Alteration of microglial metabolism and inflammatory profile contributes to neurotoxicity in a hiPSC-derived microglia model of frontotemporal dementia 3

(


Introduction
FTD is the second leading cause of early-onset dementia, accounting for approximately 10-20% of all cases. FTD3 is a rare sub-form of FTD, caused by a point mutation in CHMP2B (Skibinski et al., 2005), an important component of the endosomal-sorting complex required for transport-III (ESCRT-III). Mutations in this gene have been observed to cause impairments of the endo-lysosomal pathway, manifesting in enlarged endosomes and impaired endosome-lysosome fusion in postmortem patient brains (Urwin et al., 2010).
We have previously shown that hiPSC-derived FTD3 neurons display accumulation of abnormally enlarged endosomes, as well as mitochondrial and respirational deficits, ultimately causing neurodegeneration (Zhang et al., 2017). More recently, we demonstrated that astrocytes play an important role in FTD3 pathogenesis. Our findings indicated altered autophagy, metabolic failure and disrupted mitochondria dynamics in hiPSC-derived FTD3 astrocytes, promoting astrocytic reactivity and secretion of inflammatory cytokines and mediators, exerting neurotoxic effects, culminating in neuronal death (Chandrasekaran et al., 2021). These findings highlight the involvement of multiple brain cell populations in FTD3 pathogenesis and progression.
Given that astrocytes interact closely via cytokines and chemokines with microglia, we consequently propose that microglial activation and responses contribute to FTD3 pathology, based on our previous observations of astrocyte reactivity. Microglia are the resident innate immune cells of the central nervous system, and normally play crucial roles in maintaining tissue homeostasis, synaptic pruning and facilitate host defense mechanisms (Hansen et al., 2018). In pathological conditions, this homeostasis is disrupted, and microglia become activated to secrete inflammatory mediators, and undergo phagocytosis to eliminate potential threats. Such pathological conditions can be based on infections, trauma, but also accumulation and aggregation of proteins and degenerating neurons (Bachiller et al., 2018). Microglia dynamically change their shape in response to stimuli to release factors in order to remove foreign substances, protein aggregates or damaged cells to restore brain homeostasis (Tang and Le, 2016). Chronic activation of microglia, which can be triggered by the inability to resolve protein aggregates or persistent cell death, can lead to neuroinflammation. Such chronic activation, commonly seen in neurodegenerative disorders, is characterized by excessive secretion of pro-inflammatory cytokines and neurotoxic mediators, which in combination with astrocytic reactivity intensifies the inflammatory response. Persistence of this stage further promotes neural death (Heneka et al., 2015;Philips and Robberecht, 2011;Tansey and Goldberg, 2010). Moreover, genome wide association studies (GWAS) have identified specific single nucleotide polymorphisms (SNPs) that can be associated with increased risk of developing FTD (Ferrari et al., 2014). Many of such risk variants are located at immune-associated loci, in genes highly expressed by microglia, and particularly genetic alterations affecting the region harboring the human leukocyte antigen (HLA) have been indicated as potential risk contributors of FTD. Together, these findings point towards an important microglial involvement in FTD pathogenesis (Broce et al., 2018).
Several lines of evidence have indicated that microglia not only change their activity and expression of cytokines and chemokines in response to stimuli, but also adapt their cell metabolism pathways necessary for microglial activity (Lauro and Limatola, 2020). The immune responses of microglia require high energy demands, and microglia have been shown to dynamically undergo changes in their metabolic profile, to meet the energy demands of the current condition (Fairley et al., 2021). Metabolic function and immunity are interlinked processes, and recent evidence demonstrate that different microglial phenotypes are linked to distinct metabolic pathways (Orihuela et al., 2016). Reactive microglia are associated with a metabolic reprogramming, from oxidative phosphorylation (OXPHOS) to favoring the glycolytic pathway, indicating changes in glucose metabolism during neuroinflammation, possibly to ensure faster ATP production to meet the increased energy needs following up-regulation of cytokine secretion and increased phagocytic activity (Lauro and Limatola, 2020). Moreover, in conditions where glucose is limited, microglia can use glutamine as an alternative fuel, to retain their critical function even in conditions where brain homeostasis is compromised (Bernier et al., 2020). This is the case in many neurodegenerative disorders, where glucose hypometabolism is evident, highlighting the metabolic flexibility of microglia (Bernier et al., 2020;Butterfield and Halliwell, 2019). Microglial metabolic changes in neurodegenerative disorders are suggested to be dependent on acute versus chronic activation. While acute activation leads to increased glycolysis and enhanced immune responses, chronic activation and prolonged reliance on glycolysis ultimately leads to metabolic defects and depleted immune functions (Aldana, 2019;Lauro and Limatola, 2020). Metabolic dysregulation could thus contribute to a shift in microglial reactivity, promoting a detrimental phenotype, either by increasing the secretion of neurotoxic mediators, or by diminishing the beneficial neuroprotective effects (Baik et al., 2019). Targeting microglia, either by stimulating or dampening microglial function, could thus be a potential therapeutic strategy. Pharmacological intervention on metabolic phenotypes represents an interesting approach to intervene into the FTD3 disease pathology. In this study, we investigate microglia-specific pathology in a hiPSC-derived microglia model of FTD3, with a heterozygous and a homozygous CHMP2B mutation. We evaluate the microglial immune function as well as metabolic activity to investigate the potential role of microglia as pro-inflammatory drivers of disease pathology.

Cell culture of hiPSC lines
In this study we used a hiPSC line derived from a wild-type healthy individual generated by episomal reprogramming of fibroblasts (Okita et al., 2011;Rasmussen et al., 2014), as well as two CRISPR/Cas9 gene edited cell lines; a mutant CHMP2B heterozygous FTD3 hiPSC line and a mutant CHMP2B homozygous FTD3 line, both established from the original healthy control, as previously described (Chandrasekaran et al., 2021). A detailed description of the cell lines can be obtained in Table S1A. All cell lines were maintained in Essential 8 (E8) media (Thermo Fisher Scientific, A1517001), with media change performed every day, and 0.5 mM EDTA (Thermo Fisher Scientific, 15575020) passaging approximately every third day, until differentiation.

Co-culture of neurons with conditioned microglia media
To evaluate detrimental neurotoxic effects of microglia, control neurons were treated with conditioned media from the different microglia groups; unstimulated wild-type microglia, unstimulated CHMP2B heterozygous FTD3 microglia, unstimulated CHMP2B homozygous FTD3 microglia, IFN-γ stimulated wild-type microglia, IFN-γ stimulated CHMP2B heterozygous FTD3 microglia, IFN-γ stimulated CHMP2B homozygous FTD3 microglia, LPS stimulated wild-type microglia, LPS stimulated CHMP2B heterozygous FTD3 microglia and LPS stimulated CHMP2B homozygous FTD3 microglia. Treatment of neurons was carried out for 24 h up to one week, and one-week exposure was selected as the optimal condition for further analysis. The one-week treatment was chosen to optimally mimic the effects of persistent microglia activation and cytokine exposure on neurons occurring in the brain. Moreover, the one-week treatment allowed for investigation of differences in neurite outgrowth, without compromising the survival of the neurons. Following, neurons were fixed for immunocytochemistry with the neuronal marker Beta-III-Tubulin, and neurite outgrowth (length) was assessed using the neurite tracer software in ImageJ (Pool et al., 2008). Data was obtained from three independent treatment experiments for each microglia condition (100 000 neurons per well).

Mesoscale neuroinflammation assessment
100 000 microglia per cm 2 were plated in 35 mm Corning Cellbind cell culture dishes (Sigma, CLS3294-210EA) and conditioned microglia media was collected from three independent microglia differentiation experiments (10 6 cells per replicate). Microglial secretion of pro-and anti-inflammatory cytokines, chemokines, vasculatory-and angiogenesis mediators was assessed using the V-PLEX Neuroinflammation Panel 1 Human Kit (MSD, K15210D), with the MESO QUICKPLEX SQ 120 Imager (MSD), coupled to the software DISCOVERY WORKBECH 4.0. Samples were run in duplicates (total of six samples per condition), and the data was normalized to the total microglia RNA concentration per sample. Results are presented as the calculated concentration mean (calc. conc. mean, pg/ml) of each cytokine measured in the six individual samples. The calculations were generated according to the manufacturer's protocol and software (MSD).

Immunocytochemistry and confocal microscopy
For immunocytochemistry (ICC), microglia (100 000 cells per cm 2 ) and neurons (50 000 cells per cm 2 ) were plated onto double-acid treated glass coverslips (12 mm), then fixed in 4% Paraformaldehyde (PFA) for 20 min at room temperature (RT), followed by a 3x5 minutes wash in PBS (Sigma, D8537). Subsequently, cells were permeabilized with a 0.2% Triton-X-100 solution for 20 min (RT), and blocked with 3% Bovine Serum Albumin (BSA, Sigma Aldrich) for 30 min (RT). Cells were then incubated with primary antibodies (Table S1B) diluted in 3% BSA, overnight at 4 • C, followed by a 3x5 min PBS wash (RT), and secondary antibody (Table S1B) incubation for 1 h in the dark (RT). Another 3x5 minutes wash with PBS was performed, then DNA was labelled with DAPI, diluted in PBS, for 7 min (dark, RT). Cells were further washed 3x5 minutes in PBS, before mounting of coverslips onto slides in DAKO fluorescence mounting solution. Samples were then imaged with a Laser Scanning confocal microscope (Zeiss LSM710) with the Zeiss Zen Black 2012 software, and obtained images were analyzed in ImageJ 2.0.0-rc-65/1.51 s.

