Modulation of synaptic plasticity, motor unit physiology, and TDP-43 pathology by CHCHD10

Mutations in CHCHD10, a gene coding for a mitochondrial intermembrane space protein, are associated with Frontotemporal dementia (FTD)-Amyotrophic lateral sclerosis (ALS) spectrum disorders, which are pathologically characterized by cytoplasmic inclusions containing TDP-43. FTD/ALS-linked CHCHD10 mutations and TDP-43 inclusions similarly induce mitochondrial defects in respiration, fusion/fission, mtDNA stability, and cristae structure, while sizeable amounts of cytoplasmic TDP-43 aggregates are found in mitochondria. However, the mechanistic link between CHCHD10 and TDP-43 pathogenesis remains unclear. In this study, we present immunohistochemical and biochemical evidence demonstrating that insoluble CHCHD10 aggregates accumulate and colocalize with phospho-TDP-43 inclusions in brains of FTLD-TDP and AD patients, and that insoluble CHCHD10 levels tightly correlate with insoluble TDP-43 levels in control and FTLD-TDP brains. In an experimental exploration of this pathological phenotype, transgenic mice neuronally expressing FTD/ALS-linked CHCHD10R15L or CHCHDS59L mutations but not CHCHD10WT transgenic mice exhibit significantly increased CHCHD10 aggregation and phospho-TDP-43 pathology, which often colocalize within the same inclusions. Such pathologies are reflected in poor functional outcomes in long-term synaptic plasticity, motor unit physiology, and behavior in CHCHD10R15L and CHCHDS59L transgenic mice. In contrast, expression of CHCHD10WT in hTDP-43 transgenic mice (TAR4;CHCHD10WT) significantly mitigates phospho-TDP-43 pathology and rescues TDP-43-induced impairments in synaptic integrity and long-term synaptic plasticity. In isolated mitochondria, the S59L mutation induces the aggregation of resident CHCHD10S59L protein as well as the aggregation and slower turnover of recombinant TDP-43 imported into mitochondria. Likewise, in an in vitro cell-free system, the S59L mutation induces the aggregation of CHCHD10S59L protein while simultaneously enhancing the aggregation of recombinant TDP-43, as evidenced by filter trap assays and atomic force microscopy. In contrast, recombinant CHCHD10WT inhibits the growth of TDP-43 aggregates. These results in human brains, transgenic mice, and in vitro systems substantiate the role of wild type and mutant CHCHD10 in modulating mitochondrial CHCHD10 and TDP-43 pathogenesis together with associated phenotypes in long-term synaptic plasticity and motor unit physiology in mice and humans.

The gene Coiled-coil-helix-coiled-coil-helix domain containing 10 (CHCHD10) encodes a small 15-kDa mitochondrial intermembrane space (IMS) protein, in which multiple mutations have been linked to the etiology of both sporadic and familial FTD-ALS spectrum disorders [21][22][23][24]. Previous studies have implicated the role of CHCHD10 in the regulation of mitochondrial respiration as well as maintenance of mitochondrial genome, cristae structure, and mitochondrial fusion [17,21,25,26]. Specifically, CHCHD10 has been identified as a component of the mitochondrial contact site and cristae organizing system (MICOS) at least in part through its interaction with the core protein mitofilin, and patient-derived fibroblasts carrying the CHCHD10 S59L mutation exhibit disrupted MICOS complex and cristae structure [27,28]. Soluble CHCHD10 levels decline in brains of FTLD-TDP patients and a mouse model of TDP-43 pathogenesis [26]. Interestingly, ectopic TDP-43 expression recapitulates the effects of FTD/ALS-linked CHCHD10 mutations in disrupting mitochondrial fusion, cristae integrity, and respiration, whereas CHCHD10 WT rescues these mitochondrial abnormalities induced by TDP-43 [26]. These changes in mitochondrial function, however, do not address whether and how CHCHD10 and TDP-43 proteinopathies per se are mechanistically related.
In this study, we found that insoluble CHCHD10 aggregates accumulate and colocalize with phospho-TDP-43 inclusions in brains of FTLD-TDP and AD patients, and insoluble CHCHD10 levels tightly correlate with insoluble TDP-43 levels in human brains. To experimentally explore this pathological finding, we assessed transgenic mice expressing wild type CHCHD10 WT or FTD/ALSlinked mutants (CHCHD10 R15L & CHCHD S59L ) under the control of the neuron-specific PrP promoter. Herein, we present the first in vivo evidence that the FTD/ALSlinked CHCHD10 R15L and CHCHD10 S59L mutations drive CHCHD10 and TDP-43 proteinopathies originating from mitochondria, which directly mirror the functional changes in long-term synaptic plasticity and motor unit physiology. In contrast, CHCHD10 WT mitigates TDP-43 pathology and rescues TDP-43-induced impairment in long-term synaptic plasticity. In vivo alterations in CHCHD10 and TDP-43 proteinopathies by CHCHD10 variants are recapitulated in primary neurons, cultured cells, isolated mitochondria, and purified recombinant proteins, which collectively indicate that CHCHD10 variants directly regulate the solubility and aggregation of CHCHD10 and TDP-43 in mitochondria.

Ethics approval
All the experiments methods and protocols involving mice in this study were approved by Institutional Animal Care and Use Committee (IACUC), and all methods were performed in accordance with the relevant guidelines and regulations have also been approved by IACUC and Institutional Biosafety Committees (IBC).

Human brain samples
Frozen tissues and floating brain sections from the frontal gyrus of FTLD-TDP, AD, and nondemented controls were obtained from Dr. Allan Levey and Dr. Marla Gearing at Emory Alzheimer's Disease Research Center (ADRC). Pathology-confirmed FTLD-tau, AD, and nondemented control tissues were matched as closely as possible for sex, age, and APOE genotypes during the procurement phase by the Emory ADRC (P50 AG025688) (Additional file 1: Fig. S1).

Mice
Wild type C57BL6, CHCHD10 WT , CHCHD10 R15L , and CHCHD10 S59L mice [26] as well as TAR4 mice expressing human wild type TDP-43 [29] were bred in the C57BL6 background. CHCHD10 variant transgenic mice and CHCHD10 expression profiles have previously been described [26]. Mice were housed together in a SPF facility with 2-4 littermates in sterile plastic cages with pelleted bedding until euthanization by isoflurane overdose and cervical dislocation or transcardial perfusion. Water and food were supplied ad libitum with 12-h light/ dark cycle and provided enrichment with a mouse igloo. Available mouse littermates were allocated to experimental groups based on age and genotypes for specific experiments. Individuals directly conducting studies on mice were blinded to their genotypes. All mice exhibited normal health without overt abnormalities until the time of euthanization. Detailed methods of behavioral testing and electrophysiological (ex vivo & in vivo) studies are described in Supplemental Methods.

Atomic force microscopy
CHCHD10 WT and CHCHD10 S59L recombinant proteins (1 μM) were incubated by themselves or with recombinant full-length human TDP-43 (1 μM) at 37 °C for 8 h with vigorous shaking (1200 rpm), after which samples were subjected to atomic force microscopy (AFM) and filter trap assay (FTA). For AFM imaging, 10 μλ protein solution was dropped on freshly cleaved mica surface. After 20 min adsorption, filtered distilled water was used to gently wash away excessive solution. The mica was incubated at room temperature until dry followed by AFM imaging. The AC-mode of AFM imaging was used on a JPK NanoWizard 4 AFM (Bruker, USA) under ambient conditions. Commercial PPP-NCHAuD cantilevers (Nanosensors, Switzerland) with nominal spring constant of 42 N/m and resonant frequency of 330 kHz were used. Particle size was estimated by measuring the area of high-contrast particles > 10 nm in height using Image J (NIH, Bethesda, MD, USA).

Recombinant proteins
Recombinant human TDP-43 was purchased from R&D Systems (Minneapolis, MN, USA). CHCHD10 WT and CHCHD10 S59L were subcloned into pET22B vector containing the HIS tag, transformed into Rosetta competent cells and grown in a 37 °C shaker at 250 rpm until OD 600 = 0.6-0.9, induce with 1 mM IPTG at 20 °C (250 rpm) for 24 h. Rosetta cells were collected by centrifugation at 4 °C and lysed in lysis buffer (Tris 20 mM, pH8.0, NaCl 150 mM, imidazole 10 mM, Gendepot 1X protease inhibitor cocktail) followed by sonication on ice. Supernatant was collected after centrifugation and shaken for 1 h at 4 °C with GE Healthcare Ni Sepharose. Bound proteins on sepharose were washed 3 times with ice-cold lysis buffer, and recombinant proteins were eluted with ice-cold elution buffer (Tris 20 mM, NaCl 150 mM, 200 mM imidazole), after which proteins were dialyzed in dialysis buffer (Tris 20 mM, NaCl 150 mM, DTT 1 mM) at 4 °C overnight.

Rotarod and grip strength
Mice were individually housed one week prior to behavioral testing. Each mouse was handled for at least 2 min in that week. On the day of behavioral testing, the mice were moved to behavior room and allowed to acclimate for at least 30 min before trials began. For the rotarod test, rotarod machine (Panlab, rotarod Model LE8205) was set with speeds of 4-40 rpm and 600 s acceleration time. Mice were brought into the room in groups of 3-5 for testing. Mice were given two minutes of practice trials at an accelerating speed starting from 4 rpm to learn the behavior. During the four trials with accelerating mode (4-40 rpm in 5 min max), speeds and times to fall were recorded. For the grip strength, the test was performed with the Ugo Basile grip strength meter (Cat.47200). Mice were put on the mesh platform on all 4 legs, and the tail was pulled slowly and gently until the mouse lost grip of the mesh, and peak gram force was recorded.

Ex vivo slice recordings
Brain slices from 10-month-old WT, CHCHD10 WT , CHCHD10 R15L , CHCHD10 S59L , TAR4 and TAR4; D10 WT mice were subjected to input-output (I-O) curves, paired-pulse facilitation (PPF), and long-term potentiation (LTP) as previously described [4]. The stimulating electrode was placed in the Schaffer collaterals of the hippocampus with the ACSF flow rate of 1 ml/min at 30 ± 0.5 °C. The recording glass electrode was positioned at the CA1 stratum radiatum below the pyramidal cell layer and the stimulating pulses were generated by the Digidata 1440 A interface (Molecular Devices, Sunnyvale, CA, USA) and a stimulus isolator (model 2200; A-M Systems, Sequim, WA, USA) controlled by Clampex 10.6 software (Molecular Devices). Field excitatory postsynaptic potentials (fEPSPs) were amplified using a differential amplifier (model 1800; A-M Systems), filtered at 1 kHz, and digitized at 10 kHz. I-O curves were generated by stepping stimulation amplitude that elicited halfmaximal fEPSP from 1 to 15 mV at the rate of 0.05 Hz. PPF was evoked by two pulses with interpulse intervals from 20 to 300 ms. Percentage of the facilitation was calculated by dividing fEPSP slope elicited by the second pulse with the fEPSP slope elicited by the first pulse. LTP was induced by theta burst stimulation (five trains of four pulses at 200 Hz separated by 200 ms, repeated six times with an inter-train interval of 10 s) and sampled for 60 min after the induction. LTP was calculated by dividing the slope of 60 min post-induction responses with the average slope of 20 min baseline responses.

In vivo sciatic nerve motor unit recordings
Mice were anesthetized with inhaled isoflurane and placed in the prone position on a thermostatic warming plate. The ring of recording electrode was placed on the skin overlying the proximal portion of the gastrocnemius muscle of the hind limb. The ring of reference electrode was placed on the skin over mid-metatarsal portion of the foot. The distance between these two ring electrodes was 10 mm. Electrode gel was applied to the skin underlying the ring electrodes to reduce impedance and maximize ring electrode-skin contact. The stimulating electrodes were two 27G monopolar needle electrodes. The cathode monopolar electrode was placed close to the sciatic never at the proximal thigh, and the anode was inserted about 10 mm proximally in the subcutaneous tissue overlying the sacrum. A monopolar needle electrode was placed on the contralateral hind limb or tail as the ground electrode. The distance between the stimulating and recording electrodes was 12-16 mm. A square-wave pulse with 0.1 ms duration and1-10 mA intensity was delivered to induce compound muscle axon potential (CMAP). Stimulus intensity was increased until the CMAP was maximized, which represents the total output from a muscle group. The constant-current stimulus of 0.1 ms duration was sent to evoke a motor response in 0.025 mA incremental steps at 0.1 Hz until a small all-or-none response was evoked. Each incremental response is stable and should be visually distinct and larger compared with the preceding response. The stimulus intensity was increased until the CMAP increased in a quantal fashion. This process was repeated for a total of 10 increments. Individual motor unit amplitudes were obtained by subtracting the amplitude of each response from that of the prior response.
The individual values were averaged to yield the single motor unit potential (SMUP Fractionation of isolated mitochondria was performed using a previously documented method [5] with minor modifications. Briefly, isolated mitochondria were washed with swelling buffer (10 mM KH2PO4, pH 7.4) followed by resuspension in swelling buffer for 20 min at 4 °C. Then the equal volume of shrinking buffer (10 mM HEPES, pH 7.4, 32% sucrose, 30% glycerol, 10 mM MgCl2) was added and resuspended for 15 min followed by centrifugation at 10,000 g for 15 min at 4 °C. Both pellet (mitoplast) including inner membrane (IM) and matrix, and supernatant including outer membrane (OM) and intermembrane space (IMS) fractions were obtained. The supernatant was centrifuged at 100,000 g for 40 min at 4 °C to separate OM and IMS fractions. The pellet (mitoplast) was washed twice with washing buffer (250 mM sucrose, 1 mM EGTA, 10 mM HEPES, pH 7.4), resuspended in swelling buffer, sonicated and subjected in centrifugation at 12,000 g for 15 min. The supernatant was centrifuged at 100,000g for 40 min at 4 °C to separate IM and matrix fractions.

Immunocytochemistry, immunohistochemistry, and fluorescent microscopy
For immunocytochemistry (ICC), primary neurons were washed with PBS and fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature. Fixed cells were washed with PBS and incubated with blocking solution (0.2% Triton X-100, 3% normal goat serum) for 1 h, incubated with primary antibody at 4 °C overnight, washed with PBST 3X, and incubated with Alexa-488 or Alexa-594 conjugated secondary IgG antibodies for 1 h at room temperature (Vector Laboratories, Burlingame, CA). Slides were then washed three times with PBST and mounted with fluorochrome mounting solution (Vector Laboratories). For immunohistochemistry (IHC), mice were perfused with PBS, and half brains/spinal cords were immediately stored at − 80℃ for biochemical analysis, and the other half brain and lumbar spinal cord were fixed with 4% paraformaldehyde at 4 ℃ for 24 h followed by cryoprotection in 30% sucrose. Prior to staining for CHCHD10, Flag-CHCHD10 (M2), mitofilin, TDP-43, or pTDP-43, antigen retrieval (10 mM sodium citrate, 0.05% Tween 20, Ph 6.0) was performed for 5 min at 95 ℃. Thirty-micron sections were blocked using normal goat serum for 1 h and subjected to primary antibodies at 4 ℃ overnight, followed by 1 h incubation with secondary antibody (Alexa-594 and Alexa-488) at room temperature prior to mounting as previously described [31]. For IHC experiments using human brain tissue, post-treatment with TrueBlack (Biotium, Fremont, CA) was performed prior to mounting. Briefly, tissues were incubated with 1X TrueBlack (diluted with 70%) ethanol for 1 min and washed with PBS three times. And then the tissues were mounted to the slides for imaging. Olympus FV10i confocal microscope (Tokyo, Japan) or ZEISS LSM880 confocal microscope was used to capture all the images, and the ImageJ software (NIH, Bethesda, MD) was used to quantify the immunoreactivities. In ICC and IHC experiments, all comparison images were acquired with identical laser intensity, exposure time, and filter. Regions of interest were chosen randomly, and investigators were blinded to genotypes of mice and experimental conditions during image acquisition and quantification. Adjustments to the brightness/contrast were applied equally to all comparison images. Intensities for synaptophysin from hippocampal CA3 and TDP-43 and pS409/410-TDP-43 from the frontal cortex were quantified using ImageJ (NIH, Bethesda, MD). Colocalizations were measured with Image J via the JACoP plugin for Manders overlap coefficient (MOC). The same thresholds for each channel were used when analyzing the coefficient.

Statistical analysis
All graphs were analyzed and made using GraphPad Prism 6.0 software (GraphPad Software, San Diego, CA, USA) using student's t-test, 1-way analysis of variance (ANOVA) followed by Tukey post hoc test, 2-way ANOVA followed by Tukey post hoc test, or linear regression and Pearson's correlation as described in figure legends. Differences were deemed significant when P < 0.05. All graphs were expressed as mean ± S.E.M (error bars).

FTD/ALS-linked CHCHD10 mutants drive CHCHD10 and TDP-43 pathogenesis in the CNS
Based on the preceding observations in post-mortem human brains, we next tested the hypothesis that FTD/ALS-linked CHCHD10 mutants, which themselves are prone to aggregation [32,33], can drive both CHCHD10 and TDP-43 proteinopathies. For this purpose, we utilized wild type (WT) and CHCHD10 transgenic mouse variants, in which N-terminally Flag-tagged wild type (CHCHD10 WT ) or FTD/ ALS-linked mutants (CHCHD10 R15L or CHCHD10 S59L ) are expressed in the central nervous system (CNS) under control of the neuron-specific mouse PrP promoter [26]. And majority of cells overexpressing CHCHD10 WT , CHCHD10 R15L or CHCHD10 S59L in these mice brain are neurons (Additional file 2: Fig. S2c). These mice (D10-WT, R15L, or S59L) express CHCHD10 transcripts at 150-180% above the endogenous mouse CHCHD10 transcript, and human CHCHD10 transgenes are expressed throughout all regions of the brain and spinal cord [26]. We have previously shown that Flag-CHCHD10 localizes nearly exclusively to mitochondria like endogenous CHCHD10 [17]. As endogenous CHCHD10 is known to be targeted primarily to the mitochondrial intermembrane space (IMS) via the Mia40 pathway [34], we tested the sub-mitochondrial localization of Flag-CHCHD10 by performing mitochondrial isolation followed by submitochondrial fractionation to separate the outer membrane (OM), intermembrane space (IMS), inner membrane (IM), and matrix from 3 Flag-CHCHD10 WT transgenic mice and a littermate WT mouse. Indeed, the vast majority of Flag-CHCHD10 (~ 72%) localized to the intermembrane space (IMS) as expected, demonstrating correct targeting and submitochondrial localization of Flag-CHCHD10 (Additional file 2: Fig. S2a,b).
To characterize pathologies in the CNS, we first examined the pattern of anti-CHCHD10 staining in the frontal cortex of 10-month-old WT, CHCHD10 WT , CHCHD10 R15L , and CHCHD10 S59L mice by immunohistochemistry (IHC). Immunofluorescence images showed that CHCHD10 R15L and CHCHD10 S59L mice frequently exhibit strong aggregate-like CHCHD10 staining in neuronal soma and punctuated staining in the neuropil (Fig. 2a). CHCHD10 WT and non-transgenic WT brains, however, exhibited largely punctuated CHCHD10 staining pattern both in the soma and neuropil (Fig. 2a; Additional file 2: Fig. S2d, control). Specifically, 19% of CHCHD10 R15L and 25% of CHCHD10 S59L expressing neurons in the frontal cortex exhibited strong CHCHD10 aggregate-like staining compared to < 5% in CHCHD10 WT containing neurons (Fig. 2b), despite more human Flag-CHCHD10 protein expression in CHCHD10 WT mice compared to CHCHD10 R15L and CHCHD10 S59L mice as seen by Western blotting (Additional file 2: Fig. S2e). Aggregate-like staining of CHCHD10 in CHCHD10 R15L and CHCHD10 S59L mice are reminiscent of the recently reported aggregation of the mouse CHCHD10 S55L knock-in mutant protein in vivo [33,35].

Long-term synaptic plasticity and motor unit deficits in CHCHD10 R15L and CHCHD10 S59L transgenic mice
Mitochondria play vital roles in synaptic plasticity and are enriched at both sides of the synapse [38][39][40]. To assess whether CHCHD10 is biochemically enriched in synapses in vivo, we isolated synaptosome versus non-synaptosome fractions from brains of WT, CHCHD10 WT , and CHCHD10 R15L mice. Both endogenous CHCHD10 and Flag-CHCHD10 variants were overwhelmingly localized to the synaptosome fraction enriched in synaptophysin, GluR2, and PSD95, despite equivalent levels of actin in the non-synaptosome fraction (Fig. 3a). As pathogenic animal models of TDP-43 demonstrate loss of synaptic integrity [17,41,42], we assessed synaptic integrity in 10-month-old CHCHD10 transgenic mouse variants and non-transgenic WT littermates by staining for synaptophysin. Examination of the CA3 region of the hippocampus showed that CHCHD10 R15L and CHCHD10 S59L brains exhibit significantly reduced immunoreactivity for synaptophysin compared to non-transgenic WT mice (Fig. 3b,c). In contrast, CHCHD10 WT mice showed significant increases in synaptophysin immunoreactivity compared to WT mice as well as CHCHD10 R15L and CHCHD10 S59L mice (Fig. 3b,c; Additional file 3: Fig. S3a, control). Assessment of astrogliosis by GFAP staining showed that CHCHD10 R15L and CHCHD10 S59L brains exhibit significantly increased GFAP immunoreactivity compared to CHCHD10 WT or WT mice at 10 months of age (Additional file 3: Fig.  S3b,d). However, Iba1 staining to assess microgliosis did not reveal significant differences among WT mice and CHCHD10 variants (Additional file 3: Fig. S3c,e).
To determine functional changes in synaptic plasticity, we next subjected 10-month-old mouse acute brain slices to short-term and long-term synaptic plasticity measures, in which the stimulating electrode was placed in the Schaffer collaterals (SC) of the hippocampus and the recording electrode at the CA1 stratum radiatum below the pyramidal cell layer (Fig. 3d) [43]. Input-output (I-O) curves of the field excitatory postsynaptic potential (fEPSP) as a function of stepwise increase in stimulus intensity did not reveal significant differences among WT mice and CHCHD10 transgenic variants, indicating that basal synaptic transmission is not impacted by CHCHD10 transgenes (Fig. 3e). We next measured paired-pulse facilitation (PPF), a measure of presynaptic efficacy. While CHCHD10 WT mice exhibited significantly increased fEPSP compared to WT mice at the 20 ms interstimulus interval condition, no differences among genotypes (WT, CHCHD10 WT , CHCHD10 R15L , & CHCHD10 S59L ) were seen from 40 to 300 ms interstimulus intervals (Fig. 3f ), indicating that CHCHD10 variants largely do not alter presynaptic facilitation. Finally, we measured basal fEPSP for 20 min followed by theta burst stimulation to elicit long-term potentiation (LTP). Both the induction and maintenance of LTP for 60 min were robust in WT and CHCHD10 WT slices, which did not differ from each other (Fig. 3g). However, CHCHD10 R15L and CHCHD10 S59L slices exhibited significant impairments in LTP throughout the course of 60 min compared to WT or CHCHD10 WT slices (Fig. 3g).
TDP-43 aggregates captured using a human TDP-43-specific antibody in FTA was significantly reduced by ~ 60% in TAR4;D10 WT versus TAR4 extracts in both RIPA-soluble and RIPA-insoluble fractions (Fig. 5c,d). As expected, no human TDP-43 aggregates were captured in WT littermate brain extracts (Fig. 5c). Staining for human TDP-43 demonstrated largely nuclear localization but also some cytoplasmic mislocalization in TAR4 mice (Fig. 5e), the latter which was significantly reduced by > 50% in TAR4;D10 WT frontal cortex (Fig. 5e,f; Additional file 5:  S5a, control) and TAR4;D10 WT lumbar spinal cords compared to those of TAR4 mice (Additional file 5: Fig.  S5c,d). We also stained for phospho-S409/S410-TDP-43, which was localized to the cytoplasm of TAR4 mice both in the frontal cortex ( Fig. 5g; Additional file 5: Fig.  S5b, control) and lumbar spinal cords (Additional file 5: Fig. S5e,f ). Phospho-TDP-43 staining was significantly reduced by 64% in TAR4;D10 WT frontal cortex (Fig. 5g,h) and ~ 50% in TAR4;D10 WT lumbar spinal cord (Additional file 5: Fig. S5e,f ) compared to those in TAR4 mice. Consistent with phospho-TDP-43 staining, Western blotting of cortical brain extracts confirmed the significant reduction of phospho-TDP-43 in TAR4;D10 WT mice compared to TAR4 littermates (Fig. 5i,j). To determine if CHCHD10 alters the sub-mitochondrial accumulation of TDP-43, we isolated mitochondria from TAR4 and TAR4;D10 WT mice and fractionated the outer membrane (OM), intermembrane space, (IMS), inner membrane (IM), and matrix. As shown in preceding results, Flag-CHCHD10 WT was highly enriched in the IMS [50] with minor localization in the matrix (Fig. 5k). TDP-43 was equally enriched in the IMS and matrix in TAR4 mitochondria (Fig. 5k). In TAR4;D10 WT mitochondria, however, TDP-43 was generally reduced in mitochondria and absent in fractions where Flag-CHCHD10 WT was most enriched (IMS & matrix) (Fig. 5k), consistent with our preceding findings. These results collectively demonstrate that restoring or elevating CHCHD10 WT expression mitigates TDP-43 pathology in vivo. Finally, we examined the functional role of CHCHD10 on synaptic integrity and synaptic plasticity in 10-monthold WT, TAR4, and TAR4;10 WT littermates. Hemizygous TAR4 mice exhibited significantly reduced immunoreactivity for synaptic markers synaptophysin in the CA3 region of the hippocampus compared to WT mice, which was completely rescued to or beyond WT levels in TAR4;D10 WT mice (Fig. 6a,b). In acute brain slice recordings from 10-month-old littermate mice, we observed little to no alterations in input-output (I-O) curves among the genotypes (Additional file 6: Fig. S6). Although no significance differences were detected in PPF among the genotypes, TAR4;D10 WT mice tended to exhibit higher fEPSP slope than WT or TAR4 mice at all interstimulus intervals (Fig. 6c). TAR4 brain slices exhibited significantly impaired LTP after theta burst stimulation compared to WT mice at all time points (Fig. 6d). Such impairment in LTP was completely rescued by CHCHD10 expression in TAR4;D10 WT mice to levels indistinguishable from WT slices (Fig. 6d), consistent with the protective effects of CHCHD10 WT against TDP-43 pathogenesis.

Mutant CHCHD10 promotes mitochondrial TDP-43 accumulation and aggregation
TDP-43 is readily imported into mitochondria [16] and misfolded or aggregation-prone proteins in the cytosol are frequently imported into mitochondria to be degraded, which otherwise could overwhelm the cytosolic proteostasis machinery [51]. As CHCHD10 is a mitochondria-resident protein localized to the intermembrane space [50] and physically interacts with TDP-43 [17], we hypothesized that FTD/ALS-linked CHCHD10 mutations promote the accumulation of misfolded TDP-43 upon mislocalization to the cytoplasm and import into mitochondria. To test the hypothesis, we isolated intact mitochondria from HEK293T cells transfected with Flag-CHCHD10 WT or Flag-CHCHD10 S59L . We then exposed mitochondria to human recombinant TDP-43 to perform mitochondrial import for 1 h at 33 °C, after which mitochondria were spun to remove unimported hTDP-43 (Fig. 7a). Resuspended mitochondria were then returned to 33 °C for 0-12 h. After each time point, mitochondria were collected by centrifugation, lysed with RIPA, and soluble supernatant and insoluble pellet were collected (Fig. 7a). Each fraction was then subjected to Western blotting and/or filter-trap assay (FTA). Western blotting for imported TDP-43 showed that CHCHD10 S59L -containing mitochondria exhibit significantly slower turnover of TDP-43 over 12 h compared to CHCHD10 WT containing mitochondria (Fig. 7b,c). Interestingly, the C-terminal 35-kDa fragment of TDP-43 was also detected in mitochondria, which exhibited similar turnover rates as full-length TDP-43 (Fig. 7b). This was accompanied by the expected accumulation of insoluble CHCHD10 S59L versus CHCHD10 WT in the pellet fraction (Fig. 7b, lower panel) and significantly increased capture of TDP-43 aggregates in CHCHD10 S59L -containing insoluble mitochondrial pellet by FTA (Fig. 7d,e). This insoluble aggregated pool of TDP-43 was seen early after import and was relatively stable over the 12 h period (Fig. 7d,e), which likely accounts for the slower turnover of imported TDP-43 in CHCHD10 S59L -containing mitochondria. Little to no TDP-43 was captured in the absence of recombinant TDP-43 import (Fig. 7d: time 0'), demonstrating the specificity of the FTA. In support of these findings, biochemically isolated mitochondria from CHCHD10 R15L and CHCHD10 S59L mouse brains exhibited significantly more endogenous TDP-43 compared to mitochondria from CHCHD10 WT mouse brains (Additional file 7: Fig S7a,b). Furthermore, in CHCHD10 transgenic primary neurons transduced with TDP-43-tomato AAV9, CHCHD10 R15L and CHCHD10 S59L neurons demonstrated significantly increased TDP-43 cytoplasmic mislocalization and colocalization with the mitochondrial marker Tom20 than CHCHD10 WT and WT neurons (Additional file 7: Fig. S7c, d).
Co-incubation of TDP-43 with CHCHD10 S59L but not CHCHD10 WT strongly increased TDP-43 aggregation as detected by FTA (Fig. 7g). Visualization of protein aggregates by AFM showed the presence of small oligomer-like CHCHD10 S59L aggregates but little to no CHCHD10 WT aggregates when incubated alone (Fig. 7h). TDP-43 alone formed high-contrast aggregates indicative of increased particle height, which appeared to arise from the low-contrast material of low particle height (Fig. 7i, left panel). Co-incubation of TDP-43 with CHCHD10 S59L altered the morphology of the particles, such that the high-height particles significantly increased in area (large spheroid morphology) at the expense of the low-height material (Fig. 7i, middle panel), reminiscent of the buoyant FTLD-TDP 'type A' aggregates [52]. Co-incubation of TDP-43 with CHCHD10 WT , however, significantly altered by morphology of protein particles by decreasing high-height particles in favor of low-height material (Fig. 7i, right panel). Quantification of high-height particles > 10 nm showed that CHCHD10 S59L significantly increases the area of high-height particle by > threefold, whereas CHCHD10 WT significantly decreases the areas of high-height particle by > 50% (Fig. 7j), clearly indicating that CHCHD10 S59L increases the size of aggregates whereas CHCHD10 WT reduces the growth of aggregates. Three-dimensional reconstruction of representative AFM images illustrates the typical height and area of the particles corresponding to each condition (Fig. 7k). These results taken together indicate that CHCHD10 S59L increases TDP-43 aggregation upon import into mitochondria, likely by directly co-aggregating with TDP-43, whereas the direct interaction of CHCHD10 WT with TDP-43 inhibits TDP-43 aggregation.

Discussion
TDP-43 inclusions are pathologically evident in more than 95% of ALS, ~ 50% of FTD (FTLD-TDP variant), and > 50% of late-onset AD [4,8,9]. As our earlier study found that detergent soluble CHCHD10 levels significantly decline in brains of FTLD-TDP patients and TDP-43 transgenic mice, we initially looked for changes  These changes mirrored the functional impairments in long-term synaptic plasticity in brain as well as deficits in motor neuron unit function, sciatic nerve action potential velocity, grip strength, and rotarod performance seen in CHCHD10 R15L and CHCHD10 S59L mice. As endogenous soluble CHCHD10 levels decline significantly in brains of FTLD-TDP patients and TDP-43 transgenic mice [26] and CHCHD10 WT inhibits TDP-43 aggregation in vitro, we also showed that restoring CHCHD10 WT in TDP-43 transgenic mice (TAR4;D10 WT ) mitigates TDP-43 pathology and rescues TDP-43-induced deficits in long-term synaptic plasticity in vivo. These results overall implicate CHCHD10-mediated modulation of TDP-43 aggregation in mitochondria as a significant factor contributing to impairments in long-term synaptic plasticity and motor unit function. Mitochondrial targeting of misfolded or aggregationprone proteins from the cytosol has recently been proposed as a pathway to handle and turnover misfolded proteins within mitochondria, which otherwise would overwhelm the cytosolic proteostasis machinery [51]. This mitochondria-mediated proteostasis mechanism was initially discovered in yeast and named 'mitochondria as guardian in cytosol' (MAGIC) [51]. While previous studies have found TDP-43 to be translocated into mitochondria with ALS-linked TDP-43 mutations exhibiting greater translocation efficiency [16], this study for the first time showed that TDP-43 is subject to CHCHD10-regulated aggregation and turnover upon import into mitochondria. Our observation that isolated mitochondria containing CHCHD10 S59L significantly increases TDP-43 aggregation upon import is consistent with the accumulation of TDP-43 in mitochondria in CHCHD10 R15L and CHCHD10 S59L transgenic primary neurons and brains in vivo. These results are also consistent with a recent study showing that the CHCHD10 S59L mutation increases insoluble mitochondrial TDP-43 in transfected HeLa cells [53]. While the authors interpreted this to indicate increased mitochondrial translocation of TDP-43 [53], our results instead indicate that CHCHD10 mutations enhance the aggregation of TDP-43 upon its translocation into mitochondria, likely via co-aggregation of both proteins. This is an important distinction, as our mechanistic model predicts that sufficient mitochondrial translocation of TDP-43 is a necessary secondary trigger allowing mutant or misfolded CHCHD10 to exacerbate TDP-43 pathology. It is notable that TDP-43 inclusions were not observed in the only autopsied patient carrying a CHCHD10 mutation to date (p.R15L) [54]. In this patient, lack of the TDP-43 pathology component due to insufficient trigger of mitochondrial TDP-43 translocation may underlie the very slow progression of disease (19 years to death) and atypical ALS symptoms [54]. In this light, it would be highly informative to correlate disease progression with the severity of TDP-43 pathology in future autopsy studies of FTD-ALS patients carrying CHCHD10 mutations. Our findings from FTLD-TDP and AD cases showing robust colocalization phospho-TDP-43 pathology with CHCHD10 inclusions (~ 75-84%) and tight correlation between insoluble CHCHD10 with insoluble TDP-43 indeed indicates a pathological link between them in human neurodegenerative diseases. Beyond TDP-43-related spectrum disorders, our mechanistic model may also help to shed light on the diversity of disease phenotypes associated with CHCHD10 mutations, such as mitochondrial myopathy [55], spinal muscular atrophy (SMA) [56], and Charcot-Marie-Tooth disease type 2 (CMT2) [57]. In such cases, secondary triggers, if any, may represent different aggregation-prone proteins expressed in disease-relevant tissues that may also be subject to CHCHD10-regulated proteostasis and aggregation.
The propensity for mutant CHCHD10 to misfold and aggregate per se represents a gain-of-function molecular phenotypes intrinsic to its protein structure [33,35]. We speculate that such misfolding could also be induced in CHCHD10 WT under pathological conditions, perhaps through excessive mitochondrial translocation of TDP-43 (i.e. FTLD-TDP or AD). In this scenario, the loss of functional CHCHD10 or gain of misfolded CHCHD10 fails to oppose or directly promotes TDP-43 aggregation, respectively, thereby generating a maladaptive feed-forward cycle of mitochondrial proteinopathy. This notion is supported by the observation that insoluble TDP-43 positively correlates with insoluble CHCHD10 in human brains. Accumulation of TDP-43 and mutant CHCHD10 in mitochondria are damaging to mitochondria and neurons that house them [17,26], and blockade of mitochondrial TDP-43 translocation also significantly attenuates CHCHD10 S59L -mediated toxicity [53]. Hence, aggregateharboring mitochondria likely also contribute to cytosolic inclusions upon mitochondrial damage and release of the aggregate content to the cytosol (Fig. 8).
The CHCHD10 R15L and CHCHD10 S59L transgenic models used in this study faithfully recapitulated many pathological phenotypes associated with FTD-ALS. These included CHCHD10 pathology and increased pTDP-43 levels associated with deficits in synaptic integrity, long-term synaptic plasticity, motor unit physiology, sciatic nerve action potential velocity, and motor behavior. Deficits in MUNE indicated loss of the functional connectivity between motor neurons and muscle, one of the key features of motor neuron disease [44][45][46]. Significantly slower sciatic nerve action potential velocity indicated the presence of axonal pathology, consistent with axonal pathology observed in the BAC CHCHD10 R15L mice [32]. These electrophysiological measures were mirrored in impaired rotarod performance and grip strength in CHCHD10 R15L and CHCHD10 S59L mice. As mitochondrial health and transport is crucial to sustain synaptic function at both sides of the synapse [38][39][40], deficits in long-term synaptic plasticity in the form of LTP in CHCHD10 R15L and CHCHD10 S59L mice indicated reduced ability to sustain synaptic efficacy, a phenotype tied to learning and memory [58]. It is notable that restoration of CHCHD10 WT in TDP-43 transgenic mice (TAR4) fully rescued synaptic integrity and LTP deficits induced by TDP-43, which was underpinned by significantly reduced TDP-43 pathology in TAR4;D10 WT mice. These transgenic models, which express CHCHD10 variants under the control of neuronspecific mouse PrP promoter and correctly target to the IMS of mitochondria, therefore, represent useful tools to study CHCHD10 function and dysfunction originating from neurons of the CNS. Other reported models such as the CHCHD10 S55L knock-in mouse model driven by the endogenous mouse CHCHD10 promoter [33,35] and the BAC transgenic mice driving CHCHD10 R15L expression with human CHCHD10 promoter [32] exhibit In healthy neurons, small amounts of misfolded proteins, such as TDP-43, accumulate in the cytoplasm, some of which are imported into mitochondria. Mitochondria containing a healthy level of soluble CHCHD10 WT inhibits TDP-43 insolubility and aggregation, thereby sustaining mitochondrial health. Healthy mitochondria at the synapse facilitates long-term synaptic plasticity and motor unit function. In FTLD-TDP brains, soluble CHCHD10 declines and instead deposits as insoluble CHCHD10 aggregates with phospho-TDP-43. Mechanistically, in diseased neurons, excessive amounts of misfolded proteins, such as TDP-43 or phospho-TDP-43, accumulate in the cytoplasm, some of which are imported into mitochondria. Mitochondria containing mutant and/or misfolded CHCHD10 enhance both CHCHD10 and TDP-43 aggregation, thereby driving mitochondrial proteinopathy and dysfunction. Aggregate-harboring mitochondria may contribute to cytoplasmic inclusions upon mitochondrial damage and release of the aggregate content to the cytosol. At the synapse, proteopathic mitochondria become dysfunctional and fail to support long-term synaptic plasticity or motor unit function