iPSC-derived microglia carrying the TREM2 R47H/+ mutation are pro-inflammatory and promote synapse loss

Abstract

Genetic findings have highlighted key roles for microglia in the pathology of neurodegenerative conditions such as Alzheimer’s disease (AD). A number of mutations in the microglial protein TREM2 (triggering receptor expressed on myeloid cells 2) have been associated with increased risk for developing Alzheimer’s disease (AD), most notably the R47H/+ substitution. We employed gene editing and stem cell models to gain insight into the effects of the TREM2 R47H/+ mutation on human iPSC-derived microglia. We found transcriptional changes affecting numerous cellular processes, with R47H/+ cells exhibiting a pro-inflammatory gene expression signature. TREM2 R47H/+ also caused impairments in microglial movement and the uptake of multiple substrates, as well as rendering microglia hyper-responsive to inflammatory stimuli. We developed an in vitro laser-induced injury model in neuron-microglia co-cultures, finding an impaired injury response by TREM2 R47H/+ microglia. Furthermore, mouse brains transplanted with TREM2 R47H/+ microglia exhibited reduced synaptic density, with upregulation of multiple complement cascade components in TREM2 R47H/+ microglia suggesting inappropriate synaptic pruning as one potential mechanism. These findings identify a number of potentially detrimental effects of the TREM2 R47H/+ mutation on microglial gene expression and function likely to underlie its association with AD.

1. Introduction

Alzheimer’s disease (AD) is a devastating neurodegenerative condition with no cure (Canter et al., 2016). The AD brain is characterized by widespread neurodegeneration, neuroinflammation, and the presence of hallmark protein aggregates composed primarily of amyloid-β (Aβ) or hyperphosphorylated tau protein (amyloid plaques and neurofibrillary tangles, respectively; De Strooper and Karran, 2016). Familial forms of AD (fAD) are well established to result from dominantly inherited mutations that affect Aβ production and/or processing by neurons, thus neuronal mechanisms have historically been the focus of AD research (Hardy and Higgins, 1992; Makin, 2018). More recently, however, large-scale genome-wide association studies (GWAS) have highlighted many non-neuronal genetic risk factors for sporadic forms of AD (sAD; Bellenguez et al., 2022; Escott-Price and Hardy, 2022; Kunkle et al., 2019; Lambert et al., 2013). Remarkably, of the >90 prioritized sAD risk genes identified thus far, almost half are more highly expressed in microglia than in any other brain cell type, indicating that microglial dysfunction can be an active driver in the disease process (Escott-Price and Hardy, 2022; Lewcock et al., 2020; Penney et al., 2020; Zhang et al., 2016). Among these microglia-specific sAD risk factors is the subject of this study, the triggering receptor expressed on myeloid cells 2 (TREM2) (Guerreiro et al., 2013; Jonsson et al., 2013).

Microglia perform numerous important functions in the developing and adult brain. As innate immune cells, they continually survey their environment and can phagocytose components of dead cells, protein aggregates, and other debris, while also regulating inflammation via cytokine release in response to various stimuli (Li and Barres, 2018). Additionally, microglia can more directly impact neurodevelopment by providing trophic support and sculpting neural circuits via synaptic pruning (Salter and Stevens, 2017). While microglial activities are essential for proper neuronal development and function, chronic inflammation due to persistent microglial activation is a hallmark of neurodegenerative disease that can directly incite neuronal damage (Lewcock et al., 2020). Synaptic pruning by microglia can also be inappropriately re-activated, contributing to synapse loss during the neurodegenerative process (Hammond et al., 2019).

In 2013, heterozygous point mutations in TREM2 were identified as risk factors for AD, most notably the TREM2 R47H/+-causing mutation which increases AD risk by roughly three-fold (Guerreiro et al., 2013; Jonsson et al., 2013; Zhou et al., 2019). More than a decade earlier, however, homozygous deletion mutations in TREM2 and its binding partner TYROBP were identified as causative mutations for polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy (PLOSL), an aggressive early onset disease characterized by bone cysts, demyelination, neurodegeneration, and dementia (Errichiello et al., 2019; Paloneva et al., 2002). Additional homozygous loss-of-function mutations, including TREM2 T66M, which prevents TREM2 glycosylation and thus cell surface expression, were subsequently shown to cause PLOSL and a frontotemporal dementia (FTD)-like syndrome (Dardiotis et al., 2017; Kleinberger et al., 2014). Cell surface expressed TREM2 can also be cleaved, liberating extracellular soluble TREM2 (sTREM2) species, and the TREM2 intracellular domain (ICD) (Kleinberger et al., 2014; Thornton et al., 2017). Elevated sTREM2 levels have been reported in AD patients compared to controls, as well as in R47H/+ carriers vs. non-carriers (Piccio et al., 2016).

TREM2 has been shown to regulate many microglial processes. TREM2 knockout mice exhibit impaired developmental synaptic pruning and impaired sensing and clearance of debris following myelin injury (Filipello et al., 2018; Nugent et al., 2020; Poliani et al., 2015). In AD mouse models, TREM2 knockout impairs the ability of microglia to sense and respond to amyloid plaques, also resulting in detrimental effects in tau models (Gratuze et al., 2020; Keren-Shaul et al., 2017; Lee et al., 2021; Leyns et al., 2017; Wang et al., 2016, 2015). Inflammatory responses by microglia in amyloid and tau models are generally dampened by loss of TREM2 (Griciuc et al., 2019; Jay et al., 2017; Leyns et al., 2017). Studies examining TREM2 R47H mutations have often reported similar effects to knockout mutations, with R47H microglia in amyloid models exhibiting impaired plaque localization and response (Cheng-Hathaway et al., 2018; Song et al., 2018; Tran et al., 2023). The apparent effects of the R47H mutation in tau mouse models have varied, with reports of increased or decreased tau seeding, microglial inflammation and synaptic/cognitive effects (Gratuze et al., 2020; Leyns et al., 2019; Sayed et al., 2021; Tran et al., 2023).

Many microglial risk genes for AD, including TREM2, are relatively poorly conserved between mouse and human, suggesting the importance of using human systems to understand their functions (Penney et al., 2020). Accordingly, human induced pluripotent stem cell (iPSC)-based models have also been used to examine the effects of TREM2 Null and R47H/+ mutations. These mutations have often, but not always, been reported to reduce the uptake of brain-relevant substrates such as Aβ (Andreone et al., 2020; Liu et al., 2020; McQuade et al., 2020; Piers et al., 2020; Popescu et al., 2022; Reich et al., 2021). Metabolic impairments have also been observed in microglia carrying these mutations, while reported effects on movement and chemotaxis have been mixed (Jairaman et al., 2022; Piers et al., 2020; Reich et al., 2021). Normal or impaired inflammatory responses in TREM2 loss-of-function iPSC-microglia have mostly been reported, while different studies have shown either increased or decreased inflammation in human cells carrying the R47H mutation (Andreone et al., 2020; Brownjohn et al., 2018; Cosker et al., 2021; Garcia-Reitboeck et al., 2018; Hall-Roberts et al., 2020; Liu et al., 2020; Piers et al., 2020). Similarly, single-nucleus and bulk RNA sequencing experiments from AD patients carrying R47H/+ mutations have observed a pro-inflammatory gene expression signature, though not in all cases (Korvatska et al., 2020; Sayed et al., 2021; Zhou et al., 2020). Thus, it remains unclear how the TREM2 R47H/+ mutation affects microglial function and leads to different clinical phenotypes.