Phagocytosis assay
Microglia (100 000 cells per cm 2 ) were cultured in a pre-coated 35 mm Corning Cellbind cell culture dish (Sigma, CLS3294-210EA) in microglia media, then supplemented with 75 μg/ml pHrodo Red E. coli BioParticles Conjugate for Phagocytosis (Thermo Fisher Scientific, P35361). Live imaging was performed using the Biostation IM Cell S1/ Cell S2 System (Nikon Instruments, 2022), and images were obtained every 10 min for 24 h. Obtained images were processed in the Nikon Biostation Imaging Software. Data was obtained from three independent terminal differentiation experiments (10 6 cells per replicate).

Transmission electron microscopy
For transmission electron microscopy (TEM), mature microglia were fixed in 3% Gluteraldehyde (Merck, 1042390250) in 0.1 M Naphosphate buffer with pH 7.2 at 4 • C for 1 h, resuspended in 0.1 M Na-phosphate buffer, centrifuged at high speed, and embedded in 2% agar, then cut into 1-2 mm 3 blocks under a stereomicroscope. Following, post-fixation was performed in 1% osmium tetroxide (EMS) in 0.1% Na-phosphate buffer for 1 h (RT). Samples were washed in MilliQ water, then a stepwise dehydration in ethanol with increasing concentration was performed. Propylene oxide (Merck) was used as an intermediate to allow overnight infiltration with Epon (TAAB, T031). The following day, the cells were embedded in pure Epon, then cured at 60 • C for 48 h. An ultramicrotome with a glass knife (Leica Ultracut, Leica Microsystems, Wetzlar, Germany) was used to cut semi-thin sections (2 μm), which were stained with 1% Toluidine blue (Millipore, 1159300025) in 1% Borax (LabChem, LC117101), to evaluate a section of interest. Ultra-thin sections (50-70 nm) were then cut on the ultramicrotome with a diamond knife (Jumdi, 2 mm), and sections were collected onto membrane-covered grids. Following, grids were contrasted with 2% uranyl acetate (Polyscience) and lead citrate (Reynolds 1963), then analyzed with a Philips CM100 transmission electron microscope operating at 60 kV, connected to Morada digital camera equipment and the iTEM software system (Olympus). Data was obtained from three independent terminal differentiation experiments (10 6 cells per replicate).

Metabolic labelling
Microglia (100 000 per cm 2 ) were plated in pre-coated 6-well Corning Cellbind multiple well plates (Sigma, CLS3335-50EA). For metabolic assessment, microglia were washed with PBS, then incubated with [U-13 C]glutamine (99 %) from Cambridge Isotope Laboratories (Tewksbury, MA, USA) for 90 min at 37 • C. Following, the incubation media was collected, and microglia were treated with ice-cold PBS, then lysed and extracted using 70% ethanol. Cells were then scraped off and centrifuged, and the cell extracts (supernatant) were lyophilized and reconstituted in water for biochemical analysis, whereas the cell pellets were used for protein content assessment using the Pierce BCA Assay with BSA as standard. Cell extracts were separated, and mapping of metabolites was performed with a gas chromatograph (Agilent 7820A chromatograph, J&W GC Column HP-5MS, parts no. 19091S-433) coupled to a mass spectrometer (Agilent, 5977E). Following, the isotopic enrichment was calculated according to (Biemann, 1962), with data presented as labelling (%) of mass of the unlabelled molecule (M) + number of labelled C-atoms (X) in a given metabolite. Data was obtained from three independent terminal differentiation experiments (3 × 10 6 cells per replicate).

RNA extraction and sequencing
RNA was extracted from the microglia using the RNeasy® Plus Micro Kit (Qiagen, 74004), according to the manufacturer's protocol, and the quality of the RNA was assessed using an Agilent 2100 Bioanalyzer system with RNA 6000 nano chip and reagents. Library preparation and sequencing (DNBseq, 2x100 nt, stranded paired-end) was performed by an external provider (Beijing Genomics Institute, BGI). RNA was collected in two batches to ensure sufficient material for RNA sequencing. In batch one (B1) RNA was collected from unstimulated (two replicates), TNF-α treated (three replicates) and LPS treated wildtype microglia (three replicates), as well as unstimulated (three replicates), TNF-α treated (three replicates) and LPS treated (three replicates) CHMP2B homozygous FTD3 microglia, and TNF-α treated (three replicates) and LPS-treated (three replicates) CHMP2B heterozygous FTD3 microglia (three replicates). In batch 2 (B2) RNA was collected from unstimulated wild-type microglia (three replicates) and unstimulated CHMP2B heterozygous FTD3 microglia (three replicates). B1 and B2 were combined, and data was thus obtained from five independent replicates of unstimulated wild-type microglia and three independent replicates for all other conditions (2 × 10 6 cells per replicate).

RNA-seq data analysis pipeline
Total RNA-seq reads were pre-processed with Cutadapt (Martin, 2011) v1.18 to remove adapter residues, reads with low quality, and reads excessively short after trimming (-q = 30, -pair-filter = any, min_length= "20{\Prime}). BBDuk v38.22 (BBMap -Bushnell B. -sourc eforge.net/projects/bbmap/) was used to remove residual ribosomal RNA reads (minimum covered fraction = 0.5, kmer size = 31) annotated in SILVA v119.1 (Quast et al., 2013). Before and after each preprocessing step FastQC v0.11.5 (https://www.bioinformatics.babraha m.ac.uk/projects/fastqc/) and MultiQC (Ewels et al., 2016) v1.9 were used to assess the reads status (data not shown). The strandness was confirmed with RSeQC (Wang et al., 2012) v4.0.0. Reads were mapped to the human genome hg38 with STAR (Dobin et al., 2013) v2.6.1d (1pass). The number of reads present before and after each of the preprocessing steps and the mapping can be found in Figure S1. The transcriptome was assembled in StringTie (Kovaka et al., 2019;Pertea et al., 2015) v2.0 and extended annotations produced by StringTie in merge mode, including both the assembled transcripts (minimum tpm = 1, minimum fpkm = 1) and a set of reference annotations derived by merging GENCODE (Frankish et al., 2019) v33 and FANTOM-CAT (Hon et al., 2017), were provided to featureCounts (Liao et al., 2014) (subread v1.6.3) for gene quantification (minimum overlap = 90 nt, minimum fraction of overlapping nt = 90%, non-chimeric mapping of the readpair fragment). The average number of read-pair fragments uniquely assigned to an annotated gene (±standard deviation) was 17.74 ± 8.48 million. Differential gene expression analysis was carried out in DESeq2 (Love et al., 2014) v1.22.2, using the following design formula to account for the two sequencing batches: ~BATCH + GENOTYPE + TREATMENT + GENOTYPE:TREATMENT. In the principal component analysis based on the DESeq2 rlog-normalized gene counts, the batch effect was removed using the removeBatchEffect function of limma (Ritchie et al., 2015) v3.50.1. StageR (Van den Berge et al., 2017) v1.16.0 was used to adjust the false discovery rate (FDR) given that multiple comparisons (between genotypes given a fixed treatment or vice versa) were performed. For this, we extracted the contrasts of each comparison from the DESeq2 model matrix and screened genes with Benjamini-Hochberg (Benjamini and Hochberg, 1995) adjusted Waldtest p-values < 0.05 in StageR to create a p-value matrix for the stagewise confirmation step, in which the individual hypotheses are assessed for the genes that pass the screening stage. Next, the familywise error rate correction was performed at the confirmation stage using the Holm's method (Sture Holm, 1979) to obtain the overall FDR (OFDR).

Overrepresentation analysis
Genes quantified by the RNA-seq data analysis pipeline were considered expressed if represented by at least 1 read in at least 3 of the samples. Expressed genes were considered differentially expressed (DEGs) in a certain genotype or treatment comparison if they presented absolute log2 fold change > 1, stage-wise adjusted overall FDR < 0.05 and mean of DESeq2-normalized read counts > 10. The results of the DEG analysis are reported in Supplementary Data 1. The overrepresentation analysis was carried out in Cytoscape v.3.9.1 (Shannon et al., 2003) using the stringApp v.1.7.1  to retrieve proteins corresponding to translated DEGs from the STRING database (Szklarczyk et al., 2019) and to perform the enrichment test using all expressed genes as background. The unfiltered results of this analysis are reported in Supplementary Data 2. For the graphical representations of the overrepresented terms, highly redundant terms (gene set overlap ≥ 50%) were filtered out in stringApp. Plots were generated using make_dotplot (https://github.com/gcorsi/enrichment-visualizer), limiting the terms shown to the top 10 most significantly overrepresented (lowest FDR) and grouping together GO Biological Processes (Carbon et al., 2017), KEGG pathways (Kanehisa et al., 2016), WikiPathways (Martens et al., 2021) Reactome Pathways (Gillespie et al., 2022).

