.45 [61] Cxcl9 chemokine (C-X-C motif) ligand 9 6.40

Differences in the macrophage transcriptome between BCG-challenged and Control mice
The comparison of gene expression profiles in macrophages between BCG-challenged and Control mice enabled us to uncover changes common to macrophage and microglia immune cells and changes unique to microglia cells. Differential gene and transcript isoform expression between BCG-challenged and Control mice in macrophages are listed in Table 1. Among the 3,851 differentially expressed genes in macrophages, 2,151 were over-expressed and 1,700 were under-expressed in the BCG-challenged relative to Control mice. The predominance of overexpressed genes (75%) in response to BCG-challenge is consistent with previous reports of transcriptome changes in the macrophages in response to BCG challenge [78] and consistent with the microglia changes previously described. Among these genes S100A9, S100A8, Il1b, interferon gamma (Ifng), nitric oxide synthase 2 inducible (Nos2), and interleukin 1 family member 9 (Il1f9) were identified. Like in microglia, many differentially expressed genes in macrophages were related with inflammatory response. Differentially expressed genes between BCG-challenged and Control in macrophages (|log 2 (fold change)| > 6, FDR-adjusted Pvalue < 3.0 x 10 −4 ) are listed in Table 2 together with supporting literature. An extended list of differentially expressed genes in macrophages is provided in S1 Table. The increase in expression of Ifng and Nos2 is a component in the macrophage response to infectious or inflammatory diseases represented in our study by the BCG challenge [61,79]. High levels of Ifng were found in patients diagnosed with major depression without pathogen stimuli [80][81][82][83][84]. Additionally, an increase in Ifng expression up-regulates Caspase 1 (Casp1), an enzyme that converts pro-Il1b into active mature Il1b [85] and that has a role in acute [86] and major depression [74,75]. Our results corroborated that the up-regulation of Ifng and Il1b (Table 2, S1 Table) is correlated with the development of depression-like behavior in BCGchallenged mice [5,77].

Transcriptome differences between BCG-challenged and Control mice shared between cell types
Among the genes differentially expressed between BCG-challenged and Control mice, 353 genes overlapped between microglia and macrophages (Table 1). Shared profiles were identified in genes belonging to families involved in immune response including: S100A8, S100A9, Il1r1, and Il1b. Kynurenine 3-monooxygenase (Kmo) was over-expressed (FDR-adjusted Pvalue < 0.05) in BCG-challenged relative to Control mice, both in microglia and macrophages (S1 Table). Kmo catalyzes the conversion of kynurenine to 3-hydroxykynurenine and modulation of KMO activity has been implicated in several neurodegenerative diseases [87]. Over-expression of Il1b and Kmo was observed in the microglia of mice that also exhibited social withdrawal after LPS injection [88]. Also, over-expression of Kmo was reported in the rat whole brain after a systemic challenge with LPS [89,90] and in human brain cells after Il1b treatment [91]. On the other hand, a study of mouse whole brain after BCG-challenge reported over-expression of indoleamine 2,3-dioxygenase 1 (Ido1) but no differential expression of Kmo [92]. Both, Kmo and Ido1 reduce the level of circulating tryptophan and production of serotonin and increase the levels of tryptophan metabolites that have cytotoxic effect [1,2]. In the present study, Kmo was over-expressed but Ido1 and the ortholog Ido2 were not differentially expressed in the microglia between BCG-challenged and Control mice. Our study supports the emerging notion that Kmo could be induced by immune challenge [93,94] and, as potential target of inflammatory cytokines in the kynurenine pathway, Kmo could be associated with behavioral disorders.

Functional analysis of microglia and macrophages transcriptomic differences between BCG-challenged and Control mice
Analysis of enriched functional categories based on the comparison of genes expression profiles between BCG-challenged and Control mice in microglia and macrophages offered insights into the biological processes and pathways associated with BCG-challenge. Table 3 summarizes the clusters of functional categories surpassing a DAVID Enrichment Score > 4. S2 Table presents the complete list of categories that exhibited an Enrichment Score > 2 and corresponding gene count.
