PolyGA targets the ER stress-adaptive response by impairing GRP75 function at the MAM in C9ORF72-ALS/FTD

ER stress signaling is linked to the pathophysiological and clinical disease manifestations in amyotrophic lateral sclerosis (ALS). Here, we have investigated ER stress-induced adaptive mechanisms in C9ORF72-ALS/FTD, focusing on uncovering early endogenous neuroprotective mechanisms and the crosstalk between pathological and adaptive responses in disease onset and progression. We provide evidence for the early onset of ER stress-mediated adaptive response in C9ORF72 patient-derived motoneurons (MNs), reflected by the elevated increase in GRP75 expression. These transiently increased GRP75 levels enhance ER–mitochondrial association, boosting mitochondrial function and sustaining cellular bioenergetics during the initial stage of disease, thereby counteracting early mitochondrial deficits. In C9orf72 rodent neurons, an abrupt reduction in GRP75 expression coincided with the onset of UPR, mitochondrial dysfunction and the emergence of PolyGA aggregates, which co-localize with GRP75. Similarly, the overexpression of PolyGA in WT cortical neurons or C9ORF72 patient-derived MNs led to the sequestration of GRP75 within PolyGA inclusions, resulting in mitochondrial calcium (Ca2+) uptake impairments. Corroborating these findings, we found that PolyGA aggregate-bearing human post-mortem C9ORF72 hippocampal dentate gyrus neurons not only display reduced expression of GRP75 but also exhibit GRP75 sequestration within inclusions. Sustaining high GRP75 expression in spinal C9orf72 rodent MNs specifically prevented ER stress, normalized mitochondrial function, abrogated PolyGA accumulation in spinal MNs, and ameliorated ALS-associated behavioral phenotype. Taken together, our results are in line with the notion that neurons in C9ORF72-ALS/FTD are particularly susceptible to ER–mitochondrial dysfunction and that GRP75 serves as a critical endogenous neuroprotective factor. This neuroprotective pathway, is eventually targeted by PolyGA, leading to GRP75 sequestration, and its subsequent loss of function at the MAM, compromising mitochondrial function and promoting disease onset. Supplementary Information The online version contains supplementary material available at 10.1007/s00401-022-02494-5.


Introduction
Silent prodromal phases, with no overt symptoms, are a conserved feature of most neurodegenerative diseases (NDs) [29,35]. Adaptive responses, through homeostatic and compensatory processes, are likely to predominate during these early phases and might well be sufficient to counteract disease-associated functional deficits and enable cells to cope with stress. However, chronic stress responses may eventually be detrimental to the cell [19,29,34,35,39].
Identification of early compensatory measures and the specific pathomechanisms impairing them is pivotal in finding valid temporal entry points to target the pathology.
During presymptomatic stages of amyotrophic lateral sclerosis (ALS), motoneurons (MNs) already exhibit early vulnerability to endoplasmic reticulum (ER) stress [57]. Initially, ER stress responses are physiologically adaptive enabling cells to cope with stress. However, the chronicity of the ER stress response is detrimental to the cell [19, 32,34,39,57]. We studied the relatively common monogenic form of ALS and frontotemporal dementia (FTD) due to a hexanucleotide (G4C2) repeat expansion within the first intron of the C9ORF72 gene [15,43,55]. In C9ORF72-ALS/FTD patient-derived MNs, decreased cell survival was linked to dysfunction in Ca 2+ homeostasis, increased ER stress, reduced mitochondrial membrane potential as well as mitochondrial dysfunction [13,14,23,37,48,66].
Three neurodegenerative mechanisms are implicated in C9ORF72-ALS/FTD pathology: loss of function due to reduced C9ORF72 transcript and proteins levels [2], a gain of function due to the formation of RNA foci [46], and a gain of toxic function, due to the generation and accumulation of five different dipeptide repeat proteins (DPRs), via repeat-associated non-AUG (RAN) translation of hexanucleotide repeat sequences from both sense and antisense strands [20]. Of the five DPR species, there is evidence for a toxic gain-of-function role for PolyGA, PolyGR, and PolyPR aggregates in promoting neurodegeneration in affected brain areas [30,36,40]. Previous studies have established that the positively charged arginine-rich PolyGR and PolyPR proteins, and the highly insoluble PolyGA aggregates are extremely toxic in neurons [44]. PolyGA is the most prominently detected DPR in cytoplasmic inclusions [68], and it accumulates in p62/ubiquitin-positive inclusions in the brain and spinal cord of C9ORF72-ALS/FTD patients [44,68]. PolyGA-overexpressing mice exhibit motor and cognitive deficits along with TDP-43 pathology, cerebellar atrophy, and astrogliosis [20]. PolyGA induces toxicity by promoting ER stress and activating caspase 3-related apoptotic pathways [44,68]. PolyGA directly connects with and inhibits the proteasome, thereby promoting TDP-43 pathology [33,68]. Moreover, PolyGA sequesters proteins, such as Unc119 [], a lipid-binding chaperon involved in vesicular and protein trafficking, contributing to selective neuronal vulnerability in C9ORF72-ALS/FTD [44]. In addition, HR23 proteins involved in proteasomal degradation together with proteins involved in nucleocytoplasmic transport were sequestered along with GA aggregates [67]. A recent study reported that PolyGA aggregates sequestered VCP, impairing autophagy [1]. Taken together, these results have suggested contrasting potential scenarios of how a C9ORF72 hexanucleotide repeat expansion leads to neurodegeneration. However, the molecular pathomechanism by which PolyGA might influence early phases of MN degeneration, and in particular, regulation of early ER-mitochondrial adaptive response and signaling pathway has not been studied in detail thus far.
We here show that the initiation of ER-stress-induced compensatory responses involve the enhanced expression of glucose-regulated protein 75 (GRP75) at mitochondriaassociated membranes (MAMs), counteracting early mitochondrial Ca 2+ imbalance in human C9ORF72-ALS/FTD and C9orf72 mouse neurons. This protracted pre-symptomatic phase is followed by the near-complete loss of GRP75 expression coinciding with the appearance of unfolded protein response (UPR) signaling, mitochondrial dysfunction, and appearance of PolyGA aggregates. PolyGA inclusions sequestered GRP75, leading to its reduction at the MAMs and eventual loss of function, and mitochondrial dysfunction. Comparable pathological sequestration of GRP75 in PolyGA aggregate-bearing neurons was observed in human C9ORF72-ALS/FTD post-mortem tissue. Sustaining high GRP75 expression in spinal C9orf72 MNs specifically prevented ER stress and normalized mitochondrial function, suggesting that impairment of GRP75 function is a critical early event in C9ORF72-ALS/FTD development and progression.

