C9orf72 hexanucleotide repeat expansion leads to altered neuronal and dendritic spine morphology and synaptic dysfunction

Frontotemporal lobar degeneration (FTLD) comprises a heterogenous group of progressive neurodegenerative syndromes. To date, no validated biomarkers or effective disease-modifying therapies exist for the different clinical or genetic subtypes of FTLD. The most common genetic cause underlying FTLD and amyotrophic lateral sclerosis (ALS) is a hexanucleotide repeat expansion in the C9orf72 gene (C9-HRE). FTLD is accompanied by changes in several neurotransmitter systems, including the glutamatergic, GABAergic, dopaminergic, and serotonergic systems and many clinical symptoms can be explained by disturbances in these systems. Here, we aimed to elucidate the effects of the C9-HRE on synaptic function, molecular composition of synapses, and dendritic spine morphology. We overexpressed the pathological C9-HRE in cultured E18 mouse primary hippocampal neurons and characterized the pathological, morphological, and functional changes by biochemical methods, confocal microscopy, and live cell calcium imaging. The C9-HRE-expressing neurons were confirmed to display the pathological RNA foci and DPR proteins. C9-HRE expression led to significant changes in dendritic spine morphologies, as indicated by decreased number of mushroom-type spines and increased number of stubby and thin spines, as well as diminished neuronal branching. These morphological changes were accompanied by concomitantly enhanced susceptibility of the neurons to glutamate-induced excitotoxicity as well as augmented and prolonged responses to excitatory stimuli by glutamate and depolarizing potassium chloride as compared to control neurons. Mechanistically, the hyperexcitation phenotype in the C9-HRE-expressing neurons was found to be underlain by increased activity of extrasynaptic GluN2B-containing N-methyl-d-aspartate (NMDA) receptors. Our results are in accordance with the idea suggesting that C9-HRE is associated with enhanced excitotoxicity and synaptic dysfunction. Thus, therapeutic interventions targeted to alleviate synaptic disturbances might offer efficient avenues for the treatment of patients with C9-HRE-associated FTLD.

and synaptic dysfunction. Thus, therapeutic interventions targeted to alleviate synaptic disturbances might offer efficient avenues for the treatment of patients with C9-HRE-associated FTLD.

Background
Neurodegenerative diseases, including frontotemporal lobar degeneration (FTLD), are common neurological diseases, leading to dementia. FTLD, a prevalent cause of early-onset dementia in individuals <65 years of age, is a heterogeneous group of clinical syndromes characterized by progressive atrophy in the frontal and temporal lobes. It is associated with behavioral and executive functional changes (behavioral variant frontotemporal dementia, bvFTD) and/or impairment of speech and language skills (primary progressive aphasias, PPAs) and/or extrapyramidal features (FTLD-plus) (Bang et al., 2015;Onyike and Diehl-Schmid, 2013). Moreover, a great amount of FTLD patients show neuropsychiatric symptoms (Panegyres et al., 2007) or motor dysfunction in keeping with the criteria of amyotrophic lateral sclerosis (ALS) (Strong et al., 2009). The major genetic cause underlying both sporadic and familial FTLD clinical syndromes as well as ALS is the GGGGCC hexanucleotide repeat expansion in the C9orf72 gene (C9-HRE) (DeJesus-Hernandez et al., 2011;Majounie et al., 2012;Renton et al., 2011). Increasing evidence suggests alterations in different neurotransmitter systems, such as glutamatergic, dopaminergic, GABAergic, and serotonergic systems in the brains of FTLD or ALS patients, which could underlie the clinical, neuropsychiatric and cognitive symptoms observed in these patients (Murley and Rowe, 2018). Moreover, induced pluripotent stem cell (iPSC)-derived motor neurons from C9-HRE-carrying patients have been reported to display enhanced susceptibility to glutamateinduced excitotoxicity (Donnelly et al., 2013), and a recent report shows significant alterations in the network activity of iPSC-derived cortical neurons from C9-HRE carriers (Perkins et al., 2021). Interestingly, in Alzheimer's disease (AD), synapses are considered the earliest site of neuronal dysfunction and altered synaptic activity is the best pathologic correlate of cognitive impairment in AD patients, suggesting that synaptic dysfunction may represent one of the earliest pathological changes in neurodegenerative brain even more widely (Coleman and Yao, 2003).
There are two major types of synapses in the central nervous system (CNS). Type I synapses are excitatory and use glutamate as their primary neurotransmitter. In contrast, type II synapses are inhibitory with γ-amino butyric acid (GABA) as their major neurotransmitter (Knafo et al., 2012). The inhibitory synapses are mainly located in dendritic shafts, whereas the excitatory synapses are primarily located on specific structures in the neurons, termed dendritic spines (Sala and Segal, 2014). These are small, actin-rich morphological specializations that protrude on the surface of dendrites of many neuron types and are the location of over 90% of all excitatory synapses (Harris and Kater, 1994). The morphology and number of dendritic spines can change within a relatively short timeframe (from milliseconds to days). Spines exhibit remarkable morphological plasticity during development as well as in the adult brain, providing a fundamental basis for neuronal plasticity related to, for instance, learning and memory (Fortin et al., 2012). Dendritic spines differ in their shape and biochemical composition, including synaptic receptor subtypes. The morphology of the spines is commonly classified in three main types, namely, mushroom, stubby, and thin spines (Gipson and Olive, 2017;Rochefort and Konnerth, 2012). A fourth group, termed filopodia, is a group of spines that is often observed in large numbers during development (Yuste and Bonhoeffer, 2004). Mushroom spines have a large head with a thin neck whereas stubby spines are presented with a large head, but no discernible neck. Thin spines have a slender, filopodia-like shape without an obvious head. Dendritic spine dynamics and synaptic function are more accurately described by spine head diameter and volume as well as neck width and length, rather than discrete morphology-based classifications (Tønnesen et al., 2014). Specifically, the spine head volume correlates to the size of the postsynaptic density (PSD) and the number of postsynaptic receptors and the pool of neurotransmitters ready to be released, whereas the length of the spine neck contributes to the biochemical and electrical properties of the spine (Arellano et al., 2007;Yuste and Majewska, 2001). Learning and strengthening of the synapse causes an enlargement of the spine head and shortening of the neck (Yuste and Bonhoeffer, 2001). Shortening and widening of the spine neck decreases the electrical resistance, which results in a larger excitatory postsynaptic potential, while increasing the head size also helps to integrate and accommodate a higher number of receptors (Yuste, 2013).
Interestingly, reduction in synapse density in FTLD brains has already been reported in 1995 (Brun et al., 1995). Even though changes in synaptic function have been of increasing interest in FTLD research, the available data remain so far limited. Current research suggests that synapse loss may be reversible due to the highly dynamic plasticity of synapses (Freeman and Mallucci, 2016). This raises a great possibility for therapeutic interventions, which could support or improve synaptic maintenance and function at early stages of disease and therefore enhance brain function and cognitive abilities of the patients. Thus, the involvement and mechanisms of synaptic alterations in FTLD pathogenesis should be investigated in detail. In the present study, we have aimed to provide novel insights into potential alterations of synaptic function and dendritic spine morphology in neurons expressing the FTLD-associated C9-HRE.