Protein-Protein physical interaction network
A STRING network of physical interactions (confidence cutoff > 0.8) between the protein VPS4 and all CHMP proteins found expressed in the microglia cells according to the RNA-seq data analysis was created in Cytoscape via stringApp. The node colors and the external donut chart were set using Omics Visualizer v.1.3.0 (Legeay et al., 2020).

qPCR validation
RNA was extracted from the microglia using the RNeasy® Plus Micro Kit (Qiagen, 74004), according to the manufacturer's protocol. cDNA synthesis was performed with 100 ng of total RNA, according to the Promega ImProm-II TM Reverse Transcription System (Promega, A3800).
Shortly, the RNA was mixed with oligoT random primer (0.5 μg/μl), heated for 5 min at 70 • C, then immediately put on ice for 5 min. Following, reverse transcription mix consisting of ImProm II buffer, dNTP mix (10 mM), Nuclease-free water, RNasin Ribonuclease inhibitor (40u/μl), MgCl 2 (2.5 mM) and ImProm II was added, and samples were incubated for 5 min (RT), then 1 h at 42 • C. The enzyme was then inactivated at 70 • C for 15 min, and preferable dilutions of cDNA (1:5) was prepared for qPCR. For qPCR analysis, a mastermix containing nuclease-free water, SYBR TM Green PCR Master Mix (Applied Biosystems, 4309155), forward-and reverse primers (10 μM, TAG Copenhagen, Table S1C) was added to each well of a qPCR plate, with a total of 8 μl per reaction. 2 μl of cDNA (1 ng/μl) was added per reaction well. The qPCR was performed in triplicates, and a non-template control was included for each target gene. The analysis was performed using the QuantStudio TM 3 Real-Time PCR system (Applied Biosystems). Data was processed with the Design & Analysis 2.6.0 software and normalized to GAPDH as well as TBP and S18 (Supplementary, Figure S7). The relative mRNA expression was calculated using the ΔΔCT method.

CHMP2B intron5 FTD3 mouse model validation
Snap frozen brains of 7-months CHMP2B wild-type and CHMP2B intron5 mice, as previously published in (Ghazi-Noori et al., 2012), were kindly provided by Adrian M. Isaacs. The mouse brains were dissected, and the brainstem and cerebellum were removed. RNA was extracted from one cerebral hemisphere and used for qPCR validation. Data was obtained from three CHMP2B wild-type mice and three CHMP2B intron5 mice.

Statistics
All statistical analyses presented in the Fig. 1D, E; 3F-G; 4A, E-H; 6A-D; 7C-F; 8A-I; S4A-D; S5A-D; S6A-G; S7A-N as well as Tables S1D, S1E, S1G, S1H and S1I were executed in GraphPad Prism Version 9.2.0, and the data is presented as mean ± standard error of the mean (SEM). Significance was determined using a Welch's t test ( Table S1H) or an area under the curve analysis followed by a one-way ANOVA with correction for multiple comparisons (Dunnett, Fig. 1D-E and Table S1G). A p-value < 0.05 was considered statistically significant; *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. The statistical analyses presented in Fig Figure S2; Figure S3; Table S1F and Supplementary Data 1 is described in the RNA-seq and overrepresentation analysis section.

Efficient Generation and characterization of FTD3 microglia from human induced pluripotent stem cells
To systematically investigate the CHMP2B function and implication of mutations on microglia phenotypes, two knock-in lines carrying either a heterozygous or a homozygous CHMP2B IVS5AS G-C mutation were generated by implementing CRISPR-Cas9 precision genome editing in a control hiPSC line from a healthy individual (K3P53) (Rasmussen et al., 2014) (Supplementary, Table S1A). The CHMP2B mutant lines and parental control lines were subsequently differentiated into microglia according to a modified protocol (Haenseler et al., 2017;van Wilgenburg et al., 2013) (Fig. 1A). All hiPSC lines demonstrated similar differentiation potential, verified by immunocytochemistry (ICC). The generated microglia showed positive expression of Ionized Calcium Binding Adaptor Molecule 1 (IBA1) and Purinergic Receptor P2Y12 (P2RY12) (Fig. 1B), which are commonly exploited as microglia-specific markers (Jurga et al., 2020). However, IBA1 expression is also seen in other types of macrophages (Kenkhuis et al., 2022). Consequently, we assessed the expression of Transmembrane Protein 119 (TMEM119), a highly specific and microglia exclusive marker (Butovsky et al., 2014). Both control and FTD3 microglia demonstrated positive TMEM119 expression, indicating a 100% pure population of microglia ( Fig. 1B).

Functional characterization of hiPSC-derived microglia
The functionality of hiPSC-derived microglia was assessed via pHrodo™ -labelled E. coli bioparticle phagocytosis assay and live imaging. Both control and FTD3 microglia were able to successfully perform phagocytosis of bacteria, indicating that these are functionally active microglia, thus preserving the properties of microglia in the brain (Fig. 1C). We quantitatively assessed the phagocytic properties over the course of 24 h using the Biostation IM Cell S1/Cell S2 system (Nikon Instruments, 2022). This analysis revealed increased phagocytic activity in the FTD3 microglia, and the excessive activity was particularly profound in the CHMP2B homozygous FTD3 line, evident by an overall significant increase in fluorescence intensity, demonstrating increased uptake of pHrodo™ -labelled particles (Fig. 1D, E). These findings indicate a state of microglial hyperactivation in FTD3. The CHMP2B heterozygous FTD3 microglia showed increased phagocytic activity at early time points (Fig. 1D), which started decreasing after approximately 20 h indicating that the CHMP2B heterozygous FTD3 microglia might exist already in a hyperactive phagocytic state early-on and become less able to respond to the stimuli and perform phagocytosis. In contrast, the CHMP2B homozygous FTD3 microglia remained in a hyperactive phagocytic state throughout the analysis, suggesting that the presented microglial phenotype is dependent on the presence of a homozygous CHMP2B mutation. FTD3 patients have so far only been described presenting with the heterozygous mutation, but we decided to include the homozygous mutation to achieve a systematic overview of the CHMP2B function in microglia response. Although FTD3 patients with a homozygous mutation have not yet been identified, investigating the role of both mutations gives valuable insights into CHMP2B function in microglial phenotype and activity.

Transcriptomic characterization of hiPSC-derived microglia
The functional characterization of our hiPSC-derived microglia indicated diverse microglia profiles depending on the underlying genotypes. To further evaluate the microglia profiles, we compared the transcriptomes of the different genotypes by high-throughput RNA sequencing. For studying the transcriptomic response of microglia to activation, we stimulated the cell lines with the cytokine tumor necrosis factor (TNF-alpha) and the bacterial toxin lipopolysaccharide (LPS). The Principal Component (PC) and the differential expression analysis of the Results are displayed as mean ± standard error of the mean (SEM) from three independent replicates. Statistical significance (D-E) was determined using an area under the curve analysis followed by a one-way ANOVA with correction for multiple comparisons (Dunnett) and significance levels are indicated p < 0.0001****. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.
Differential gene expression between FTD3 genotypes and treatments. A. Principal component (PC) analysis of RNA-seq data for all samples. The first two PCs account for 68% of the expression variance in the RNA-seq data from the 29 investigated samples. The samples are labelled by genotype (Wild-type, Heterozygous, Homozygous), treatment (Vehicle, LPS, TNF-α) and batch number (B1, B2). B. Venn diagrams of the sets of genes found differentially expressed in the comparisons between treatments for each CHMP2B FTD3 genotype. C. Venn diagrams of the sets of genes found differentially expressed in the comparisons between CHMP2B FTD3 genotypes for each treatment. The size of the circles reflects the size of a set; the intersection "(Het/WT ∩ Hom/WT) -Hom/Het" in the untreated cells (Vehicle) was annotated manually for visibility (n = 40 genes). The data was obtained from a minimum of three independent replicates of each condition. For unstimulated wild-type microglia five independent replicates were analyzed. ∩ = intersection of sets.
(caption on next page) H. Haukedal et al. RNA sequencing data revealed differences in the microglia transcriptomes between the genotypes as well as the treated (TNF-α or LPS) vs. unstimulated (veh) conditions (Fig. 2). The largest variance explained by a single PC (43%) mainly captured the differences between unstimulated microglia and microglia treated with TNF-a or LPS, whereas 25% of the gene expression variance can be explained by the different genotypes (Fig. 2A). The differential expression analysis of the RNA sequencing revealed distinct sets of differentially expressed genes (DEGs) within the given genotypes in response to stimuli (Fig. 2B), particularly in pathways related to immune responses (Supplementary Data 1 and 2). Moreover, comparison of the wild-type and FTD3 microglia in each condition (TNF-α-treated, LPS-treated and unstimulated (vehicle)) revealed clear differences (Fig. 2C). In line with the observation from the functional microglia characterization, the transcriptomic profile of CHMP2B homozygous FTD3 microglia was very different from the wild-type microglia (4854 differentially expressed genes), but the difference to CHMP2B heterozygous FTD3 microglia was even bigger (8240 DEGs). The heterozygous FTD3 and wild-type microglia were most similar in their transcriptome (3940 DEGs), likely due to the presence of one common allele. Microglia activation showed a dramatic change of gene expression (to the largest extent overlapping between the two treatments) compared to unstimulated conditions, confirming a functional microglia population with changes in reactivity profile in response to stimuli (Fig. 2B). The discrepancy between the transcriptome profiles of heterozygous and homozygous FTD3 microglia was further evident in the poor concordance of the expression changes affecting genes differentially expressed in both homozygous and heterozygous microglia compared to wild type, particularly in unstimulated conditions (Supplementary, Fig. S2B).

FTD3 microglia demonstrate altered degradation of phagocytosed particles
The phagocytosis assessment indicates that our FTD3 microglia are functionally active and able to take up foreign particles. However, defects in protein degradation have been observed in FTD3, with impaired autophagy and accumulation of cargo-filled vesicles (Chandrasekaran et al., 2021). CHMP2B and ESCRT-III play a key role in the endo-lysosomal pathway, and evidence imply an important involvement of this complex in the process of phagocytosis (Avalos-Padilla et al., 2018;Pareja et al., 2017). Therefore, we investigated whether our hiPSCderived microglia were able to degrade the cargo engulfed by phagocytosis. Transmission electron microscopy (TEM) was performed to evaluate the presence of vesicles and cargo-filled structures at resting state, as well as after treatment with pHrodo™ -labelled E. coli bioparticles. Interestingly, the microglia carrying a heterozygous CHMP2B FTD3 mutation demonstrated accumulation of cargo-filled structures at an unstimulated state, suggesting alteration of the degradation pathway, similar to our previous findings in astrocytes (Chandrasekaran et al., 2021). Qualitative TEM analysis demonstrated large, empty vesicles in the wild-type microglia and CHMP2B homozygous FTD3 microglia, indicating successful clearing of cargo (Fig. 3A). These findings suggest that heterozygous and homozygous mutations subsequently result in alternate phenotypes. Following microglia treatment with bacteria particles, wild-type microglia appeared to efficiently clear the phagocytic uptake, whereas enhanced accumulation of cargo-filled structures could be observed in the CHMP2B heterozygous FTD3 microglia, largely taking up the cellular space (Fig. 3B). The enhanced presence of cargo-filled vesicles indicates that although the CHMP2B heterozygous FTD3 microglia can perform phagocytosis, they are not able to sufficiently clear the cargo uptake, which might explain why the phagocytic activity depletes over the course of time (Fig. 1D). Strikingly, we see an increased accumulation of cargo-filled vesicles in the CHMP2B homozygous FTD3 microglia following bacteria treatment, indicating that the phagocytic activity is too excessive for the cells to be able to clear the uptake (Fig. 3B). A number of these structures appear substantially more electron-dense compared to the vesicles observed in the CHMP2B heterozygous FTD3 line. This could be an indicator of lysosome build-up and suggests that different stages of the endo-lysosomal pathway might be implicated depending on the state of the CHMP2B mutation. Moreover, shedding of vesicles seems to be intact in both wildtype and CHMP2B homozygous FTD3 microglia, whereas these structures accumulate within the CHMP2B heterozygous FTD3 microglial cells (Fig. 3C).

Altered expression of autophagy-associated genes in FTD3 microglia
Differential gene expression analysis, based on the RNA-seq data, provided further indication of altered protein degeneration and autophagy in FTD3 microglia. We report a significant (defined in methods) increased expression level of autophagy associated genes such as SQSTM1, ATP13A2, MAP1LC3A and LAMPs in unstimulated CHMP2B heterozygous FTD3 microglia compared to wild-type microglia (Fig. 3D, Supplementary Data 1). These genes are implicated in autophagosome formation and the increase in autophagosome production may relate to the observed abnormal accumulation of cargo-filled structures. On the contrary, the same genes presented significantly lower expression levels in unstimulated CHMP2B homozygous FTD3 microglia (Fig. 3D, Supplementary Data 1), where no accumulation was identified, highlighting the finding of two alternate microglial phenotypes (Fig. 3E). Interestingly, loss of ATP13A2 has further been associated with lysosomal storage disorder (Bras et al., 2012;van Veen et al., 2014). The expression of autophagy-related genes was however increased in LPS-treated CHMP2B homozygous FTD3 microglia (Supplementary Data 1), which correlates with the TEM analysis of microglia exposed to bacteria (Fig. 3B). Additionally, the differential expression analysis revealed significant increased expression level of genes associated with mitophagy and damaged mitochondria such as BNIP3, OPTN and PINK1 in unstimulated CHMP2B heterozygous FTD3 microglia compared to the unstimulated wild-type microglia. In contrary, the expression levels of the same set of genes were decreased in unstimulated CHMP2B homozygous FTD3 microglia compared to the wild type (Fig. 3D, Supplementary Data 1), as well as to the heterozygous microglia (Fig. 3E,  Supplementary Data 1). These findings were confirmed by qPCR, which demonstrated significant increased levels of SQSTM1 in unstimulated heterozygous microglia and significant decreased expression levels in unstimulated homozygous microglia, compared to the wild type (Fig. 3F). This may further contribute to alterations in mitochondria dynamics and dysregulation of fission and fusion (Fig. 4). To validate our findings in an in vivo model system, we implemented a Fig. 3. FTD3 microglia display altered protein degradation. A. TEM evaluation of cargo-filled vesicles in resting microglia conditions. Scale bar 1 μm. B. TEM evaluation of cargo-filled structures in microglia following phagocytosis assay, with treatment of E. coli bioparticles. Vesicles are indicated by arrows. Scale bar 1 μm. C. TEM evaluation of vesicle shedding in microglia following phagocytosis assay, with treatment of E. coli bioparticles. Scale bar 1 μm. D. Overview of significantly differentially expressed genes associated with protein degradation and autophagy in CHMP2B heterozygous and homozygous FTD3 microglia compared to wild-type microglia. E. Overview of significantly differentially expressed genes associated with protein degradation and autophagy in CHMP2B homozygous FTD3 microglia compared to CHMP2B heterozygous FTD3 microglia (absolute log2FC > 1; statistical significance as described in the Methods and reported in Supplementary Data 1). The data was obtained from five independent replicates of wild-type microglia and three independent replicates of heterozygous and homozygous microglia. F. qPCR validation of SQSTM1 expression. G. qPCR validation of Optn, Sqstm1 and Lamp1 in the CHMP2B intron5 FTD3 mouse model compared to CHMP2B wild-type mice. Results are displayed as mean ± standard error of the mean (SEM) from three independent replicates. Statistical significance was determined using a Welch's t test (F-G) and significance levels are indicated by p < 0.05* and p < 0.0001****.
(caption on next page) H. Haukedal et al. CHMP2B intron5 FTD3 mouse model in the study. qPCR analysis of FTD3 mice brain tissue confirmed increased expression of the autophagyrelated genes Optn, Sqstm1 and Lamp1 (Fig. 3G), comparable to our findings in the human heterozygous FTD3 microglia. Interestingly, the CHMP2B intron5 FTD3 mouse model is maintained as a homozygous line (Ghazi-Noori et al., 2012). Accumulation of cargo-filled phagosomes is only observed in our homozygous microglia after bacteria exposure (Fig. 3B). This is accompanied with an increase in expression of autophagy-associated genes (Supplementary Data 1). The FTD3 mice thus resemble our bacteria-stimulated homozygous microglia, which can be correlated with the reactivity and gliosis identified in the CHMP2B intron5 FTD3 mouse model (Clayton et al., 2017). Importantly, the mouse model is a homozygous model, and is therefore not a true reflection of the patient situation.