Enriched functional categories shared by both cell types included immunological response, chemotaxis, cell migration and apoptosis among others biological processes and molecular functions ( Table 3, S2 Table). These clusters with highest enrichment overlap between BP~positive regulation of immune system process~regulation of leukocyte activation~regulation of cell activation 4.01 1 Each row corresponds to a cluster of Functional Annotation Tool (FAT) GO categories. 2 The three GO terms exhibiting most significant enrichment P-value in each cluster are listed, separated by "~". Additional information in each cluster is provided in S2 Table. doi:10.1371/journal.pone.0150858.t003 microglia and macrophages, corroborating the persistence of systemic response 7 days after BCG-challenge. The enrichment of apoptosis-related processes observed in this and previous studies [95] is associated with the over-expression of Nos2 in both cell types (Table 2). Likewise, the enrichment of monocyte and leukocyte migration processes (Table 3) reported here and by others [96] is in correspondence with the over-expression of S100A9 and S100A8 (Table 2) in both cell types. Higher leukocyte migration, activity of the pro-inflammatory cytokines [97], and apoptosis [98] were all observed in patients with major depression disorders. Additionally, the observed enrichment of protein metabolism including inhibition of enzymes and peptidases is compatible to changes in protein metabolism and downregulation of proteolysis reported in patients with bipolar depression [99,100] and schizophrenia [101].
Complementary functional analysis of the microglia and macrophages profiles using GSEA (Tables 4 and 5, respectively) confirmed the previous gene list enrichment and further enabled the discrimination of enrichment among over-and under-expressed genes in BCG-challenged Table 4. Gene Set Enrichment Analysis (GSEA) categories enriched among transcript isoforms overexpressed (FDR-adjusted P-value < 5.0 x 10 −4 > 10 transcript isoforms) and under-expressed (Pvalue < 0.05 > 10 transcript isoforms) in BCG-challenged relative to Control mice in microglia. relative to Control mice. S3 and S4 Tables include additional categories enriched at Pvalue < 0.05 and with a minimum of 10 genes in microglia and macrophages, respectively. A remarkable GSEA finding in microglia was the downregulation of inositol metabolism because this result is consistent with changes in this metabolite in schizophrenia, bipolar, and depression disorders [101,102].

Gene networks of microglia and macrophages transcriptomic differences between BCG-challenged and Control mice
Visualization of networks of genes differentially expressed between BCG-challenged and Control mice in microglia (Fig 1) and macrophages (Fig 2) augmented our understanding of the molecular relationships among genes. Networks including more than 5 connected genes are discussed. The predominance of over-expressed genes relative to under-expressed genes in both networks is in agreement with their predominance among the most differentially expressed genes in Table 2 and with the enriched categories detected using GSEA (Tables 4  and 5). These networks also exhibit a predominance of genes related to immune response. The microglia network (Fig 1) includes 57 gene nodes and 84% of the genes were overexpressed in BCG-challenged relative to Control (most at FDR-adjusted P-value < 0.001). In this network, Janus kinase 2 (Jak2), signal transducer and activator of transcription 3 (Stat3), signal transducer and activator of transcription 1 (Stat1), and nuclear factor erythroid derived 2 (Nfe2) were well-connected gene nodes linking to 11, 7, 5, and 5 other gene nodes, respectively. These genes were over-expressed in BCG-challenge relative to Control mice. Stat1 was affiliated to the enriched category "response to bacterium" in functional cluster 1 (Table 3) whereas nuclear factor erythroid derived 2 (Nfe2) was also over-expressed in BCG-challenge relative to the Control mice, although the corresponding categories were enriched below the set threshold.