Adeno-associated virus constructs
To express the pathological C9-HRE, a 66 x GGGGCC repeat construct (66R) in pAM/CBA-pl-WPRE-BGH containing inverted repeats of serotype 2 adeno-associated virus (AAV) was used in the transfection. As a control, a construct containing 2 x GGGGCC (2R) was used. These constructs were kindly provided by Dr. Leonard Petrucelli (Mayo Clinic, Jacksonville, FL, USA) and they have been described previously (Chew et al., 2015). The constructs were packaged in AAV9 virus particles at the BioCenter Finland National Virus Vector Laboratory, University of Eastern Finland (UEF), Kuopio, Finland and used for transduction of mouse primary hippocampal neurons.

Primary mouse hippocampal neuron culture, virus vector transduction, and transient transfection
Primary mouse hippocampal neuron cultures were established as previously described (Kurkinen et al., 2016) under the ethical permission by the Laboratory Animal Center of the University of Eastern Finland, Kuopio, Finland. Briefly, the hippocampi were harvested from 18-day-old (E18) JAXC57BL/6J mouse embryos. Single cell suspension was prepared from dissected hippocampi by dissociation in papaindissociation buffer [10 mg DL-Cysteine-HCl (Sigma), 10 mg bovine serum albumin (BSA, Sigma), 250 mg glucose in Dulbecco's Phosphate-Buffered Saline (DPBS, Lonza) containing 10 mg/ml papain (Sigma) and 10 mg/ml DNAse I (Sigma)] at +37 • C for 5 min. The tissue was pelleted and triturated in trituration medium containing Ca 2+ -and Mg 2+ -free Hank's buffered salt solution (HBSS, Lonza), 100 mM sodium pyruvate, 1 M HEPES (pH 7.2), and 10 mg/ml DNase I. The single cell suspension was centrifuged at 800×g for 5 min and the hippocampal cells were resuspended in hippocampal neuron feeding medium containing Neurobasal medium (Gibco) supplemented with 1 x B27 (Gibco), 0.5 mM Lglutamine (Lonza), 100 U/ml penicillin, and 100 μg/ml streptomycin (Lonza). Cells were plated at an appropriate density on culture plates (with or without glass coverslips) coated with 10 μg/ml Poly-D-Lysine (PDL, Sigma) and 30 μg/ml laminin (Sigma). The cells were cultured in hippocampal neuron feeding medium at +37 • C in 5% CO 2 . Half of the culture medium was changed every fourth day.
The cells were cultured for five days in vitro (DIV) before AAV transduction with multiplicity of infection (MOI) of 5000 of either control AAV-2R or pathological AAV-66R vectors. The culture medium was exchanged with fresh culture medium every fourth day. The cells were matured until DIV 21 before sample collection or fixation. Transduction efficacy, based on the number of AAV-66R transduced cells harboring RNA foci in the cultures, was estimated to be approximately 70% ( Supplementary Fig. 1).
For immunostaining and neurite branching analysis, the cells were co-transfected on DIV 19 with 2R or 66R plasmids and pLVX-IRES-ZsGreen1 vector (Clontech Laboratories) to detect transfected cells based on Zoanthus sp. green fluorescent protein 1 fluorescence (ZsGreen1). For dendritic spine analysis, the cells were co-transfected with enhanced green fluorescent protein (EGFP) and 2R or 66R plasmids. To transfect the cells, the culture medium was removed and the cells were pre-treated with 10 mM MgCl 2 (Thermo Scientific) in hippocampal neuron feeding medium for 1 h at +37 • C. A total of 0.8 μg plasmid DNA and 2 μl Lipofectamine 2000 reagent (Thermo Scientific) was used per transfection according to manufacturer's instructions. The conditioned culture medium was changed back to the cells 3 h post transfection. At 48 h after transfection on DIV 21, the cells were washed two times with warm DPBS and fixed with 4% (v/v) formaldehyde (Thermo Scientific) in DPBS for 20 min at RT, washed five times for 2 min in DPBS, and stored at +4 • C. The transfection efficiency was low (a few cells/per coverslip with 400,000 cells plated), but enabled examination of individual transfected cells in the immunostaining, FISH, dendritic branching, and dendritic spine analyses.

Immunocytochemistry and fluorescence in situ hybridization
For fluorescence in situ hybridization (FISH) diethyl pyrocarbonatecontaining Dulbecco's Phosphate-Buffered Saline (DEPC-PBS) and DEPC-treated H 2 O were used. FISH was performed in an RNase-free environment as described previously (Chew et al., 2015). Fluorescently labelled locked nucleic acids (LNA) TYE™ 563-(CCCCGG) 3 probe (C 4 G 2 ) was used to detect the sense foci and TYE™ 563-(CAG) 6 (CAG) was used as a negative control probe (Exiqon). Briefly, the cells on coverslips were permeabilized in 0.2% Triton X-100 for 15 min at RT and washed three times in DEPC-PBS. Next, pre-hybridization was performed in hybridization buffer, containing 10% dextran sulfate, formamide, 20 x SSC, and 1 M sodium phosphate buffer, pH 7, in DEPC-H 2 O, at +66 • C for 30 min. The probes were denatured at +85 • C for 1 min 15 s and added to hybridization buffer at a concentration of 40 nM C 4 G 2 or CAG probe and incubated with the coverslips at +66 • C for at least 16 h. Then, after washing in washing buffer 1 (0.1% Tween in 2 X SSC) for 5 min at RT and washing buffer 2 (0.2 X SSC) for 10 min at +60 • C, the coverslips were mounted with Vectashield Vibrance Antifade mounting medium containing 4, 6-diamidino-2-phenylindole (DAPI, Vector Laboratories) and imaged with Zeiss LSM800 Airyscan confocal microscope.