FTD3 microglia demonstrate altered metabolism
The inflammatory profile of microglia, as well as their capability to phagocytose and degrade particles is highly dependent on their metabolic activity. Moreover, metabolic defects have been identified in both neurons and astrocytes in FTD3 (Aldana et al., 2020;Chandrasekaran et al., 2021). Therefore, we investigated glucose and glutamine metabolism in our hiPSC-derived microglia via dynamic metabolic mapping with stable isotope labelling. Metabolic reprogramming is commonly observed in reactive microglia, and metabolic defects following activation can ultimately result in a chronic microglial state, with diminished immune responses (Baik et al., 2019), similar to what we observe in our CHMP2B heterozygous FTD3 microglia. No changes in glucose metabolism were however evident in the CHMP2B homozygous FTD3 microglia. Instead, glutamine mapping revealed an increased activity in the CHMP2B homozygous FTD3 microglia, demonstrated by a significant increase in citrate, malate and aspartate (Fig. 4A), indicating a shift within the tricarboxylic acid (TCA) cycle, that was not observed in the CHMP2B heterozygous FTD3 microglia. Labeling in succinate derived from labeled glutamine was significantly increased in both CHMP2B homozygous and heterozygous FTD3 microglia.
Moreover, differential expression analysis revealed altered expression of genes associated with mitochondria and OXPHOS, displaying significantly lower expression levels in CHMP2B heterozygous FTD3 microglia compared to wild-type microglia in unstimulated conditions. This includes MT-ATPs which are a part of the mitochondria protontransporting ATP synthase, MT-Cos which form the catalytic core of Fig. 4. FTD3 microglia display altered metabolism and mitochondria dynamics. A. Assessment of glutamine metabolism, via metabolic mapping with [U-13 C] glutamine and gas chromatography coupled to mass spectrometry (GCMS) presented as 13C-enrichment in selected metabolites. B. Overview of significantly differentially expressed genes associated with metabolism and mitochondria in CHMP2B heterozygous and homozygous FTD3 microglia compared to wild-type microglia. C. Overview of significantly differentially expressed genes associated with metabolism and mitochondria in CHMP2B homozygous FTD3 microglia compared to CHMP2B heterozygous FTD3 microglia (absolute log2FC > 1; statistical significance as described in the Methods and reported in Supplementary Data 1. The data was obtained from five independent replicates of wild-type microglia and three independent replicates of heterozygous and homozygous microglia. D. TEM evaluation of mitochondria. Scale bar 1 μm. E. Quantitative TEM morphometry assessment of relative mitochondria to cytoplasm ratio. F. Quantitative TEM morphometry assessment of relative individual mitochondria size. G. qPCR validation of Mt-nd4l expression. H. qPCR assessment of Mt-nd4l in the CHMP2B intron5 FTD3 mouse model compared to CHMP2B wild-type mice. Results are displayed as mean ± SEM from three independent replicates. Statistical significance was determined using a one-way ANOVA with correction for multiple comparisons (Tukey, A) or a Welch's t test (E-H) and significance levels are indicated by p < 0.05*, p < 0.01**, p < 0.001*** and p < 0.0001****.

Fig. 5.
Overrepresented terms enriched for the differentially expressed genes found by comparing (left) TNFa-treated cells with unstimulated cells ("Vehicle") or (right) LPS-treated cells with unstimulated cells ("Vehicle"). The terms represent Gene Ontology (GO) Biological Processes, KEGG Pathways, WikiPathways, or Reactome Pathways, in which the differentially expressed genes participate. The background for the overrepresentation analysis comprised all expressed genes. The top 10 overrepresented terms are reported for each of the three genotypes in separate rows. The data was obtained from a minimum of three independent replicates of each condition. For unstimulated wild-type microglia five independent replicates were analyzed. Results are displayed as mean ± SEM from six independent replicates. Statistical significance was determined using a one-way ANOVA with correction for multiple comparisons (Tukey, A-D) and significance levels are indicated by p < 0.05*, p < 0.01**, p < 0.001*** and p < 0.0001****. Representative ICC images of neurite outgrowth analysis of healthy neurons, expressing MAP2, after treatment with conditioned microglia media from various conditions. Scale bar 50 μm. C. Quantitative assessment of neurite outgrowth after exposure of conditioned microglia media (from unstimulated microglia). D. Quantitative assessment of neurite outgrowth after exposure of conditioned microglia media (from IFN-γ treated microglia). E. Quantitative assessment of neurite outgrowth after exposure of conditioned microglia media (from LPS treated microglia). F. Quantitative assessment of effect of IFN-γ and LPS treatment within the individual cell lines. Results are displayed as mean ± SEM from three independent replicates. Statistical significance was determined using a two-way ANOVA with correction for multiple comparisons (Tukey, C-F) and significance levels are indicated by p < 0.05*, p < 0.01**, p < 0.001*** and p < 0.0001****. Overview of microglia pro-inflammatory secretion in response to IFN-γ stimulation. K. Overview of microglia pro/anti-inflammatory secretion in response to IFN-γ stimulation. L. Overview of microglia anti-inflammatory secretion in response to IFN-γ stimulation. Results are displayed as mean ± SEM from three independent replicates. Statistical significance was determined using a two-way ANOVA with correction for multiple comparisons (Dunnett, A-G) and significance levels are indicated by p < 0.05*, p < 0.01**, p < 0.001*** and p < 0.0001****. cytochrome c oxidase, and MT-NDs, a part of the NADH dehydrogenase (Fig. 4B, Supplementary Data 1). In contrary, these genes demonstrated increased expression levels in unstimulated CHMP2B homozygous FTD3 microglia compared to wild-type microglia (Fig. 4B, Supplementary Data 1), as well as to heterozygous microglia (Fig. 4C, Supplementary Data 1), once again confirming an alternate phenotype depending on the mutation status. This was validated by qPCR, demonstrating significant decrease in MT-ND4L expression in heterozygous FTD3 microglia compared to wild-type microglia, opposed to a significant increased expression in homozygous microglia (Fig. 4G). These analyses thus support the findings from the metabolic assessment, indicating an increased alternative metabolic activity in CHMP2B homozygous FTD3 microglia, potentially indicating mitochondria alterations. The CHMP2B intron5 FTD3 mouse model demonstrated a mitochondria profile comparable to the CHMP2B homozygous FTD3 microglia, validated by qPCR, showing significantly increased expression of mt-nd4l (Fig. 4H).

FTD3 microglia demonstrate altered mitochondria dynamics
The observation of metabolic alterations as well as altered expression of mitochondria associated genes in FTD3 microglia led us to investigate the mitochondria ultrastructure and distribution within the cells using TEM (Fig. 4D). CHMP2B homozygous FTD3 microglia displayed smaller mitochondria, accumulating in the perinuclear regions, likely due to the increased energy demands necessary to support the observed hyperactivity. These findings suggest an imbalance in mitochondria fission and fusion dynamics, favoring mitochondrial fission. Moreover, quantitative analysis revealed a non-significant trend towards increased relative mitochondria to cytoplasm ratio in these cells (Fig. 4E), as well as a significant reduction of the relative individual mitochondria size, confirming an increased presence of smaller mitochondria (Fig. 4F). In contrary, the CHMP2B heterozygous mutant showed the opposite trend, reflecting a non-significant decreased relative ratio (Fig. 4E) and a significant increase of the relative individual mitochondria area (Fig. 4F), indicating fewer, large mitochondria. These findings correlate well with the alternate activity we observe resulting from a heterozygous and homozygous CHMP2B mutation.

Enrichment of inflammatory pathways in response to stimuli
To evaluate changes within the transcriptomic microglial profile in response to stimuli, RNA sequencing and differential gene expression analysis of microglia stimulated with TNF-α and LPS was performed. The treatments were compared to unstimulated (vehicle) microglia for each genotype. Overrepresented terms relative to the differentially expressed genes were identified based on an enrichment analysis, representing Gene Ontology (GO) Biological Processes, KEGG Pathways, WikiPathways or Reactome Pathways using Cytoscape, as described in detail in the method section. The enrichment analysis was performed on the top DEGs with all expressed genes as background. Many of the top overrepresented terms, both in TNF-α stimulated (left) and LPS-stimulated (right) conditions, were related to immune system response pathways. This indicates a change in the microglial reactivity in response to stimuli, resembling functional microglia of the CNS. Immune-related pathways were particularly overrepresented in microglia treated with LPS, as expected with a stronger stimulus (Fig. 5).

FTD3 microglia display altered inflammatory profile
To further assess the inflammatory profile of our hiPSC-derived microglia, we evaluated the secretion of cytokines, chemokines, angiogenesis factors, and vasculature mediators using mesoscale analysis. Interestingly, CHMP2B homozygous FTD3 microglia demonstrated increased secretion levels of key cytokines in an unstimulated state, compared to wild-type microglia. In contrast, CHMP2B heterozygous FTD3 microglia showed reduced levels of cytokine and inflammatory mediators compared to the control (Fig. 6). IFN-γ treatment, which is expected to initiate a similar response as TNF-α, showed a similar trend, whereas LPS treatment overall seemed to have a stronger effect in the wild-type microglia (Supplementary, Table S1E). The differential gene Fig. 9. Functional redundancy. Network of high-confidence physical interactions between VPS4 and all CHMP proteins found expressed according to the RNAsequencing analysis, obtained from STRING (confidence cutoff > 0.8). The internal node color shows the log2 fold change of the gene's expression between unstimulated CHMP2B FTD3 heterozygous cells (left circle side) or unstimulated CHMP2B FTD3 homozygous cells (right circle side) and the unstimulated wild-type microglia. The external node color shows the significance of the expression change (overall FDR). The data was obtained from five independent replicates of wildtype microglia and three independent replicates of CHMP2B heterozygous and homozygous FTD3 microglia. expression profile of inflammatory cytokines and mediators in FTD3 microglia, however, showed few changes compared to the wild-type microglia. Most of these were not significantly changed, and overall low agreement in terms of increased or decreased expression was detected (Supplementary, Table S1F). This indicates that the changes we observe in secreted cytokines are likely caused by translational regulation.
In an unstimulated state, CHMP2B homozygous FTD3 microglia displayed a more profound pro-inflammatory profile, compared to the wild-type microglia, with elevated secretion of key pro-inflammatory cytokines such as IFN-γ and IL-1β (Fig. 6A). Microglial secretion of IFN-γ is controversial, particularly in unstimulated conditions (Monteiro et al., 2017). However, the secretion of IFN-γ was only measured in unstimulated and LPS-treated microglia, with no IFN-γ supplementation. Our results thus clearly indicate that the cytokine is in fact produced and released by the microglia, and not a consequence of media contamination. The concentration of IFN-γ secreted by the unstimulated wild-type microglia was however substantially lower compared to the other pro-inflammatory cytokines, suggesting that IFN-γ is only secreted in low amounts when microglia are not exposed to a stimulus. CHMP2B heterozygous FTD3 microglia instead demonstrated a very limited secretion of cytokines, compared to the other two cell lines, suggesting a potential unresponsive state, which could be an indicator of chronic activation. The same tendency could be observed for cytokines which have been described to have both pro-and anti-inflammatory effects (Fig. 6B), as well as key anti-inflammatory cytokines (Fig. 6C). Interestingly, significantly increased levels of IL-15 secretion were observed in the CHMP2B heterozygous FTD3 microglia. This cytokine has been suggested to play a role in the crosstalk between microglia and astrocytes. Up-regulation could thus potentially be a compensatory mechanism in the brain, where microglia are not able to generate an appropriate response, and increased interactions with astrocytes are needed (Shi et al., 2020). Similarly, chemokine secretion was significantly low in the CHMP2B heterozygous FTD3 mutant, whereas no consistent trend towards increased or decreased secretion levels could be seen for the CHMP2B homozygous FTD3 microglia (Fig. 6D). IFN-γ treatment had little effect on the microglial profile, but overall seemed to lower the secretion of pro-and anti-inflammatory cytokines for both the wild-type and CHMP2B homozygous FTD3 microglia, indicating that more stimuli is necessary to induce a major shift in reactivity. However, IFN-γ appeared to have a stimulatory effect on the CHMP2B heterozygous FTD3 microglia, leading to up-regulated secretion (Supplementary, Figure S4). Not surprisingly, LPS treatment led to the most severe changes in inflammatory profile for all cell lines (Supplementary, Figure S5). In general, wild-type microglia showed the strongest response to LPS stimuli, potentially indicating a compromised microglial response in the mutated cell lines. These findings further support the indication that the FTD3 microglial phenotype is dependent on the mutation status, with the homozygous CHMP2B mutation demonstrating higher reactivity accompanied with an inflammatory profile.

CHMP2B homozygous FTD3 microglia exert neurotoxic effects
To assess whether FTD3 microglia exert detrimental effects on healthy neurons, we exposed wild-type cortical glutamatergic neurons to conditioned media from CHMP2B heterozygous and homozygous FTD3 microglia. The neurons were first characterized and verified by ICC, showing positive expression of the neuronal marker Microtubule Associated Protein 2 (MAP2). A small astrocytic population was detected by Glial Fibrillary Acidic Protein (GFAP) labelling, and neuronal subpopulations were assessed with the Vesicular GABA Transporter (VGAT) marker for GABAergic neurons and Vesicular Glutamate Transporter (VGLUT) for glutamatergic neurons, indicating that the majority of neurons within the cultures were in fact glutamatergic neurons (Fig. 7A). Following, control neurons were treated with conditioned media from both unstimulated and activated FTD3 and control microglia to evaluate the neurotoxic effects of the microglia. Neurite retraction is a common event in toxic conditions and an early sign of neurodegeneration. Microglia-mediated inflammation has been shown to induce neurite retractions, and reduced neurite length can thus be an indicator of microglial neurotoxicity (Münch et al., 2003). Neurite outgrowth was therefore assessed through ICC (Fig. 7B) using Neurite Tracer (Pool et al., 2008) in ImageJ 2.0.0-rc-65/1.51 s. The average neurite length was significantly reduced after treatment with conditioned CHMP2B homozygous FTD3 media, suggesting a neurotoxic effect of the increased cytokine secretion (Fig. 7C). In contrary, exposure to CHMP2B heterozygous FTD3 media had no significant effect, which could reflect the low levels of cytokine secretion detected in the conditioned heterozygous microglia media. Interestingly, IFN-γ activation of the CHMP2B heterozygous FTD3 microglia appeared to have a beneficial effect on neurite outgrowth, both compared to the wild-type (Fig. 7D) and conditioned media from non-stimulated CHMP2B heterozygous FTD3 microglia (Fig. 7F). In contrary, conditioned media from IFN-γ activated CHMP2B homozygous FTD3 microglia, induced severe neurotoxic effects, reflected by significant reduction of neurite outgrowth (Fig. 7D, F). No changes could be observed after IFN-γ activation in wild-type microglia, suggesting a well-balanced microglial response, with no detrimental effects on surrounding neurons. Additionally, conditioned media from both CHMP2B heterozygous and homozygous FTD3 microglia, stimulated with LPS, reduced the neurite outgrowth (Fig. 7E). No significant differences in cell death were observed at the analysed time point, and the neural cell count after conditioned microglia media exposure was similar for all conditions (Supplementary, Fig. S6A). Therefore, we did not perform additional assessments of the presence of necrotic or apoptotic cells. As microglial neurotoxicity can be mediated by the production and release of nitric oxide (iNOS) and superoxide (NADPH oxidase), we assessed the expression level of these factors in our unstimulated and stimulated microglia. Expression of iNOS was substantially low or not detected in the samples. Although NADPH oxidase was identified and displayed similar trends as the cytokine assay in unstimulated conditions (Supplementary, Fig. S6B), thus potentially contributing to neurotoxicity, no significant changes were observed in the IFN-γ stimulated microglia (Supplementary, Fig. S6C) or the LPS-stimulated heterozygous microglia (Supplementary, Fig. S6D). Moreover, the effects of the treatments on the NADPH level did not correlate with the changes observed in the neurite outgrowth analysis (Supplementary, Fig. S6E-G), thus indicating that the cytokine secretion is more likely accountable for the microglial impact on the neurons. Importantly, the findings of a beneficial effect of IFN-γ stimulation of CHMP2B heterozygous FTD3 microglia, resulting in increased neurite outgrowth, indicate a potential beneficial effect of mild immune-boosting in specific FTD3 conditions.

IFN-γ treatment can boost unresponsive microglia and promote beneficial effects
The observation of a potential beneficial effect of IFN-γ activated CHMP2B heterozygous FTD3 microglia on neurite outgrowth led us to investigate the effect of IFN-γ treatment within the individual cell lines. We observe very limited effects of the treatment for wild-type microglia, both for pro-and anti-inflammatory cytokines (Fig. 8A, D and G), which is reflected by the un-altered effect on surrounding neurons. Strikingly, IFN-γ leads to a significant response within the CHMP2B heterozygous FTD3 microglia, with increased secretion levels of both pro-and antiinflammatory cytokines (Fig. 8B, E and H). In contrary, the CHMP2B homozygous FTD3 microglia demonstrate overall reduced levels of both pro-and anti-inflammatory cytokines in response to IFN-γ treatment (Fig. 8C, F and I). This reduction is consistent with the observation that the CHMP2B homozygous FTD3 microglia demonstrate a hyperactive state already in unstimulated conditions, and IFN-γ is not sufficient to boost the secretion. The effect of the IFN-γ treatment thus seems to have an opposite trend for the CHMP2B heterozygous FTD3 mutant, compared to both the wild-type and the CHMP2B homozygous FTD3 mutant, with increased secretion levels of pro-inflammatory cytokines (Fig. 8J), cytokines displaying both pro-and anti-inflammatory effects (Fig. 8K) and key anti-inflammatory cytokines (Fig. 8L). The increase in anti-inflammatory cytokine secretion could underlie the beneficial effect on neurite outgrowth after IFN-γ treatment. Importantly, IL-10 secretion is significantly up-regulated after activation of CHMP2B heterozygous FTD3 microglia, a cytokine that has been observed to have direct protective effects on cortical neurons and promote neurite outgrowth (Chen et al., 2016). These findings imply that IFN-γ supplementation could potentially restore beneficial immunological functions in specific pathological conditions.

Discussion
FTD3 is a rare sub-form of FTD, caused by a mutation in CHMP2B, first identified in a large Danish pedigree (Brown, 1998). We have previously described neuron-and astrocyte-specific phenotypes linked to this mutation (Chandrasekaran et al., 2021;Zhang et al., 2017). Astrocytes and microglia are the resident glial cells of the central nervous system, and together, they play a crucial role in maintaining a healthy brain homeostasis. Our findings of astrocytic reactivity in FTD3 as well as evidence of neuroinflammation in various FTD3 model systems indicate a role of microglia in disease pathology. In this study, we therefore investigated implication of microglia in disease pathogenesis as potential pro-inflammatory drivers, using a hiPSC-derived microglia cell model of FTD3, established by CRISPR/Cas9 gene editing. We demonstrate successful differentiation of microglia from all hiPSC lines, verified by ICC, highlighting the potential of our in-vitro model system. FTD3 patients carry heterozygous mutations in CHMP2B. Nevertheless, to gain insights into the biological function of CHMP2B in microglia responses and activity we included both the heterozygous and homozygous CHMP2B mutation in this study. This nuanced investigation of both the mono-and bi-allelic mutation state is important to achieve a systemic assessment of the CHMP2B function in microglia.
Remarkably, we identify an alternate phenotype of CHMP2B heterozygous FTD3 microglia, compared to CHMP2B homozygous FTD3 microglia, suggesting that the response and profile of microglia is highly dependent on the underlying genetic composition. Although these findings are surprising, similar observations have been described for other FTD-linked mutations, such as GRN. Studies of heterozygous (GRN +/-) and homozygous (GRN -/-) knockout mice indicate a dosagedependent effect that ultimately causes distinct pathologies, with heterozygous mutations causing FTD pathology, oppose to a pathology resembling lysosomal storage pathology in homozygous mice. Moreover, no gliosis was observed in these heterozygous mice, but increased reactivity and inflammatory effects were evident in homozygous GRN conditions (Smith et al., 2012). Recent studies have further confirmed that such genetic variants can cause different phenotypic outcome dependent on their mono-or bi-allelic state (Huin et al., 2020), supporting our findings of an alternate heterozygous and homozygous FTD3 microglia phenotype.
Interestingly, phagocytic activity is dramatically increased in CHMP2B homozygous FTD3 microglia, indicating hyperactivation, confirmed by increased secretion of inflammatory mediators. Moreover, these microglia display successful protein degradation at unstimulated states, in contrary to CHMP2B heterozygous FTD3 microglia, where we observe accumulation of cargo-filled vesicles, similar to our findings in FTD3 astrocytes. This becomes even more profound following phagocytosis assay with pHrodo™ labelled E. coli bioparticles. However, following phagocytosis assay, even CHMP2B homozygous microglia display insufficient clearing of engulfed bacteria, demonstrated by accumulation of cargo-filled lysosomal structures. The ESCRT-III complex is a key component of cellular degradation processes, such as autophagy, and increasing evidence imply that it also plays a role in phagocytosis (Vietri et al., 2020). Consequently, mutations in CHMP2B affecting the efficiency and function of the ESCRT-III machinery account for the observed accumulation of intracellular material. The unexpected differences in phenotypes between heterozygous and homozygous CHMP2B conditions could indicate that multiple mechanisms are involved in phagocytic degradation, and that the hyperactivity seen for the CHMP2B homozygous microglia is a compensatory mechanism to improve recycling of cargo. This hyperactivation could potentially explain why the CHMP2B homozygous FTD3 microglia are able to successfully degrade particles in unstimulated conditions, but after exposure of additional particles, these properties are diminished, evident as accumulation of structures containing intracellular material. The hypothesis that CHMP2B and ESCRT-III impairments affect phagocytosis and protein degradation through multiple mechanisms, is further supported by the variety of the accumulated structures in heterozygous versus homozygous CHMP2B conditions. In the CHMP2B heterozygous FTD3 microglia, we observed vesicles resembling early stages of the endo-lysosomal pathway, whereas dark, electron-dense lysosome-like structures are evident for the CHMP2B homozygous FTD3 microglia. Interestingly, these findings correlate with studies on other FTD-linked mutations showing that homozygous mutations in Progranulin (GRN) lead to abnormal lysosome storage disorders characterized by build-up of toxic materials within the lysosomes, whereas heterozygous mutants display FTD pathology affecting earlier endo-lysosomal structures (Smith et al., 2012). This is further in line with studies showing a phenotype resembling lysosomal storage pathology in a homozygous CHMP2B mouse model (Clayton et al., 2015;Götzl et al., 2014;Smith et al., 2012). In summary, we demonstrate a hyperactive homozygous FTD3 microglia phagocytic profile with sufficient protein degradation in unstimulated conditions, compared to a tolerant heterozygous microglia profile with impaired protein degradation displayed by vesicle accumulation. These findings highlight the alternate phenotype caused by a mono-or bi-allelic CHMP2B mutation state.
The functional role of CHMP2B in the endo-lysosomal fusion processes is still not fully explored. CHMP2B recruits and interacts with VPS4 via its C-terminal MIT-interacting motif (MIM), which is essential for the formation of multivesicular bodies (MVBs) destined to fuse with lysosomes. If this interaction is inhibited, abnormal and enlarged MVBs are formed, which are not capable of fusion with lysosomes (Ugbode and West, 2021). This function is clearly impaired in the CHMP2B heterozygous FTD3 mutation but seems to be less impaired in the CHMP2B homozygous FTD3 mutation. This could potentially be due to a dominant negative effect exerted by the c-terminally truncated CHMP2B through interference with the wild-type protein. The wild-type protein can still interact in this complex with the VPS4 protein, sequester it and inhibit the interaction with functional homologues of CHMP2B. In contrary if both alleles are mutated the interaction with VPS4 is completely abolished, which could then allow functional paralogous proteins to interact with VPS4. Based on STRING (Szklarczyk et al., 2019) database analyses of the genes detected in the transcriptome of FTD3 microglia, functional paralogues of CHMP2B interacting with VPS4 could be CHMP1A, CHMP1B, CHMP5, CHMP2A, CHMP4A and CHMP4B (Fig. 9). Such functional redundancies are common and can explain why loss of function or in this case interaction might present with less severe phenotypes than heterozygous mutations (Nowak et al., 1997). Interestingly, the differential gene expression analysis of the RNA sequencing data revealed a significantly increased expression level of CHMP4A in unstimulated CHMP2B homozygous FTD3 microglia compared to unstimulated wild-type microglia (Fig. 7B). Therefore, we propose that CHMP4A could compensate for the loss of function in CHMP2B homozygous mutants. This finding was specific for the homozygous microglia and might reflect a compensatory mechanism which could explain the hyperactive CHMP2B homozygous FTD3 microglial profile. We thus propose that the alternate microglia phenotypes is a consequence of a dominant negative effect of the heterozygous CHMP2B mutation, which is counteracted by functional redundancy in a homozygous state.
The opposing effects of a heterozygous and homozygous CHMP2B mutation are further demonstrated by the inflammatory cytokine profiling assay and metabolic mapping. CHMP2B heterozygous FTD3 microglia display an "unresponsive" state, similar to tolerant microglia, that have been described in conditions of chronic activation. This could potentially be linked to metabolic impairments, which have been suggested to contribute to a reactive switch, from a protective to a detrimental microglial profile (Aldana, 2019). CHMP2B heterozygous FTD3 microglia demonstrate metabolic reprogramming, from OXPHOS to glycolysis, which is not evident in the CHMP2B homozygous FTD3 line. Such a shift in metabolism has been described for reactive microglia, but prolonged metabolic defects following microglial activation have been demonstrated to ultimately result in a chronic tolerant phase, with diminished immune responses (Baik et al., 2019). This is thus in line with the low activity and cytokine secretion reflected in our CHMP2B heterozygous FTD3 microglia. In contrary, CHMP2B homozygous FTD3 microglia display a shift in the TCA cycle, with altered glutamine metabolism. Microglia have been shown to be dynamically flexible and can rapidly adapt their uptake of metabolic substrates to meet energy needs. Glutamine can act as an alternative fuel if glucose is absent, and allows for microglia to retain their crucial function when brain homeostasis is compromised (Bernier et al., 2020). The increased glutamine uptake might thus be a compensatory mechanism, that explains how CHMP2B homozygous FTD3 microglia are able to maintain a highly functional phagocytic activity and increased secretion of cytokines, which would demand high energy production. These findings correlate with our mitochondria observations, where the increased energy demands lead to a hyperactive microglial state, resulting in increased mitochondria fission and accumulation. Although the TEM analysis indicated such an imbalance in mitochondrial dynamics, the observations should be investigated in future studies, by evaluating the expression of fission/fusion mitochondrial markers.
It has been proposed that glutamine metabolism is used to promote succinate synthesis in activated macrophages (Viola et al., 2019). This was supported by observations in macrophages, where succinate increased cytokine secretion (Tannahill et al., 2013). Furthermore, in macrophages the enzyme succinate dehydrogenase may be implicated in driving the activation response by reprogramming mitochondria to increase radical oxygen species (ROS) and decrease OXPHOS (Mills and O'Neill, 2016;Tannahill et al., 2013). Together, the combination of increased glycolysis and succinate dehydrogenase upregulation may sustain the activated phenotype in macrophages. Taking these findings of peripheral macrophages into consideration, it can be speculated that the increase in labeled succinate derived from incubation with [U-13 C] glutamine, may support a metabolic reprogramming in microglia where glutamine may be used to produce succinate to sustain the acute and chronic activation states in both CHMP2B homozygous and heterozygous FTD3 microglia, respectively. In summary, our metabolic assessment and cytokine assay demonstrate a tolerant heterozygous FTD3 microglia state, with low secretion of cytokines, possibly due to prolonged metabolic defects. In contrast, homozygous FTD3 microglia present a hyperactive state with high cytokine secretion, potentially mediated by increased glutamine metabolism.
Moreover, conditioned CHMP2B homozygous FTD3 microglia media demonstrated neurotoxic effects on healthy neurons, whereas no change could be observed for the heterozygous condition. This is most likely a reflection of the low cytokine secretion detected in the microglia media. These findings indicate that microglia exert detrimental effects on neurons through multiple mechanisms. The lack of neurotoxicity from the heterozygous FTD3 microglia could indicate that the detrimental effects of microglia in the heterozygous FTD3 condition are a result of a tolerant microglia state, with depleted function, thus also a loss of the protective microglial properties. This would not be reflected solely by conditioned media exposure and should thus be further investigated in more advanced co-culture model systems. Interestingly, FTD3 microglia respond differently to activation stimuli, proposing a potential beneficial effect of IFN-γ in CHMP2B heterozygous FTD3 microglia. This was particularly evident for anti-inflammatory cytokines, such as IL-10 presenting increased secretion levels, which have been observed to have direct protective effects on cortical neurons, promoting neural outgrowth (Chen et al., 2016). Other studies have shown a similar trend in defective, tolerant microglia, where chronic exposure to amyloid beta caused a metabolic reprogramming, that ultimately could be rescued with IFN-γ boosting, indicating a beneficial effect (Baik et al., 2019). These findings highlight the potential of using mild immune-triggering stimuli as a treatment strategy to ameliorate FTD3 pathology.
Our findings clearly indicate that IFN-γ-stimulation can trigger a change in microglial phenotype and their effects on neurons. However, the action of IFN-γ in microglia activation and reactivity is complex and opposing effects of IFN-γ have been described in the CNS, presenting either neurotoxic or neuroprotective effects. Evidence suggests that the action of IFN-γ is dose-dependent, and that the cytokine concentration can mediate the balance of pro-and anti-inflammatory microglial profiles. Low dosage has been shown to enable the microglia to exert neuroprotective functions, whereas high IFN-γ exposure initiates neurotoxic microglial actions (Ottum et al., 2015). Moreover, the microglial ability to produce and secrete IFN-γ is highly controversial, particularly in unstimulated condition, and the major source of IFN-γ is described to be lymphoid cells (Monteiro et al., 2017). However, microglia stimulation with IL-12 and IL-18 have been shown to induce IFN-γ production and release (Kawanokuchi et al., 2006). Our data indicate that microglia secrete IFN-γ also in unstimulated conditions (Fig. 6A). The secretion of IFN-γ was only measured in unstimulated and LPS-treated conditions. Furthermore, the microglia culture media was not supplemented with IFN-γ according to the manufacturer's media composition (Advanced DMEM/F12, Thermo Fisher Scientific, 12634010). This indicates that the detected IFN-γ was a result of microglial secretion and is not a consequence of media contamination. The concentration of IFN-γ secreted by the unstimulated wild-type microglia was however substantially lower compared to the other pro-inflammatory cytokines, suggesting that IFN-γ is only secreted in low amounts when microglia are not exposed to a stimulus. Furthermore, the unstimulated wild-type and CHMP2B homozygous FTD3 microglia secreted higher amounts of IL-12 compared to the heterozygous microglia, which could potentially explain the IFN-γ production. IL-12p70 and IL-12/IL-23p40 secretion was substantially increased in the unstimulated CHMP2B homozygous FTD3 microglia (Supplementary, Table S1D), thus correlating with the IFN-γ levels. Our findings indicate that microglia can exert neurotoxic effects either by gain or loss of functions. In the heterozygous FTD3 condition tolerant microglia could lack important neuroprotective properties, whereas hyperactivity and excessive production of inflammatory mediators is demonstrated in the homozygous FTD3 condition. Both could have detrimental effects on surrounding neurons. Moreover, we propose that both the dosage of immunomodulating treatments as well as disease stage and inflammation profile should thus be considered when establishing an appropriate therapeutic strategy.
In this study we are using hiPSC-derived microglia as a model of FTD3. This is thus a simplified model system, focusing specifically on microglia-specific phenotypes in the disease, and more advanced model systems such as co-cultures or brain organoids should be included in future studies to verify our findings. Moreover, several studies have indicated differences between primary-and in vitro microglia-like cells (Butovsky et al., 2014;Melief et al., 2016). Artificial culture conditions as well as the absence of functional neurite networks and astrocyte syncytia could potentially affect the microglia phenotypes in vitro (Biber et al., 2014;Kann et al., 2022). To investigate our findings in an in vivo setting, we therefore performed qPCR on brain tissue from a transgenic CHMP2B intron5 FTD3 mouse model (Ghazi-Noori et al., 2012). We were able to validate our findings of increased expression of autophagyrelated genes such as Optn, Sqstm1 and Lamp1 in the FTD3 mice compared to the wild-type mice (Fig. 4G). Although the CHMP2B intron5 FTD3 mice are maintained as a homozygous line, the expression profile of autophagy-related genes matched our unstimulated heterozygous FTD3 microglia and differed from the phenotype of our homozygous FTD3 microglia in unstimulated conditions. However, when stimulated with bacteria the CHMP2B homozygous FTD3 microglia revealed autophagy-impairments and aggregation similar to the mouse model (Clayton et al., 2015). These findings could however correlate with the reactive gliosis demonstrated in the CHMP2B intron5 FTD3 mice and a similar hyperactive, pro-inflammatory profile of homozygous FTD3 microglia has been observed in the CHMP2B intron5 FTD3 mouse model (Clayton et al., 2017). Some studies have however shown that although both astrogliosis and microgliosis become apparent with age in the CHMP2B intron5 FTD3 mouse model, astrocytic activation is more prominent (Vernay et al., 2016). Interestingly, an inflammatory profile similar to the CHMP2B intron5 FTD3 mouse model was observed in postmortem FTD3 patient brains, which present with the heterozygous CHMP2B mutation. The profound inflammation in these models was however observed at later disease time points, highlighting the importance of considering not only mutation-dosage but also disease stage when establishing therapeutic strategies. Moreover, although the in vivo model systems allow for investigation of cellular populations and interactions in a complex environment, the brain tissues and thus obtained data includes diverse cell types. We have previously shown that FTD3 neurons display different mitochondria phenotypes compared to our FTD3 astrocytes and microglia, whereas alteration in endo-lysosomal/ autophagy pathways are evident in all cell populations (Chandrasekaran et al., 2021;Zhang et al., 2017). Additionally, it has been well documented that both immune response and microglia phenotypes and activity are divergent between mice and men (Mestas and Hughes, 2004). The physiology of commonly used animal models, such as rodents, deviates from human physiology. This is especially apparent in terms of immune regulatory pathways. An additional point of concern is the artificial mutation-doses required to induce phenotypes observed in humans. In the case of the CHMP2B intron5 FTD3 mouse model a homozygous mutation is mandatory to observe the phenotypes present in FTD3 heterozygous patients. Whilst human postmortem brains are instrumental to understand disease pathology, they can only reflect the late/end-stage disease pathology. Therefore, the human FTD3 in vitro hiPSC-derived models provide an advantage to investigate microgliaspecific FTD3 phenotypes, which was the scope of the present study. The FTD3 in vitro hiPSC-derived model presented in this study demonstrate transcriptomic similarity to primary human microglia (Haenseler et al., 2017). Consequently, this model can provide additional insights into disease pathology and close potential knowledge gaps between animal models and postmortem analyses. Additionally, the FTD3 in vitro hiPSC-derived model is a powerful tool to gain valuable insights into early human FTD3 pathology and provides an important basis for future investigations and treatment strategies.

Conclusion and future perspective
In summary, our findings demonstrate an alternate microglial phenotype between a heterozygous and homozygous FTD3-linked CHMP2B mutation. Moreover, we suggest that metabolic defects play a key role in microglial reactivity, and that certain FTD3 conditions might benefit from mild immune boosting, thus providing an interesting potential therapeutic strategy. knock-in lines. A.C. and S.C. contributed to qPCR. All the authors read and approved the final manuscript.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.

Ethical statement
Approval of the study have been acquired from the Ethics Committee of the Capital Region of Denmark (H-4-2011-157), and written informed consent was obtained from the individual who donated samples for hiPSC generation.

Data availability
The datasets supporting the conclusions of this article are included within the article and the additional supplementary data files. The dataset generated for the RNA sequencing analysis has been deposited in the NCBI GEO database with the accession number GEO: GSE215241.