Abstract
Dysregulation of microglial function contributes to Alzheimer’s disease (AD) pathogenesis. Several genetic and transcriptome studies have revealed microglia specific genetic risk factors, and changes in microglia expression profiles in AD pathogenesis, viz. the human-Alzheimer’s microglia/myeloid (HAM) profile in AD patients and the disease-associated microglia profile (DAM) in AD mouse models. The transcriptional changes involve genes in immune and inflammatory pathways, and in pathways associated with Aβ clearance. Aβ oligomers have been suggested to be the initial trigger of microglia activation in AD. To study the direct response to Aβ oligomers exposure, we assessed changes in gene expression in an in vitro model for microglia, the human monocyte-derived microglial-like (MDMi) cells. We confirmed the initiation of an inflammatory profile following LPS stimulation, based on increased expression of IL1B, IL6, and TNFα. In contrast, the Aβ1-42 oligomers did not induce an inflammatory profile or a classical HAM or DAM profile. Interestingly, we observed a specific increase in the expression of metallothioneins in the Aβ1-42 oligomer treated MDMi cells. Metallothioneins are involved in metal ion regulation, protection against reactive oxygen species, and have anti-inflammatory properties. In conclusion, our data suggests that Aβ1-42 oligomers may trigger a protective response both in vitro and in vivo.
Introduction
Alzheimer’s disease (AD) is a neurodegenerative disorder that is clinically characterized by progressive memory loss and impairment in cognitive functions (Querfurth and LaFerla, 2010). Hallmarks of AD pathology are the aggregation of amyloid-beta (Aβ) in extracellular plaques, intraneuronal hyperphosphorylated tau tangles (Hardy and Allsop, 1991; Selkoe, 1991), and reactive gliosis (Itagaki et al., 1989; Kato et al., 1998). Aβ plaques are surrounded and infiltrated by reactive microglia (Itagaki et al., 1989; Rozemuller et al., 1986).
Microglia are the resident immune cells of the brain (Ginhoux et al., 2013, 2010). In neurodegenerative diseases, including AD, microglia adopt an activated phenotype and undergo major morphological and functional changes (Wolf et al., 2017). Genome-wide association studies identified several AD risk loci that are found in or near genes predominantly expressed in microglia (APOE, TREM2, and CD33) (Hemonnot et al., 2019; Johnson et al., 2020; Lambert et al., 2013). The identification of a distinct AD-pathology-associated gene expression profile in microglia by recent transcriptome studies support the important role of microglia in AD (Mathys et al., 2019; Srinivasan et al., 2019). The AD-pathology-associated microglia profile was enriched in immune and inflammatory pathways, as well as in Aβ clearance pathways (Mathys et al., 2019). Srinivasan et al., 2019 referred to their Human Alzheimer’s Microglia/Myeloid cells profile as the HAM signature (Srinivasan et al., 2019). Single-cell RNA sequencing analysis of sorted microglia from the 5xFAD AD mouse model revealed a distinct microglia subtype, referred to as disease-associated microglia (DAM) (Keren-Shaul et al., 2017). Some well-known AD risk factors, including ApoE, Ctsd, Lpl, TYROBP, and TREM2 are upregulated in DAM (Keren-Shaul et al., 2017).
In AD, the acute activation of microglia by Aβ can be beneficial, as it increases phagocytosis and clearance of Aβ by microglia (Prokop et al., 2013; Rogers et al., 2002). However, when the activation of microglia becomes chronic, the cells contribute to neurotoxicity by the release of pro-inflammatory cytokines and mediate synapse loss (Bamberger et al., 2003; Sarlus and Heneka, 2017). Aβ in the form of monomers, oligomers, and fibrils, has multiple biological effects, including neurotoxicity and the activation of microglia (Lue et al., 2019; Walker et al., 2006, 2001). The Aβ oligomers are considered to be the main culprit in AD pathogenesis events (Esparza et al., 2013; Selkoe and Hardy, 2016).
In this study, we investigated the direct effect of Aβ1-42 oligomers on the microglia transcriptome. We decided not to use microglia from aged human brain tissue as these are likely to already have been exposed to Aβ. Furthermore, the isolation procedure of microglia from post-mortem human brain tissue changes its activation state, leading to the loss of classical microglial marker expression once cultured (Gosselin et al., 2017). Therefore, we used the human monocyte-derived induced microglia-like (MDMi) cell model as a source of human microglia to investigate the transcriptomic changes induced by human Aβ1-42 oligomer stimulation. The MDMi cell model is based on established protocols to differentiate human monocytes towards a microglia-like phenotype (Leone et al., 2006; Ohgidani et al., 2015; Ryan et al., 2017), with minor adaptations (Ormel et al., 2020). In short, monocytes isolated from healthy controls are differentiated in MDMi cells using several factors that are important for microglia development. Studies have shown that the MDMi cell models display characteristics of central nervous system resident microglia and can be used for transcriptome analysis and functional assays (Leone et al., 2006; Ohgidani et al., 2015; Ryan et al., 2017; Sellgren et al., 2019). These findings confirm that MDMi cells can be used as a model to study human microglia function in health and disease.