Neuronal viability and lactate dehydrogenase assays
For Map2-based viability assay, the hippocampal neurons on 48-well plates, fixed in 4% PFA, were permeabilized and the endogenous peroxidase activity was blocked by 0.3% hydrogen peroxide (H 2 O 2 ) in methanol for 10 min at RT. After permeabilization, the cells were incubated in blocking solution, containing 1% BSA and 10% horse serum (Vector Labs) for 20 min at RT, followed by primary antibody staining with Map2 antibody (1:2000, Sigma, M9942) overnight at +4 • C. Then, the cells were washed 3 × 10 min with DPBS followed by secondary antibody staining with biotinylated horse anti-mouse (1:500, Vector Labs) for 1 h at RT. After washing for 3 × 10 min with DPBS, tertiary antibody staining with ExtrAvidin-HRP (1:500, Sigma) for 1 h at RT and washing for 3 × 10 min in DPBS were performed. Then, TMB substrate kit for peroxidase (Vector Labs) was used for colour development according to user guide. The absorbance was measured at 650 nm and the average absorbance from negative control wells (without Map2 antibody) was subtracted from the absorbance of the sample wells.
To measure cytotoxicity at different times during the culture of the hippocampal neurons transduced on DIV 5 with AAV-2R or AAV-66R, lactate dehydrogenase (LDH) assay was performed from medium samples at DIV 7, 14, and 21 according to kit instructions (Cytotoxicity Detection Kit (LDH, Roche, 11644793001).

Glutamate excitotoxicity
To determine glutamate-induced excitotoxicity, AAV-2R or AAV-66R transduced hippocampal neurons were cultured until DIV 20 and treated with L-glutamate (Sigma). To this end, the culture medium from the wells was removed and half amount of fresh culture medium was added. The removed culture medium was sterile-filtered and used for glutamate stock preparation. Final concentration of L-glutamate in the well was 200 μM. After a 24-h treatment, the cells were fixed in 4% PFA followed by Map2 staining and TMB substrate detection to determine neural viability as described above.

Synaptic protein extraction and Western blotting
Synaptic protein extraction from AAV-2R or AAV-66R-transduced hippocampal neurons at DIV 21 was performed according to supplier's instructions using Syn-PER reagent (Thermo Scientific). The cell culture medium was removed from the wells and the cells were washed twice with ice-cold DPBS. Then, 400 μl of Syn-PER Reagent supplemented with protease and phosphatase inhibitors (Thermo Scientific) was added, the cells were scraped, and the cell lysate was centrifuged at 1200×g for 10 min at +4 • C. After centrifugation, the supernatant was centrifuged at 15,000×g for 20 min at +4 • C. Then the supernatant was removed and the synaptosome pellet was resuspended in 40 μl of Syn-PER reagent and kept on ice until protein concentration measurement. The cell pellet obtained in the first centrifugation step was resuspended in 1% sodium dodecyl sulphate (SDS, Sigma) extraction buffer containing 10 mM Tris-HCL (Sigma), 2 mM EDTA (Sigma), and supplemented with protease and phosphatase inhibitors, and heated at +85 • C for 7 min before protein concentration measurement.

Dendritic branching analysis
Hippocampal neurons were co-transfected with 2R or 66R and ZsGreen1 plasmids at DIV 19 and fixed 48 h after transfection on DIV21. Then, the ZsGreen1-positive neurons were imaged with LSM 800 Airyscan confocal microscope. Neuronal branching was analyzed from Zstack images of 2R-and 66R-transfeced neurons using Vaa3D software and its Neuron-Tracker-App2 (Version 2.921, Vaa3D Plugin Interface 2.12) (Peng et al., 2014;Peng et al., 2010). The threshold was set to auto-thresholding and image analysis was performed with the same settings in all experimental groups. The image analysis was performed blinded and the code was only opened after quantification. The length and number of branches were quantified.

Dendritic spine analysis
For dendritic spine analysis, hippocampal neurons were plated on PDL-coated glass 4-well chamber slides (LabTek), co-transfected with 2R or 66R and EGFP plasmids at DIV 19, and fixed 48 h after transfection. Dendritic spines from EGFP-positive neurons were imaged with LSM 800 Airyscan confocal microscope (Zeiss). Serial Z-stacks of optical sections from dendritic segments were analyzed by NeuronStudio software (Computational Neurobiology and Imaging Center Mount Sinai School of Medicine, New York, Version 0.9.92 64-bit). Settings were adjusted as following: Volume: pixel dimensions were set to X = 0.066 μm (pixel width), Y = 0.066 μm (pixel height), and Z = 0.390 μm (voxel depth); Dendrite detection: attach ratio was set to 1.5, minimum length to 3 μm, discretization ratio to 1, and realign junctions to yes; Spine detection: maximum height of spines was set to 5.0 μm and minimum height to 0.2 μm, maximum width to 3 μm, maximum stubby size to 10 voxels, and minimum non-stubby size to 5 voxels; Spine classifier: neck ratio (headneck ratio) was set 1.1, thin ratio to 2.5, and mushroom size to 0.35 μm.
To obtain a clear image, the image filter Blur-MP was run before the analysis The spines were sub-grouped according to their morphology to mushroom, stubby, or thin spines according to Neuron studio's automatic settings using the Rayburst algorithm. The image analysis was performed blinded and the code was only opened after quantification.

