Sigma-1 receptor maintains ATAD3A as a monomer to inhibit mitochondrial fragmentation at the mitochondria-associated membrane in amyotrophic lateral sclerosis

Organelle contact sites are multifunctional platforms


Introduction
Organelle contacts between the endoplasmic reticulum (ER) and mitochondria, known as the mitochondria-associated membranes (MAMs), are multifunctional platforms for maintaining cellular homeostasis, including cholesterol synthesis, Ca 2+ transfer and autophagy induction (Fujimoto and Hayashi, 2011;Markovinovic et al., 2022;Vance, 2014). MAM integrity is generally compromised in amyotrophic lateral sclerosis (ALS), a fatal neurodegenerative disease characterized by a selective loss of motor neurons (Paillusson et al., 2016). We previously reported that MAM disruption is a general pathomechanism in ALS based on studies employing ALS model mice and the MAM reporter system named MAMtrackers (Sakai et al., 2021;Watanabe et al., 2016). Impaired MAM integrity is involved in other neurological disorders, such as Alzheimer's disease (AD) (Area-Gomez and Schon, 2017), Parkinson's disease (Gomez-Suaga et al., 2018), and Huntington's disease (HD) (Maity et al., 2022). Accumulating evidence has revealed that MAM integrity is essential for mitochondrial homeostasis, especially in mitochondrial dynamics such as fission and fusion. Various proteins associated with mitochondrial dynamics are accumulated at the MAM. For instance, mitofusin-2 (MFN2) , which tethers the ER to mitochondrial outer membranes, interacts with apoptosis regulator protein Bax to promote mitochondrial fusion (Cerqua et al., 2010). Syntaxin-17 (STX17), a SNARE protein crucial for the fusion of autophagosomes and lysosomes, regulates mitochondrial division with dynamin-related protein 1 (Drp1) at the MAM (Arasaki et al., 2015).
Moreover, mitochondrial fission protein 1 (Fis1), an essential factor of mitochondrial fission, is also accumulated at the MAM to form the ER-mitochondria tethering complex with B-cell receptor-associated protein 31 (BAP31) (Iwasawa et al., 2011). Mitochondrial fragmentation associated with mitochondrial dysfunction is observed in the various neurodegenerative diseases (Knott et al., 2008); thus, targeting MAM is a promising strategy to prevent mitochondrial dysfunction.
Sigma-1 receptor (σ1R) is a MAM-specific chaperone protein. Σ1R is best known as a regulator of Ca 2+ transfer from the ER to the mitochondria, which is mediated by inositol triphosphate aggregation, leading to the loss-of-function of σ1R, was observed in the motor neurons of sporadic ALS cases (Prause et al., 2013). Furthermore, σ1R agonist administration extended the survival times of ALS model mice (Mancuso et al., 2012;Ono et al., 2013). These studies demonstrate that σ1R is essential for MAM integrity and that σ1R deficiency is closely associated with ALS pathogenesis.
On the other hand, recent studies revealed that another MAM-associated protein, AAA ATPase domain-containing protein A3 (ATAD3A), is involved in the pathogenesis of neurological diseases such as AD (Zhao et al., 2022) and HD (Zhao et al., 2019). ATAD3A is a nuclear-encoded mitochondrial membrane protein that accumulates at the MAM (Gilquin et al., 2010). ATAD3A reportedly contributes to maintaining mitochondrial DNA stability and supports mitochondrial protein synthesis (He et al., 2012;He et al., 2007;Peralta et al., 2018). Moreover, ATAD3A scaffolds mitochondrial inner membrane structures, including OXPHOS complexes (Arguello et al., 2021). An increased level of oxidative ATAD3A dimer induces Drp1 translocation, leading to mitochondrial fission, and inhibition of Drp1 binding to ATAD3A dimer ameliorates mitochondrial dysfunction in HD model cells or mice (Zhao et al., 2019). Similar observations were also confirmed in AD (Zhao et al., 2022). In these neurodegenerative diseases, mitochondrial fragmentation related to mitochondrial dysfunction was associated with oxidative ATAD3A dimerization. Preventing mitochondrial fragmentation induced by the ATAD3A-Drp1 axis is a potential therapeutic strategy for AD and HD. However, the role of ATAD3A in ALS pathogenesis has yet to be uncovered.
Thus, in this study, we aimed to determine whether the deregulation of ATAD3A dimer/monomer state is involved in ALS pathology and is associated with the loss of σ1R function that causes inherited ALS. Using our MAM reporter system, MAMtracker-Luc, we identified ATAD3A as a novel inducer of the MAM that interacts with σ1R. Indeed, ATAD3A was crucial for MAM induction by σ1R. Moreover, σ1R retained ATAD3A as a monomer, which was associated with mitochondrial fusion in cultured cells. Conversely, in Sigmar1 (a gene coding σ1R)-knockout (Sigmar1 − / − ) or ALS-linked mutant SOD1 transgenic mice, oxidative ATAD3A dimerization was
Genotyping of mice was performed as described previously (Watanabe et al., 2016). C57BL/6J (nTg) mice were obtained from CLEA Japan Inc. (Tokyo, Japan). All the mice were maintained under a standard specific pathogen-free environment (12 h light-dark-cycle; 23 ± 1 ºC; 50 ± 5 % humidity) with free access to food and water. Mice were treated in compliance with the guidelines established by the Institutional Animal Care and Use Committee of Nagoya University. The experiments using genetically modified animals and organisms were approved by the Animal Care and Use Committee and the recombinant DNA experiment committee of Nagoya University.

Quantification of the MAM using MAMtracker-Luc
Neuro2a cells stably expressing MAMtracker-Luc were established, maintained, and measured using a NanoGlo Live cell assay kit (Promega Corp., Madison, WI, USA) as described elsewhere (Sakai et al., 2021). Briefly, the cells were seeded at 1.0 × 10 4 /well before the day of transfection.

Co-immunoprecipitation and immunoblotting
Protein concentration in each fraction was measured using a Bio-Rad Bradford protein assay kit Neuro2a cells seeded on a poly-D-lysine coated 100 mm culture dish were fractionated as previously described (Watanabe et al., 2016). Briefly, the cells were homogenized in isotonic buffer (10 mM HEPES, 250 mM sucrose, pH 7.4) supplemented with cOmplete protease inhibitor and centrifuged at 500×g, 4 ºC for 5 min. The pellet was collected as P1 (nuclei and debris). The supernatant was further centrifuged at 10,300×g, 4 ºC for 20 min. The pellet was collected as crude mitochondria by resuspending in ice-cold resuspension buffer (5 mM HEPES, 250 mM D-mannitol, J o u r n a l P r e -p r o o f Journal Pre-proof collected and ultracentrifuged at 100,000×g, 4 ºC for 1 hour to separate P3 (pellet, microsomes) and cytoplasm, respectively. MAM fraction was obtained from the crude mitochondria overlayed on 30 % Percoll (Cytiva, Tokyo, Japan) solution by centrifugation at 95,000×g, 4 ºC for 30 min.
To detect oxidized ATAD3A dimers, cells or tissues are lysed in ice-cold RIPA buffer supplemented with cOmplete protease inhibitor and PhosSTOP. The lysates were incubated with equal volumes of 2 × SDS-PAGE loading buffer with or without 2.5 %(v/v) 2-ME for 3 min at 95 ºC.

Immunofluorescence
Mice at indicated age were deeply anesthetized and transcardially perfused with phosphate-buffered saline (PBS) following 4 %(w/v) paraformaldehyde (PFA) in 0.1 M phosphate buffer for 10 minutes, respectively. After incubation with 30 %(w/v) sucrose in PBS, dissected lumbar spinal cords were embedded in Tissue-Tek OCT compound medium (Sakura Finetek, Tokyo, Japan) and frozen at −80 ºC until use. HeLa cells were seeded at 5.0× 10 4 /well on 4 well slide chambers (LabTek II from Thermo Fisher Scientific), and were fixed with 4 %(w/v) PFA in 0.1 M phosphate buffer for 15 min at room temperature. For σ1R staining, 12 µm-sliced spinal cord sections or the fixed cells were incubated in 10 mM Tris-HCl (pH 9.5) and 6 M urea at 85 ºC for 10 min (Hayashi et al., 2011). After blocking, the sections or cells were incubated with primary antibodies overnight at 4 ºC. Bound primary antibodies were detected with Alexa Fluor 488 or Alexa Fluor 546-conjugated secondary antibodies at the concentration described above. Images were obtained by confocal laser scanning microscopy (LSM-700; Carl Zeiss AG, Oberkochen, Germany) and the equipped software (Zen; Carl Zeiss). For co-localization analyses, Pearson's correlation coefficient was calculated using the Zen software.

Measurement of mitochondrial morphology
Confocal immunofluorescent images of random 5 square fields consisting of 100 µm on a side taken by LSM-700 were analyzed using Fiji software (Schindelin et al., 2012). Mitochondrial morphology was semi-automatically measured using Mitochondrial Morphology macro (https://imagejdocu.list.lu/plugin/morphology/mitochondrial_morphology_macro_plug-in/) (Dagda et al., 2009). Deconvolution from 5 continuous z-stacked images was performed on the Zen software where indicated.

Measurement of relative intracellular ATP levels
HeLa cells were seeded on a 96 well plate at 1.0 × 10 4 /well, treated with 10 µM PRE-084 or BD-1063 for 24 hours, and then analyzed by an ATP assay kit for cells (TOYO B-net, Tokyo, Japan).
Briefly, after replacing the medium with a fresh one, 100 µL the ATP measurement reagent in the kit was added and stirred for 1 min. After further incubation at 23 ºC for 10 min, luminescence was measured using a plate reader Infinite 200 PRO. The relative intracellular ATP levels were normalized to the DMSO-treated controls.

Statistics
All semiquantitative immunoblotting data, cell viability data, and immunofluorescent images were analyzed by one-way ANOVA followed by multiple comparison tests with Sidak's correction for three or more groups and Welch's t-tests for two groups. All statistical analyzes were performed using GraphPad Prism software (GraphPad Software, La Jolla, CA, USA). No randomization or blinding was used in this study.

ATAD3A is crucial for σ1R-mediated MAM induction
First, to examine whether σ1R and ATAD3A interact with each other, we performed a co-immunoprecipitation assay using mouse neuroblastoma Neuro2a cells. ATAD3A bound to both overexpressed and endogenous σ1R ( Figure 1A & B), suggesting that σ1R and ATAD3A contribute tethering of the ER and mitochondria, which leads to MAM formation. We also tried to examine the interaction between endogenous σ1R and ATAD3A; however, we were not able to immunoprecipitate enough amount of target proteins by anti-σ1R antibody used in this study J o u r n a l P r e -p r o o f Journal Pre-proof (Supplementary Figure S3), possibly due to its low efficiency of the immunoprecipitation. In our previous study, we developed a split luciferase-based MAM reporter, MAMtracker-Luc, which enables high-throughput quantification of the MAM in living cells ( Figure 1C) (Sakai et al., 2021).
The two components of luciferase are anchored to mitochondrial or ER membranes, respectively, and complimented when the ER and mitochondria are in proximity (< 200 nm) at the MAM. Thus, using Neuro2a cells stably expressing MAMtracker-Luc, we examined whether σ1R and ATAD3A require each other for MAM induction. As shown in Figure 1D, σ1R overexpression induced the MAM, which was consistent with previous studies, including ours (Prause et al., 2013;Sakai et al., 2021;Watanabe et al., 2016). Similarly, ATAD3A overexpression induced the MAM. However, the impact on MAM induction was much greater for ATAD3A [1.32 (σ1R) Figure 2B). CC1 or CC2 deletion mutant, i.e., ∆CC1 or ∆CC2, was not co-immunoprecipitated with σ1R ( Figure 2C), indicating that CC1 and CC2 are essential for interacting with σ1R. On the other hand, AAA lacking (∆AAA) mutant maintained its ability to interact with σ1R, indicating that mitochondrial localization of ATAD3A is not necessary for the interaction with σ1R. We confirmed that ∆AAA mutant was not able to localize at the mitochondria as previously reported (Gilquin et al., 2010), whereas ∆CC1 and ∆CC2 mutants retained their J o u r n a l P r e -p r o o f Journal Pre-proof mitochondrial localization ( Figure 2D and E). Consistent with our co-immunoprecipitation results, ∆AAA mutant colocalized with σ1R ( Figure 2F). Furthermore, ∆AAA mutant was mainly fractionated into the MAM fraction, whereas wild-type ATAD3A was predominantly fractionated into the mitochondrial fraction ( Figure 2G). These observations suggest that ∆AAA mutant is unable to induce the MAM, at least partially, due to a loss of its proper localization at mitochondrial membranes.

σ1R and ATAD3A are associated with mitochondrial fragmentation in vitro and in vivo
ATAD3A did not require σ1R for MAM induction, implying that the σ1R and ATAD3A may play another physiological role. In human cervical carcinoma HeLa cells, σ1R partially colocalized with ATAD3A at edges of the mitochondria ( Figure 3A). Similarly, σ1R made contact with the edges of Collectively, these results indicate that σ1R and ATAD3A interaction is associated with mitochondrial fragmentation.

σ1R prevents ATAD3A oxidative dimerization to inhibit mitochondrial fragmentation
Given that oxidative ATAD3A dimerization reportedly induces mitochondrial fragmentation via HeLa cells. Because the oxidative ATAD3A dimer was dominant in HeLa cells, no further increase of the oxidative ATAD3A dimer by σ1R depletion was observed. In contrast to our observations in HeLa cells, oxidative ATAD3A dimerization was almost absent in mouse spinal cord lysates ( Figure   4D), and σ1R deficiency markedly induced oxidative ATAD3A dimerization ( Figure 4E). Total ATAD3A levels were not altered in either HeLa cells or mouse spinal cord lysates ( Figure 4C and F), suggesting that σ1R did not affect total levels of ATAD3A. Taken together, these results indicate that σ1R maintains ATAD3A as a monomer to inhibit mitochondrial fragmentation.

Pharmacological activation of σ1R maintained ATAD3A as a monomer, which is associated with mitochondrial elongation
Next, we examined whether the pharmacological activation or inactivation of σ1R affects the ATAD3A dimerization state and mitochondrial morphology. PRE-084, an agonist of σ1R, induced mitochondrial elongation after 24 hours of the treatment ( Figure 5A-C). On the other hand, BD-1063, an antagonist of σ1R, temporarily elongated the mitochondrial length, which is consistent with the previous report (Bernard-Marissal et al., 2015), but then resulted in mitochondrial fragmentation ( Figure 5A-C). Moreover, PRE-084 increased intracellular ATP levels, whereas BD-1063 decreased (Fig. 5D). Although PRE-084 also affects glycolysis (Motawe et al., 2020), this observation implies the possibility that PRE-084-induced mitochondrial elongation is associated with increased ATP synthesis in the mitochondria. ATAD3A dimer was apparently decreased after 24 hours of the PRE-084 treatment, whereas BD-1063 did not affect the ATAD3A dimer/monomer ratio at all ( Figure 5E and F). Consistent with the above observations, ATAD3A knock-down prevented the mitochondrial elongation induced by PRE-084 ( Fig. 5G-I), suggesting that the pharmacological activation of σ1R leads to mitochondrial elongation via ATAD3A. In contrast, BD-1063-induced mitochondrial fragmentation was independent of ATAD3A (Supplementary Figure S5). Thus, the pharmacological activation of σ1R maintained ATAD3A as a monomer, which is associated with mitochondrial elongation, whereas the pharmacological inactivation of σ1R led to fragmentation of the mitochondria independent of ATAD3A.

ATAD3A monomerization and localization in mitochondria are decreased in SOD1 G93A mice
In our previous studies, we revealed that the MAM is generally disrupted in inherited ALS (Sakai et al., 2021;Watanabe et al., 2016). Therefore, we investigated whether the disrupted J o u r n a l P r e -p r o o f Journal Pre-proof interaction between σ1R and ATAD3A was involved in mitochondrial alternation observed in ALS.
In the spinal motor neurons of SOD1 G93A ALS model mice, we found that the number of σ1R and ATAD3A contacts was decreased ( Figure 6A and B). The levels of the oxidative ATAD3A dimer were also increased ( Figure 6C and D). Simultaneously, the total level of ATAD3A was decreased in SOD1 G93A mouse motor neurons ( Figure 6A, C and E). Similar to the Sigmar1 − / − mice, the mitochondria, indicated by ATAD3A, were highly fragmented ( Figure 6F-H) in SOD1 G93A mice.
These findings are well consistent with our observations in Sigmar1 −/− mice, suggesting that ATAD3A dysregulation is associated with mitochondrial fragmentation in ALS.

Discussion
In the present study, we identified ATAD3A as a novel interactor of σ1R, which is essential for σ1R-dependent MAM induction. Although ATAD3A induced the MAM without σ1R, a deficit of the σ1R-ATAD3A interaction was associated with mitochondrial fragmentation and oxidative ATAD3A dimer formation. Collectively, our findings indicate that σ1R contributes to the homeostasis of the MAM and mitochondria via its interaction with ATAD3A.
σ1R is reportedly crucial for MAM integrity (Fujimoto et al., 2012;Su et al., 2010;Weng et al., 2017). Our findings indicate that σ1R induces tethering of the ER to the mitochondria with the help of ATAD3A; however, because of the limitation of our co-immunoprecipitation studies, we cannot address whether σ1R directly binds to ATAD3A. Recent studies have proposed that σ1R has a topology oriented toward the ER lumen (Lee et al., 2016;Sharma et al., 2021). Thus, it seems reasonable that σ1R and ATAD3A do not directly interact with each other. Because σ1R binds to IP 3 R3 and stabilizes the complex formed with mitochondrial voltage-dependent anion channel 1 (VDAC1) and glucose-regulated protein 75 (GRP-75) (Szabadkai et al., 2006), stabilizing the complex including ATAD3A to tether the ER and mitochondria would be a possible mechanism of the σ1R-mediated MAM induction. Proteomic analyses would help to identify the molecule on the ER that is responsible for forming σ1R-ATAD3A complex in the future. Despite such a limitation of our study, the σ1R-ATAD3A interaction is an interesting mechanism underlying MAM induction. As we have previously reported, in the central nervous system, IP 3 R3 expression is strictly limited to motor neurons in the brain stem and spinal cord (Watanabe et al., 2016). In contrast, σ1R and ATAD3A are expressed throughout the brain. Thus, the σ1R may also contribute to the maintenance of the MAM in neurons lacking IP 3 R3 via ATAD3A.

Journal Pre-proof
To date, various sets of proteins are known to tether the ER and mitochondria, including vesicle-associated membrane protein-associated protein B (VAPB)-regulator of microtubule dynamics protein 3 (RMDN3/PTPIP51) (Stoica et al., 2014), mitofusin-2 (Mfn2) (de Brito and Scorrano, 2009), B-cell receptor-associated protein 31 (BAP31)-mitochondrial fission protein 1 (FIS1) (Iwasawa et al., 2011), and PDZ domain-containing protein 8 (Hirabayashi et al., 2017). The σ1R and ATAD3A form a novel set of proteins comprising a MAM tethering complex. Interestingly, the loss of either σ1R or ATAD3A affects only about 30 % of the total MAM, suggesting that the MAM is an extremely diverse intracellular structure. This notion is supported by our previous electron microscopy observations, which revealed that the MAM is only partially reduced in Sigmar1 −/− or ALS model mice (Watanabe et al., 2016). Future studies focused on the MAM diversity might contribute to revealing specific physiological roles of the MAM dependent on each tethering complex.
Previously, the unfolded PR domain of ATAD3A was predicted to be responsible for interaction at the MAM (Baudier, 2018); however, we found that CC1/2 domains rather than PR domain were required for interaction with the ER and MAM induction. It is still unclear whether the CC1/2 domains of ATAD3A are really located on mitochondrial outer membranes. According to some previous studies, ATAD3A N-terminal domains are localized in intermembrane spaces (Arguello et al., 2021;Baudier, 2018); however, our finding that ∆AAA mutant is localized at the MAM suggests that ATAD3A N-terminal domains can directly bind to ER proteins. Given that ATAD3A N-terminal domain locates in intermembrane spaces when ATAD3A forms an oxidative dimer (Zhao et al., 2019), one possible interpretation is that ATAD3A alters its conformations and/or localization depending on its oligomerization status or interacting partners.
Members of the AAA ATPase family are known to form hexameric ring complexes (Miller and Enemark, 2016) and ATAD3A has been shown to oligomerize without any oxidation, which is crucial for the physiological function of ATAD3A (Miller and Enemark, 2016;Peralta et al., 2018).
Conversely, oxidative ATAD3A dimerization is reportedly involved in HD (Zhao et al., 2019) and AD (Zhao et al., 2022), which leads to mitochondrial dysfunction. Mitochondrial impairment has also been documented in various ALS models (Smith et al., 2019). Indeed, in our previous study, MAM disruption decreased ATP supply from mitochondria (Watanabe et al., 2016). Mitochondrial fragmentation is also a pathological feature observed in ALS motor neurons (Magrané et al., 2014;Vinsant et al., 2013). Moreover, mitochondrial fragmentation is observed in various J o u r n a l P r e -p r o o f Journal Pre-proof neurodegenerative diseases, including AD and HD (Knott et al., 2008). Therefore, the prevention of mitochondrial fragmentation is a possible protective mechanism.
Although the MAM is important for mitochondrial homeostasis, especially in mitochondrial dynamics of fission and fusion, and various proteins associated with the mitochondrial division are accumulated at the MAM, it is unclear how MAM alteration affects mitochondrial function and morphology. In this study, we found that MAM induction by σ1R-ATAD3A axis prevented mitochondrial fragmentation, and σ1R deficit led to MAM deficiency and was associated with mitochondrial fragmentation, which is observed in ALS models. On the other hand, mitochondria are also reportedly fragmented in AD models (Medala et al., 2021), in which the MAM is excessively induced (Area-Gomez et al., 2012). One possible interpretation is that the impact of MAM induction on mitochondrial morphology is dependent on the molecules involved. For example, BAP31 prevents Fis1-dependent Drp1 recruitment (Iwasawa et al., 2011); thus, Fis1 and BAP31-mediated MAM formation prevents mitochondrial fission. On the other hand, STX17-associated MAM promotes mitochondrial fission by recruiting Drp1 (Arasaki et al., 2015). In the case of ATAD3A, ATAD3A dimer induces Drp1 recruitment (Zhao et al., 2019); that is, ATAD3A dimer at the MAM promotes mitochondrial fragmentation. In our observations, σ1R, which induced the MAM, decreased the dimer/monomer ratio of ATAD3A; thus, MAM induction by σ1R-ATAD3A axis promoted mitochondrial elongation rather than fragmentation. Therefore, it is likely that mitochondrial fragmentation in various neurodegenerative diseases is differently associated with MAM alternations specific to each disease.
Another question is how σ1R prevents ATAD3A dimerization. Because the MAM, including σ1R-containing MAM, regulates Ca 2+ transfer into the mitochondria (Raeisossadati and Ferrari,    major axis length (K) of mitochondria were measured to quantify mitochondrial morphology. All the data are expressed as mean ± SEM with p-values, analyzed by one-way ANOVA followed by multiple comparison tests with Sidak's correction for three or more groups and Welch's t-tests for two groups. Scale bar: 20 µm. Representative immunoblotting images of 6 months-old wild-type or Sigmar1 −/− mouse spinal cord lysates are shown in (D). Relative ATAD3A dimer/monomer ratio (E) and total ATAD3A levels (F) were quantified. Data are expressed as mean ± SEM with p-values (n = 3 per group), analyzed by one-way ANOVA followed by Welch's t-tests.