We first describe the generation of the MDMi cell model and investigate their response to the classical inflammatory stimulus lipopolysaccharide (LPS), which confirmed the initiation of an inflammatory profile, based on an increased expression of IL1B, IL6, and TNFα. We then exposed the MDMi cells to Aβ1-42 oligomers to study changes in the transcriptome and determined whether the expression of the most significant upregulated genes was also increased in human primary microglia isolated from AD compared to non-demented control (NDC) cases. We also compared three well-known AD microglia gene profiles (Keren-Shaul et al., 2017; Mathys et al., 2019; Srinivasan et al., 2019) to our stimulated and unstimulated MDMi cells. Our results showed that the Aβ1-42 oligomers did not induce an inflammatory profile or a classical HAM or DAM profile. Instead, we observed a specific increase in the expression of metallothionein genes following Aβ1-42 oligomer treatment. Our work suggests that Aβ1-42 oligomers may initially induce a protective response in MDMi cells.
Materials and methods
Data and code availability
The bulk RNA sequencing dataset generated in this study have been deposited in NCBI’s Gene Expression Omnibus (accession number:). All the analysis is described in the Methods.
Differentiation of monocyte-derived microglia-like (MDMi) cells
Monocyte isolation
Blood samples of 10 healthy controls were obtained from the Dutch blood bank (Sanquin, https://www.sanquin.nl/en) (Supplementary Table 1). Peripheral blood mononuclear cells (PBMCs) were enriched by density gradient separation using Ficoll (Ficoll-Paque™ plus, GE Healthcare). Monocytes were isolated from PBMCs with anti-CD14 conjugated magnetic microbeads (130-050-201, Miltenyi Biotec) according to the manufacturer’s protocol. Monocytes were immersed in 45% fetal calf serum (FCS, 10500 ThermoFisher Scientific), and 10% dimethyl sulfoxide (DMSO) in RPMI culture medium (RPMI 1640, Gibco with 100 Units/mL penicillin, 100 μg/mL streptomycin and 2mM L-glutamine) and stored in liquid nitrogen until further use.
Monocyte-derived microglia-like cell establishment
The protocol for generating MDMi cells is based on established protocols (Leone et al., 2006; Ohgidani et al., 2015; Ryan et al., 2017), with minor adaptations to optimize the expression of TREM2, TYROBP and PROS1 (Ormel et al., 2020). To generate MDMi cells, monocytes were thawed on ice-cold RPMI medium and plated at a density of 600.000 cells/well or 200.000 cells/well in a 48- or 96-well plate for transcriptome and phagocytosis assays, respectively. The wells were coated with poly-L lysine (PLL, Sigma Aldrich) at 37 °C in 5% CO2 for 30 minutes. Cells were left to adhere to the wells under standard humified culture conditions at 37 °C in 5% CO2 for at least 1 hr, then cells were washed with PBS (Invitrogen) and the medium was replaced by 25% ACM (Astrocyte conditioned medium SCC1811, Sanbio) in RPMI medium. To induce the differentiation towards MDMi cells, the 25% ACM/RPMI medium was replaced on day four and day eight with RPMI containing 25% ACM, 10 ng/ml human M-CSF (130-096-491, Miltenyi Biotec), 10 ng/mL GM-CSF (130-093-862, Miltenyi Biotec), 1 ng/ml TGFβ (130-095-067, Miltenyi Biotec), 12.5 ng/ml IFN-γ (130-096-872, Miltenyi Biotec), and 100 ng/mL IL-34 (130-108-997, Miltenyi Biotec).
Stimulation of MDMi cells with Aβ1-42 oligomers or lipopolysaccharide
At day 10 in culture, MDMi cells were treated with 500 nM stable human Aβ1-42 oligomers (180222EMH, gift from Crossbeta) or 100 ng/mL lipopolysaccharide in PBS (LPS, L4391-1MG, E. coli 0111:B4, Sigma-Aldrich). As a control for the oligomer and LPS condition, the same volume of vehicle (20 mM HEPES, 150 mM NaCl and 200 mM sucrose, pH 7.2) or PBS was added, respectively. After 24 hrs of stimulation, MDMi cells were washed with PBS and collected for RNA isolation (48-well plates) in Trizol (Ambion, Life Technologies, Carlsbad, CA, USA) or used for phagocytosis assay.
RNA isolation, library preparation, and sequencing
RNA isolation was performed using the miRNeasy Mini kit (217004, Qiagen) according to the manufacturer’s protocol including the DNase treatment. The RNA concentration was determined using a Nanodrop 2000 spectrophotometer (Thermo Scientific, Waltham). cDNA libraries were prepared according to the CelSeq2 protocol (Hashimshony et al., 2012; Simmini et al., 2014) by Single Cell Discoveries, Utrecht (Muraro et al., 2016). Briefly, for each sample a custom-made primer was added to the RNA, denatured at 70 °C for 2 min, and immediately cooled down. mRNA was reverse transcribed into cDNA using clean-up beads (ThermoFisher Scientific, Ambion, Waltham, USA). The purified cDNA was transcribed in vitro to obtain amplified RNA using the MegaScript kit, which was followed by purification with Agencourt RNAclean XP (RNAse free) beads (Beckman Coulter). Amplified RNA was fragmented with fragmentation buffer and purified using the RNA cleanXP beads. Quality of the amplified RNA was determined with the RNA 6000 Pico chips (Agilent) on an Agilent 2100 bioanalyzer. Next, the amplified RNA was reverse transcribed and amplified by PCR. Libraries were labelled with a 4 bp unique molecular identifier (UMI) that was added to the primer. Samples from the different stimulation and control conditions were pooled in libraries for each donor (see Supplementary Table 1). The quality of the libraries was determined with the DNA High Sensitivity chips on an Agilent 2100 bioanalyzer (Agilent). Libraries were sequenced on the Illumina NextSeq 500 platform using paired-end sequencing (75 bp) with a depth of 10M reads per sample.
Bulk RNA sequencing analysis
Libraries were de-multiplexed and raw reads were aligned to the hg19 human RefSeq transcriptome with Burrows-Wheeler Aligner (BWA) (Li and Durbin, 2010). Duplicate reads and reads that mapped equally well to multiple locations were discarded. The quality control, normalization, and identification of differentially expressed genes were done with DESeq2, an R (version 3.6.3) based package (Love MI et al 2014). Samples that had more than 10M reads were discarded (see Supplementary Table 1). Read counts were normalized to transcripts per million (TPM). The gene expression was corrected for the covariate sex. Genes were considered differentially expressed with an adjusted (adj.) p-value of less than 0.05 and a Log2 fold change of at least 2 for the comparison LPS and PBS and a Log2 fold change of at least 0.5 for Aβ1-42 oligomers and vehicle stimulation.
Gene ontology enrichment and heatmap analysis
The gene ontology (GO) (GO Molecular Function 2018) and Panther pathway analyses (Panther 2016) were performed on the list of differentially expressed genes. The lists of genes used for analysis are provided as Supplementary Table 2 and 3. We loaded these genes in EnrichR (http://amp.pharm.mssm.edu/Enrichr/), a web-based tool for enrichment analysis (Chen et al., 2013; Kuleshov et al., 2016). Heatmaps were created using the Morpheus Broad Institute Software (https://software.broadinstitute.org/morpheus/).
Comparison with previously published datasets
Srinivasan et al., 2019 – Human Alzheimer’s microglia/myeloid cells (HAM) profile - Differential gene expression results of myeloid cells of AD and control subjects were downloaded from http://research-pub.gene.com/BrainMyeloidLandscape/BrainMyeloidLandscape2/#study/study/GSE125050/studyReport.html (see Supplementary Table 4). Briefly, Srinivasan et al., 2019 performed RNA sequencing on FAC-sorted cells from frozen post-mortem brain tissue. Cells were isolated from fusiform gyrus tissue from 10 AD and 15 control cases (Srinivasan et al., 2019).
Mathys et al., 2019 – Differential gene expression results of microglia were obtained from Supplementary Table 2 and 7 (Mathys et al., 2019). Mathys et al., 2019 performed single-nuclear RNA sequencing on nuclei from frozen post-mortem tissue of 24 AD and 24 no-pathology cases. A total of 1920 microglia nuclei were obtained (Mathys et al., 2019).
Keren-Shaul et al., 2017 – Disease-associated microglia (DAM) profile - The differential gene expression data of isolated cortical cells from three 5xFAD transgenic and three wild-type 6-month-old mice were obtained from Supplementary Table 2 (Keren-Shaul et al., 2017). This study identified 500 differentially expressed genes that make-up the disease-associated microglia (DAM) profile with single-cell RNA sequencing (Supplementary Table 4) (Keren-Shaul et al., 2017).
Phagocytosis assay
Fluosphere carboxylate-modified microspheres (2 μm, yellow-green fluorescent, F8827, ThermoFisher) were used for the phagocytosis assay. After 10 days of differentiation, MDMi cells were stimulated with Aβ1-42 oligomers or LPS for 24 hrs. In the unstimulated control conditions vehicle (20 mM HEPES, 150 mM NaCl and 200 mM sucrose, pH 7.2) or PBS was added to the culture medium. Next, the MDMi cells were cultured with the uncoated fluorescent beads (3 beads/cell) under standard humified culture conditions at 37 °C in 5% CO2 for 1 hr. For each condition, three cell culture wells were used per donor. After 1 hr incubation with the beads, the cells were washed twice with cold PBS to remove non-phagocytized beads. The cells were detached using Trypsin 1x (15090, ThermoFisher Scientific) and 0.5 M EDTA, pH 8.0 in PBS under standard humified culture conditions at 37 °C in 5% CO2 for 5 min. Fluorescence activated cell sorting (FACS, BD FACSCanto™ II, software version 8.0.1) was used to analyze >1000 living cells, gated by FSC and SSC. Phagocytosis of fluorescent beads was gated by FSC and the blue 1-a (laser 488, filter 530/30) channel. The gates were set by first sorting MDMi cells without beads, monocytes, and beads only. These gate settings were later used to sort the positive MDMi cells.
Immunocytochemistry
Monocytes were plated on PLL-coated coverslips in a 24-well plate. After 10 days of differentiation, MDMi cells were washed with PBS containing 137 mM NaCl, 1.8 mM KH2PO4, 5.96 mM Na2HPO4.2H2O, 2.7 mM KCl, pH 7.4 and fixed in 4% PFA. MDMi cells were incubated with a primary antibody against Iba1 (1:1,000; 019-19744, FUJIFILM Wako Chemicals) in blocking buffer (2% bovine serum albumin, 0.1% TritonX100 (Merck, Darmstadt, Germany), and 5% normal donkey serum (017-000-121, Jackson ImmunoResearch) in PBS). Following washes with PBS, the secondary antibody donkey-anti-rabbit 488 (1:700; 715-585-150, Jackson ImmunoResearch) and Hoechst (1:1,000; H3569, ThermoFischer Scientific) as a nuclear staining were added. Finally, MDMi cells were washed with PBS and embedded with Mowiol (0.1 M Tris-HCl, pH 8.5, 25% Glycerol, 10% w/v Mowiol 4-88). Images were taken on a Zeiss AxioScopeA1 microscope using an EC Plan-Neofluar 20x/0.50 M27 objective, an AxioCam MRm camera (Zeiss), and Zen 2011 software. Images were taken with a resolution of 1388 × 1040 pixels. The phagocytosis assay was imaged with a Zeiss LSM 880 confocal laser microscope using a 40x/1.3NA oil DICII objective (EC PlnN), and Zen black Z.1SP3 software. Images were taken with a z-step of 0.7 μm and a resolution of 1024 × 1024 pixels. The phagocytosis assay was imaged with a Zeiss LSM 880 confocal laser microscope using a 40x/1.3NA oil DICII objective (EC PlnN), and Zen black Z.1SP3 software. Images were taken with a z-step of 0.7 μm and a resolution of 1024 × 1024 pixels.
Isolation of human microglia
Human brain tissue was obtained from the Netherlands Brain Bank (NBB, www.brainbank.nl). Permission for brain autopsy and the use of brain tissue and clinical information for research purposes were obtained per donor ante-mortem by the NBB. The donor identity was pseudo-anonymized by the NBB. For some donors, the amyloid and Braak scores still have to be confirmed by a neuropathologist (Supplementary Table 5).
Our microglia isolation protocol was adapted from Melief et al., 2016. Briefly, the isolation of microglia started 6 to 24 hrs after autopsy. Approximately 4-10 grams of tissue of the gyrus temporalis superior (GTS1-3) was collected in cold Hibernate-A medium. GTS1-3 tissue was mechanically dissociated with a scalpel in ice-cold GKN-BSA containing 11.1 mM D-(+)-Glucose monohydrate (14431-43-7, Sigma Aldrich) and 50 mM bovine serum albumin (BSA, A450-3, Sigma-Aldrich) in PBS, pH 7.4 (10010031, Invitrogen). To obtain a cell suspension, tissue was enzymatically dissociated with Collagenase type I (370 units/ml, LS004196, Worthington Biochemical) and DNase I (100 μg/ml, 10104159001, Sigma-Aldrich) at 37 °C in a shaking incubator (140-170 rpm) for 1 hr. Dissociated GTS tissue was washed and centrifuged at 1,800 rpm at 4 °C, and the pellet was resuspended in GKN-BSA. The cell suspension was filtered through a 100 μm cell strainer (Corning, New York, USA). Percoll (GE Healthcare) gradient centrifugation was used to separate the different cellular fractions. Percoll was added dropwise to the cells in GKN-BSA and centrifuged at 4000 rpm at 4 °C for 30 min. The middle turbid layer (cellular fraction) was carefully collected and washed with an equal volume of GKN-BSA, centrifuged and suspended in MACS buffer containing 2 mM EDTA and 1% FBS in PBS 1x (Invitrogen). Microglia were isolated with anti-CD11b conjugated magnetic microbeads (130-049-601, Miltenyi Biotec GmbH) according to manufacturer’s protocol, using MS columns (130-042-201, Miltenyi Biotec) placed in a magnetic field. The eluted CD11b positive cells were stored in Trizol until RNA isolation for qPCR.
Isolation of mouse microglia
All experiments were performed in line with institutional guidelines of the University Medical Center Utrecht, approved by the Animal Ethics Committee of Utrecht University (AVD1150020174314), and were conducted in agreement with Dutch laws (wet op de Dierproeven, 1996) and European regulations (Guidelines 86/609/EEC). Animals were housed under standard conditions with access to water and food ad libitum. We used the well-established AD mouse model, APPswePS1dE9 double transgenic line (Jankowsky et al., 2001; Orre et al., 2014). This line has been backcrossed to C57BL/6 mice for more than 20 generations (Kamphuis et al., 2015), since the genetic background is known to influence AD pathogenesis (Hyman and Tanzi, 2019; Song et al., 2011; Tahara et al., 2006). We isolated microglia from three pooled cortices of 4-month-old AD mice and wild-type littermates as controls (resulting in N = 12 samples). Genotype was confirmed by performing real-time PCR with primers targeted to the two transgenes expressed by the APP/PS1 mice — human/mouse chimeric APP with K595N/M596L Swedish mutation and human PS1 carrying the Exon 9 deletion. For further details on this transgenic line see The Jackson Laboratory (B6C3-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax;StockNo: 34829; https://www.jax.org/strain/004462). Mice were anesthetized by an overdose of Pentobarbital and transcardially perfused with HBSS (14175-053, Gibco). Cortical regions were dissected on ice and immediately processed for transcriptome analysis. Briefly, cortical tissue was mechanically dissociated and subjected to enzymatic dissociation using Papain (final concentration of 8 U/ml, Worthington) in combination with 100 μg/ml DNase I in Pipes based buffer containing 1 mM Pipes (P1851, Sigma Aldrich), 25 mM L-Cystein HCL, and 5 mM EDTA. Tissue was enzymatically dissociated at 37 °C in a shaking incubator (Incu-shaker mini, Benchmark) for 50 min. Subsequently, DNase I in GKN/BSA was added and the tissue suspension was incubated for another 15 min. A 90% Percoll gradient centrifugation was used to collect the cellular fraction before cells were incubated with anti-CD11b conjugated magnetic microbeads according to the manufacturer’s protocol. The eluted CD11b positive cells were stored in Trizol until further use.
RNA isolation, cDNA synthesis, and quantitative real-time PCR for human and mouse microglia
For RNA isolation, samples were thawed and total RNA was isolated with TRIzol (Life Technologies) according to manufacturer’s protocol, and the RNA was subsequently precipitated in 2-propanol and 20 μg/μl glycogen (Roche) overnight at −20 °C. Samples were centrifuged (12,000 x g for 30 minutes) at 4 °C, washed twice with cold 75% ethanol, and the RNA pellet was dissolved in MilliQ. The RNA concentration was determined using a Nanodrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA).
Total RNA was treated with DNaseI (gDNA wipe out buffer, Qiagen) at 42 °C for 2 min, and cDNA was synthesized using Quantic Reverse Transcription kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. Quantitative PCR (qPCR) was performed using the Quantstudio 6 Flex (Applied Biosystems, Life Technologies, USA). For the qPCR reaction 1 μl of 1:20 diluted cDNA in MilliQ was used with 1 μl primer mix (forward and reverse primers, 2 pmol/μl, Supplementary Table 7), 5 μl FastStart Universal SYBR Green Master mix (Roche, Basel, Switzerland) and 3 μl MilliQ. The following cycling conditions were used: 2 min 50 °C, 10 min 95 °C, and 40 cycles of 15 sec 95 °C and 1 min 60 °C. A dissociation curve was obtained by ramping the temperature from 60 °C to 90 °C. Applied Biosystems 7500 Real-Time PCR software (Applied Biosystems) was used for the analysis of amplification curves. Gene expression was normalized to three of the following reference genes β-actin, S18, hypoxanthine phosphoribosyltransferase (HPRT) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH).
Statistics
Data was analyzed with GraphPad Prism8. A paired two-sided t-test was used for groups with equal variances or its nonparametric equivalent, the Wilcoxon signed rank test was used to determine the effect of the different stimulations on MDMi cells. For the statistical analysis of mouse microglia and human microglia data, the unpaired two-sided t-test for two-group comparison for groups with equal variances was used. The one-way ANOVA followed by Bonferroni’s post-hoc test was used to assess the effect between three groups. Normal distribution was tested with the D’Agostino & Pearson omnibus normality test. For data that were not normally distributed the Mann-Whitney test or Kruskal-Wallis test was used followed by Dunn’s tests for multiple comparisons. Outliers were identified using the Robust regression and Outlier removal (ROUT) method with the coefficient set to 1, available in GraphPad.
Results
Characterization of the monocyte-derived microglia-like cell (MDMi) model
Human monocytes from 10 healthy individuals were cultured for 10 days and differentiated into a microglia-like phenotype with M-CSF, GM-CSF, IL-34, TGFβ, and IFN-γ, which are important factors in the development of microglia (Ryan et al., 2017). This led to the development of a microglia-like morphology, i.e. a ramified appearance and the expression of Iba1 (Fig. 1a-b). These monocyte-derived microglia-like (MDMi) cells had a distinct gene expression profile compared to monocytes based on increased expression of the common microglial markers, APOE, TREM2, and C1QA (Fig. 1c). Other microglial markers, such as P2YR12, were not significantly upregulated in MDMi cells when compared to monocytes. We also compared the gene expression profile of primary human microglia (pMG) to MDMi cells. Expression levels of IRF8, TREM2, and C1QA were similar in MDMi and pMG. These findings are in line with previous work, showing that unsupervised clustering based on the expression of a microglial marker panel resulted in clustering of cultured pMG with MDMi cells and less with monocytes (Ormel et al., 2020). Based on both morphology and gene expression, MDMi cells resemble primary human microglia and therefore can be used as a model to study human microglia function in health and disease. To characterize the MDMi cells further, we determined the phagocytic function of MDMi cells and compared that to pMG. Uncoated fluorescent beads were taken up by MDMi cells after 1 hr of incubation with the beads (Fig. 1d). Fluorescence activated cell sorting (FACS) was used to determine the percentage of phagocytizing cells, i.e. the phagocytosis index (Fig. 1e). Our results show that MDMi cells, like microglia, can phagocytize, although significantly less than pMG (24.5% ± 5.5 vs 10.3% ± 1.1; Fig. 1f).
Inflammatory response of monocyte-derived microglia-like cells
To assess whether LPS stimulation induced changes in the transcriptome we performed bulk RNA sequencing on stimulated and unstimulated MDMi cells. We first confirmed that MDMi cells expressed genes that encode for proteins that are essential for an LPS-induced immune response, such as the toll-like receptor-4 (TLR4), CD14, and Ly96 (Fig. 2a). Next, we used the phagocytosis assay to assess the effect of LPS stimulation on the phagocytic ability of MDMi cells. The 24 hrs stimulation with 100 ng/ml LPS increased the phagocytosis index of the MDMi cells by 7% (PBS 10.3% ± 1.1; LPS 17.9% ± 1.8; Fig. 2b). Whole transcriptome analysis showed an inflammatory response in MDMi cells after LPS stimulation. A principal component analysis (PCA) identified two clusters corresponding to the LPS stimulated and unstimulated MDMi samples (PBS; Fig. 2c). Differential gene expression analysis identified 675 genes that were significantly upregulated (Log2 FC > 2 and adj. p-value < 0.05) and 413 genes that were downregulated (Log2 FC > −2 and adj. p-value < 0.05) after LPS stimulation (Supplementary Table 2).
Gene ontology (GO) and Panther pathway analyses on the 200 most significantly upregulated differentially expressed genes after LPS stimulation, identified enrichment for genes involved in cytokine, chemokine, and interleukin signaling pathways (Fig. 2e). The genes enriched in these pathways, such as IL1B, IL6, TNFα (Fig. 2f), and several chemokines are known to be upregulated in primary microglia after stimulation by LPS (Melief et al., 2016; Sneeboer et al., 2019). Moreover, the GO analysis also revealed changes in “transition metal ion binding” due to upregulation of the expression of SOD2, IL1α, and two isoforms of the metallothionein family, MT1 and MT2.
The expression of some microglia homeostatic markers, including CX3CR1 and P2RY12, were downregulated by LPS stimulation, while TMEM119 remained unaffected (Fig. 2g). This is in line with previous studies showing a downregulation of homeostatic microglial genes after stimulation (Butovsky and Weiner, 2018; Keren-Shaul et al., 2017). Overall, LPS stimulation of MDMi cells resulted in an immune activated phenotype.
Transcriptomic changes induced by Aβ1-42 oligomer stimulation
We performed bulk RNA sequencing on stimulated and unstimulated MDMi cells to assess the direct effect of Aβ1-42 oligomer stimulation on changes in the transcriptome. We confirmed that the MDMi cells express several genes, including CD33, TLRs, formyl peptide receptor 2 (FPR2), and TREM2, encoding for proteins involved in Aβ clearance and in triggering an inflammatory response (Fig. 3a) (Doens and Fernández, 2014; Salminen et al., 2009; Tejera and Heneka, 2016). In line with this, stimulating the MDMi cells with 500 nM Aβ1-42 oligomers for 24 hrs increased the phagocytosis index of the MDMi cells with 5.7% (Vehicle 9.8% ± 1.4; Oligomers 15.5% ± 1.9; Fig. 3b).
A PCA plot showed that the Aβ1-42 oligomer-stimulated and unstimulated samples (vehicle) do not form separate clusters. Instead, samples from the same donor clustered together (Fig. 3c), indicating a larger difference between donors than induced by the Aβ treatment. Differential gene expression analysis identified 20 genes that were upregulated and two genes that were downregulated (with adj. p < 0.05 and a Log2 FC > 0.5) in Aβ1-42 oligomer-stimulated MDMi samples (Fig. 3d and Supplementary Table 3). Several subtypes of the MT1 isoform were upregulated after Aβ1-42 oligomer stimulation (Fig. 3e). The top five upregulated genes with the highest log2 FC were all subtypes of metallothionein 1 (MT1), a family of cysteine-rich proteins. GO on the 20 differentially expressed genes between vehicle and Aβ1-42 oligomer-stimulated MDMi cells identified changes in “metal ion binding”, “transition metal ion binding”, and “zinc ion binding” (Fig. 3f). The Panther pathway analysis suggested changes in “hypoxia and oxidative stress response”, and in the “AD presenilin pathway”. However, these pathways were identified based on the upregulation of only one gene, viz. TXN and CD44, respectively. The top five upregulated transcripts (MT1, MT2, SOD2, AKR1B1, and C15ORF48) were validated by qPCR (Supplementary Fig. S1). These results indicated that the transcriptomic changes induced by human Aβ1-42 oligomers are predominantly related to metal binding and control of oxidative stress.
Aβ1-42 oligomer response in human primary microglia
Four subtypes of MT1 (MT1E, MT1G, MT1L, and MT1X) that are in the top five upregulated genes after Aβ1-42 oligomer stimulation in MDMi cells were shown by Bossers et al., 2010 to increase with disease progression and were highest in end-stage AD (Bossers et al., 2010). However, the gene expression profile from Bossers et al., 2010 was not specific to the microglia population (Bossers et al., 2010). Therefore, to determine whether the expression of MT1, MT2, SOD2, AKR1B1, and C15ORF48 is increased in pMG from AD-cases, a qPCR analysis was done on primary microglia isolated from AD-cases microglia and NDC. MT1 expression was significantly increased in AD-cases (Fig. 4a). The expression of a selection of genes associated with late-onset AD (CD33, TREM2, TYROBP, ITGAM, and ApoE) was not increased in AD-cases (Fig. S2).
To determine whether Aβ1-42 oligomer stimulation in pMG recapitulated the gene expression profile found in MDMi cells, a selection of pMG isolated from NDC-cases was cultured for 24-48 hrs and subsequently stimulated with Aβ1-42 oligomers for 24 hrs. This stimulation did not induce an upregulation of MT1, MT2, SOD2, AKR1B1, and C15ORF48 (Fig. 4b).
The Aβ1-42 oligomer-induced expression profile in MDMi cells is not detected in isolated microglia of APPswePS1dE9 mice
A previously published dataset showed that Mt1 (0.8 Log2 FC), Mt2 (1.4 Log2 FC), and Sod2 (0.8 Log2 FC) were increased in cortical microglia isolated from 15- to 18-month-old APPswePS1dE9 mice (Orre et al., 2014). We next determined whether these upregulated genes were also increased in cortical microglia from 4-month-old APPswePS1dE9 mice, when increased levels of amyloid and the first plaques are detected (Garcia-Alloza et al., 2006; Van Tijn et al., 2012). qPCR analysis showed no increase in mRNA levels of these genes in 4-month-old APPswePS1dE9 mice (Fig. S3).
AD pathology-associated microglia profiles
Next, we compared our stimulated and unstimulated MDMi transcriptome datasets to three well-known AD microglia profiles (Keren-Shaul et al., 2017; Mathys et al., 2019; Srinivasan et al., 2019). The studies by Srinivasan et al., 2019 and Mathys et al., 2019 investigated transcriptome changes in microglia of healthy control donors and AD patients (Mathys et al., 2019; Srinivasan et al., 2019). The first study refers to their novel human Alzheimer’s microglia/myeloid cells profile as the HAM-profile (Srinivasan et al., 2019). The HAM-profile contains 66 differentially expressed genes between AD and control subjects, from which we detected 50 genes in the expression dataset of Aβ1-42 oligomer stimulation and 51 genes in the expression dataset of LPS stimulated MDMi cells. However, only Kcnj5 was differentially expressed in both the HAM signature and our dataset from Aβ1-42 oligomer-stimulated MDMi cells (Fig. 5a). In our dataset, this gene was downregulated instead of upregulated (Fig. 5b). The comparison of the HAM signature and our LPS-stimulated MDMi cells identified 20 genes with a p-value < 0.05, of which eight genes were upregulated and one was downregulated in both HAM and LPS-stimulated MDMi cells (Fig. 5b). Overall, the HAM signature is not present in our MDMi cell model after Aβ1-42 oligomer stimulation (Fig. 5c). Only a small overlap was detected in the LPS-stimulated MDMi cells (Fig. 5a-b).
The AD-pathology associated microglia profile described by Mathys et al., 2019 contains 77 differentially expressed genes (Mathys et al., 2019). From this profile, two genes, Acsl1 and Fth1, were also upregulated in Aβ1-42 oligomer-stimulated MDMi cells (Fig. S4a-b). In the LPS-stimulated MDMi cells 19 genes from the 77 differentially expressed genes were upregulated. Overall, the AD-pathology associated microglia signature is not present in our MDMi cell model after Aβ1-42 oligomer stimulation. Only a partial overlap was detected in the LPS-stimulated MDMi cells (Fig. S4c-d).
The study by Keren-Shaul et al., 2017 identified the so-called disease-associated microglia (DAM), in the 5xFAD AD mouse model (Keren-Shaul et al., 2017). The DAM profile is to a certain degree conserved in mice and humans (Mathys et al., 2019). The comparison of the DAM profile and our Aβ1-42 oligomer-stimulated MDMi cells identified 17 genes with a p-value < 0.05, from which nine were upregulated and three downregulated in both DAM and Aβ1-42 oligomer-stimulated MDMi cells.
Discussion
In this study, we performed bulk RNA sequencing on unstimulated and stimulated MDMi cells to investigate the effect of LPS and human Aβ1-42 oligomers on their transcriptome. We first characterized the MDMi cell model on morphology and gene expression and concluded that they recapitulate some of the key aspects of microglial phenotype and function. In accordance with previous findings, LPS stimulation induced an immune activated phenotype in the MDMi cells (Ormel et al., 2020). Both LPS and Aβ1-42 oligomer stimulation increased the phagocytosis index of MDMi cells. Furthermore, the Aβ1-42 oligomer stimulation resulted in the upregulation of 20 genes, including several MT1 subtypes, and the downregulation of two genes. Overall, Aβ1-42 oligomers induced less changes in MDMi cells than the strong inflammatory stimulus LPS.
Some of the classical AD markers as identified in GWAS studies, including CD33, TREM2, and ApoE, were not upregulated by Aβ1-42 oligomer stimulation in MDMi cells. Differentially expressed genes from stimulated MDMi samples showed little overlap with AD microglial profiles described in recent transcriptome studies (Keren-Shaul et al., 2017; Mathys et al., 2019; Srinivasan et al., 2019). These findings indicate that stimulation with Aβ1-42 oligomers induces a distinct gene expression profile in MDMi cells, different from the gene profile of microglia from end-stage AD pathology, characterized by a high amyloid burden, and, in humans, also increased neurofibrillary tangles. The profile identified in MDMi cells is specific for Aβ1-42 oligomer stimulation and not influenced by the effect of other cells in the central nervous system, previous activation of the cells, tau pathology, and/or other amyloid isoforms. Whether an end-stage AD pathology profile could be induced in MDMi cells by co-culturing with other cells, stimulation with tau and/or other amyloid isoforms should be investigated to determine the synergistic effect of these factors.
Several subtypes of a specific family of metalloproteins, the metallothioneins (MTs), were upregulated after stimulation with Aβ1-42 oligomers. MTs are low molecular weight cysteine-rich metal-binding proteins. MTs consist of four subfamilies, with MT1 and MT2 being the most predominant isoforms present in most tissues (Vašák and Meloni, 2011). In the central nervous system, MT1 and MT2 are predominantly expressed by astrocytes. However, expression in neurons, endothelial cells, and microglia has also been shown (Pedersen et al., 2009). Although the primary role of MTs remains unclear, increasing evidence indicates that these proteins have multiple functions, including maintenance of zinc and copper homeostasis, anti-inflammatory properties, regulating the biosynthesis and activity of zinc-binding proteins, and protection against reactive oxygen species (Hozumi et al., 2004; Manso et al., 2012; Waller et al., 2018). Their expression is increased in response to a variety of stimuli, including oxidative stress, neuroinflammation, and toxic levels of metal ions (West et al., 2008). In mice, it was found that MT1 and MT2 were upregulated in the brain five hrs after LPS injection (Searle et al., 1984). Also in MDMi cells stimulated with LPS, an upregulation of several MT1 subtypes and MT2A was found. In several neurodegenerative disorders, including AD, MT1 and MT2 are upregulated (Adlard et al., 1998; Duguid et al., 1989; Zambenedetti et al., 1998) and associated with Aβ plaques in several AD animal models (Carrasco et al., 2006; Hidalgo et al., 2006). The upregulation of MTs could have a neuroprotective function against the oxidative stress and neuroinflammation involved in AD pathogenesis (Adlard et al., 1998; Nunomura et al., 2006). Also, some other genes encoding for proteins known for their anti-oxidant response were upregulated in the Aβ1-42 oligomer stimulated MDMi, such as SOD2 (Massaad et al., 2009) and NQO1 (Raina et al., 1999; SantaCruz et al., 2004).
Bossers et al., 2010 showed that the expression of several MT1 isoforms is upregulated during the progression of AD. However, this gene expression profile was obtained from the grey matter of the frontal cortex and was not specific to the microglia population (Bossers et al., 2010). Interestingly MT1 was also significantly upregulated in isolated microglia from AD-compared to NDC-cases. Based on the mRNA expression of both MT2 and C15ORF48, two groups of donors seem to be present within the AD-cases. However, no significant association was detected with the known confounding variables age, sex, post-mortem delay, pH of the cerebral spinal fluid or with Braak or amyloid scores (Supplementary Table 5 and 6). The expression of a selection of genes associated with late-onset AD (CD33, TREM2, TYROBP, ITGAM, and ApoE) were not upregulated in the isolated microglia of AD-cases. Although several recent studies found differences between the gene expression profile of microglia from AD- and NDC-cases (Grubman et al., 2019; Mathys et al., 2019; Srinivasan et al., 2019), this was not found in all studies (Alsema et al., 2020).
Culturing primary microglia from a selection of NDC-cases for 24-48 hrs and subsequent 24 hrs stimulation with Aβ1-42 oligomers did not induce the upregulation of the top upregulated genes detected in MDMi cells. This is in contrast with the study by Walker et al., 2006 where stimulation for 24 hrs with oligomeric amyloid resulted in the activation of an inflammatory response and the upregulation of MT1, MT2 and SOD2 in cultured microglia from the superior frontal cortex (Walker et al., 2006). The genes were categorized as anti-inflammatory and anti-oxidant proteins. Besides the upregulation of these three genes, KYNU, also present in the Aβ1-42 oligomer profile, was increased as well (Walker et al., 2006). Differences in the concentration of Aβ1-42 oligomers (i.e. 500 nM or 2 μM) used, isolation procedures or the duration of the microglia in culture prior to the stimulation (i.e. 24-48 hrs or 12-24 days) could explain the different response to Aβ1-42 oligomer stimulation. We waited 48 hrs before Aβ1-42 oligomer stimulation, which might have been too short to bring the microglia to a nonactivated state after isolation. However, as microglial markers are downregulated in a culture environment (Gosselin et al., 2017), culturing the microglia for a longer period also introduces other limitations.
In summary, in this study we investigated the transcriptomic changes following stimulation of MDMi cells with inflammatory stimulus and Aβ1-42 oligomers. LPS stimulation induced an immune activated phenotype, and stimulation with stable Aβ1-42 oligomers induced a specific more anti-inflammatory transcriptome profile in MDMi cells. Several anti-oxidant genes were upregulated by Aβ1-42 oligomer stimulation, most notably metallothionein subtypes, which are known to be involved in metal ion regulation, protection against reactive oxygen species and have anti-inflammatory properties. In conclusion, the role of metallothioneins in AD pathogenesis and their induced upregulation after acute exposure to Aβ1-42 oligomers should be investigated further to determine whether the acute activation by oligomeric amyloid could trigger a protective response.
Supplementary data
Acknowledgements
This work was supported by a TAS-ZonMw grant (40-41400-98-16020), ZonMw grant (733050505) and a Memorabel-ZonMw grant (733050816) to E.M.H. We are grateful to the Netherlands Brain Bank (www.brainbank.nl) who provided us with the post-mortem human brain tissue. We are also grateful to Guus Scheefhals from Crossbeta for providing the Aβ1-42 oligomers used in this study. We thank Y. He, M. P. Boks, and M. Litjens for the isolation of monocytes, C. San Martin Paniello for assistance with setting up the MDMi cell model, and R.D. van Dijk and M.A.M. Sneeboer for their help with setting up the microglial isolations and phagocytosis assays using FACS.