Calcium imaging
Primary hippocampal neurons transduced with 2R and 66R at DIV 21 were loaded with the calcium indicator Fluo-4 AM for 30 min at 37 • C (1×, Fluo-4 Direct Calcium Assay Kit, Invitrogen, USA). Then, the cells were washed for 10 min at 37 • C and 10 min at RT in basic salt solution (BSS) containing in mM: 152 NaCl, 10 HEPES, 10 glucose, 2.5 KCl, 2 CaCl 2 , 1 MgCl 2 and pH adjusted to 7.4, and transferred to TILL Photonics imaging system (TILL Photonics GmbH, Germany). In the recording chamber, the cells were constantly perfused with BSS (rate 2 ml/min). Setup was equipped with Rapid Solution Changer RSC-200 (BioLogic Science Instruments, Grenoble, France) that allows fast drug applications with exchange time ~ 30 ms. The cells were sequentially treated with short (2 s) applications 100 μM glutamate, 100 μM GABA, 50 mM KCl (a depolarizing agent) and 5 μM ionomycin (a membrane-permeable calcium ionophore). KCl was applied as a marker for neurons (Simonetti et al., 2006) and ionomycin responses were used for signal normalization.
The neurons were also treated with three different NMDAR inhibitors: 10 μM MK-801 (M107, Sigma), 5 μM Ifenprodil (I2892, Sigma), or 10 μM Sigma). For this, after the first application of glutamate (2 s), NMDAR inhibitors were applied for 180 s followed by second application of glutamate together with the inhibitor (2 s). As control, glutamate was applied twice following the same sequence. Neurons growing in clusters were detected with 10× objective on Olympus IX-70 microscope (Tokyo, Japan), illuminated with the excitation light (480 nm) and the emission light passed through the FITC filter set (Chroma Technology Inc., USA). The fluorescence images were recorded with CCD camera (SenciCam, PCO imaging, Kelhaim, Germany).

Statistical analyses
The data are shown as mean ± standard deviation (SD) or median ± interquartile range. Statistical significance between the groups was tested using independent sample t-test (normally distributed) or Mann-Whitney U test (not normally distributed) depending on whether the data fulfilled the assumptions for parametric tests. Shapiro-Wilk test was used to determine if the data points were normally distributed. One-way analysis of variance (ANOVA) followed by Sidak's multiple comparison (normally distributed) or Kruskal-Wallis test (not normally distributed) was used for data with more than two groups. All statistical analyses were performed using GraphPad Prism 8.3.1 software and a threshold for statistical significance was set at p ≤ 0.05. Graphs were created using the GraphPad Prism software.

Results
In this study, we have examined the effects of the pathological C9-HRE on the morphology and function of mouse primary hippocampal neurons in order to elucidate the molecular mechanisms of potential synaptic dysfunction and neurodegeneration involved in the C9-HREassociated disease pathogenesis. Specifically, we concentrated on the effects of the C9-HRE on neuronal and dendritic spine morphology, molecular composition of the synapses, and synaptic activity.

Characterization of mouse primary hippocampal neuron cultures expressing C9orf72 expanded repeats
First, we characterized the primary mouse hippocampal cultures used in this study by ICC. The neuronal cultures were found to encompass Map2-positive neurons as well as Gfap-positive astrocytes after 21 days in vitro (Fig. 1A). Moreover, Nmdr1-positive glutamatergic (Fig. 1B) and Gad65-positive GABAergic neurons were present in the cultures (Fig. 1C). Next, we examined if expression of a stretch of 66 GGGGCC repeats (66R) in mouse primary hippocampal neurons leads to the production of the typical gain-of-function pathological hallmarks of C9-HRE in comparison to the control neurons expressing two GGGGCC repeats (2R). The same 66R construct has previously been shown to lead to the production of RNA foci and DPR proteins in mouse CNS in vivo (Chew et al., 2015) as well as in neuronal and microglial cell lines (Rostalski et al., 2020) and cultured primary mouse cortical neurons in vitro in our own previous studies (Leskelä et al., 2021). The neurons were concomitantly co-transfected with ZsGreen1-or EGFP-encoding plasmids to visualize the transfected cells. FISH analysis using the TYE™ 563-(CCCCGG) 3 probe recognizing the sense strand-derived expanded GGGGCC repeat-containing RNA revealed that the hippocampal neurons expressing the 66R contained nuclear RNA foci (Fig. 1E). The RNA foci were detected only in the 66R-expressing neurons, but not in neurons expressing the 2R control construct (Fig. 1D). The RNA foci in the 66Rexpressing neurons were not detected with a negative control TYE™ 563-(CAG) 6 probe (data not shown), indicating specificity of the TYE™ 563-(CCCCGG) 3 probe against the expanded GGGGCC repeats. Then, we assessed by ICC if the C9-HRE-derived DPR proteins were produced in the 66R-expressing neurons. These investigations indicated that the 66R-expressing neurons produced only the poly-GP DPR proteins, which The 66R neurons did not show increased cytotoxicity compared to 2R control neurons. n = 6 from 3 independent experiments, Unpaired t-test, not significant. (C) The 66R neurons displayed significantly decreased viability after 24-h treatment with 200 μM glutamate as compared to vehicle treatment or glutamate-treated 2Rexpressing neurons. The neuronal viability was assessed using Map2-TMS-based neuronal viability assay similarly to A. n = 14-16 from 3 independent experiments, One-way ANOVA), p-values are shown above the bars.
were localized in the cytoplasm of the neuronal soma (Fig. 1G). Other DPR proteins poly-GA, poly-GR, poly-PR, and poly-PA remained undetectable by ICC. We did not observe cytoplasmic TDP-43 inclusions in the 66R-expressing neurons (data not shown), which are other typical neuropathological hallmarks in the brain of C9-HRE-associated FTLD patients and shown to be present in vivo in 66R-expressing mouse brain (Chew et al., 2015). Taken together, expression of the 66R construct in primary hippocampal neurons leads to the typical gain-of-toxic-function Fig. 3. C9orf72 expanded repeats decrease the length and branching of the dendrites in mouse primary hippocampal neurons. Neurons were cotransfected at 19 days in vitro (DIV) with either 2R or 66R plasmid and Zsgreen1 (green) and analyzed at DIV 21. Neurons expressing 66R show a simpler morphology and decreased dendritic branching (B) compared to control 2R neurons (A). Digital reconstruction (shown in white in A and B) of the neurons with Vaa3D program was used to quantitatively analyze the morphology of the neurons. Total length of the dendrites (C), and total number of branches (D) were significantly decreased in 66R-expressing neurons compared to control 2R neurons. n = 70-74 from 3 independent experiments, unpaired t-test (total length, normally distributed data), Mann-Whitney U test (total branches, not normally distributed data), p-values are shown above the bars. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) hallmarks of the C9-HRE, namely the RNA foci and accumulation of DPR proteins, and yields therefore a valid model to study the effects of the C9-HRE on neuronal and synaptic morphology and function.