Mouse strains
T h e C 9 -5 0 0 BA C m o u s e l i n e ( F V B / N J -Tg(C9orf72)500Lpwr/J) described in [41] were purchased from Jackson Laboratory (https:// www. jax. org/ strain/ 029099) and maintained in heterozygosis crossed with (noncarrier) mice FVB/NJ (Janvier labs, SC-FVBN-F). Within our colony, an acute phenotype is observed in 25-30% of females with a median life span of 105 days. The remaining female and male mice exhibit a slow progressing phenotype [53]. Note: some labs were not able to observe the original phenotype [51]. For all immunochemical analyses, only males were included and both males and females were examined together as from P125, to eliminate the acute phenotype, and to include only the slow progressive phenotype. All survival and behavior cohorts included both genders and slow and fast progressing females. Long range PCR was routinely performed to define repeat length-matched cohorts. All behavior and survival assays were performed in repeat length-matched mice. Adult C57BL/6 J mice were transduced with lentiviral-PolyGA. Animal care, ethical usage and procedures were in accordance with the Swiss Veterinary Law guidelines, and the study was approved by the animal commission of Canton of Bern, Switzerland.

Imaging and image analysis
Confocal images were acquired using SP8 (Leica Microsystems) fitted with a 63 × oil objective, or FV1000 (Olympus) microscope, fitted with a 20 × , 40 × air objective or 60 × oil objective. All images were processed using Imaris software version 7.6.3 or Fiji. For the analysis of BiP, P i -eif2α, GRP75, labeling intensities, data were acquired using identical confocal settings, with signals at the brightest cells being non-saturated, and background levels outside MN pools still detectable. Images were analyzed quantitatively using FiJi or Imaris. Signal intensity values for the antigen of interest were calculated over several consecutive lumbar spinal cord Z-stack spaced 0.5 μm, after background subtraction from every channel. Lowest signals had values of below 50 and high-intensity neurons exhibited labeling values up to 255. Signal values below 50 in the case of BiP in WT animals were counted as basal expression. To count MNs numbers, cell counter plugin from Fiji was used. Imaris software was used to reconstruct the 3D isosurface for PolyGA aggregates volume. Fiji software was employed to analyze the mitochondria-ER contacts and sphericity for 3view EM images. SBF SEM volume reconstruction were done using the Volume viewer plug in from Fiji and IMOD software was used for the 3D segmentation of mitochondria.

Human post-mortem tissue
The human post-mortem brain, hippocampus and lumbar spinal cord samples fixed in buffered formalin were obtained from the archives of the Department of (Neuro)Pathology (Amsterdam UMC, University of Amsterdam, The Netherlands), and included four C9ORF72-ALS/FTD, and four age-matched control cases. They were selected from a retrospective searchable neuropathological database that was reviewed independently by two neuropathologists (Eleonora Aronica and Dirk Troost), and included cases with consent for post-mortem brain/spinal cord autopsy and use of the autopsy tissue and their medical records for research purposes. All of the C9ORF72-ALS/FTD patients had shown clinical signs and symptoms of lower and upper MN disease. All the patients fulfilled the diagnostic criteria for ALS. The controls included in the present study were adult individuals without any history of neurological disease, based on their last clinical evaluation (Demographic information in Supplementary Table 2, online resource).

Diaminobenzidine immunohistochemistry of human post-mortem tissue
Paraffin sections (3-4 µm) were placed on poly-l-lysinecoated slides and dried for 3-4 h on a heating plate at 40 °C. The sections were deparaffinized in xylene for 20 min and rehydrated in 100%, 95% and 70% EtOH for 5 min each, followed by endogenous peroxidase quenching (0.3% H 2 O 2 in methanol) for 20 min. For antigen retrieval, the sections were heated in citrate buffer, pH 6 (Dako) for 20 min in a pressure cooker. After PBS washes, the sections were incubated with primary antibody (Supplementary Table 3, online resource) for 1 h at RT or at 4 °C overnight. After washing in PBS, they were incubated with the polymeric horseradish peroxidase linker secondary antibody (IL Immunologic) for 30 min at RT. For light microscopy, DAB reagent (DAB kits; DCS Innovative Diagnostic System) was used to stain the sections, and counter-stained with 6% haematoxylin for 3 min. Images of the DAB-stained sections were taken with a Zeiss Axioplan microscope equipped with a 40 × , 63 × objective and an Axiocam 506 color camera (Zeiss). Quantification was done manually using a 20 × lens on the DAB-stained sections from each case by counting the no. of α-MNs. Semi-quantitative assessment of GRP75 immunolabeling was performed on serial sections. Three lumbar spinal cord sections each from the four C9ORF72-ALS/FTD cases as well as four normal controls were examined. The level of GRP75 immunoreactivity in the majority (more than 95%) of the MNs of control cases were assigned as normal levels. Using a 10 × , 20 × objective, the number of MNs displaying GRP75 immunoreactivity were counted, and quantified based upon the observed GRP75 levels using the following scheme, A = MNs showing normal levels of GRP75, B = MNs showing moderate reduction of GRP75 levels, C = MNs showing strong reduction of GRP75 levels.

Immunofluorescence labeling of human post-mortem tissue
Immunofluorescence labeling was performed as described previously [17,64]. Briefly, deparaffinized tissue sections were boiled in citrate buffer, pH 6 (Dako) for 20 min in a pressure cooker for antigen retrieval. Sections were blocked with 10% normal goat serum (Life Technologies), for 1 h at RT, and incubated with the required primary antibody (Supplementary Table 3, online resource), at a dilution of 1:100 for each antibody at RT for 1 h or at 4 °C overnight. After washing in TBST for 10 min the sections were incubated with Alexa Fluor secondary antibodies (dilution 1:500 in PBS) at RT for 2 h. Sections were washed in TBST (2 × 10 min), and stained for 10 min with 0.1% Sudan Black in 80% EtOH to suppress endogenous lipofuscin auto-fluorescence and washed for 5 min in TBST and mounted with Vectashield mounting medium (Vector Laboratories) containing DAPI. Images were obtained with a LSM 700 (Zeiss) microscope using 40 × and 63 × objectives. Images were acquired by averaging 4 scans/area of interest resulting in an image size of 1024 × 1024 pixels. The laser intensity was kept constant for all the sections examined. Images were analysed using Adobe Photoshop CS5 and ZEN (Blue edition) 2009 software. For quantification of GRP75 immunofluorescence intensity in MNs, ROIs were manually defined and average intensity was measured using Fiji software. GRP75 intensities were binned into three classes (normal, medium reduction, and strong reduction) based on pixel intensity and graphs were plotted as percentage of MNs expressing different GRP75 levels. GRP75 sequestration by PolyGA in hippocampal dentate gyrus neurons was measured using IMARIS, wherein neurons harbouring PolyGA aggregates were analysed for GRP75 immunoreactivity either surrounding or colocalizing within aggregates. Graph was plotted as percentage of neurons, harbouring PolyGA aggregates exhibiting sequestration or devoid of GRP75 sequestration.

Statistical analysis
Analysis were done using GraphPad Prism 6.0. Statistical significances were evaluated by two-tailed, unpaired Student's t test and one way ANOVA. Post ANOVA Sidak test was used to evaluate statistical significance as indicated in the respective figure legend. Values are expressed as mean ± standard error of the mean (SEM) or standard deviation (SD) as indicated in the corresponding legends. *P < 0.05, **P < 0.01, ***P < 0.001 throughout the manuscript.
Additional methods are described in detail in Supplementary Information.
To examine whether the increase in GRP75 mRNA levels was a direct response to ER stress, we treated 1-week-old iMNs, which at this point lack ER stress, with tunicamycin . c SBF-SEM Q.A. of mitochondria numbers in 2-week-old iMNs, showing increased number of mitochondria in all C9ORF72 patient lines compared to controls or corresponding isogenic (Unpaired t test: Ctrl(1) vs C9ORF72(1) t = 7.556, P < 0.0001***; Ctrl(2) vs C9ORF72(2) t = 6.815, P < 0.0001***; Ctrl(1) vs C9ORF72(3), t = 6.678, P < 0.0001***; iso-C9ORF72 (5) vs C9ORF72(5), t = 6.750, P < 0.0001***). Combined graph for Ctrls lines (Ctrl 1-2, Iso-C9 (5)) and C9 (C9 1-2-3-5), Unpaired t test Ctrls vs C9 t = 12.53, P < 0.0001, ***. d Representative images of mitochondria contacts with ER for Iso-C9ORF72 (5) and C9ORF72 (5). Linear regression between the no. of contacts per mitochondria and their sphericity in 2-week-old iMNs (Ctrl (1) , we performed a mild treatment (TU: 1 µg/mL) for 18 h. TU treatment augmented GRP75 mRNA levels in both control and mutant 1-week-old iMNs, suggesting that higher GRP75 transcripts reflected ongoing ER stress in iMNs (Fig. 1c). Of note, mutant iMNs exhibited a stronger response to TU treatment than healthy control iMNs. Moreover, other MAM molecules remained largely unaltered in response to the mild TU treatment: no significant alteration of twofold or more was observed ( Supplementary Fig. 2b, online resource). We further validated GRP75 protein levels in iMNs at 1, 2, and 4 weeks by immunostaining. Corroborating the qPCR data, no change in BiP and GRP75 protein levels was observed in 1-week-old iMNs, whereas all C9ORF72-ALS/FTD patient lines exhibited higher expression of the ER stress marker BiP and GRP75 compared with healthy control iMNs and isogenic control lines after 2 weeks (Fig. 1d) and 4 weeks of differentiation (Supplementary Fig. 2c, online resource). As additional evidence, immunoblotting performed on 2-week-old iMNs confirmed elevated BiP and GRP75 expression in C9ORF72-ALS/FTD patient-derived iMNs (Fig. 1e). As shown in (Supplementary Fig. 1c, d, online resource), we detected no TDP-43 mislocalization or PolyGA aggregates in 2-week-old iMNs, suggesting that both ER stress and increased GRP75 levels are observed in iMNs before the appearance of these major C9ORF72-linked pathological hallmarks. Enhanced ER-mitochondria coupling via GRP75 promotes optimal mitochondrial calcium uptake and bioenergetics. a Baseline (0-60 s) and stimulated mitochondrial calcium (Ca 2+ ) uptake (80 s onward) traces from Ctrls(1 and 2) and C9ORF72 (1,2,3) iMNs and Iso-C9ORF72(4 and 5) and C9ORF72(4 and 5) iMNs at 1-week post-differentiation when there is no ER stress or GRP75 upregulation. All patient lines exhibited decreased mitochondrial Ca 2+ transients compared to controls or respective isogenic controls. We next validated these observations using three C9ORF72-ALS/FTD patient-derived and three healthy control neuronal cell lines which were generated via direct conversion of fibroblasts to neurons (dNeus) [50], thereby preserving hallmarks of cellular aging (see Supplementary Fig. 3a, b, online resource for phenotypic description and quality of differentiation). Two-week-old C9ORF72 dNeus showed increased levels of BiP and CHOP mRNA, and concomitantly higher protein levels of GRP75 was detected by immunostaining Supplementary Fig. 3c, d, online resource). Comparable to iMNs, dNeus also exhibited higher expression of ITPR3 transcripts ( Supplementary  Fig. 3c, d, online resource), suggestive of a conserved ER stress response and associated changes in GRP75 expression in C9ORF72-ALS/FTD patient neurons.
Enhanced ER-mitochondria association are specific to C9ORF72 iMNs GRP75 serves as a functional linker protein between the ER and the mitochondria. Thus, any changes in its expression pattern would influence both cellular compartments. Therefore, we examined whether mitochondria in C9ORF72 iMNs display early structural or functional alterations. To this end, we performed serial block-face scanning electron microscopy (SBF-SEM) on four C9ORF72, two healthy and one isogenic control iMNs. 3D reconstruction of images revealed that in all four C9ORF72 iMNs, other than in the healthy or isogenic controls, a large fraction of mitochondria were rounded in shape (Fig. 2a). To obtain a quantitative measure for the observed mitochondrial shape changes, we measured mitochondrial sphericity and plotted the values as a frequency distribution [61]. Applying a cutoff from 0.86 to 0.98, where values closer to 1.0 indicate a perfectly spherical object, all four C9ORF72 iMNs presented a significantly higher proportion of rounded mitochondria compared to healthy/isogenic controls (Fig. 2b). Subsequently, we assessed mitochondrial numbers, which were also on average higher in all four C9ORF72 iMNs compared to the control iMNs (Fig. 2c). Notably, the isogenic control Iso-C9ORF72(5) again had reverted the mitochondrial phenotype displayed by the mutant C9ORF72(5) line. Measuring the number of ER contacts that each mitochondrion made revealed that spherical mitochondria in mutant MNs on average exhibited a higher number of contact points (Fig. 2d). As GRP75 physically tethers IP 3 R-VDAC1, thereby promoting mitochondrial Ca 2+ uptake, we quantified the IP 3 R-VDAC1 interaction via proximity ligation assay (PLA). All C9ORF72 iMNs at 2 weeks presented increased interaction between the two partners (Fig. 2e).
We next assessed whether the changes in mitochondrial structure were present within C9ORF72 iPSCs and might reflect mutation-related alterations. SBF-SEM was performed on two C9ORF72 and two healthy control iPSCs. Both C9ORF72 iPSCs displayed higher proportion of spherical mitochondria compared to controls (Fig. 2f). Notably, the number of ER-mitochondria contacts remained unchanged between mutant and control-iPSCs (Fig. 2g). Further, measurement of GRP75 and CHOP transcripts revealed no change in expression levels, suggestive of those expression changes being specific to iMNs (Fig. 2h). We evaluated key parameters of mitochondrial function using the Seahorse assay and found reduced basal respiration and consequently lower ATP production in C9ORF72 iPSCs (Fig. 2i). Taken together, these data indicate that intrinsic mitochondrial deficits are present in C9ORF72 cells and that specifically C9ORF72 iMNs undergo early increase in ER-mitochondria associations in parallel to the observed ER stress.

Increased ER-mitochondria association via GRP75 normalizes Ca 2+ uptake and ATP generation in C9ORF72 iMNs
GRP75 creates a physical link between the ER membrane and the outer mitochondrial membrane by facilitating the interaction between ER-bound IP 3 Rs and mitochondrial Fig. 4 Age-dependent reduction in GRP75 expression in human and rodent C9ORF72 CNS tissue. a Representative images showing differential GRP75 immunoreactivity in human lumbar spinal cord anterior horns of C9ORF72 ALS patients compared to the normal controls. The variable immunofluorescence intensity was further scored as low, normal, medium and high as depicted. Images are shown from two C9ORF72 patients and rendered for expression intensity. Right: Q.A. of GRP75 average fluorescence intensity: unpaired t test Control vs C9ORF72, t = 11.61, P < 0.0001***; Q.A. of GRP75 immunoreactivity: normal (C9ORF72 18%), moderate reduction (C9ORF72 40%), strong reduction (C9ORF72 42%), Chisquare analysis P < 0.0001***. (Number of MNs analyzed, Control: 29; C9ORF72: 77), n = 4 C9ORF72-fALS patients and n = 4 agematched controls. Note that in Control α-MNs, GRP75 immunoreactivity was uniform, whereas more α-MNs from C9ORF72-ALS/ FTD cases were examined in detail to account for the high degree of variability in GRP75 staining. b Representative double immunofluorescence labeling of human C9ORF72-ALS/FTD α-MNs within the lumbar spinal cord using antibodies against GRP75 and pTDP-43. Arrow: pTDP-43-positive inclusion. c Representative double immunofluorescence labeling of C9ORF72-ALS/FTD patient hippocampal dentate gyrus neurons compared to normal control. Note the overall reduced labeling of GRP75 and its focal sequestration with PolyGA aggregates in C9ORF72-ALS/FTD patient dentate gyrus hippocampal neurons, where 71.50% ± 3.6 of the neurons showed GRP75 sequestration within PolyGA aggregates. Sixty PolyGA aggregate bearing neurons were analyzed. d Representative confocal images of spinal MNs stained for GRP75 from WT and C9-500 animals at different ages. 3D rendering of MNs based on the intensity level of GRP75 staining show increased levels of GRP75 expression at P80 and decreased levels of GRP75 expression in P200 C9-500 animals compared to WT. e Longitudinal Q.A. of GRP75 expression in WT and C9-500 animals showing MNs with high GRP75 levels at P80 and P120 (41-48%) with subsequent drop in expression after P150, where GRP75 levels are below WT expression levels. f Representative confocal immunofluorescence imaging of GRP75 expression in WT and C9-500 cortex showing increased GRP75 levels at P150 and below WT levels of expression at P240. g Representative WB of brain extracts from WT and C9-500 animals at P240. MAM isolation using sucrose gradient revealed decreased GRP75 levels in C9-500 animals compared to WT, while no difference was detected for the loading controls MFN2. Quantitative analysis of the intensity of GRP75 at the MAM compared to WT (Unpaired t test: t = 6.263, P = 0.0033**). N = 3 mice, repeated thrice as separate experiments. Scale bars: a 15 µm, b 60 µm, c 8 µm, d 30 and 10 µm (zoom), f 30 µm ◂ VDAC1 to promote core mitochondrial Ca 2+ uptake and optimal mitochondrial bioenergetics [62]. Dysregulation of mitochondrial Ca 2+ homeostasis and mitochondrial Ca 2+ overload has been linked to neuronal death in neurodegenerative disorders [10,59,60]. Therefore, we measured mitochondrial Ca 2+ uptake in C9ORF72 and control iMNs after 1 and 2 weeks of differentiation. Fluo-4AM was combined with an intracellular buffer that eliminated cytosolic and ER Ca 2+ signals [47], thereby enabling specifically the measurement of mitochondrial Ca 2+ uptake. Unexpectedly, all 1-week-old C9ORF72 iMNs exhibited significant mitochondrial Ca 2+ uptake deficits when compared with healthy or isogenic control iMNs, even though iMNs at 1 week do not yet manifest ER stress or GRP75 expression changes (Fig. 3a). Not only were Ca 2+ transients lower within the mitochondria, but also the mitochondrial membrane potential, which is the driving force for Ca 2+ uptake, was significantly reduced when measured simultaneously using ΔΨM probe tetramethylrhodamine methyl ester (TMRM) dye ( Supplementary Fig. 4a, b, online resource). To follow up on this, we evaluated key parameters of mitochondrial function using the Seahorse assay and found reduced basal and maximal respiration, and consequently lower ATP production, suggesting that mitochondrial Ca 2+ uptake deficits observed in 1-week-old C9ORF72 iMNs likely precede the ER stress response (Fig. 3b).
We next measured mitochondrial Ca 2+ uptake in 2-weekold C9ORF72 iMNs, which exhibit ER stress and increased GRP75 expression. Upon Ca 2+ release from the ER, the maximum mitochondrial Ca 2+ uptake and mitochondrial membrane potentials were not significantly altered in these neurons compared with their healthy/isogenic controls ( Fig. 3c and Supplementary Fig. 4c, online resource). The mitochondrial Ca 2+ transients thus detected are a cumulative readout of the release of Ca 2+ from ER stores not only by IP 3 R, but also via ryanodine receptors and sphingolipid Ca 2+ release-mediating proteins of the ER. Therefore, to reveal the contribution of increased GRP75 to mitochondrial Ca 2+ uptake, iMNs were treated with MKT-077, an established pharmacological inhibitor of GRP75 [26], at different concentrations to establish an optimum dosage curve for the inhibition of mitochondrial Ca 2+ uptake specifically in iMNs ( Supplementary Fig. 4d, online resource). 5 µM MKT-077 treatment led to reduced mitochondrial Ca 2+ transients within both healthy and isogenic controls. However, in all C9ORF72 iMNs mitochondrial Ca 2+ transients were significantly impaired (Fig. 3c), suggesting that elevated GRP75 levels likely promoted optimal ER-mitochondrial association, thereby reversing mitochondrial Ca 2+ uptake deficits in C9ORF72 iMNs. Notably, the ameliorated mitochondrial Ca 2+ uptake within 2-week-old iMNs was also reflected by almost normalized mitochondrial respiration and ATP production in C9ORF72 iMNs ( Fig. 3d and Supplementary  Fig. 4e, online resource).
As further evidence for the involvement of ER stressmediated GRP75 increase in normalizing mitochondrial Ca 2+ uptake, we treated 2-week-old iMNs harboring higher Diminished GRP75 expression and mitochondrial impairments coincide with UPR. a Immunolabeling for the ER stress marker BiP in WT and C9-500 spinal cord reveals an increase in BiP intensity in mutant MNs at P60. As early as P60, high BiP expression is observed in ∼30%, which spreads to more than 70% of MNs by P150 and is subsequently downregulated at P200. n = 3 mice/genotype/age. b Immunolabeling for GRP75 and BiP in WT and C9-500 spinal cord sections. Bottom: Q.A. indicates a direct correlation between higher levels of BiP and increased expression intensity of GRP75 in C9-500 MNs at P125: Pearson correlation WT r 2 = 0.01412, P = 0.3167, C9-500 r 2 = 0.6486, P < 0.0001 (n = 73 MNs for WT and 86 MNs for C9-500, from 3 mice/genotype). c Appearance of advanced UPR in mutant MNs as measured by the presence of phosphorylated eIF2α (P i -eIF2α) as from P150 (78% ± 3.403) and P i -PERK. Right: number of ChAT + ve MNs expressing P i -eIF2α analyzed at P80 WT n = 68 and C9-500, n = 73, P150 WT, n = 76 and C9-500, n = 71, P200 WT, n = 192 and C9-500, n = 151). d Representative confocal images of GRP75 and P i -eIF2α immunofluorescence staining, depicting high levels of GRP75 levels in C9-500 animals at P80 when there is no UPR activation, while at P150, when ~ 78% of MNs are positive for P i -eIF2α, GRP75 levels are lower compared to WT.  (Fig. 3e). Sal treatment led to reduced mitochondrial Ca 2+ transients within C9ORF72 iMNs, but had no effect on control/isogenic control iMNs (Fig. 3f). Our data suggest that C9ORF72 iMNs exhibit early mitochondrial impairments, which are neutralized by ER stress-mediated elevated GRP75 expression, suggestive of an early adaptive response crucial for sustaining mitochondrial function.

Reduced GRP75 levels in C9ORF72-ALS/FTD post-mortem tissue and C9-500 rodent neurons
We next examined lumbar spinal cord specimens of four C9ORF72-ALS/FTD and four control cases by immunofluorescence and DAB immunohistochemistry. We found homogeneous, strong cytoplasmic GRP75 immunoreactivity of numerous large and small anterior horn neurons in the control cases. In contrast, a large fraction of the remaining neurons in C9ORF72-ALS/FTD cases showed a considerable reduction in average GRP75 immunoreactivity ( Fig. 4a and Supplementary Fig. 5a, online resource). Many α-MNs in the lumbar spinal cord of the C9ORF72-ALS/FTD patients contained characteristic pTDP-43 aggregates of varying morphology (dash-or dot-like/granular, skein-like, dense/ globular, Supplementary Fig. 5b, online resource), probably depending upon their stage of maturation [52]. We consistently observed that α-MNs harboring large, compact globular or skein-like pTDP-43 aggregates showed reduced cytoplasmic GRP75 immunoreactivity in comparison to the adjacent α-MNs with high levels of GRP75 that were either completely devoid of pTDP-43 aggregates or harbored only minor amounts of small, dispersed, granular pTDP-43 microaggregates. Of note, we also observed rare α-MNs, in which intense GRP75 immunoreactivity coincided with larger amounts of mostly granular pTDP-43 (Fig. 4b, and Supplementary Fig. 5c, online resource, bottom image).
Similarly, an overall reduction in cytoplasmic GRP75 immunoreactivity of hippocampal dentate gyrus neurons in the same C9ORF72-ALS/FTD patients as studied above was observed ( Supplementary Fig. 5d, online resource). Notably, this was prominent in nearly all pTDP-43 aggregatebearing dentate gyrus neurons ( Supplementary Fig. 5d, online resource, red arrows). In line with this observation, PolyGA aggregate-bearing dentate gyrus neurons generally showed reduced GRP75 immunoreactivity in comparison to the adjacent neurons devoid of PolyGA aggregates (Fig. 4c). On the other hand, in those neurons which harbored PolyGA aggregates, the remaining GRP75 immunoreactivity often showed a tendency to preferentially co-localize with PolyGA accumulations (Fig. 4c).
Reprogramming resets the epigenetic age of induced pluripotent stem cells (iPSCs), [49]; therefore, we assume that iPSC-derived neurons used in this study most closely resemble those in "young" mutation carriers. This renders them suitable tools for identifying the molecular mechanism by which "young" MNs might be still able to protect themselves against abnormal ALS proteins. Thus, for the ◂ assessment of GRP75 expression relevant to the chronic course and later stages of the disease, we examined GRP75 levels in the C9-500 BAC mouse model. These mice display RNA foci, pTDP43 aggregates and robust accumulation of DPRs, all of which are pathological hallmarks of C9ORF72-ALS/FTD [9,41]. We first found an early transient increase in GRP75 immunoreactivity in C9-500 spinal cord MNs compared to WT. Notably, this increase was elevated within MN pools, but variable levels of GRP75 expression were observed (Fig. 4d). Interestingly, this increase in GRP75 expression (classified as medium-high to high) was transient from postnatal (P) days 60-125 and abruptly reduced to below WT levels from P150 to beyond 200 days (Fig. 4e). Within the cortex, we found a similar temporary increase in GRP75 expression, albeit slightly shifted in time compared with spinal MNs (Fig. 4f), thus corroborating both our previous observations made in iMNs and post-mortem tissue. Next, we investigated whether the overall decline in GRP75 expression was also reflected by its specific reduction at the MAM. To this end, we isolated MAM fractions from the brain of P240 C9-500 mice. Immunoblotting MAM fractions confirmed lower levels of GRP75 localized at the MAM, indicating that this reduction might negatively affect mitochondrial Ca 2+ uptake and downstream function (Fig. 4g).

Loss of GRP75 expression and mitochondrial dysfunction coincides with the onset of UPR
Since GRP75 expression was modulated by ER stress in C9ORF72-ALS/FTD patient iMNs, we addressed the early phase of ER stress by labeling the C9-500 mouse spinal cord sections with antibodies against the luminal ER protein BiP/GRP78. From P60, a gradual increase in BiP expression in mutant C9-500 rodent MNs was observed. From P80 onward, 66-72% of spinal MNs exhibited signs of ER stress as demonstrated by high BiP levels (Fig. 5a). A similar increase in BiP levels was also observed in the cortex of the mice ( Supplementary Fig. 6, online resource). Moreover, MNs displaying ER stress also exhibited higher expression levels of GRP75, and a significant direct relationship between high BiP levels and augmented GRP75 expression was observed within spinal MNs at P125 (Fig. 5b, bottom).
To better understand the relationship between ER stress and modulation of GRP75 expression, we induced mild ER stress in WT mice for 3 days with TU (0.1 µg/g) IP injections, followed by laser dissection of spinal MNs and qPCR analysis. High levels of BiP and Chop mRNA indicated ongoing ER stress due to TU treatment. Notably, severalfold upregulation of Grp75 transcripts was observed in WT MNs, reconfirming that GRP75 levels are modulated by ER stress in MNs ( Supplementary Fig. 7a, online resource). Elevated immunoreactivities for BiP and GRP75 proteins were observed in TU-treated ventral spinal MNs, and a correlation between high BiP levels and elevated GRP75 levels was present ( Supplementary Fig. 7b, online resource). Interestingly, spinal MNs reacted most strongly to TU-induced ER stress, as both BiP and GRP75 levels in other spinal neurons (ChAT-negative) increased only very moderately (Supplementary Fig. 7c, online resource), again emphasizing the higher vulnerability of spinal MNs to ER stress.
Next, we focused our attention on P150, a time point when an abrupt loss in GRP75 expression is observed. P i -eIF2α immunoreactivity, which was hardly noticeable at P80, was prominent at P150 (78.2 ± 3.4%) and P200 (80.9 ± 4.317), and restricted in expression within ChAT-positive spinal ventral horn MNs (Fig. 5c). Moreover, the increased expression of phosphorylated PERK (P i -PERK) in C9-500 MNs confirmed UPR signaling mediated via the PERK pathway (Fig. 5c, bottom right). Notably, when MNs transited to UPR signaling, as observed in P150 C9-500 mice, those MNs also displayed negligible GRP75 expression and an inverse correlation was observed between higher P i -eIF2α expression and GRP75 levels (Fig. 5d), suggesting an ER stress statedependent modulation of GRP75 expression.
We examined the impact of reduced GRP75 expression on mitochondrial function by a colorimetric assay specifically assessing electron transport chain (ETC) activity. At P80, complex I and IV exhibited normal activity comparable to WT MNs. However, at P150 coinciding with UPR, both complexes were strongly reduced within MNs (Fig. 5e). Further, we confirmed by seahorse measurements deficits in key mitochondrial respiration states within P150 C9-500 lumbar neurons (Fig. 5f). Importantly, the timing of mitochondrial dysfunction coincided with the expression levels of GRP75, which are transiently increased between P60-P125, followed by a significant reduction at P150. As GRP75 physically interacts with both IP 3 R and VDAC1, we assessed this interaction via proximity ligation assay (PLA). While at P30, no difference in the number of IP 3 R-VDAC1 positive puncta was observed in C9-500 MNs compared with WT MNs, a significant increase in IP 3 R-VDAC1 association was observed at P125, followed by a dramatic reduction in interaction at P240 in C9-500 spinal MNs (Fig. 5g). As GRP75 is mainly localized to mitochondria, we evaluated its interaction at the MAM, which it normally achieves by binding to IP 3 R, located on the ER membrane. PLA measurements for GRP75-IP 3 R association revealed a similar time course of interaction as observed for IP 3 R-VDAC1 (Fig. 5g).
We next performed ultrastructural analyses of mitochondria in MNs by SBF-SEM and 3D-reconstruction of the images. Focusing on mitochondrial morphology in both conditions, we found that the WT MN soma presented mainly tubular and elongated mitochondria, whereas such tubular and elongated mitochondria were negligibly present in mutant MNs at P125 (Supplementary Fig. 8a, online resource). Instead, nearly spherical mitochondria were abundantly seen in mutant P125 C9-500 MNs. In fact, C9-500 MNs presented an overall higher number of mitochondria, and, in line with our iMN data (see Fig. 2a, b), a significant percentage of those mitochondria were rounded in shape, when the sphericity of mitochondria was examined at P60 and P125 ( Supplementary Fig. 8b, online resource). In line with our observations in iMNs (Fig. 2d), we found that spherical mitochondria on average exhibited increased contact lengths as well as a higher number of contact points with the ER membrane in mutant MNs (Supplementary Fig. 8c, d, online resource). Analyzing mitochondrial integrity revealed that despite being rounded, mitochondria in P125 C9orf72 MNs presented no increase in abnormal cristae compared to WT MNs. However, by P220, an age coinciding with drastically reduced GRP75 levels, nearly 80% of all mitochondria lacked intact cristae in mutant MNs, thus indicating impairment in mitochondrial integrity and function (Fig. 5h).

PolyGA sequesters GRP75 and impairs its localization at the MAM
Several studies have implicated the link between toxic DPRs, UPR markers, and neurodegeneration [14,68] and recently highlighted the presence of both DPRs and UPR markers in human C9ORF72-ALS/FTD post-mortem brain [22]. Given that our data had revealed that in the human postmortem hippocampus, PolyGA co-localizes with GRP75, indicating the possibility of aberrant GRP75 sequestration by PolyGA, we assessed PolyGA expression in C9-500 MNs. At P150, but not at P80, C9-500 MNs presenting the UPR marker P i -eIF2α also displayed an accumulation of PolyGA (Fig. 6a). Furthermore, as previously shown [11], we found a direct correlation between large PolyGA aggregates (volume > 1 µm 3 ) and the expression of P i -eIF2α in MNs, which became highly significant by P150 in mutant MNs coinciding with widespread UPR in MNs compared with P80, when highest GRP75 expression is observed. Notably, longitudinal measurement of PolyGA aggregate size within ChAT positive neurons revealed that between P150-P200 nearly 64-90% of all MNs harbored large PolyGA aggregates (Fig. 6b). Consequently, we examined the relationship between PolyGA aggregates and reduced GRP75 expression at P150. We found that spinal MNs with large PolyGA aggregates mainly showed lower GRP75 levels when compared to an age (P80) when GRP75 levels are elevated (Fig. 6c, bottom graph). Strikingly, cortical neurons of aged (P240) C9-500 mice exhibited PolyGA aggregates co-localizing with GRP75 (Fig. 6c, right). Similarly, in both human postmortem ALS tissue (Fig. 6d) and C9-500 rodent neurons (Fig. 6c, lower right panel) GRP75 was prone to surround PolyGA aggregates.
To further demonstrate that these observations were not mouse model-specific, we generated a lentiviral construct expressing GFP-tagged PolyGA (LV::PolyGA-GFP) harboring 149 repeats under synaptophysin promoter and injected the virus into the cortical layer V and the hippocampus CA1 region of WT animals (see scheme Fig. 6e). Analysis after 2 months, revealed robust expression of PolyGA both in the layer V upper MNs and in the hippocampus (Supplementary Fig. 9a, online resource). Both cortical and hippocampal CA1 neurons expressing PolyGA aggregates presented higher BiP expression levels compared with non-infected WT neurons, as well as displayed UPR (78.3 ± 5.2% of PolyGA transduced neurons), measured by P i -eIF2α immunoreactivity (Fig. 6f). Notably, PolyGA-expressing neurons were largely devoid of cytoplasmic GRP75 and PolyGA aggregates were surrounded by the remaining GRP75, consistent with sequestration of GRP75 ( Fig. 6g and Supplementary Fig. 9b, online resource). To exclude that changes in BiP or GRP75 expression were due to lentiviral transduction or expression of GFP, cortical neurons were infected with LV::GFP, which did not lead to any significant alteration in GRP75 or BiP expression, suggesting that the observed changes were PolyGA-dependent ( Supplementary Fig. 9c, online resource).

PolyGA impairs GRP75 function at the MAM, thereby compromising mitochondrial Ca 2+ uptake
Drawing from these observations, we assessed the influence of PolyGA on GRP75-mediated normalization of mitochondrial Ca 2+ uptake in iMNs. Because we could not detect PolyGA aggregates in our iMN cultures until 4 weeks of differentiation, we decided to infect iMNs with LV::PolyGA-GFP, LV::PolyPR-GFP, LV::PolyGR-GFP, or LV::GFP (Fig. 7a for experimental design). We transduced 2-week-old iMNs (transduction efficiency of more than 85%) (Fig. 7a), as at this stage, high GRP75 expression levels as well as almost normally functioning mitochondria are present. In line with a previous study [68], only PolyGA, but not PolyPR or PolyGR-transduced iMNs displayed robust UPR signaling via the appearance of P i -eIF2α (Fig. 7b). Notably, PolyGA-expressing iMNs showed significantly reduced GRP75 expression compared with PolyPR or Pol-yGR aggregate-bearing iMNs. Importantly, this reduction was irrespective of whether they were healthy/isogenic controls or C9ORF72-derived iMNs (Fig. 7c). As our data suggested that PolyGA expression negatively impacted GRP75 levels, we subsequently, performed mitochondrial Ca 2+ imaging on LV::PolyGA-GFP transduced iMNs. We could not use LV::GFP as mitochondrial Ca 2+ measurements involve two classical fluorophores measuring simultaneously mitochondrial Ca 2+ uptake as well as its membrane potential. Strikingly, PolyGA expression led to impaired mitochondrial Ca 2+ uptake in healthy iMNs as well as in all C9ORF72 iMN lines compared with non-transduced 1 3 iMNs (Fig. 7d). To understand how PolyGA affects mitochondria at a structural level, SBF-SEM was performed on healthy control iMN lines, which were transduced with LV::PolyGA-GFP. Large PolyGA aggregates harboring various organelle-like structures were observed (yellow box); rounded mitochondria were atypically clustered around the nucleus (red box), with depletion of mitochondria from the cytoplasmic region of the iMN ( Supplementary Fig. 10, online resource), indicating abnormal distribution of mitochondria in PolyGA aggregate bearing neurons.

Overexpression of GRP75 rescues ER stress and mitochondrial deficits
While early ER stress-mediated adaptive responses induce elevated expression of GRP75, thereby aiding in sustaining mitochondrial function, however, chronic ER stress eventually leads to the initiation of apoptosis [25]. Thus, we assessed whether the overexpression of GRP75 would support ER-mitochondrial association, thereby promoting ER and mitochondrial function and decreasing PolyGA load. To this end, neonatal C9-500 mice were infected with AAV6-GRP75 via i.c.v. injection. As the cloning vector had a myc-tag, we probed for exogenous GRP75 expression by immunostainig for myc. A significant fraction (84.6 ± 2.4%) of lumbar MNs were immunoreactive for myc and expressed elevated amounts of GRP75. Further, MAM isolation and immunoblotting for GRP75 revealed higher amounts of GRP75 at the MAM (Fig. 8b). Overexpression of AAV6-GRP75 in mutant MNs from birth on led to enhanced IP 3 R-VDAC1 interactions in GRP75 overexpressing C9-500 MNs. To confirm that increased IP 3 R-VDAC1 association was due to the interaction of AAV6-GRP75 with IP 3 R, we measured Myc-IP 3 R associations via PLA, which were prominently increased in C9-500 MNs. (Fig. 8c). Next, we examined ETC levels as a readout for mitochondrial function, and found that GRP75 overexpression in symptomatic C9-500 MNs restored levels of Complex I and Complex IV compared with age-matched untreated C9-500 MNs (Fig. 8d). Further, to assess the impact of GRP75 overexpression on mitochondrial Ca 2+ transients, we overexpressed GRP75 in cultured cortical neurons from C9-500 mice. As expected, mitochondrial Ca 2+ uptake was strongly impaired in non-transduced cortical neurons, but GRP75 overexpression led to normalized Ca2 + transients (Fig. 8e).
To gain more insights into the ameliorated mitochondrial phenotype at a structural level, SBF-SEM was performed on C9-500 spinal cord at P220, when endogenous GRP75 expression is significantly reduced. C9-500 MNs at P220 presented clusters of rounded mitochondria lacking cristae (red arrows), whereas C9-500 MNs overexpressing GRP75 exhibited both elongated (white arrows) and rounded mitochondria (yellow arrows) at comparable levels (Fig. 8f). Importantly, a significant proportion of mitochondria presented cristae abnormalities or their complete loss in C9-500 MNs when compared with AAV6-GRP75 overexpressing C9-500 MNs (Fig. 8g). We also examined mitochondrial cristae at P125, when higher endogenous GRP75 expression is observed; at this time point, mitochondrial cristae in C9-500 MNs were unchanged compared to WT MNs. Subsequently, we analyzed ER stress and detected normalized BiP levels and absence of UPR signaling in mutant MNs overexpressing GRP75 (Fig. 8h). Moreover, we found that GRP75 overexpression also diminished the number of large PolyGA aggregates in C9-500 MNs (Fig. 8i). Consequently, we determined whether restoring ER and mitochondrial homeostasis via GRP75 overexpression is neuroprotective. Counting MN numbers at P220 in C9-500 spinal cord expressing AAV6-GRP75 revealed unchanged MN numbers compared with WT mice. In comparison, a 30% reduction (P = 0.0024) in MN numbers, indicative of ongoing MN degeneration, was observed in untreated C9-500 spinal cord (Fig. 8j), suggesting that high GRP75 expression provides neuroprotection in C9ORF72-ALS/FTD.

Discussion
Identifying early disease coping mechanisms conferring neuronal resilience is expected to lead to effective treatments for NDs. In this study, employing patient-derived MNs and the C9-500 mouse model of ALS, we found that the transient upregulation of GRP75, a mitochondrial associated membrane (MAM) protein, is an ER stress-induced adaptive response. Mechanistically, this GRP75 increase facilitates ER-mitochondrial coupling, thus sustaining mitochondrial Ca 2+ uptake and efficient function ( Supplementary Fig. 12, online resource). It is well-documented that GRP75 expression is modulated by ER stress, as both BiP and GRP75 genes contain ER stress response element (ERSE) on their promoters, these can simultaneously undergo transcription in response to ER stress signals [38]. Notably, previous studies indicate that the function of GRP75 likely depends upon intrinsic neuronal vulnerabilities and the type of stressor. Under physiological conditions, GRP75 inhibition activates mitochondrial stress responses [7]. However, during pathological disease states in Parkinson's disease, increased GRP75 expression in neuronal cells prevented mitochondrial dysfunction and cell death [8], whereas in human dopaminergic neurons, GRP75 overexpression enhanced the cytotoxic effects of the mitochondrial complex I inhibitor rotenone [31]. Similarly, in cells exhibiting oxidative stress, GRP75 potentiated cell death via mitochondrial Ca 2+ overload [26]. Our results indicate that elevated endogenous expression of GRP75 promotes mitochondrial function via normalized mitochondrial Ca 2+ uptake in C9orf72 neurons (Fig. 8d, e), suggesting a key protective role for GRP75 during the initial phase of the disease. Our data also highlight the notion that "younger" neurons with relatively preserved ER-MAM can utilize GRP75-mediated specific molecular mechanisms to protect themselves, whereas in older neurons, reduced GRP75 expression compromises this ability.
An important finding from our study was the identification of distinct early impairments in C9ORF72 mitochondrial ability to uptake Ca 2+ and of overall reduced ATP levels in mutant MNs (Fig. 3). These impairments occurred before the appearance of pathological hallmarks, suggesting that they might define an intrinsic vulnerability pathway. Conserved mitochondrial functional deficits are observed within C9ORF72 iPSCs, hinting towards C9ORF72 mutation as a key component of this intrinsic vulnerability. In addition, oxidative stress, impaired basal and maximal mitochondrial respiration and the direct binding of toxic PolyGR to Atp5a1, causing mitochondrial functional deficits, have recently been demonstrated in C9ORF72-ALS/ FTD; however, it is not clear how early these deficits manifest in human patients [12,42,48]. We found that PolyGA aggregate-bearing human postmortem C9ORF72-ALS/FTD patient hippocampal dentate gyrus neurons display reduced expression of GRP75. In human C9ORF72-ALS/FTD patient lumbar spinal cord, where PolyGA inclusions are less abundant and pTDP-43 pathology predominates, MNs harboring pTDP-43 aggregates displayed reduced GRP75 immunoreactivity. These findings are interesting considering the variability of TDP-43 aggregate pathology [4][5][6].
Elevated levels of GRP75 together with its co-localization mostly with the small granular, early pTDP-43 aggregates suggest that GRP75 is probably still actively involved in fostering neuronal protein quality control (PQC) mechanisms, counteracting incipient aggregation, whereas GRP75 is probably no longer active in protecting against the large globular or skein-like inclusions because of decline in overall PQC mechanisms. Thus, the observed early compensatory responses mediated by GRP75 may become targets of toxic inclusions, as we discovered that even in healthy iMNs, PolyGA expression likely impaired GRP75 function at the MAM, leading to deficits in mitochondrial Ca 2+ uptake. This is likely to be detrimental to neuronal survival in the long run due to inefficient ATP generation, and thus compromised bioenergetics.
The precise mechanism leading to PolyGA-mediated sequestration of GRP75 remains unclear. GRP75 belongs to the HSP70 family of chaperones based on sequence homology. It is the only member of this protein family mostly localized and functioning within mitochondria [18]. Therefore, we also investigated the physical localization of PolyGA within MAMs and were unable to detect PolyGA biochemically in the MAM fractions (data not shown). Recently, UBQLN2 was implicated in recognizing HSP70 ubiquitination, which facilitated UBQLN2-HSP70-GA complex formation and promoted PolyGA degradation [66], thus, further studies are required to decipher whether ubiquitination is involved in sequestering GRP75.
Several previous studies have also shown the deleterious effects of DPRs other than PolyGA such as PolyGR or PolyPR in mediating ER-mitochondrial toxicity [12,42,58]. Interestingly, in this regard, our data show specific reduction in GRP75 expression in association with PolyGA expression, but not other DPRs. Considering our findings of reduction of GRP75 expression in human C9ORF72 neurons, PolyGA may actively participate in mitochondrial dysfunction via sequestration of GRP75 in C9ORF72-ALS/ FTD. Although this requires further study, it is possible that after translation, GRP75 is transported primarily into the mitochondria, and either inefficient mitochondrial translocation complex or insufficient ATP required for the transport might make GRP75 vulnerable to cytoplasmic sequestration by PolyGA. Alternatively, PolyGA might sequester GRP75 within the cytosol, thereby hindering its proper localization to the mitochondria and, thus diminishing its association at the MAMs.
We found that transient overexpression of GRP75 restored mitochondrial function by normalizing complex I and IV and reduced ER stress in C9-500 spinal MNs. As the ER lacks an internal ATP generation machinery, it crucially relies on the mitochondrial ATP supply, and any mitochondrial ATP deficits coupled with aggregation-prone proteins might negatively influence ER homeostasis. Most likely, ER homeostasis critically depends on a permanent supply of ATP, which is essential for optimal protein folding [24] and for the clearance of aggregated proteins in proteinopathies, such as NDs [21,27]. Because GRP75 enhances ER-mitochondria associations, it should be involved in alleviating ATP deficits, thus promoting the dissipation of adverse effects of ER stress. Notably, overexpression of GRP75 decreased levels of large PolyGA inclusions. These data are highly indicative of an interplay between GRP75 and PolyGA. Nevertheless, this relationship requires further study, as does the mechanism by which GRP75 attenuates ER stress and PolyGA aggregation. In addition, increased expression and co-localization of GRP75 with pTDP-43 might be indicative of GRP75 trying to manage the PQC pathways in those MNs, while decreased expression of GRP75 could be indicative of the failure of such PQC mechanism organized by GRP75. It is also reasonable to speculate that GRP75, a molecular chaperone of the HSP70 family, might also be involved in TDP-43 phase separation, which requires ATPase-dependent activity of HSP70 chaperones [65].
Whether exogenous overexpression of GRP75 rescued disease pathology by primarily restoring ER homeostasis, thereby reducing PolyGA aggregation, or whether restoring mitochondrial function was also a key factor in mitigating PolyGA toxicity, remains to be determined. Taken together, our work indicates that GRP75 serves as an early compensatory response in C9ORF72-ALS/FTD, and its pathological targeting by PolyGA aggregates not only disrupts this adaptive response involving ER-mitochondrial cross talk, but also affects mitochondrial function and neuronal survival.