Transcriptional response of murine microglia in Alzheimer’s disease and inflammation

Alzheimer’s disease (AD) is a neurodegenerative disorder and is the most common cause of late-onset dementia. Microglia, the primary innate immune cells of the central nervous system (CNS), have a complex role in AD neuropathology. In the initial stages of AD, microglia play a role in limiting pathology by removing amyloid-β (Aβ) by phagocytosis. In contrast, microglia also release pro-inflammatory cytokines and chemokines to promote neuroinflammation and exacerbate AD neuropathology. Therefore, investigating microglial gene networks could identify new targets for therapeutic strategies for AD. We identified 465 differentially expressed genes (DEG) in 5XFAD versus wild-type mice by microarray, 354 DEG in lipopolysaccharide (LPS)-stimulated N9 microglia versus unstimulated control cells using RNA-sequencing (RNA-seq), with 32 DEG common between both datasets. Analyses of the 32 common DEG uncovered numerous molecular functions and pathways involved in Aβ phagocytosis and neuroinflammation associated with AD. Furthermore, multiplex ELISA confirmed the induction of several cytokines and chemokines in LPS-stimulated microglia. In summary, AD triggered multiple signaling pathways that regulate numerous genes in microglia, contributing to Aβ phagocytosis and neuroinflammation. Overall, these data identified several regulatory factors and biomarkers in microglia that could be useful in further understanding AD neuropathology.

could lead to new therapies to treat AD. Several studies have already highlighted the importance of specific genes involved in microglial metabolism [7] and response to Aβ plaque pathology [8]. In this study, we investigated the transcriptional response of microglia in an AD versus non-AD state using microarray and RNA-sequencing (RNA-seq). We identified 465 differentially expressed genes (DEG) in 5XFAD versus wild-type mice, and 354 DEG in lipopolysaccharide (LPS)-stimulated N9 microglia versus unstimulated control cells, with 32 DEG common to both experiments. Of the 32 DEG, functional enrichment analyses identified numerous processes and pathways in which microglia are potentially involved during AD development. Furthermore, multiplex ELISA confirmed the induction of several cytokines and chemokines in LPS-stimulated microglia that were also differentially expressed in the microarray and RNA-seq datasets. Overall, these data identify novel potential regulatory factors and biomarkers in the microglial response to AD.
We compared this data to publicly available data derived from sorted microglia from female 8-month-old wild-type and 5XFAD mice, a mouse model of AD which accumulates Aβ plaques [9]. Transcriptional analysis of microglia from 5XFAD versus wild-type mice was performed by microarray [10]. A total of 465 significant gene transcripts (FC > 2, FDR-adjusted P-value < 0.05) were identified with 337 up-regulated gene transcripts and 128 down-regulated gene transcripts ( Fig. 1B and Additional File 2). Glycoprotein (transmembrane) nmb (Gpnmb) was A Scatter plot of significantly altered genes (log 2 FC > 1.5, FDR-adjusted P-value < 0.05) by RNA-seq in N9 microglia stimulated with LPS (1 µg/ml) for 6 h versus unstimulated control cells. B Scatter plot of significantly altered transcripts (FC > 2, FDR-adjusted P-value < 0.05) by microarray in microglia isolated from the brains of female 5XFAD mice versus wild-type mice (8 months old). For both scatter plots, up-regulated genes are shown in red and down-regulated genes are shown in green. C Venn diagram showing the 354 significantly altered genes in N9 microglia by RNA-seq, 465 significantly altered transcripts in murine microglia by microarray, and 32 significantly altered genes found in both the RNA-seq and microarray data  the most up-regulated gene transcript (FC = 46.43) and X-linked lymphocyte-regulated 4B (Xlr4b) was the most down-regulated gene transcript (FC = -9.76) ( Table 2). In total, 32 genes overlapped between the N9 RNA-seq experiment and 5XFAD microarray dataset (Fig. 1C). Of the 32 total genes, 31 were up-regulated and only one was downregulated ( Table 3). The majority of the most up-regulated genes were cytokines and chemokines involved in inflammation (Cxcl10, Cxcl2, Il1β, Tnf). Transthyretin (Ttr) was the only down-regulated gene common to both datasets.   2C). KEGG analysis was performed on the 32 common genes found in both the microarray and RNA-seq datasets to identify pathways associated with AD. KEGG identified a total of seven pathways (P < 0.05) associated with the 32 genes (Fig. 3). The pathways included toll-like receptor signaling (Tlr2, Tnf, Il1β, Cxcl10), TNF signaling (Cxcl10, Cxcl2, Il1β, Tnf), phagosome (C3, Itga5, Olr1, Tlr2), NOD-like receptor signaling (Cxcl2, Il1β, Tnf), cytokine-cytokine receptor interaction (Cxcl10, Cxcl2, Il1β, Tnf), complement and coagulation cascades (C3, Plaur, Serpine1), and NF-kappa B signaling (Bcl2ald, Il1β, Tnf).

Pathway and functional prediction of microglia in AD
To further understand the interactions of the DEG, we performed protein-protein interaction (PPI) analysis on the 32 genes found in both the RNA-seq and microarray datasets using STRING. For the analysis, text mining, experiments, and databases were chosen for the interaction sources, and the high confidence value of 0.700 was selected as the minimum required interaction score threshold. Of the proteins encoded by the 32 identified DEG, two distinct networks emerged with 15 proteins in one network (IL1β, Irak3, Il1rn, Cxcl2, Cxcl10, Tnf, Tlr2, Olr1, Cd83, Cd69, Cd274, Serpine1, Plaur, Itga5, Mmp12), three proteins in the second network (Ifit2, Ifit2, Oasl2), and 13 of the proteins not clustering (Fig. 4). The database did not recognize predicted gene 1966 (Gm1966). The results suggest that the given proteins were highly enriched (P < 1 × 10 -16 ). Pathways are shown in descending order based on -log 10 P-values. The number of genes associated with each GO term is shown above each bar. Only GO terms with a P-value < 0.05 are shown

Discussion
A thorough understanding of the molecular characteristics and regulatory pathways of microglia is essential to develop therapeutic strategies for AD. Accordingly, numerous studies have performed transcriptome analyses to elucidate the role of specific genes in microglia important in AD neuropathology, including Trem2 and ApoE [7,[10][11][12][13]. Other studies have investigated the transcriptome of microglia under distinct biological activities, such as Aβ plaque phagocytosis [14], tau pathology [15] and metabolism [16]. In this study, we used a two-pronged approach to determine DEG in AD and inflammation. First, we determined the DEG in microglia from 5XFAD mice versus wild-type mice by microarray. We then determined the DEG in LPS-stimulated microglia using RNA-seq. Altogether, 32 DEG overlapped between the two datasets.
Neuroinflammation by microglia is hypothesized to exacerbate AD neuropathology [17]. In our study, several immunological pathways associated with inflammation were identified in AD-associated microglia that correlated with DEG of well-known pro-inflammatory cytokines and chemokines, such as IL-1β, TNF, CXCL10, and CXCL2. Caspase-1 dependent IL-1β secretion occurs during NLRP3 inflammasome activation, and activation of the inflammasome has emerged as an important mechanism in chronic neuroinflammation in AD [18]. Furthermore, immunotherapies blocking inflammasome activation are being investigated for use in AD and other inflammatory diseases [19][20][21]. BP GO analysis also identified Olr1 as an inflammatory and immune system response gene. Several studies suggest a role for Olr1 in lipid metabolism, and genetic variation in Olr1 as a risk factor for AD [22][23][24], but little is known about Olr1 in AD-associated neuroinflammation. Furthermore, Olr1 was recognized as a gene involved in the phagosome pathway by KEGG, suggesting an important role in multiple AD pathogenic processes.
Genes not previously characterized in AD were identified by KEGG to be involved in molecular signaling pathways important in AD development. The B cell leukemia/ lymphoma 2 related protein A1d (Bcl2a1d) was identified as being involved in the nuclear factor kappa-light-chainenhancer of activated B cells (NF-κB) signaling pathway. Activation of NF-κB is an important mechanism in chronic neuroinflammation that significantly increases AD pathology [25]. Furthermore, the plasminogen activator, urokinase receptor (Plaur) was recognized as being associated with complement and coagulation cascades. Complement and coagulation cascades are associated with blood-brain barrier dysfunction and AD progression in mice [26] and humans [27]. Characterization of these genes in the context of AD could further our Biomarkers can be a useful tool to predict and diagnosis a variety of neurological disorders, including AD [28]. In our study, the only down-regulated gene amongst the 32 DEG was transthyretin (Ttr). Indeed, Ttr has been implicated in numerous processes associated with AD, including Aβ binding [29], Aβ transport at the bloodbrain barrier (BBB) [30], toxicity [31,32], and neuroprotection by interfering with Aβ formation [29,33,34]. Ttr has been previously identified as being decreased in the cerebral spinal fluid (CSF) [35] and blood [36][37][38] of AD patients, and is being investigated as a blood biomarker for AD [38,39]. Several other DEG in our study could also potentially be used as biomarkers due to their highlevel of differential expression between groups (greater than threefold in both assays), including Ifit3, Cxcl10, Cxcl2, Il1β, Oasl2, Ifit2, and Slfn5. Further investigation will be needed, as clinically useful biomarkers for AD should be inexpensive, non-invasive, reliably detectable and able to distinguish AD from other forms of dementia [40].
There are several limitations of our study which are common to most studies directed at identifying dynamic alterations in gene expression networks in AD. Microgliosis is a hallmark of AD neuropathology, resulting in large numbers of microglia concentrated around Aβ plaques [41]. Since the role of microglia in AD is not fully understood, it is difficult to ascertain whether the altered genes identified in our study are involved in disease progression or in the neuroprotective microglia response. Another caveat to our study is that murine microglia were used in both experiments. Although the value of murine microglial research in AD cannot be overstated, the transcriptional signature of microglia in human AD is drastically different than that of murine microglia [11,42]. When comparing human and mouse AD models, transgenic mice overexpress Aβ in a non-physiological manner, resulting in rapid Aβ accumulation and a higher plaque burden when compared to human AD [43,44]. Although the 5XFAD mouse model does not perfectly recapitulate human disease studies, comparing the microglial response in humans and mice have been mixed, with some indicating that the response may be different [11,42] and some studies indicating that the microglial response in both humans and mice may be similar [7,45]. This is likely due to differences in the complexity of human disease where both plaques and neurofibrillary tangles are present compared to the 5XFAD mice in which only plaques form. Therefore, in vitro and in vivo models of AD remain powerful tools to begin to dissect underlying microglial gene regulatory mechanisms in early AD pathology.

Conclusions
In summary, we identified 465 DEG in 5XFAD versus wild-type mice by microarray, 354 DEG in LPS-stimulated N9 microglia versus unstimulated control cells using RNA-seq, with 32 DEG common between both data sets. Analyses of the 32 DEG uncovered numerous molecular functions and pathways involved in Aβ phagocytosis and neuroinflammation associated with AD that may be further investigated. Furthermore, multiplex ELISA confirmed the induction of several cytokines and chemokines in LPS-stimulated microglia. Overall, these data identified several regulatory factors and biomarkers in microglia that could be useful in further understanding AD neuropathology.

Ethical approval and consent to participate
All methods were carried out in accordance with relevant local and University of Wisconsin guidelines and regulations. All animals were handled in accordance with the Animal Research: Reporting of in vivo Experiments (ARRIVE) guidelines and the University of Wisconsin's Institutional Animal Care and Use Committee policies and our approved protocols.

Cell culture assays
BMDM were prepared as previously described [19]. Briefly, tibias were removed from wild-type C57BL/6J mice (6-10 months old) (The Jackson Laboratory), flushed with complete RPMI media supplemented with 20% L-cell conditioned media (LCCM) [19], and cultured for 4-7 days prior to use. Differentiated BMDM were seeded at a cell density of 400,000 cells/well in a 24-well tissue culture plate. Cells were stimulated with ultrapure LPS (50 ng/ml) from Escherichia coli O111:B4 Microglia were prepared from wild-type C57BL/6J mice (5-8 months old) (The Jackson Laboratory) as previously described [7]. Microglia were isolated using Magnetic Activated Cell Sorting (MACS, Miltenyi Biotec) according to manufacturer's instructions. Briefly, mice were perfused with cold PBS containing 0. . LS columns were washed three times with MACS buffer (PBS containing 0.5% BSA and 1 mM EDTA) before elution. This method allows for preparation of microglial cultures of high purity (> 95%). Purity of isolated microglia are routinely confirmed by flow cytometry. Approximately 50,000 microglia/well were seeded on poly-L-lysine coated 24-well plates in complete RPMI media supplemented with 20% LCCM [19] and human TGF-β (10 ng/ml) (PeproTech, Cat. No. 100-21). Media was changed three days after plating, and cells were used in the week following the media change. Microglia were stimulated with LPS (50 ng/ml) for 6 h. Supernatants were flash frozen on dry ice and stored at -80℃ until use.
Immortalized murine N9 microglia were cultured as previously described [46]. N9 microglia were seeded at a cell density of 250,000 cells/well in a 24-well tissue culture plate. Cells were stimulated with LPS (1 µg/ml) [47] for 6 h. Supernatants were flash frozen on dry ice and stored at -80℃ until use.

Fig. 6
Cytokines and chemokines secreted by N9 microglia following LPS stimulation. N9 microglia were stimulated with LPS (1 µg/ml) for 6 h. Supernatants were collected and assayed for 22 cytokines and chemokines using a multiplex assay. Experiments were performed in triplicate. Results are from three independent experiments. Data are shown as mean ± SEM. ****P ≤ 0.0001, ***P ≤ 0.001

RNA-seq
RNA was extracted from immortalized N9 microglia stimulated with LPS (1 µg/ml) [47] for 6 h using a RNeasy Plus Mini Kit (Qiagen, Cat. No. 74134). The N9 microglial cell line is derived from mouse brain and shares numerous phenotypic traits with primary mouse microglia [48]. Quality and quantity of RNA was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies) and a Nanodrop spectrophotometer (Thermo Scientific). All samples had an RNA integrity number (RIN) of 9.7 or higher. RNA library preparation and transcriptome sequencing were performed by Novogene using the Illumina NovaSeq 6000 Sequencing System. Bioinformatics analysis was performed by Novogene with differential expression analysis performed using the DESeq2 R package (1.20.0). The resulting P-values were adjusted using the Benjamini and Hochberg method for controlling the false discovery rate [49]. Genes with FDR-adjusted P-values < 0.05 and log2FC > 1.5 were considered differentially expressed.

Microarray
The microarray has been published in a previous study [10] and the publicly available dataset was used (GSE65067). Briefly, microglia from female 8 month old wild-type (n = 3) and 5XFAD (n = 5) mice (The Jackson Laboratory) were FACS-sorted directly into RTL-plus lysis buffer. RNA extraction from microglia was performed using a RNeasy Plus Micro Kit (Qiagen, Cat. No. 74034). Microarray hybridization (Affymetrix MoGene 1.0 ST array) and data processing were performed at the Washington University Genome Center. Raw data were normalized using the Robust Multi-Array (RMA) method and genes were pre-filtered for expression value greater than or equal to 120 expression units. This method provides a cut-off above which genes have a 95% chance of expression demonstrated in Immgen dataset, which uses the same array platform [10]. P-values were calculated using a Student's t-test and FDR-adjusted P-values were calculated using the Benjamini and Hochberg method [49]. Transcripts with FDR-adjusted P-values < 0.05 and FC > 2 were considered differentially expressed.

Functional enrichment analyses
Genes found to be differentially expressed in both the RNA-seq and microarray experiments were selected for biological function and pathway analysis. The gene list was uploaded into the Database for Annotation, Visualization and Integrated Discovery (DAVID, v. 6.8) [50,51] for GO and KEGG pathway analysis. GO gene count thresholds of 4, 2, and 2 were used for BP, CC, and MF respectively. Each GO and KEGG pathway with a P-value < 0.05 was considered significant. Additionally, PPI analysis was performed to identify interactions of the selected proteins based on their gene IDs using the STRING database [52]. For the analysis, text mining, experiments, and databases were chosen for active interaction sources using the high confidence (0.700) threshold setting.

Statistical analyses
Statistical analysis was performed using Prism 9.0.0 (GraphPad). Data are presented as mean ± SEM. Comparison between two groups was performed using a Student's t-test. A P-value ≤ 0.05 (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 and ****P ≤ 0.0001) was used as the significance cutoff. The Venn diagram demonstrating overlap in DEG amongst the two datasets was generated using InteractiVenn [53].