Mouse hippocampal neurons expressing C9orf72 expanded repeats do not show decreased neuronal viability under basal conditions but exhibit increased sensitivity to glutamate-induced excitotoxicity
According to previous reports, the C9-HRE might lead to cell toxicity (Stopford et al., 2017). In order to determine if expression of the pathological 66R construct causes altered neuronal viability and cell toxicity, we performed Map2 antibody-based viability and LDH-based toxicity assays, respectively. The Map2 viability assay was performed on 21-dayold cultures of hippocampal neurons that were transduced with 2R or 66R. The LDH release was assessed in medium samples collected at day 7, 14 and 21. The Map2 assay revealed that the 66R expression did not influence cell viability and there was no significant difference in the neuronal viability between the 2R and 66R neurons under basal culture conditions ( Fig. 2A). Similarly, the LDH cytotoxicity assay did not indicate significant differences between the 2R and 66R neurons, further suggesting that 66R expression does not lead to toxicity at any of the time points investigated (Fig. 2B).
Previous studies have shown that iPSC-derived neurons from C9-HRE carriers are susceptible to glutamate-induced excitotoxicity (Donnelly et al., 2013). Therefore, we next tested whether the 66R-expressing neurons showed evidence of glutamate excitotoxicity (Fig. 2C). There was no difference in neuronal viability between vehicle-treated 2R-and 66R-expressing neurons under basal conditions. Moreover, the 2Rexpressing neurons did not show increased neuronal loss upon treatment with 200 μM glutamate as compared to vehicle-treated cells.
However, a significant decrease in cell viability was observed in 66R neurons when treated with 200 μM glutamate as compared to vehicletreated 66R neurons or 2R neurons treated with glutamate, indicating increased sensitivity to glutamate-mediated excitotoxicity.

Expression of C9orf72 expanded repeats leads to decreased length and impaired branching of dendrites in mouse hippocampal neurons
To further study the effects of the C9-HRE on neurons, we examined the dendritic branching of hippocampal neurons co-expressing 2R or 66R and ZsGreen. Observation of the neurons under a confocal microscope revealed a drastic difference in dendritic branching between 2R (Fig. 3A) and 66R neurons (Fig. 3B). We then utilized the Vaa3D program to reconstruct the neurons into 3D images and analyzed the length and branching of their dendrites. The data showed a significant reduction in the total length of the dendrites (Fig. 3C) in the 66R neurons as compared to 2R neurons. The length of the dendrites was approximately three times shorter than in the control 2R neurons. In addition, there was a significant decrease in the total number of branches in the 66R neurons. The 2R neurons exhibited 3.5 times more branches than the 66R neurons (Fig. 3D). These findings suggest that the C9-HRE leads to a decreased morphological complexity of the neurons.

C9orf72 expanded repeats cause altered dendritic spine morphologies in mouse hippocampal neurons
We next explored whether the 66R-expressing neurons showed morphological changes in dendritic spines. The hippocampal neurons were co-transfected on DIV 19 with either 2R or 66R and EGFP and analyzed on DIV 21 ( Fig. 4A and B). Quantitative analysis of the number and morphology of the dendritic spines in 2R-and 66R-expressing neurons indicated that the total number of dendritic spines was not affected by the 66R expression (Fig. 4C). However, the 66R-expressing neurons exhibited a significant reduction in the number of mushroomtype spines and there were approximately 35% less mushroom spines compared to 2R neurons (Fig. 4C). In contrast, 66R-expressing neurons displayed a significant increase in the number of thin and stubby spines. There were 17% more thin and 21% more stubby spines in the 66R neurons when compared to 2R neurons (Fig. 4C). There were no significant differences in the spine head diameter of the morphologically different spine types between the control 2R and 66R neurons (Fig. 4D). These results indicate significant alterations in the dendritic spine morphologies upon expression of the C9-HRE in mouse hippocampal neurons. Furthermore, the decreased dendritic complexity and altered dendritic spine morphologies in the 66R neurons altogether suggest that the C9-HRE might cause alterations in the structure and function of synapses.

C9orf72 expanded repeat-expressing mouse hippocampal neurons do not show significant alterations in the molecular composition of synapses
We then extracted synaptic proteins from primary hippocampal neurons transduced with 2R or 66R. The neuronal synaptosomes were confirmed to contain neuroligin 1 and Psd95, both post-synaptic proteins, as well as the pre-synaptic protein synaptophysin. However, no differences in the levels of these proteins were observable between the 2R and 66R-expressing neurons ( Fig. 5A and C). The C9rant antibody, which recognizes poly-GP DPR proteins, was able to detect the presence of a smear of poly-GP in the synaptosomal fraction of 66R expressing cells (Fig. 5A). There were no differences in the levels of phosphorylated and total Ca 2+ /calmodulin-dependent protein kinase II (CamKII), a major kinase at the synapses involved in the mechanisms of learning and memory, between the 66R and 2R control neurons ( Fig. 5B and C). Moreover, no significant alterations were observable in the levels of the actin-binding proteins LIM domain kinase 1 (Limk1) nor cofilin or their phosphorylated forms ( Fig. 5B and C). Finally, the levels of the SNARE protein Vamp2 (synaptobrevin-2) in neurons expressing the 66R remained unchanged when compared to 2R neurons ( Fig. 5A and C).