Here, we use gene editing and human iPSC-based in vitro and in vivo microglia models to characterize the effects of the TREM2 R47H/+ mutation on microglial function and gain insight into its mechanism of action. These analyses reveal, transcriptional changes associated with multiple cellular processes and highlighted by a pro-inflammatory signature in R47H/+ microglia. TREM2 R47H/+ microglia also exhibit exaggerated cytokine release and gene expression changes in response to inflammatory stimuli. In functional assays, the R47H/+ mutation impairs the uptake of multiple brain-relevant substrates. We also find impaired responses of R47H/+ microglia in a novel laser-induced neuronal injury model. Further, in xenotransplantation experiments in mice, we find that synaptic density is reduced by TREM2 R47H/+ iPSC-microglia. Thus, we identify numerous functional and gene expression alterations in human microglia resulting from the TREM2 R47H/+ mutation, that may underlie its association with AD.

2. Materials and Methods

2.1. Gene editing, induced pluripotent stem cell and microglia culture

Stem cell culture was performed in the MIT Picower Institute iPSC core facility. iPSCs from an unaffected 75-year-old female were used as the parental line (Coriell #AG19173). Cells were cultured on hES-Qualified MatrigelR (Corning)-coated tissue culture plates in mTeSR1 media (STEMCELL). CRISPR-Cas9-mediated gene editing to generate the TREM2 R47H/+ mutation was performed as previously described (Lin et al., 2018). Potential edited clones were screened by Sanger sequencing (Azenta Life Sciences), after which we selected two clones carrying the R47H/+ mutation for further analysis. An additional clone from the CRISPR screening that remained un-edited at the TREM2 locus, as well as the parental iPSC line, were used as isogenic controls. We also isolated an iPSC clone carrying a 1 bp insertion, causing a frameshift mutation and early stop codon in the TREM2 open reading frame, which we refer to as TREM2 Null. Cells were karyotyped to identify any chromosomal abnormalities and the top 4 predicted CRISPR off-target sites were sequenced to rule out off-target mutations. Differentiation to iPSC-microglia was performed according to McQuade et al. (2018) with modified maturation supplementation using 25 ng/mL human M-CSF and 100 ng/mL human IL-34 (PeproTech) (McQuade et al., 2018). In vitro experiments were performed using DIV 40-60 microglia.

2.2. Quantitative PCR

RNA was extracted from microglia using the QIAGEN RNeasy Plus Mini Kit. cDNA synthesis was performed with RNA to cDNA EcoDry Premix (Oligo dT) (Clontech). qPCR was performed with SsoFast EvaGreen Supermix (Bio-Rad) using a C1000 Thermal Cycler and a C96 Real-Time System (Bio-Rad). Target genes were normalized using β-Actin (ACTB).

2.3. Western blotting

Protein was extracted using RIPA buffer. Western blots were performed using PVDF membranes (Millipore) following standard methods. Antibodies used included: mouse-β-actin (Sigma), rabbit-TREM2 (Cell Signaling Technology) rabbit-p-AKT (S473, Cell Signaling Technology), mouse-AKT (Cell Signaling Technology), rabbit-pan-p-PKC (T514, Cell Signaling Technology), rabbit-p-ERK (Cell Signaling Technology), rabbit-ERK (Cell Signaling Technology), rabbit-p-p38 (T180/182, Cell Signaling Technology), rabbit-p38 (Cell Signaling Technology).

2.4. Inflammatory treatments and ELISA

ELISAs were performed using culture media from control and TREM2 mutant microglia grown in 96-well plates (20k/well). Soluble TREM2 was quantified using the human TREM2 ELISA Kit (Abcam). For IL6 and TNFα ELISAs, cells were either left untreated, or treated for 2 days with 100ng/mL LPS or 50 ng/mL IFNγ before collection of culture media. Human IL-6 or TNFα Quantkine ELISA Kits (R+D Systems) were used.

2.5. RNA Sequencing and analysis

RNA samples from microglia lines without stimulation (Figure 1) were collected in triplicate from clone A and clone B of each genotype, resulting in 6 replicates each from control and R47H/+ microglia. For cytokine-induced expression changes (Figure 2), LPS (100 ng/mL) or IFNγ (50 ng/mL) treatments were performed for 2 hours prior to RNA isolation. Sequencing for cytokine-induced changes was performed on 2 RNA samples each from clone A and clone B for each treatment condition (a total of 4 replicates per genotype, per condition). RNA was extracted using the QIAGEN RNeasy Plus Mini Kit and subject to QC using an Advanced Analytical-Fragment Analyzer before library preparation using NEB Ultra II RNAseq library preparation kit. Libraries were pooled for sequencing using Illumina NextSeq500 or NovaSeq6000 platforms at the MIT BioMicro Center. Reads were aligned to the human genome reference GRCh38/hg38 using the R software package Rsubread (Liao et al., 2019). Mapped reads were converted to gene level counts using the featureCounts function of Rsubread with “strandSpecific” parameter set to 0. Differential analysis and principal component analysis were performed using DESeq2 (Love et al., 2014). Principal component analysis was performed using the plotPCA function in DEseq2 after transforming the count data using the variance stabilizing transformation (vst function in DESseq2). Default settings were used for the plotPCA function which uses the top 500 genes selected by highest row variance. Differential expression for inflammatory treatments was performed by comparing baseline to LPS and IFNγ conditions within genotypes. Poor read quality was obtained for one LPS-treated control sample (clone A), thus this sample and a corresponding untreated control sample were excluded from the LPS differential analysis. Cutoffs for differentially expressed genes in all analyses were set at >0.3 log2 fold change and p-value <0.05. Our initial analysis of control and R47H/+ microglia filtered for genes expressed among the top 5000 expressed genes in our iPSC-microglia, while the top 10,000 expressed genes were used in our LPS and IFNγ-induced gene analysis because many cytokine genes are expressed at low levels without induction. ToppGene was used for gene ontology analysis with the most highly enriched non-redundant gene ontology terms presented (Chen et al., 2009). Morpheus was used to generate heatmaps of the fold change Z-Score for expression across genotypes and/or treatments (Broad Institute, https://software.broadinstitute.org/morpheus). RNA-sequencing data is available at NCBI Gene Expression Omnibus, accession number GSE241858.

Figure 1. The TREM2 R47H/+ mutation alters microglial gene expression and induces a pro-inflammatory signature.