The macrophages network (Fig 2) included 129 gene nodes and 66% of the genes were overexpressed in BCG-challenged relative to Control (FDR-adjusted P-value < 0.001). Casitas B- Table 5. Gene Set Enrichment Analysis (GSEA) categories enriched among transcript isoforms overexpressed (FDR-adjusted P-value < 1.5 x 10 −2 > 10 transcript isoforms) and under-expressed (FDRadjusted P-value < 5 x 10 −4 > 10 transcript isoforms) in BCG-challenged relative to Control mice in macrophages. lineage lymphoma (Cbl) was the most connected gene node, linked to 13 other nodes. This gene was over-expressed in BCG-challenged relative to Control mice yet was not affiliated to the functional categories enriched beyond the set threshold. Discs large homolog 4 (Dlg4) and tumor necrosis factor receptor-associated factor 1 (Traf1) were over-expressed genes in BCGchallenged relative to Control mice and each gene was connected to 7 gene nodes. Dlg4 was affiliated to enriched categories including behavior, endocytosis, and structural molecule activity across 3 functional clusters whereas Traf1 was affiliated to enriched categories related to apoptosis across 2 functional clusters (Table 3). Caspase 3 (Casp3) was linked to 5 other gene nodes and was also over-expressed in BCG-challenged relative to Control mice. Casp3 was affiliated to several enriched categories including apoptosis, kinase activities, and cells activation across 6 functional clusters (Table 3).

Transcriptome differences between microglia and macrophages from BCG-challenged mice
The study of transcript isoforms and genes differentially expressed between microglia and macrophages 7 days after BCG challenge offered additional insights into the roles of both cell types after immune challenge (Table 1). Over 9,780 transcript isoforms corresponding to 9,117 genes were differentially expressed between microglia and macrophages in BCG-challenged mice. The genes most differentially expressed between microglia and macrophages from BCG-challenged mice (log 2 (fold change) > |6|, FDR-adjusted P-value < 2.0 x 10 −4 ) are listed in Table 6 together with supporting literature review. An extended list of differentially expressed genes between microglia and macrophages is provided in S5 Table. The list of genes over-expressed in microglia relative to macrophages from BCG-challenged mice confirms the similar molecular mechanisms shared between neurological disorders and response to immune challenge that also share depressive behaviors. Among the differentially expressed genes, collagen type VIII alpha 1 (Col8a1) and potassium voltage-gated channel Iskrelated subfamily gene 2 (Kcne2) were over-expressed meanwhile arginase liver (Arg1) and Nos2 were under-expressed in microglia relative to macrophages from BCG-challenged mice. Col8a1 is up-regulated during repair processes in the mouse brain and collagen subunits have been linked to axonal guidance, synaptogenesis and Schwann cell differentiation [111,112]. Genomic rearrangements involving this gene have been linked to Tourette syndrome, a neuropsychiatric disorder that can encompass depression symptoms [113]. In addition to Col8a1, other genes over-expressed in microglia relative to macrophages from BCG-challenged mice have been linked to neuropsychiatric disorders that encompass depression symptoms (Table 6) including: leucine rich repeat protein 1 neuronal (Lrrn1), tetraspanin 7 (Tspan7), early B cell factor 3 (Ebf3), neuron navigator 3 (Nav3), and WW domain-containing oxidoreductase (Wwox). Over-expression of Lrrn1 was linked to autism spectrum disorder and Tourette syndrome [114]. Tspan7 was associated with Huntington's chorea, fragile X syndrome, and myotonic dystrophy [115]. Changes in the expression of Ebf3 was observed in obsessivecompulsive disorder [116]. Nav3 and WW domain-containing oxidoreductase (Wwox) have been implicated in Alzheimer's disease [117,118]. Nav3 was also associated with amyotrophic lateral sclerosis, Parkinson's disease [118], and was up-regulated in microglia compare to macrophages in response to hypoxia after stroke [106].
Kcne2 was among the genes over-expressed in microglia compared to macrophages in BCG-challenged mice. Expression of K+ channel genes follows nitric oxide changes after LPS stimulation indicating that K+ channels are involved in microglia activation [104]. Also, sialic acid binding Ig-like lectin H (Siglech) is considered a microglia signature gene [105] and G protein-coupled receptor 34 (Gpr34) is highly expressed in microglia and regulates the function, morphology, and phagocytosis of microglia during neuroinflammation [119,120].
Among the genes under-expressed in microglia relative to macrophages from BCG-challenged mice, the enzymes encoded by Arg1 and Nos2 compete for the same substrate: L-arginine [121,122]. The over-expression of Nos2 in macrophages is controlled by cytokines in response to a pathogen or inflammatory diseases [79] and also up-regulation of Nos2 and Arg1 were found in mouse and rat macrophages treated with LPS [123]. MicroRNA 155 (Mir155) was also over-expressed in macrophages relative to microglia (S5 Table). This microRNA is a crucial regulator of apoptosis and cell fate decisions in BCG-challenged macrophages and over-expression triggers mitogen-activated protein kinases (MAPK) cascades [124]. The genes under-expressed in microglia relative to macrophages after BCG challenge offer insights into the molecular mechanisms that are differentially regulated due to their potential negative impact in the brain.

GO cluster 1 Terms 2 ES
BP~response to wounding~inflammatory response~defense response 8.72 BP~regulation of cell activation~regulation of leukocyte activation~positive regulation of immune system process 6.72 BP~cell migration~localization of cell~cell motility 6.55 BP~regulation of cytokine production~regulation of cytokine biosynthetic process~positive regulation of cytokine biosynthetic process 6.44 MF,BP~purine nucleotide binding~ribonucleotide binding~purine ribonucleotide binding 6.14 MF,BP~taxis~chemotaxis~chemokine activity 6.04 BP~cell death~death~programmed cell death 5.92 BP~regulation of apoptosis~regulation of programmed cell death~regulation of cell death 5.53 BP~regulation of cell activation~regulation of leukocyte activation~regulation of lymphocyte activation 5.26 BP~vasculature development~blood vessel development~blood vessel morphogenesis 4.59 BP~hemopoietic or lymphoid organ development~immune system development~hemopoiesis 4.59 BP~regulation of cytokine production~regulation of interferon-gamma production~positive regulation of cytokine production 4.26 MF~GTP binding~guanyl ribonucleotide binding~guanyl nucleotide binding 4.07 1 Each row corresponds to a cluster of Functional Annotation Tool (FAT) GO categories. 2 The three GO terms exhibiting most significant enrichment P-value in each cluster are listed, separated by "~". Additional information in each cluster is provided in S6 complete list of categories surpassing Enrichment Score > 2 and corresponding gene counts. Functional categories encompassing genes differentially expressed between cell types included leukocyte regulation and activation, chemokine and cytokine activities, MAP kinase activity, and apoptosis (Table 7). These results are consistent with reports of enriched categories in LPS-challenged microglia [125] and BCG-challenged macrophages [124]. Enrichment of apoptosis (Table 7) is associated with the over-expression of Nos2 and Arg1 in macrophages because these genes regulate the production of peroxynitrites that in turn induce apoptosis [126,127]. Enrichment of categories within over-or under-expressed genes in microglia relative to macrophages detected using GSEA offered confirmatory and complementary information. Table 8 lists GSEA findings at FDR-adjusted P-value < 0.01 and including > 10 transcript isoforms and S7 Table includes a more extensive list of categories enriched at FDR-adjusted Pvalue < 0.05 and including > 10 transcript isoforms per category. The enrichment of categories associated with immune response, ribosome and cytokine activity among genes underexpressed in microglia relative to macrophages could be related to the potential detrimental neurological effect of microglia activation [128]. Among the enriched categories including genes over-expressed in the microglia, dysregulation of the tight junction pathway has been related to anxiety behaviors and altered signaling in the brain [129]. Similarly enriched categories included brain, central and nervous system development categories are consistent with reports of functional signatures in microglia [105]. Alternative splicing between microglia and macrophages from BCGchallenged mice Significant evidence of alternative splicing was detected in the comparison between cell types relative to challenge levels. Table 9 summarizes genes with at least nine transcript isoforms and at least two differentially expressed (FDR-adjusted P-value < 0.05) between microglia and macrophages. S8 Table lists all genes that match the broader alternative splicing definition of at least one transcript isoform differentially expressed among multiple isoforms. Alternative splicing events differentiating cell types were identified in 387 genes including 1,607 transcript isoforms. Confirming our findings, at least 72 of these genes have recorded alternative splicing events in the microglia as described in the database of alternative splicing of brain cells [130]. The alternative splicing cassette characterized by the inclusion or exclusion of an exon was the most frequent type of splicing event with 68% of occurrences. Alternative 3 0 and alternative 5 0 splicing events characterized by alternative usage of a splicing site on the 3 0 and 5 0 end of an exon, respectively were the second most common and the less common events amounting to 25% and 7% of the transcript isoforms, respectively. Intron retention (the inclusion or exclusion of a segment previously annotated to be an intron); tandem cassette (the Table 9. Genes exhibiting an alternative splicing event between microglia and macrophages in BCG-challenged mice including at least nine transcript isoforms and at least two over-or under-expressed (FDR-adjusted P-value < 0.05) transcript isoforms between cell types. inclusion or exclusion of two or more tandem exons); and mutually exclusive exons (the inclusion of one exon in one transcript and inclusion of a different exon in another transcript) were identified in 19%, 13%, and 11% of the 387 genes investigated, respectively. More than one alternative splicing mode was observed in some genes including WNK lysine deficient protein kinase 1 (Wnk1), tripartite motif-containing 33 (Trim33), and SWI/SNF related matrix associated actin dependent regulator of chromatin subfamily c member 2 (Smarcc2). Among the genes exhibiting alternative splicing events in the microglia, several genes have been associated with neurological disorders. Schizophrenia has been associated with microtubule-actin crosslinking factor 1 (Macf1) [131,132], Wnk1 [131], and Fused in sarcoma (Fus) [133]. Our results confirm alternative splicing events that are shared between neurological disorders and microglia after immune challenge and both linked to depressive behaviors.
Gene networks of transcriptomic differences between microglia and macrophages from BCG-challenged mice Additional understanding of the relationship among genes in microglia and macrophages from BCG-challenged mice was gained from networks visualization. A network was depicted considering the 2,487 most extreme differentially expressed genes between microglia and macrophages (log 2 (fold change) > |2|, FDR-adjusted P-value < 2.0 x 10 −4 ). From these, subnetworks connecting more than 5 genes are presented to facilitate visualization and interpretation. A gene network was constructed that included 146 genes with 39% of the genes overexpressed in microglia and the rest under-expressed (Fig 3). The predominance of genes under-expressed in microglia relative to macrophages suggests that gene dysregulation leading to negative neurological effects is more contained and quickly resolved in microglia than in macrophages after a BCG challenge. Janus kinase (Jak2) was the most connected gene in the network, linked to 11 other genes (Fig 3). This gene was under-expressed in microglia relative to macrophages and was affiliated to the enriched categories of protein kinase activity, nucleotide binding, apoptosis, cell differentiation, and regulation of phosphorylation in four functional clusters (Table 7). Kit oncogene (Kit) was under-expressed in microglia relative to macrophages, was linked to eight other genes, and was affiliated to the enriched categories cell migration, chemotaxis, nucleotide binding, phosphorylation, protein kinase activity, apoptosis, leukocyte proliferation and differentiation distributed across 11 functional clusters (Table 7). Diaphanous homolog 1 (Diap1) was also under-expressed in microglia, was connected to seven gene nodes, and was affiliated to the functional categories actin cytoskeleton and protein organization that were enriched below the threshold.
The most connected genes over-expressed in microglia relative to macrophages (Fig 3) included Rous sarcoma oncogene (Src), ubiquitin specific peptidase 2 (Usp2), and transforming growth factor, beta receptor I (Tgfbr1) and these genes were linked to 9, 7, and 5 other genes, respectively. Src is affiliated to enriched categories including cell migration, protein kinase activity, nucleotide binding, apoptosis, and cell adhesion in six clusters (Table 7). Ups2 is affiliated to enriched categories including phosphate and phosphorus metabolic process in one cluster and Tgfbr1 is affiliated to enriched categories including cell migration, nucleotide binding, phosphorylation, protein kinase activity, angiogenesis and apoptosis in seven clusters ( Table 7). The genes most connected in the network enriched apoptosis, cell migration, chemotaxis, and cell adhesion categories. This finding is in agreement with reports of the functional response of macrophages [124,134] and microglia [135] after a bacterial challenge.
This research studied the transcriptome of whole-brain microglia. However, microglia from different brain regions is likely to contribute unequally to depressive behaviors. Likewise, some brain regions are more responsive to periphery immune challenges than others. Additional studies that can parse the differential effect of periphery BCG challenge across brain regions and the corresponding changes in the transcriptome will offer more precise insights into the association between the transcriptome changes in response to immune challenge within brain regions and corresponding depressive-like behaviors.

Conclusions
The impact of immune challenge on the microglia after recovery from sickness and the implications on depressive behaviors was studied. The transcriptome of microglia 7 days after BCG challenge was compared to the corresponding transcriptome from unchallenged Control mice and to macrophages from the same mice. The number of genes differentially expressed between Network of genes differentially expressed between microglia and macrophages in BCGchallenged mice. Red (green) nodes denote genes over-(under-) expressed in BCG-challenged relative to Control mice. All genes were differentially expressed at FDR-adjusted P-value < 0.0001. Edges denote known relationships between genes in the SysBiomics repository. Framed genes (squares) are discussed in the manuscript. BCG-challenged and Control mice suggests the capacity of microglia to restrain or quickly resolve transcriptomic dysregulation relative to macrophages. The differential expression of Kmo in microglia between BCG-challenged and Control mice suggests that Kmo is a potential target of pro-inflammatory cytokines in the kynurenine pathway and a potential factor of depressive-like symptoms that remain after sickness symptoms subside. The over-expression of a number of genes, including Ifng and Il1b, in microglia relative to macrophages offered further evidence of transcriptome conditions associated with depression-like symptoms. Functional analysis highlighted the enrichment of categories including immune response and chemotaxis by genes over-expressed in BCG-challenged compared to Control mice in both cell types. Network visualization uncovered the key role of Jak2, Stat3, Stat1, and Nfe2 as hub genes, dysregulated and connected to other dysregulated genes in the microglia 7 days after BCG challenge.
The large number of differentially expressed genes between cell types from BCG-challenged mice speaks to the response of these cells to immune challenge. Also, a number of genes exhibiting differential splicing events between microglia and macrophages in this study have been linked to neurological disorders. Network visualization depicted the capability of microglia to exhibit transcriptome dysregulation after sickness recovery from immune challenge, albeit lower than macrophages. The present study provides ample evidence that the microglia transcriptome dysregulation after BCG challenge is shared with neurological disorders that also exhibit depressive behaviors.
Supporting Information S1 Table. Differentially expressed genes (FDR-adjusted P-value < 0.05) between BCG-challenged and Control mice within cell type and supporting literature review. (DOCX) S2 Table. Functional cluster (DAVID Enrichment score ES > 2) of categories enriched among differentially expressed genes between BCG-challenged and Control mice within cell type. (DOCX) S3 Table. Gene Set Enrichment Analysis (GSEA) categories enriched among transcript isoforms over-expressed (FDR-adjusted P-value < 0.05 and > 10 transcript isoforms) and under-expressed (Nominal P-value < 0.05 > 10 transcript isoforms) in BCG-challenged relative to Control mice in microglia. (DOCX) S4 Table. Gene Set Enrichment Analysis (GSEA) categories enriched among transcript isoforms over-expressed (FDR-adjusted P-value < 0.05 and > 10 transcript isoforms) and under-expressed (FDR-adjusted P-value < 0.05 > 10 transcript isoforms) in BCG-challenged relative to Control in macrophages. (DOCX) S5 Table. Differentially expressed genes (FDR-adjusted P-value < 2.0 x 10 −4 ) between microglia and macrophages in BCG-challenged mice and supporting literature review.  Table. Gene Set Enrichment Analysis (GSEA) categories enriched among transcript isoforms under-expressed (FDR-adjusted P-value < 0.05 and > 10 transcript isoforms) and over-expressed (P-value < 0.01 and > 10 transcript isoforms) in microglia relative to peripheral macrophages in BCG-challenged mice. (DOCX) S8 Table. Genes exhibiting an alternative splicing event between microglia and peripheral macrophages in BCG-challenged mice including at least two transcript isoforms and at least one over-or under-expressed (FDR-adjusted P-value < 0.05) transcript isoforms between cell types.