Mouse hippocampal neurons expressing the C9orf72 expanded repeats show enhanced responsiveness to glutamate stimulation, which is mediated by extrasynaptic NMDA receptors
Finally, we investigated if the structural changes in the dendrites and dendritic spines as well as the enhanced sensitivity to glutamate excitotoxicity of the 66R neurons associated with functional changes in the neurons. To this end, we utilized calcium imaging of live neurons and treated the 2R and 66R neurons by short application of glutamate, GABA, KCl (a depolarizing agent) and ionomycin (a calcium ionophore) in a sequence. Fig. 6A shows representative responses of 2R and 66R neurons to these stimuli. The 2R and 66R neurons displayed highamplitude responses to glutamate and KCl, indicating a high responsiveness to these excitatory stimuli, whereas responses to GABA were almost negligible. This suggests that the neurons have lost depolarizing responses to GABA, which are typical for immature neurons, but have still not yet fully matured as no clear inhibitory response to GABA could be detected. In a population of 94 2R and 126 66R neurons, respectively, we found that the amplitude of the response to both glutamate and depolarizing KCl was significantly higher in the 66R neurons compared to control 2R neurons. Moreover, the response remained sustained in the 66R neurons, while it decayed faster in the 2R neurons (Fig. 6B). This hyperresponsiveness to glutamate in 66R neurons was consistent with its excitotoxic effect observed in the viability assay (Fig. 2C).
To shed light on the underlying molecular mechanism mediating the glutamate-induced hyperexcitation in 66R neurons, we used different NMDA receptor (NMDAR) inhibitors to investigate whether they affect the responses of the neurons to glutamate treatment. MK801 is a general NMDAR inhibitor, whereas TCN-201 shows selectivity to GluN2Acontaining NMDARs and ifenprodil to GluN2B-containing NMDARs (Bettini et al., 2010;Williams, 1993). The GluN2A-containing NMDARs preferentially associate with Psd95 at the postsynaptic densities, whereas GluN2B subunits are enriched in extrasynaptic NMDARs (caption on next page) N. Huber et al. (Zamzow et al., 2013). Following the first application of glutamate, glutamate was applied a second time after incubating the cells for 3 min with or without the different inhibitors (control). The amplitude of the first glutamate response in 2R and 66R neurons, respectively, was set to 100%. MK801 was found to strongly block the glutamate response in both 2R and 66R-expressing neurons (Fig. 6C) with a similar percentual reduction for both. However, the inhibitory effect of ifenprodil was significantly stronger in the 66R neurons as compared to the 2R neurons. TCN-201 inhibited the glutamate response in a similar manner in both 2R and 66R neurons. These data altogether suggest that the hyperexcitation phenotype observed in the 66R neurons was largely mediated by extrasynaptic GluN2B-containing NMDARs, which are known as the main drivers of excitotoxicity (Liu et al., 2007).