(A) Schematic of CRISPR mutagenesis and generation of TREM2 R47H/+ mutant microglia. (B) Diagram of TREM2 protein and downstream effectors. Locations of the R47H/+ mutation is indicated. Chromatogram traces showing successful editing of the R47H sequence in TREM2. (C) Quantification of TREM2 mRNA level in TREM2 R47H/+ and isogenic control microglia. n=6 for each of CTRL-A, CTRL-B, R47H-A and R47H-B. (D) Images and (E) quantification of western blots for different forms of TREM2. TREM2 null iPSC-microglia (see Methods) were included to control for antibody specificity. Actin serves as loading control. n=6 for each of CTRL-A, CTRL-B, R47H-A and R47H-B. (F) Quantification of ELISA for soluble TREM2 levels in culture media from control and TREM2 R47H/+ microglia. n=3 for each of CTRL-A, CTRL-B, R47H-A, R47H-B and null. ***p<0.001. Student’s t-test. (G) PCA plot for control and TREM2 R47H/+ biological and technical replicates. (H) Volcano plot and (I) enriched GO terms for genes differentially expressed between control and TREM2 R47H/+ microglia. (J) Heatmap of the gene expression Z-Score for genes in the ‘Inflammatory response’ GO term in control and TREM2 mutant microglia that showed significantly increased expression in one or both of TREM2 R47H/+ and T66M microglia.

Figure 2. TREM2 R47H/+ microglia show exaggerated responses to inflammatory stimuli.

(A) Quantification of IL6 ELISAs from culture media following treatment with 100 ng/mL LPS (left) or 50 ng/mL IFNγ (right). n=8 for CTRL-A, CTRL-B, R47H-A and R47H-B for all baseline conditions; n=9 for CTRL-A, CTRL-B, R47H-A and R47H-B for all LPS and IFNγ treated conditions. ***p<0.001. 1-way ANOVA with Sidak’s test. (B) Quantification of TNFα ELISAs from culture media following treatment with 100 ng/mL LPS (left) or 50 ng/mL IFNγ (right). n=8 for CTRL-A, CTRL-B, R47H-A and R47H-B for all baseline conditions; n=9 for CTRL-A, CTRL-B, R47H-A and R47H-B for all LPS and IFNγ treated conditions. *p<0.05, ***p<0.001. 1-way ANOVA with Sidak’s test. (C) Volcano plots for genes differentially expressed by LPS treatment in control and TREM2 R47H/+ microglia. (D) The number of differentially expressed genes and enrichment p-value (inset) for LPS-differentially expressed genes from the ‘Response to LPS’ and ‘Cell activation’ GO terms. (E) Volcano plots for genes differentially expressed by IFNγ treatment in control and TREM2 R47H/+ microglia. (F) The number of differentially expressed genes and enrichment p-value (inset) for IFNγ-differentially expressed genes from the ‘Type II interferon’ and ‘Cell activation’ GO terms. (G) Heatmap of the gene expression Z-Score for selected cytokine and chemokine genes induced by LPS and/or IFNγ in control and TREM2 R47H/+ microglia. Note that one CTRL-A LPS-treated sample (and thus also a corresponding untreated control) was removed from analysis due to poor read quality.

2.6. Uptake, immunostaining, imaging and flow cytometry

Synaptosomes were isolated as previously described (Penney et al., 2017). Myelin was isolated using a modified protocol from Larocca and Norton (Larocca and Norton, 2006). Briefly, wild-type mouse brains were homogenized in 0.32 M sucrose with a loose pestle. Homogenate was layered on top of 0.85 M sucrose and spun at 42,865 g for 60 min. The interphase was added to ultrapure water (Thermo Fisher) and spun at 42,865 g for 15 minutes. The pellet was then washed twice in ultrapure water (6915 g, 15 minutes), resuspended in 0.32 M sucrose, layered on top of 0.85 M sucrose and spun at 42,865 g for 30 minutes. The interphase was subsequently transferred and washed in 10 mL ultrapure water (42,865 g, 15 minutes). Amyloid-β₄₂-HiLyte Fluor 488 was obtained from AnaSpec. Uptake for immunostaining experiments was performed in 24-well plates containing glass coverslips. 20 μg/mL synaptosomes, 40 μg/mL myelin or 200 ng/mL Aβ₄₂-488 were added to culture media. 3 hours later cultures were washed and fixed with 4% paraformaldehyde (Electron Microscopy Solutions) in phosphate-buffered saline (PBS; Thermo Fisher) for 10 minutes. After PBS washes, cells were blocked 1 hour in PBS with 5% normal goat serum (Millipore) and 0.2% Triton-X 100 (Sigma). Staining was performed overnight at 4°C. Neuron-microglia co-cultures were fixed similarly; co-cultures and xenotransplant brain sections were stained using the same protocol. Antibodies used included: mouse-SVP38 (Sigma), chicken-MBP (Millipore), guinea pig-IBA1 (Synaptic Systems), chicken-VGlut1 (Synaptic Systems), rabbit-PSD95 (Abcam), mouse-β-Tubulin 3 (TUJ1; GeneTex), rabbit-β-Tubulin 3 (BioLegend) and mouse-STEM101 (Takara Bio). Imaging was performed using Zeiss LSM880 or LSM900 confocal microscopes. Synaptic puncta quantifications were performed on small 63x image stacks (5 μm total). Projections of each channel were thresholded and overlapping signal was detected using the Fiji Image Calculator function. The 3D Objects Counter function was then used to quantify overlapping puncta 30-200 voxels in size. For flow cytometry, synaptosomes or myelin were conjugated with pHrodo iFL Green STP Ester (Invitrogen) at a 1:10 pHrodo:synaptosome/myelin ratio, incubated 30 minutes, pelleted, then washed 2X with PBS, pelleting each time. 20 μg/mL pHrodo-synaptosomes, 40 μg/mL pHrodo-myelin or 200 ng/mL Aβ₄₂-488 was added the media of microglia plated in 96-well plates (20,000/well) and incubated 3 hours. Flow cytometry was performed using a FACS Celesta HTS-1 sorter (Koch Institute Flow Cytometry Core at MIT) or a BD Fortessa (Whitehead Institute Flow Cytometry Core at MIT).

2.7. Live imaging

Live-imagining was performed using a Zeiss LSM900 equipped with a humidity and CO₂ controlled, heated chamber kept at 37°C. For live uptake experiments, microglia were labeled with Vybrant DiD (Molecular Probes) and plated at 200k/well in 12-well plates. 200 ng/mL Aβ₄₂-488 was added to the media and images were collected every 10 minutes for 2 hours. Aβ₄₂-488 signal co-localizing with Vybrant DiD-labeled microglia was quantified across time to monitor uptake.

2.8. In vitro laser injury model

NGN2-induced neurons were initially grown on Matrigel-coated plates, then transferred to poly-D-lysine (Sigma-Aldrich) coated plates at day 7 of induction, at which point they were grown in BrainPhys Neuronal Media (STEMCELL) supplemented with 2% B27 (Life Technologies), 1% N2 (Life Technologies), 1% NEAA (Sigma-Aldrich). 1% GlutaMAX (Thermo Fisher), 200 nM L-Ascorbic acid, 20 ng/mL hBDNF (PeproTech), 20 ng/mL hGDNF (PeproTech). Neuron-microglia co-cultures were grown in the same media but supplemented additionally with 25 ng/mL hM-CSF. DIV21 or older NGN2 cultures (~250k cells/24-well plate) were overlaid with ~250k Vybrant DiO (Molecular Probes)-labeled iPSC-microglia in 200 μl 25% Matrigel:75% BrainPhys media. After 30 minutes, warm BrainPhys media (without Matrigel) was returned to the cultures. Following 7-10 days of co-culture, Vybrant DiO microglia were live-imaged on a Zeiss LSM900 with a temperature/CO₂ controlled chamber to capture baseline state. Laser injury was induced at 6X zoom by 3 minutes live scanning at 100% laser intensity, focused to the middle of a neuron cell body cluster. A photobleached area, corresponding to the laser-induced injury, was evident following laser scanning injury. Images of Vybrant DiO microglia signal were captured every 5 minutes post-injury for 90 minutes. Microglia movement in the image series was tracked using the Fiji TrackMate plugin (Tinevez et al., 2017). The vectors for x-y displacement of each tracked cell between each time point were used to determine total microglial movement. Directed movement towards (or away from) the injury site was calculated as the difference between the initial and final distance of a given cell from the center of the injury site, for each cell that could be tracked for the 90-minute session.

2.9. Xenotransplant experiments

Rag2 knockout, IL-2rγ knockout, human CSF1-expressing mice were housed in the MIT Division of Comparative Medicine (DCM) SCID facility. Experiments were performed following DCM and Committee on Animal Care (CAC) guidelines. Transplants were performed as described by Hasselmann and colleagues (Hasselmann et al., 2019). Briefly, following HPC induction, ~500k microglia precursors were injected transcranially into early postnatal (P0-P4) Rag2 knockout, IL-2rγ knockout, human CSF1 pups using a Hamilton syringe. DPBS-injected pups served as controls. At ~3 months of age, transplanted mice were anaesthetized, perfused with PBS, then brains were drop-fixed in 4% paraformaldehyde overnight. After washing with PBS, brains were sectioned at 40 μm thickness on a Leica VT1000S vibratome. Brain sections were then stained and imaged as described in the ‘Uptake, immunostaining, imaging and flow cytometry’ section above.

2.10. Statistical analysis

Data are presented as mean +/− SEM and were analyzed using Prism 9 (GraphPad). T-test or 1-way ANOVA followed by post hoc Tukey’s test was used for most analyses, with Sidak’s multiple comparison test used to compare baseline and LPS/IFNγ-induced conditions within genotype, as well as LPS/IFNγ-induced conditions between genotypes. In all cases *p<0.05 was considered significant.

3. Results

3.1. TREM2 R47H/+ microglia exhibit a pro-inflammatory gene expression profile

Beginning with an iPSC line derived from a healthy 75-year-old female, we performed CRISPR/Cas9-mediated gene editing (see Methods; Cong et al., 2013) to introduce point mutations corresponding to the R47H substitution in TREM2 protein (Figures 1A and 1B). Following screening of putative gene edited iPSC clones, we moved forward with two clones carrying the heterozygous R47H-causing mutation (R47H/+) associated with AD (Figure 1B). An additional clone from the CRISPR screening that remained un-edited at the TREM2 locus, as well as the parental iPSC line, were used as isogenic controls. We then used a modified protocol from McQuade et al. to differentiate these four iPSC lines into iPSC-derived microglial cells for further validation and analysis (see Methods; McQuade et al., 2018).

We first examined the purity of our microglia, finding that differentiated cells from all four of our lines exhibited ~99% positivity for both CD11B and CD45 by flow cytometry (Figure S1A). We then examined TREM2 mRNA and protein expression in these cells, finding similar expression of TREM2 transcript in control and R47H/+ microglia (Figure 1C). For this and all future experiments, an equal number of samples from clone A and clone B of each genotype were analyzed unless otherwise noted. Results for each clone are presented, while statistical analysis was performed using pooled samples for each genotype.

We next used western blotting to examine whether TREM2 protein levels and glycosylation/maturation were affected by the R47H/+ mutation. Lysates from both control and R47H/+ cells show a main band at ~27 kDa, corresponding to unmodified full length TREM2 protein, as well as higher molecular weight species corresponding to glycosylated forms of TREM2 and the shorter TREM2 intracellular domain (ICD) (Figure 1D). Lysates from TREM2 Null iPSC-microglia (see Methods), included as a control for antibody specificity, showed no signal for any TREM2 species (Figure 1D; Kleinberger et al., 2014). Quantification demonstrated similar levels of each TREM2 species in lysates from control and TREM2 R47H/+ microglia (Figure 1E). We then used ELISA to quantify sTREM2 levels in media from microglial cultures carrying our TREM2 mutations (Figure 1F). We observed a significant increase in sTREM2 produced from R47H/+ cells relative to controls, consistent with observations from AD patients carrying R47H/+ mutations (Figure 1F; Piccio et al., 2016).

To assess the effect of the R47H/+ mutation on TREM2 function, we then examined cell signaling pathways known to lie downstream of TREM2 (Figures 1B and S1B-S1E). We found no significant differences in the levels of phosphorylation of AKT (Ak strain transforming), PKC (protein kinase C), or the MAPKs (mitogen-activated protein kinases) ERK (extracellular signal-regulated kinase) or p38 in TREM2 R47H/+ microglia compared to controls (Figures S1B-S1E). These results suggest that the major signaling modules downstream of TREM2 are not appreciably affected by the R47H/+ mutation in unstimulated microglia.

We next sought to characterize the functional effects of the R47H/+ mutation more broadly through whole-transcriptome profiling of microglial mRNA (see Methods). Following sequencing we first performed principal component analysis (PCA) on our RNA sequencing samples using the DEseq2 package (See Methods; Figure 1G). We found that the control and R47H/+ samples clustered separately along the PC1 axis (81% variance) of the PCA plot (Figure 1G). We next performed differential gene expression analysis, identifying 1009 genes as differentially expressed in R47H/+ microglia compared to controls (see Methods; Figures 1H and 1I and Table S1). Gene ontology analysis revealed an enrichment for downregulated genes involved in biological processes affecting lipid metabolism, ribosome biogenesis, cell mobility, the cytoskeleton and apoptosis (Figures 1H and 1I). In contrast, genes found to be upregulated in R47H/+ cells showed immune and inflammatory responses and cell activation as the most enriched biological processes, with phosphorylation and cellular transport also being affected (Figure 1I).

The presence of a pro-inflammatory gene expression signature in TREM2 R47H/+ microglia was further underscored by examination of the expression levels of genes mapping to the biological process ‘inflammatory response’ that were upregulated by the R47H/+ mutation (Figure 1J). This R47H/+-associated gene expression signature included C1QA and C3 (encoding the complement components 1QA and 3), CX3CR1 (encoding the fractalkine receptor), CLEC7A (encoding a C-type lectin pattern recognition receptor) and PLCG2 (associated with AD risk and encoding phospholipase C gamma 2), all with well-established roles in microglia (Figure 1J; Andreone et al., 2020; Fuhrmann et al., 2010; Hong et al., 2016; Wang et al., 2022). Thus, our results indicate that the R47H/+ mutation induces a pro-inflammatory gene expression signature in iPSC-microglia even in the absence of inflammatory stimuli.

3.2. TREM2 R47H/+ microglia are hyper-responsive to inflammatory challenge

We next sought to examine the effects of the R47H/+ mutation on microglial inflammatory responses to address conflicting findings from previous mouse, iPSC and patient studies (Cosker et al., 2021; Liu et al., 2020; Piers et al., 2020; Sayed et al., 2021; Zhou et al., 2020). We treated iPSC-derived microglia for 2 days with one of two different pro-inflammatory stimuli, the bacterial component lipopolysaccharide (LPS, 100 ng/mL) or the cytokine interferon γ (IFNγ, 50 ng/mL). We then collected culture media and performed ELISAs to examine levels of the pro-inflammatory cytokines interleukin 6 (IL6) and tumor necrosis factor α (TNFα). LPS treatment significantly increased IL6 levels in media from both control and TREM2 R74H/+ microglia, with the R47H/+ increase being significantly greater (Figure 2A). IFNγ treatment resulted in more modest increases in IL6 levels, which were again significantly higher in media from R47H/+ cells than controls (Figure 2A). Significant increases in TNFα levels were only seen in media from R47H/+ cells after treatment with LPS or IFNγ (Figure 2B). These observations show that microglia carrying the TREM2 R47H/+ mutation exhibit a more pronounced response to pro-inflammatory stimuli than microglia carrying wild-type TREM2 protein.

We then used gene expression profiling to gain more insight into how the R47H/+ mutation impacts cellular responses to LPS and IFNγ (see Methods). Principal component analysis showed clustering of replicates by genotype and treatment condition (See Methods; Figure S2A). We found that LPS treatment for 2 hours induced similar numbers of differentially expressed genes (DEGs) in control and R47H/+ cells (Figure 2C and Tables S2 and S3). As expected, DEGs from both genotypes were highly enriched for the biological process ‘response to LPS’ (Figure 2D). We noted, however, that ~40% more genes from the ‘response to LPS’ GO category were significantly differentially expressed in R47H/+ cells compared to controls, and that R47H/+-altered genes also exhibited considerably stronger enrichment for this biological process than control-altered genes (Figure 2D). Furthermore, we found that LPS treatment caused differential expression of 87 genes from the ‘cell activation’ biological process in R47H/+ microglia, compared to only 50 in control microglia (Figure 2D). These results indicate a stronger transcriptional response to LPS in R47H/+ microglia than in control cells.

IFNγ treatment of our microglia for 2 hours induced considerably more DEGs than a similar duration of LPS treatment (Figure 2E and Tables S4 and S5). ~1450 genes were differentially expressed in R47H/+ microglia, compared to only ~850 in control microglia, suggesting a more robust IFNγ response in R47H/+ cells (Figure 2E). Again, as expected, these DEGs were strongly enriched for the pathway ‘type II interferon’, with more pathway genes altered in R47H/+ microglia than control cells (Figure 2F). As with LPS, we also observed differential expression of ~35-40% more ‘cell activation’ genes in TREM2 R47H/+ microglia following IFNγ treatment compared to controls (Figure 2F).

Consistent with our ELISA analysis of cytokine levels in microglia media, our RNA sequencing analysis showed stronger transcriptional induction of many cytokine- and chemokine-encoding genes following LPS and/or IFNγ treatment in R47H/+ microglia compared to control cells (Figure 2G). Taken together, these results indicate that in addition to a pro-inflammatory gene signature under baseline conditions (Figure 1), TREM2 R47H/+ microglia also exhibit exaggerated transcriptional responses and cytokine release following challenge with inflammatory insults.

3.3. Microglial uptake of multiple substrates is impaired by the TREM2 R47H/+ mutation

Surveillance and removal of debris from the cellular environment is a fundamental function of innate immune cells such as microglia (Li and Barres, 2018). Thus, we next sought to examine the ability of control and R47H/+ microglia to take up brain-relevant substrates such as synaptic and myelin debris and Aβ peptides. TREM2 has been shown to play a role in developmental synaptic pruning in the mouse brain, and microglia have been reported to engage in inappropriate synaptic pruning in adult Alzheimer’s disease model mice (Filipello et al., 2018; Hong et al., 2016; Salter and Stevens, 2017). Microglia also help to clear damaged myelin from the brain, a process that in mice is impaired by TREM2 loss-of-function mutations (Nugent et al., 2020; Poliani et al., 2015).

We thus purified synaptosomes (preparations of synaptic membrane) and myelin from wild-type mouse brains (see Methods). We added synaptosomes, myelin or 488-labeled Aβ₄₂ (Aβ-488) to microglia mono-cultures for 3 hours, followed by immunostaining to examine uptake of each substrate (see Methods; Figure 3A-3C). Uptake of all three substrates appeared to be reduced in R47H/+ microglia as compared to control cells (Figures 3A-3C). To more accurately quantify synaptosome and myelin uptake, we conjugated our synaptosome and myelin preparations to the pH sensitive dye pHrodo-green, which fluoresces more strongly in low pH cellular compartments such as lysosomes (see Methods). Following incubation of microglia with pHrodo-synaptosomes, pHrodo-myelin or Aβ-488 for 3 hours, cells were collected, washed and subjected to flow cytometry analysis (Figures 3D-3F). Quantification of these experiments showed significantly reduced uptake of all three target substrates by TREM2 R47H/+ microglia relative to controls (Figures 3D-3F).

Figure 3. The TREM2 R47H/+ mutation impairs microglial uptake of brain-relevant substrates.

(A) Images of TREM2 R47H/+ and isogenic control microglia stained for IBA1 (red) and SVP38 (green) following 3 hours uptake of mouse synaptosomes. A no-synaptosome condition using control cells is included to verify SVP38 specificity. Scale bar = 10 μm. (B) Images of TREM2 R47H/+ and control microglia stained for IBA1 (red) and MBP (green) following 3 hours uptake of mouse myelin. A no myelin condition using control cells is included to verify MBP specificity. Scale bar = 10 μm. (C) Images of TREM2 R47H/+ and control microglia stained for IBA1 (red) following 3 hours uptake of Aβ₄₂-488 (green). A no-Aβ₄₂-488 condition using control cells is included to verify signal specificity. Scale bar = 10 μm. (D) Quantification of pHrodo-synaptosome signal in flow cytometry experiments using TREM2 R47H/+ and control microglia. n=6 for CTRL-A, CTRL-B, R47H-A and R47H-B. *p<0.05. Student’s t-test. (E) Quantification of pHrodo-myelin signal in flow cytometry experiments using TREM2 R47H/+ and control microglia. n=7 for CTRL-A, CTRL-B, R47H-A and R47H-B. ***p<0.001. Student’s t-test. (F) Quantification of Aβ₄₂-488 signal in flow cytometry experiments using TREM2 R47H/+ and control microglia. n=6 for CTRL-A, CTRL-B, R47H-A and R47H-B. ***p<0.001. Student’s t-test. (G) Live imaging of microglial Aβ₄₂-488 (green) uptake. Microglia are labeled with Vybrant-DiD (red). Scale bar = 10 μm. (H) Quantification of Aβ₄₂-488 uptake over the 2-hour time series. n=>18 for CTRL-B and R47H-B. Representative of 2 independent experiments, one using each clone ***p<0.001. Student’s t-test.

Next, we performed live imaging experiments to track Aβ uptake over time. We first labeled our microglial cells with the cell permeant dye Vybrant DiD to allow live tracking (Figure 3G). Following addition of Aβ-488 to the culture media, we then collected confocal images of the microglia every 10 minutes for 2 hours (Figures 3G-3H; Supplemental movies 1 -2). The initially weakly diffuse Aβ-488 signal in the media gradually accumulated in Vybrant DiD-labeled microglia (Figure 3G and 3G’). Quantification of the fluorescent signal intensity per cell revealed a significantly reduced Aβ-488 signal from R47H/+ cells compared to controls as early as 20 minutes after Aβ-488 addition that persisted throughout the course of the 2-hour experiment (Figures 3H). Taken together, these findings demonstrate that the TREM2 R47H/+ mutation causes significant impairments in the uptake of multiple physiologically relevant substrates by microglia.

3.4. TREM2 R47H/+ microglia have an impaired response to laser-induced injury

In the rodent brain, microglia exhibit robust responses to various types of neuronal injury (Li and Barres, 2018). In one common injury model, microglia extend processes and migrate toward sites of laser-induced damage, a process that is impaired in heterozygous and homozygous TREM2 knockout mice (Mazaheri et al., 2017; Sayed et al., 2018). We thus sought to examine whether the TREM2 R47H/+ mutation affects the microglial injury response. To that end, we established an in vitro laser injury response model using iPSC-derived cells. We generated co-cultures of wild-type Neurogenin 2 (NGN2) induced neurons and our control and TREM2 R47H/+ microglia (see Methods; Figures 4A and 4B). For neuronal injury, microglia were live-labeled with Vybrant DiO and then encapsulated in 25% Matrigel and overlaid on >3-week NGN2 cultures (see Methods; Figure 4C). Under these conditions microglia concentrated around aggregates of neuronal cell bodies (Figure 4C, left panel). After 7-10 days of co-culture, we induced neuronal injury using high intensity confocal laser scanning (corresponding to the photobleached and boxed area in Figure 4C, middle panel), and then monitored microglial responses using time lapse imaging (Figure 4C, right panel; Supplemental movies 3-4).

Figure 4. TREM2 R47H/+ microglia show an impaired response following injury in an in vitro model.

(A) Schematic of laser injury experiment. (B) Images of NGN2 induced-neuron co-cultures with control and TREM2 R47H/+ microglia stained with β-Tubulin 3 (TUJ1) (red) and IBA1 (green). Scale bar = 50 μm. (C) Images of live NGN2 induced-neuron co-cultures with control and TREM2 R47H/+ microglia. Microglia are labeled with Vybrant-DiO dye (white). Imaged before neuronal injury (left), immediately following neuron injury (middle) and 90 minutes after injury induction (right). Scale bar = 50 μm. (D) Vectors showing movement of tracked cells corresponding to the time courses in (C). Arrows represent direction and net distance traveled by each tracked cell. Scale bar = 50 μm. (E) Quantification of directed movement for each cell toward the center of the injury site over the 90-minute session. >1000 cells tracked for both genotypes. n=7 time series for CTRL-A, 4 for CTRL-B, 3 for R47H-A, 6 for R47H-B. **p<0.01. Student’s t-test. (F) Quantification of total movement (in any direction) for each cell between time points. >25000 cell movement events tracked for both genotypes, from the same injury response time series as in (E). ***p<0.001. Student’s t-test. (G) Quantification of P2RY12 mRNA levels in TREM2 R47H/+ and isogenic control microglia. n=6 for each of CTRL-A, CTRL-B, R47H-A and R47H-B. ***p<0.001. Student’s t-test. (H) Images and (I) quantification of western blots for P2RY12 protein from control and TREM2 R47H/+ microglia. Actin serves as loading control. n=4 for each of CTRL-A, CTRL-B, R47H-A and R47H-B. *p<0.05. Student’s t-test.

Using cell tracking software, we then quantified the movement of every microglial cell in a field of view following neuronal injury (see Methods; Figures 4D and 4E). Control and R47H/+ microglia both exhibited a net directed movement towards the injured area over 90 minutes (Figure 4D and 4E), but R47H/+ microglia moved on average only half as far toward the injury site as control microglia (Figure 4E).

While R47H/+ microglia also showed significant reductions in total movement compared to control cells – i.e., movement towards, away from, or at oblique angles to the injury site –, this effect was qualitatively much smaller (~9%) than the reduction observed in directed movement towards injury sites (~50 %; compare Figures 4E and 4F). Thus, it is likely that mechanisms beyond impaired cellular movement per se contribute to the reduced neuronal injury response of R47H/+ microglia. Since microglial injury responses in mice are known to be critically dependent on signaling via the purinergic P2RY12 receptor (Haynes et al., 2006), we examined expression of P2RY12 and other purinergic receptors in our microglia cultures (Figure S2A). P2RY12 was found to be significantly downregulated in R47H/+ microglia in our original RNA sequencing analysis, as well as by both RT-qPCR and western blotting analyses (Figure S3A and Figure 4G-I). Together, these findings suggest that TREM2 R47H/+ microglia show an impaired injury response and that this impairment is at least partially mediated by reduced ATP/ADP sensing.

3.5. Xenotransplanted TREM2 R47H/+ microglia reduce synaptic density in the mouse brain

To examine our control and TREM2 R47H/+ microglia in vivo, we next performed xenotransplantation experiments in immunodeficient hCSF1 (human colony stimulating factor 1)-expressing mice (see Methods; Hasselmann et al., 2019). Microglia were transcranially injected into the hippocampi of P0-P4 pups, then mice were sacrificed at roughly 3 months of age. Immunostaining of sections from transplanted brains demonstrated co-localization of the microglial marker IBA1 and the human nuclei-specific marker STEM101 in a subset of microglia (Figure 5A). Endogenous STEM101-negative mouse microglia intermingled with transplanted STEM101-positive cells, with human microglia exhibiting stronger IBA1 signal than their mouse counterparts (Figure 5A). The density of transplanted microglia in area CA1 of the hippocampus was similar for control and R47H/+ microglia (Figure 5B and 5C).

Figure 5. Transplanted TREM2 R47H/+ microglia reduce synapse density in the mouse brain.

(A) Tile-scan image of representative hippocampus following xenotransplant using control iPSC-microglia, stained with DAPI (blue), the neuronal marker β-Tubulin 3 (TUJ1, red), the microglia marker IBA1 (green) and the human nuclei-specific marker STEM101 (white). Scale bar = 100 μm, inset scale bar = 10 μm. (B) Images of transplanted control and R47H/+ microglia in hippocampal area CA1, stained for DAPI (blue), IBA1 (green) STEM101 (red). Scale bar = 50 μm. (C) Quantification of the number of STEM101 positive cells in brain sections from mice transplanted with control and TREM2 R47H/+ microglia. n=6 for CTRL-A, 5 for CTRL-B, 5 for R47H-A, 4 for R47H-B. (D) Images of DAPI (blue) and MBP (red) in hippocampi from mice transplanted with PBS or control or TREM2 R47H/+ microglia. Scale bar = 50 μm. (E) Images of DAPI (blue) and SVP38 (green) in hippocampi from mice transplanted with PBS or control or TREM2 R47H/+ microglia. Scale bar = 50 μm. (F) Quantification of MBP signal in hippocampi of transplanted mice. n=12 from 6 mice for CTRL-A, 10 from 5 mice for CTRL-B, 10 from 5 mice for R47H-A and 8 from 4 mice for R47H-B. (G) Quantification of SVP signal in hippocampi of transplanted mice. n=12 from 6 mice for CTRL-A, 10 from 5 mice for CTRL-B, 10 from 5 mice for R47H-A and 8 from 4 mice for R47H-B. *p<0.05, **p<0.01. 1-way ANOVA with Tukey’s test. (H) High magnification images of PSD95 (green) and VGlut1 (white) in hippocampi of transplanted mice. Scale bar = 5 μm. (I) Processed images of synaptic puncta as defined by co-localization of the pre- and post-synaptic markers VGlut1 and PSD95 corresponding to the images in (H) (see Methods). (J) Quantification of the number of synaptic puncta per field of view from mice injected with PBS, control or TREM2 R47H/+ microglia. n=6 for CTRL-A, 5 for CTRL-B, 5 for R47H-A, 4 for R47H-B. *p<0.05, **p<0.01. 1-way ANOVA with Tukey’s test. (K) Heatmap of the gene expression Z-Score in control and TREM2 mutant microglia for complement system genes from our original RNA-seq experiments. Only complement system genes within the top 10,000 highest expressed genes in our iPSC microglia were included.

We then examined possible impacts of transplanted iPSC-microglia on myelin integrity and synapse density in the mouse hippocampus, using PBS injected mice as controls. Staining for myelin basic protein (MBP), we did not detect any changes in myelin between sections from control and R47H/+ injected mice compared to those injected with PBS (Figures 5D and 5F). However, staining for the pre-synaptic protein synaptophysin (SVP38) revealed a significant reduction in synaptophysin signal in brains that had been transplanted with R47H/+ microglia compared to PBS or control microglia-injected brains (Figures 5E and 5G). This reduction in synaptophysin signal was suggestive of reduced synapse density, thus we next performed co-staining for the pre-synaptic marker vesicular glutamate transporter 1 (VGlut1) and the post-synaptic marker post-synaptic density protein-95 (PSD95) to evaluate synaptic puncta in our transplanted mice (Figure 5H). Quantification of co-localized VGlut1 and PSD95 puncta confirmed a reduction of synaptic puncta in brains transplanted with R47H/+ microglia compared to those injected with PBS or control microglia (Figure 5H-5J).

These results were surprising since our in vitro experiments indicated reduced uptake of isolated synaptosomes by R47H/+ microglia (Figure 3). This difference between our in vitro and in vivo results likely reflects the distinct mechanisms that govern synapse loss in vivo (Hammond et al., 2019). Inflammatory insults, activation of uridine signaling and complement system activation are all associated with synapse loss in the mouse brain (Dundee et al., 2023; Propson et al., 2021; Rao et al., 2012). In addition to the pro-inflammatory effect that TREM2 R47H/+ mutations confer on microglia (Figures 1 and 2), our original RNA sequencing experiments also indicated an upregulation of multiple complement pathway components in R47H/+ microglia (Figure 5J). Indeed, complement activation can itself promote inflammation as well as synaptic pruning (Lee et al., 2019; Propson et al., 2021). In particular, C1QA, C1QB, C1QC and C3 were all among the top 150 highest expressed genes in our iPSC-microglia and also exhibited increased expression in TREM2 R47H/+ microglia compared to control cells (Figure 5K). Similar effects were seen for C2 and the complement receptors CR1 and C1QR1/CD93 in R47H/+ microglia, though these genes were not as highly expressed (Figure 5K). Furthermore, the purinergic receptor P2RY6, which mediates synapse loss in aging, was also up-regulated in microglia carrying TREM2 mutations (Figure S3A; Dundee et al., 2023). Thus, these observations indicate that TREM2 R47H/+ microglia can promote synapse loss in vivo, likely mediated by some combination of inflammatory mechanisms, increased P2RY6 signaling, and/or activation of complement-dependent synaptic pruning.

4. Discussion

Taken together, our observations indicate numerous effects of the TREM2 R47H/+ mutation on microglial gene expression and function. We found transcriptional effects on biological processes including lipid metabolism, the cytoskeleton, cellular movement and cellular transport. The most prominent gene expression changes, however, among upregulated genes and affected cell activation and inflammatory pathways. These gene expression changes were mirrored at the functional level, with R47H/+ microglia exhibiting enhanced cytokine release and gene expression changes following inflammatory insult, impaired cellular uptake and impaired responses to laser-induced injury. We also observed that TREM2 R47H/+ microglia induced synapse loss when transplanted into mouse brains. While there is still not consensus in the field, our findings support a growing body of evidence that TREM2 R47H/+ mutations do not act as simple TREM2 loss-of-function mutations (Cheng-Hathaway et al., 2018; Liu et al., 2020; Piers et al., 2020; Sayed et al., 2021; Song et al., 2018).

Our observations also clearly support recent single-nucleus RNA sequencing and functional studies showing a pro-inflammatory effect of the TREM2 R47H/+ mutation (Liu et al., 2020; Sayed et al., 2021). Importantly, other genetic factors such as APOE4, the most common cause of AD, have also been shown to modulate intrinsic inflammatory responses by brain cells (Ennerfelt and Lukens, 2020; Lin et al., 2018; Vitek et al., 2009). Thus, genetic factors likely synergize with various other triggers including pathogenic protein aggregates and neuronal damage in contributing to the chronic neuroinflammation that characterizes neurodegenerative disease (De Strooper and Karran, 2016; Li and Barres, 2018).

Synapse loss is another key event in AD progression. It is the pathological feature that correlates most closely with cognitive decline, and often precedes neuronal loss (Canter et al., 2016). While TREM2 knockout mice exhibit impaired developmental synaptic pruning (Filipello et al., 2018), our transplantation experiments show that human microglia carrying the TREM2 R47H/+ mutation can promote synapse loss in the mouse brain (Figure 5). These findings are broadly consistent with recent observations that TREM2 R47H mutations can promote neuron and synapse loss in vitro and in aged mice (Popescu et al., 2022; Tran et al., 2023). It should be noted, however, that Propescu and colleagues (2022) also observed increased in vitro synaptosome uptake by TREM2 R47H/+ iPSC-microglia, in contrast to our findings (Figure 3). The potential reasons for this distinction are unclear, though differences in microglia differentiation protocol and/or synaptosome preparation methodology could be responsible. Inflammatory insults are known to promote the loss of synaptic elements in vivo, while complement-mediated synaptic pruning can also become inappropriately activated in a number of neurological conditions (Hammond et al., 2019; Hong et al., 2016; Lee et al., 2019; Propson et al., 2021; Rao et al., 2012). Indeed, the complement receptor 1 (CR1) is a risk factor for AD, while polymorphisms affecting complement components C3 and C4 are associated with schizophrenia risk (Lambert et al., 2009; Lee et al., 2019; Propson et al., 2021; Sekar et al., 2016). Furthermore, C1QC and C3 were among the genes found upregulated in microglia from the brains of AD patients carrying the R47H/+ mutation compared to wild-type TREM2 carriers (Sayed et al., 2021). In addition to the inflammatory and complement system alterations we observed in TREM2 R47H/+ iPSC microglia that could impact synapse integrity (Figures 1, 2 and 5), we also found that the purinergic receptor P2RY6 was upregulated by the R47H/+ mutation (Figure S3). P2RY6 senses uridine diphosphate produced by stressed neurons and can mediate synapse loss in aging and neurodegeneration models (Dundee et al., 2023; Puigdellívol et al., 2021).

In addition to promoting inflammation and synapse loss, we found that the R47H/+ mutation also impaired a number of basic microglial functions. The uptake of a range of brain-related substrates – synaptosomes, myelin and Aβ - was impaired in TREM2 R47H/+ microglia (Figure 3). We also found that TREM2 R47H/+ microglia exhibited an impaired response to laser-induced injury (Figure 4), mimicking microglial responses to tissue damage in vivo. It is notable that despite these functional effects, we found no significant alterations in the phosphorylation levels of various signaling proteins engaged downstream of TREM2 activation (Supplemental Figure 1). A number of studies have reported increased or decreased signaling due to the TREM2 R47H mutation in response to various ligands (Kober et al., 2016; Piers et al., 2020; Song et al., 2017; Wang et al., 2015). It is thus likely that similar treatments of our iPSC-microglia would reveal R47H/+-dependent differences, however, additional experiments will be required to test this possibility directly.

Multiple studies have reported impaired uptake of various substrates by microglia carrying TREM2 mutations, however, in most cases the underlying mechanisms have not been clearly defined (Andreone et al., 2020; Liu et al., 2020; McQuade et al., 2020; Piers et al., 2020; Reich et al., 2021). PPARγ (Peroxisome proliferator- activated receptor γ) signaling was found altered in iPSC-microglia carrying TREM2 mutations and a PPARγ agonist was able to rescue impaired Aβ uptake by these cells (Piers et al., 2020). PPARG expression was reduced in our TREM2 R47H/+ iPSC-microglia (data not shown), thus it is possible that a similar mechanism underlies the reduced uptake of Aβ and other substrates by R47H/+ carrying cells that we observed (Figure 3). TREM2 has also been reported to directly bind Aβ, with the R47H/+ mutation impairing the interaction (Zhao et al., 2018), suggesting the possibility that impaired binding to Aβ could ultimately lead to reduced internalization. Furthermore, the sensing of various lipid species by TREM2 has been reported to be impaired by the R47H/+ mutation and could likewise result in reduced interaction with, and reduced uptake of, substrates such as myelin, neuronal/synaptic membrane and/or lipidated Aβ species (Song et al., 2017; Wang et al., 2015; Yeh et al., 2016).

Impaired microglial responses to laser-induced injury have been reported in the brains of TREM2 heterozygous and homozygous KO mice (Mazaheri et al., 2017; Sayed et al., 2018). To our knowledge, however, injury responses by TREM2 R47H microglia, or by human microglia carrying any TREM2 mutation, have not been explored. In our novel in vitro injury model we observed significant impairments in injury responses by TREM2 R47H/+ microglia, as well as qualitatively smaller impairments in total cellular movement (Figure 4). These observations are consistent with the downregulation of cell motility genes we observed in our transcripome analysis (Figure 1). As the impairment in directed movement towards injury sites was qualitatively much larger (~50%) than the reduction in total movement (~10%) in TREM2 R47H/+ microglia, however, we belive that additional mechanims also contribute to the injury response deficit. While expression of KCNK13, encoding the THIK1 potassium channel required for microglial surveillance (Madry et al., 2018), was not altered in our TREM2 R47H/+ microglia, we did observe dysregulated expression of multiple purinergic receptors (Figure S3). In particular, P2RY12, which is required for neuronal injury responses by mouse microglia (Haynes et al., 2006), exhibited reduced mRNA and protein expression in our human TREM2 R47H/+ microglia. Further studies will be required to fully understand the mechanisms underlying the uptake and laser injury deficits caused by the TREM2 R47H/+ mutation, but whatever the exact mechanisms, our observations suggest that the TREM2 R47H/+ mutation leads to both gain- and loss-of-function phenotypes affecting microglial function.

While the use of gene editing is a powerful tool to isolate the effects of a specific mutation in an otherwise isogenic genetic background, it should be noted that differing genetic backgrounds can impact cellular phenotypes (Afshar Saber and Sahin, 2020). Moreover, microglia in vivo, and stem cells in vitro, can also display sex-specific differences (Guneykaya et al., 2018; Waldhorn et al., 2022). As the iPSC lines used in our study were all generated from stem cells derived from a single individual – a healthy 75-year-old female, future studies will be required to confirm that the TREM2 R47H/+-dependent phenotypes we report remain consistent across different genetic backgrounds and across sexes.

The findings presented here improve our understanding of TREM2 biology, and uncover numerous gene expression and functional effects that arise from the TREM2 R47H/+ mutations in human microglia. These include impairments in microglial uptake processes and responses to cellular damage, core innate immune cell functions across tissues. Perhaps more importantly, we also found a strong pro-inflammatory phenotype, and an activation of complement pathway components, both of which may contribute to microglial R47H/+-mediated synapse loss. Our findings highlight multiple effects of the TREM2 R47H/+ mutation likely to underlie its association with AD risk and suggest new nodes that could be exploited for therapeutic intervention.

Supplementary Material

2023 Penney Glia Fig S1

Supplemental Figure 1. iPSC-microglia quality controls and TREM2-regulated signaling pathways. (A) Plots of flow cytometry experiments using control and TREM2 R47H/+ microglia showing forward and side scatter (top), DAPI signal (middle) and CD11B and CD45 positivity (bottom). Blots and quantification from control and TREM2 R47H/+ microglia of (B) phospho-AKT levels, n=8 for CTRL-A, CTRL-B, R47H-A and R47H-B, (C) phospho-PKC levels, n=4 for CTRL-A, CTRL-B, R47H-A and R47H-B, (D) phospho-ERK levels, n=7 for CTRL-A, CTRL-B, R47H-A and R47H-B and (E) phospho-p38 levels. n=7 for CTRL-A, CTRL-B, R47H-A and R47H-B.

2023 Penney Glia Fig S2

Supplemental Figure 2. PCA plot for cytokine treated iPSC-microglia. (A) PCA plot for control and TREM2 R47H/+ RNA-seq samples, either under baseline conditions or treated with 100 ng/mL LPS or 50 ng/mL IFNγ for 2 hours.

2023 Penney Glia Fig S3

Supplemental Figure 3. Purinergic receptor expression in TREM2 microglia. (A) Heatmap of the gene expression Z-Score in control and TREM2 R47H/+ microglia for purinergic receptor genes that were among the top 10,000 highest expressed genes in our iPSC microglia.

Main Points:

• The TREM2 R47H/+ mutation has multiple effects on human iPSC-microglia

• TREM2 R47H/+ impairs microglia uptake and response to injury while also promoting a pro-inflammatory phenotype and leading to synapse loss in vivo

Data availability statement:

RNA-sequencing data is available at the NCBI Gene Expression Omnibus: accession number GSE241858.