Discussion
In the present study, our aim was to elucidate the influence of the C9-HRE on the composition and function of synapses in an in vitro neuronal model system expressing the C9-HRE. Most previous studies related to the C9-HRE have focused on the toxic gain-of-function effects, namely those of the RNA foci and DPR proteins, or the loss of normal C9orf72 function. In addition, other pathological mechanisms involving e.g., specific neurotransmitter deficits or synaptic or dendritic spine dysfunction might play a significant role in C9-HRE-associated FTLD pathogenesis. These events might be independently affected during disease pathogenesis but could also result from the presence of the RNA foci and/or DPR proteins or decreased expression of C9orf72 in the C9-HRE carriers. Recent studies have pointed towards an involvement of several neurotransmitter system alterations, such as those in the dopaminergic and glutamatergic systems, in FTLD pathogenesis, and especially glutamate excitotoxicity might play a key role (Bettini et al., 2010;Donnelly et al., 2013). Many symptoms associated with FTLD, such as aggressive and compulsive behavior, agitation, and eating disorders, could be rationalized by disturbances in the different neurotransmitter systems (Murley and Rowe, 2018). However, specific molecular-level changes in these neurotransmitter systems in the C9-HRE carriers are not well understood as of now.
Our studies demonstrate that mouse hippocampal neurons expressing the 66R construct display the typical C9-HRE-related cell pathological hallmarks, the nuclear RNA foci and DPR proteins. Poly-GP proteins were found to be the only expressed DPR protein species in these neurons and they were found present in the cytoplasm in the neuronal soma. Different DPR proteins have been shown to exhibit cellto-cell transmission properties in various cell models, including primary cortical neurons and human iPSC-derived neurons (Westergard et al., 2016). Previous studies have also reported cell toxicity either caused by the RNA foci or accumulation of the DPR proteins (Freibaum et al., 2015;Haeusler et al., 2014;Jovičič et al., 2015;Stopford et al., 2017;Zhang et al., 2015). Here, the expression of 66R did not lead to evident cell toxicity as indicated by unaltered neuronal viability under basal culture conditions. Interestingly, however, treatment of the primary hippocampal neurons expressing the 66R with a high concentration of glutamate triggered excitotoxicity and significantly decreased neuronal viability when compared to control neurons. This finding suggests that C9-HRE renders the neurons more vulnerable to cell death under stress conditions. It also is in accordance with previous data showing enhanced sensitivity of iPSC-motoneurons from C9-HRE carriers to glutamatemediated excitotoxicity (Donnelly et al., 2013). Glutamate excitotoxicity is a key pathogenic event in neurodegenerative diseases (Lewerenz and Maher, 2015) and it is believed to be a major disease mechanism contributing to the degeneration of motor neurons in ALS (Taylor et al., 2016). The excitotoxic mechanisms include dysfunction of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) or NMDA glutamate receptors, leading to excess Ca 2+ influx and cell death. On the other hand, glutamate is needed for normal synaptic transmission and it instigates functional and structural synaptic plasticity-related changes by activating receptors coupled to Ca 2+ influx and their downstream kinases and transcription factors (Takamori, 2006).
Interestingly, the C9-HRE-containing RNA has been found to localize in transport vesicles in neurites in patient brains as well as iPSC-neurons derived from C9-HRE carriers, leading to deficits in neuronal branching (Burguete et al., 2015). Also, overexpression of the poly-GA DPR protein leads to reduced neurite branching and a less complex neuronal network (May et al., 2014). The dendritic morphology of neurons has important functional implications in determining which signals the neurons receive and how these signals are integrated (Jan and Jan, 2010), suggesting that branching deficits lead to disturbed neuronal communication. We observed significantly defective neuronal branching in the 66R neurons, which expressed both the C9-HRE RNA-containing RNA foci and DPR protein poly-GP. The total length of the dendrites and the total number of dendritic branches were significantly reduced in neurons expressing the 66R when compared to control neurons. In these neurons, poly-GP was found to be the only DPR protein species expressed, suggesting that also this DPR protein may be associated with neuronal branching deficits. Moreover, the aggregation or overexpression of TDP-43 can lead to compromised neuronal connectivity, including deficits in neurite branching (Takamori, 2006;Taylor et al., 2016). However, we did not observe TDP-43 inclusions in the 66R neurons in our present study. It is therefore possible that this in vitro model represents an early phase of C9-HRE-associated pathogenesis and the TDP-43 pathology might require more time to develop. Deficient branching and disrupted synaptic homeostasis has also been reported in FTLD associated with other genetic mutations in addition to the C9-HRE (Sephton et al., 2014). One study, focusing on the FTLD risk factor TMEM106B and its binding partner MAP6, reported branching deficits and altered dendrite morphology  due to misbalanced lysosomal trafficking. All in all, the defective dendritic branching, suggesting a severely diminished morphological complexity of the neuronal network, might be associated with altered synaptic function in FTLD.
Indeed, an altered complexity of neuronal network and decreased length and branching of the dendrites ultimately lead to fewer synapses in the network. Synapses are the major sites of information input located on the dendritic spines, which are small, specialized protrusions arising along dendrites. We found that the total number and spine head diameter of the dendritic spines were not affected by the 66R expression in primary hippocampal neurons. However, the proportion of spines of different morphological subtypes was significantly shifted, suggesting marked alterations in the dendritic spines. The number of morphologically mature, mushroom-type spines was significantly reduced, whereas the number of thin and stubby spines was increased in primary mouse hippocampal neurons expressing 66R. Dendritic spine alterations have been observed in various neuropsychiatric disorders, including autism Fig. 4. The C9orf72 expanded repeats drastically alter dendritic spine morphologies in mouse primary hippocampal neurons. Neurons were co-transfected at 19 days in vitro (DIV) with either 2R or 66R plasmid and EGFP (green) and analyzed at DIV 21. Representative images of dendrites and their spines in 2R control neurons (A) as well as in 66R-expressing neurons (B) are shown. The neurons were imaged with Zeiss LSM 800 confocal microscope and analyzed with NeuronStudio program. Dendritic spine analysis shows that the total spine number is not altered (C) but the number of the morphologically different spine types is significantly changed. The 66R neurons exhibit significantly fewer mushroom-type spines but a significantly increased number of thin as well as stubby spines as compared to control 2R neurons. The number of each spine type was normalized to total spine number (C). The head diameter was not altered in any dendritic spine type (D). n = 58-75 from 3 independent experiments, unpaired t-test (mushroom, thin, and stubby spine density and head diameter of thin and stubby spines, normally distributed data), Mann-Whitney U test (total spine number, head diameter of mushroom spines, not normally distributed data), p-values are shown above the bars. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) (caption on next page) N. Huber et al. Fig. 5. The C9orf72 expanded repeats do not alter synaptic protein levels in mouse primary hippocampal neurons. Representative images of Western blots of the synaptosomal extracts of 2R control neurons and 66R neurons are shown. A smear of DPR proteins (poly-GP) is detected in the 66R neuron synaptosomal extracts by the C9rant antibody. The asterisk (*) indicates unspecific signals. No significant differences in the levels of neuroligin 1, Psd95 or synaptophysin are observed (A). Also, phosphorylated and total CamKII levels and the levels of the actin-binding proteins Limk1 and cofilin or their phosphorylated forms or the SNARE protein Vamp2 remain unaltered in 66R neurons when compared to 2R control neurons (B). All protein levels were normalized to β-actin levels of the corresponding sample. The quantifications of the protein levels are shown in (C). n = 10-19 from 3 independent experiments, unpaired t-test (p-Cofilin/Cofilin, neuroligin 1, Psd95, normally distributed data), Mann-Whitney U test (p-Limk1/Limk, p-CamKII/CamKII, synaptophysin, Vamp2, not normally distributed data), not significant.  . Response to the first glutamate application (shown in B) (before vehicle or inhibitor application) was set as 100%, Mann Whitney U test (data not normally distributed) (C) Data are shown as median ± interquartile range; Kruskal-Wallis test (data not normally distributed). Pvalues shown above the bars show comparison to the corresponding vehicle treatment, except for comparison between 2R and 66R ifenprodil treatment,which is indicated above the bars by the bracket. spectrum disorders, schizophrenia, and Down syndrome (Penzes et al., 2011). In addition, the role of dendritic spine alterations has been wellcharacterized in AD and synapses are considered the earliest site of dysfunction, correlating well with the cognitive impairment in patients (Coleman and Yao, 2003). The morphology of the dendritic spines can rapidly change through activity-dependent and independent mechanisms (Rochefort and Konnerth, 2012). The number, structure, and shape of the dendritic spines can be affected by neuronal loss or reduced dendrite length, shrinkage or enlargement of the spine head, or specific pathological changes, such as membranous changes, a hypertrophied spine apparatus, or pathological protein inclusions (Herms and Dorostkar, 2016). The morphological changes of the spines are tightly linked to the biochemical properties of the spine and the spine size correlates with the strength of the synaptic transmission (Tashiro and Yuste, 2003). Therefore, spine morphology directly reflects the stability, strength, and function of excitatory synaptic connections.
Dendritic spines are largely built by the actin cytoskeleton where globular (monomeric) actin (G-actin) polymerizes into filamentous actin (F-actin) (Gipson and Olive, 2017). There is also evidence that the reorganization of the actin cytoskeleton is tightly linked to synaptic efficacy (Cingolani and Goda, 2008). Actin is important for maintaining and regulating synaptic vesicle pools (Dillon and Goda, 2005). The spine head, which is isolated from the dendrites by the spine neck, harbors the protein-rich PSD (Rochefort and Konnerth, 2012). The PSD proteins are directly or indirectly involved in synaptic communication and thus in the regulation of synaptic strength (Tashiro and Yuste, 2003). Therefore, changes in the levels and composition of the PSD proteins may lead to long-term changes in synapses and neuronal circuits. Embedded in the PSD are also NMDA and AMPA receptors, which mediate excitatory synaptic transmission (Rochefort and Konnerth, 2012;Tashiro and Yuste, 2003). Our data indicate that the spine head volume was not affected in 66R neurons, agreeing with the findings that the levels of key pre-and postsynaptic proteins, including Psd95, neuroligin 1 or synaptophysin, remained unchanged based on the biochemical analyses of synaptosomal extracts. However, the significantly reduced number of mushroom-type dendritic spines might reflect changes in the synaptic contacts and subsequent potential alterations in the synaptic function.
Another protein of interest was the most abundant synaptic kinase, CaMKII, which regulates synaptic plasticity, including NMDA receptordependent, activity-induced spine outgrowth, by phosphorylating several binding partners (Franchini et al., 2020). However, there were no changes in phosphorylated or total CamKII levels detectable in the 66R neurons. There were also no significant alterations in the levels of the SNARE protein Vamp2, which mediates the fusion and exocytosis of synaptic vesicles containing neurotransmitters (Salpietro et al., 2019). Finally, as the regulation of the actin cytoskeleton dynamics is important for dendritic spine morphology, we investigated the levels of key actin cytoskeleton modulators in the synaptosomal extracts. An important actin-depolymerizing protein is cofilin and its regulator Limk1, and changes in these proteins have been shown to lead to alterations in spine morphology and synaptic plasticity (Meng et al., 2002;Rust et al., 2010). However, there were no changes in the levels of these actin regulators in neurons expressing the 66R. On the other hand, despite the fact that dendritic spine changes can be influenced by the reorganization of the actin cytoskeleton, which might alter synaptic vesicle mobility (Cingolani and Goda, 2008), the organization of actin in dendritic spines cannot be only explained by alterations in actin-binding proteins (Bertling et al., 2016). We found here that the DPR proteins generated from the C9-HRE were also present at the synaptosomal fractions in the 66Rexpressing hippocampal neurons, suggesting synaptic localization. It is therefore possible that the presence of these pathological proteins leads to the observed alterations in the morphology of dendrites and dendritic spines in the 66R-expressing neurons.
Although the molecular composition of the synapses did not show major changes upon the expression of the 66R in the hippocampal neurons, we found that the morphological changes in the dendritic branches and dendritic spines as well as the presence of the DPR protein poly-GP at the synaptosomal fractions were accompanied by significantly altered neuronal activity.
Our Ca 2+ imaging results suggest that 66R-expressing hippocampal neurons show a hyperexcitation phenotype as their responses to glutamate or depolarizing KCl application were significantly stronger than those in the 2R-expressing hippocampal neurons. The absence of a pronounced inhibitory response to GABA application suggested that the 2R and 66R-expressing neurons were not fully mature as the change to an inhibitory GABA response had not yet taken place. Treatment of the 2R and 66R-expressing neurons with MK801, an NMDAR inhibitor, led to a strong decrease in the glutamate response, suggesting that the glutamate response was mostly mediated by the NMDARs and not AMPA/kainate receptors.
GluN2A-containing NMDARs predominantly localize at synaptic sites in mature neurons (Franchini et al., 2020;Zamzow et al., 2013). Treatment with the GluN2A-selective inhibitor TCN-201 reduced the response to glutamate in 2R and 66R-expressing neurons to a similar extent, suggesting that synaptic transmission is unaltered in the 66R neurons. Treatment with the GluN2B-selective inhibitor ifenprodil also reduced the glutamate response in both 2R and 66R-expressing neurons but the reduction in the 66R-expressing neurons was significantly stronger. GluN2B-containing NMDARs are enriched in extrasynaptic sites but can also be found in synaptic NMDARs of immature neurons before the switch from GluN2B-to GluN2A-containing NMDARs (Hardingham and Bading, 2010;McKay et al., 2012;Zamzow et al., 2013). As the maturation state of the 2R and 66R-expressing neurons did not significantly differ (neither showed a strong inhibitory response to GABA), it is likely that the stronger reduction of the glutamate response in the ifenprodil-treated 66R neurons is due to an increased extrasynaptic transmission, which can lead to hyperexcitation and negatively affect neuronal survival. This finding is also in line with the observed excitotoxic effects after glutamate treatment in the 66R but not the 2R neurons.

Conclusion
Increasing evidence suggests that FTLD patients exhibit alterations in different neurotransmitter systems, such as glutamatergic, GABAergic, dopaminergic and serotonergic systems, which could account for the behavioral changes and other symptoms observed in these patients. Moreover, such neurotransmitter system deficits and altered neuronal network activities may also associate with different genetic forms of FTLD. Our data indicate that the C9-HRE significantly alters the synaptic structures and function, including dendritic spine morphology as well as neuronal branching. These morphological changes associate with a hyperresponsiveness to excitatory stimuli-induced synaptic transmission mediated via extrasynaptic NMDARs, leading to strongly augmented and prolonged Ca 2+ influx, and enhanced susceptibility to glutamate-induced excitotoxicity. Our present findings together with the studies by others suggest that C9-HRE leads to synaptic dysfunction and that targeting synaptic transmission mediated by the different neurotransmitters might offer promising novel therapeutic avenues to treat FTLD.

Ethics approval and consent to participate
The primary mouse hippocampal neuron cultures were prepared from E18 mouse embryos under the ethical permission by the Laboratory Animal Center of the University of Eastern Finland, Kuopio, Finland, permission number: EKS-006-2019.

Consent for publication
All authors have approved the final version of the manuscript and give their consent for publication.

Availability of data and materials
Results generated and analyzed during the current study are included in this published article. The AAV-2R and AAV-66R constructs were a kind gift from Dr. Leonard Petrucelli and were used in this study under a material transfer agreement between the Mayo Clinic, Jacksonville, FL, USA, and A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland.