TDP-43 oxidation and PP1 crosstalk at RNA granule-mitochondria contact sites

Abstract

Inter-organelle contact sites are key hubs for organelle bidirectional crosstalk. However, how mitochondria and RNA granules interact at contact sites and its regulation by mitochondrial oxidative phosphorylation (OXPHOS) remain unclear. Here, using Super-Resolution live microscopy, we identify RNA granule-mitochondria contact site formation in OXPHOS conditions. Reactive oxygen species (ROS) generated by mitochondrial OXPHOS promotes TDP-43 localization to cytoplasmic RNA granules via TDP-43 cysteine oxidation at Cys173/Cys175. Mechanistically, RNA granule-mitochondria contact tethering is mediated by TDP-43 on RNA granules binding to GADD34 on mitochondria, while contact untethering is regulated by TDP-43 oxidation. Functionally, this allows for GADD34 and its binding partner PP1 to regulate TDP-43 RNA granule dynamics, and conversely, for TDP-43 oxidation to regulate the ability of the phosphatase PP1 to form granules. Finally, disease-associated mutant TDP-43 misregulates this pathway, ultimately leading to PP1 granules lacking TDP-43. This dynamic crosstalk between TDP-43 oxidation and PP1 has significant consequences for TDP-43-associated diseases including Amyotrophic Lateral Sclerosis (ALS) and Frontotemporal Dementia (FTD).

TDP-43 localization is disrupted in ALS and FTD. Here, authors find that TDP-43 oxidation via mitochondrial OXPHOS mediates its localization to RNA granules, regulating RNA granule-mitochondria contact sites and its crosstalk with GADD34 and PP1.

Introduction

Mitochondria are highly dynamic organelles that undergo oxidative phosphorylation (OXPHOS)^1,2, which generates mitochondrial reactive oxygen species (mtROS) as a cellular signal for protein oxidation^1,2. Mitochondria also form inter-organelle contact sites with other organelles for bidirectional crosstalk^3–5, but the regulation by mitochondrial OXPHOS remains to be explored.

RNA granules are important membraneless organelles, but how RNA granules are mechanistically tethered to mitochondria at contact sites is unclear. RNA granules contain RNA-binding proteins such as TDP-43, a key regulator of RNA^6–8. TDP-43 traffics between the nucleus and cytoplasm, and accumulates in pathological inclusions in neurodegenerative diseases, including Amyotrophic Lateral Sclerosis (ALS) and Frontotemporal Dementia (FTD)^9–11. TDP-43 is also oxidized by ROS at specific cysteine residues¹², but the role of TDP-43 oxidation in regulating mitochondrial crosstalk with RNA granules is unknown.

GADD34 is a key protein that can localize to the outer mitochondrial membrane (OMM)¹³ and binds TDP-43 upon elevated ROS levels¹⁴. GADD34 is also involved in the integrated stress response¹⁵ by binding and acting as a co-factor for PP1 phosphatase activity^16,17. However, the crosstalk between the GADD34-PP1 complex and TDP-43 dynamics, and its regulation of RNA granules and mitochondria, has not been studied and has significant implications for cellular homeostasis.

Here, we found that RNA granules and mitochondria form dynamic contact sites under mitochondrial OXPHOS. Mechanistically, mtROS generated by mitochondrial OXPHOS is sufficient to promote TDP-43 localization to cytoplasmic RNA granules and is mediated by direct oxidation of TDP-43. RNA granule-mitochondria contact site tethering is further regulated by TDP-43 binding to GADD34 on mitochondria and modulated by TDP-43’s oxidation status. Functionally, this pathway mediates their bidirectional crosstalk, allowing the GADD34-PP1 complex to directly modulate TDP-43 dynamics, and conversely, for TDP-43 oxidation status to regulate PP1 phase separation dynamics. Together, these findings highlight an important functional crosstalk between TDP-43 oxidation and the GADD34-PP1 complex, which is misregulated by disease-associated mutant TDP-43.

Results

RNA granules form dynamic inter-organelle contact sites with mitochondria

Inter-organelle contact sites are important hubs regulating organelle dynamics and cell signaling^3–5. Here, using Super-Resolution microscopy, we identified the dynamic formation of contact sites between RNA granules and mitochondria in live HeLa cells under mitochondrial OXPHOS conditions. RNA granules (GFP-G3BP1) and mitochondria (mApple-TOM20) formed stable contacts with one another (white arrows) in OXPHOS media (Fig. 1a, b and Supplementary Fig. 1a, b). RNA granule-mitochondria contact sites remained stably tethered over time (white arrows), prior to subsequent untethering events (yellow arrow) (Fig. 1c–e; Supplementary Fig. 1c–e; Supplementary Video 1). 63% of RNA granules were in contact with mitochondria at any given time (Fig. 1f, g; Supplementary Fig. 1f), and RNA granule-mitochondria contacts remained tethered for an average of 53.9 ± 6.8 s prior to untethering (n = 42 contact sites; Fig. 1h, i).

Fig. 1. RNA granules form dynamic inter-organelle contact sites with mitochondria.

a, b Super-Resolution Airyscan2 live microscopy examples of RNA granules (GFP-G3BP1; orange) tethered to mitochondria (mApple-TOM20; blue) at RNA granule-mitochondria contact sites (white arrows) in OXPHOS media in HeLa cells. Corresponding linescans of RNA granule-mitochondria contact (n = 42 contacts from 9 cells). c–e Representative Super-Resolution Airyscan2 time-lapse live microscopy of dynamic RNA granule-mitochondria contact site tethering (white arrows) prior to untethering event (yellow arrow). Corresponding linescans of RNA granule-mitochondria contact at t = 0 s and untethered at t = 28 s (n = 42 examples from 9 cells). f Schematic of RNA granule-mitochondria contact site. Quantification of the percentage of RNA granules in contact with mitochondria per cell (contact tethered >9 s) (n = 90 examples from 9 cells) (g) and the minimum average RNA granule-mitochondria contact site duration (n = 42 contacts from 9 cells) (h) and corresponding histogram (i). Data are means ± s.e.m. Scale bars: 2 µm (insets: 1 µm) (a, b); 5 µm (inset: 2 µm) (c); 1 µm (d). Source data are provided as a Source data file.

RNA granule-mitochondria contacts were further observed in transmission electron microscopy studies, which show a membraneless RNA granule (RG) in contact with mitochondria (M) (black arrow) in OXPHOS conditions (Supplementary Fig. 1g), and by Super-Resolution Lattice SIM² live microscopy of RNA granules (G3BP1) and mitochondria (TOM20) in OXPHOS conditions (Supplementary Fig. 1h, i). In addition, mitochondria (TOM20) form contacts with RNA granules (G3BP1) positive for TDP-43 (Supplementary Fig. 1j–l), resulting in TDP-43 RNA granules in stable contact with mitochondria in OXPHOS conditions (Supplementary Fig. 1m). Thus, RNA granule-mitochondria contacts (Fig. 1f) are a key cellular hub that may mediate the crosstalk of both organelles.

Mitochondrial OXPHOS regulates TDP-43 localization to cytoplasmic RNA granules

Next, we investigated whether mitochondrial OXPHOS might mechanistically regulate RNA granule dynamics through mtROS generation. We conducted Super-Resolution live microscopy of RNA granule dynamics in OXPHOS conditions (Fig. 2a), and examined the role of mtROS generated by inhibiting mitochondrial OXPHOS Complex III with Antimycin A (AA)¹⁸ (t = 0–45 min) (Fig. 2b). We first confirmed that mitochondrial OXPHOS Complex III inhibition (AA treatment) increased mtROS levels (mitoHyPer7)¹⁹, and resulted in decreased mitochondrial membrane potential (TMRE) (Supplementary Fig. 2i–k). Under OXPHOS conditions, TDP-43 primarily localized to the nucleus (Fig. 2c, d and Supplementary Fig. 2a). In contrast, mtROS induction induced the striking localization of TDP-43 to cytoplasmic RNA granules over time (white arrow; Fig. 2e–g and Supplementary Fig. 2b). Indeed, while the majority of TDP-43 localized to the nucleus at baseline (Pre-treatment) (Supplementary Fig. 2c), mtROS induction led to TDP-43 localizing to cytoplasmic granules and the increased formation of G3BP1 granules (Supplementary Fig. 2d, e, l, m). Furthermore, we confirmed that treatment with Antimycin A did not significantly alter ATP levels, as measured with a live ATP dye (Supplementary Fig. 2n), further suggesting that mtROS production rather than loss of ATP production may be responsible for recruiting TDP-43 to RNA granules. Importantly, TDP-43 was clearly recruited to RNA granules within 30 min of mitochondrial OXPHOS Complex III inhibition (Supplementary Fig. 2f–h; Supplementary Video 2), demonstrating that mtROS generation by mitochondrial OXPHOS inhibition is sufficient to drive TDP-43 localization to cytoplasmic RNA granules.

Fig. 2. Mitochondrial OXPHOS regulates TDP-43 localization to cytoplasmic RNA granules.

a, b Schematic of Super-Resolution live microscopy of RNA granule dynamics in OXPHOS or control conditions. Mitochondrial OXPHOS Complex III was inhibited over time by Antimycin A (AA treatment, t = 0–45 min), leading to mtROS generation. c–g Super-Resolution Airyscan2 live microscopy example of wild-type TDP-43 (TDP-43 WT-Halo; purple) and RNA granules (mCherry-G3BP1; blue) in OXPHOS media, with corresponding linescans. TDP-43 RNA granules do not form in the cytoplasm at baseline (t = 0 min, Pre-treatment), but TDP-43 localizes to cytoplasmic RNA granules over time after mitochondrial OXPHOS Complex III inhibition (white arrows; t = 45 min, Antimycin A) (n = 63 cells (Pre AA), n = 51 cells (Post AA)). h–l Super-Resolution Airyscan2 live microscopy example of TDP-43 WT and RNA granules in Control media, with corresponding linescans. TDP-43 remains localized to the nucleus and TDP-43 does not localize to cytoplasmic RNA granules either at baseline (t = 0 min, Pre-treatment), or after AA treatment (white arrows; t = 45 min, Antimycin A) (n = 63 cells (Pre AA), n = 53 cells (Post AA)). m TDP-43 WT is significantly localized to cytoplasmic granules in OXPHOS media after AA treatment (N = 3 independent experiments; n = 63 cells (Pre AA), n = 51 cells (Post AA)). Paired two-tailed t-test, *p = 0.034. n AA treatment in control media does not lead to the formation of cytoplasmic TDP-43 granules (N = 3 independent experiments; n = 63 cells (Pre AA), n = 53 cells (Post AA)). Paired two-tailed t-test, ns not significant. o TDP-43 granules preferentially form upon AA treatment in OXPHOS media compared to control media (N = 3 independent experiments). Unpaired two-tailed t-test, *p = 0.013. p Model of mitochondrial OXPHOS generating ROS which promotes the dynamic localization of TDP-43 to cytoplasmic RNA granules. Data are means ± s.e.m. Scale bars: 5 µm (insets: 2 µm). Source data are provided as a Source data file.

Furthermore, to confirm the role of mitochondrial OXPHOS in regulating these dynamics, we cultured cells under glycolytic metabolic conditions (Fig. 2a). Control media containing glucose was used to promote glycolysis as the primary energy source, in contrast to OXPHOS media, which lacked glucose and was supplemented with galactose (OXPHOS condition) to promote OXPHOS^20,21. In control conditions, TDP-43 primarily localized to the nucleus (Pre-treatment) (Fig. 2h, i and Supplementary Fig. 3a, c) but was not altered by mitochondrial OXPHOS inhibition (Fig. 2j–l and Supplementary Fig. 3b, d–h; Supplementary Video 3). Thus, mtROS production is sufficient to promote TDP-43 localization to cytoplasmic RNA granules in OXPHOS conditions (58.2 ± 10% of cells; Fig. 2m) as compared to control conditions (17.5 ± 3.7% of cells; Fig. 2n). Together, these findings demonstrate that TDP-43 RNA granule dynamics can be directly modulated by mitochondrial OXPHOS (Fig. 2o) and that a mtROS pathway generated via mitochondrial OXPHOS regulates TDP-43 localization to cytoplasmic RNA granules (Fig. 2p).

In addition, we examined how this pathway might be further regulated. We found that pre-treatment with the non-thiol antioxidant Butylated hydroxyanisole (BHA)²² was sufficient to reduce TDP-43 localization to cytoplasmic RNA granules induced by mitochondrial OXPHOS inhibition (AA) (Supplementary Fig. 4a). In contrast, pre-treatment with the thiol antioxidant N-acetylcysteine (NAC) only reduced TDP-43 localization to cytoplasmic RNA granules induced by direct H₂O₂ treatment, but did not alter TDP-43 localization induced by mitochondrial OXPHOS inhibition (AA) (Supplementary Fig. 4b), suggesting that NAC pre-treatment may not be sufficient to counteract mtROS generated by Antimycin A. Together, these data suggest that non-thiol antioxidants (ex. BHA) rather than thiol antioxidants (ex. NAC) may have a preferential role in decreasing mtROS generated via mitochondrial OXPHOS inhibition²³. Moreover, we found that inhibition of nuclear export (Leptomycin B) was sufficient to prevent TDP-43 from further localizing to cytoplasmic RNA granules upon mitochondrial OXPHOS inhibition (Post AA) (Supplementary Fig. 4c). We also investigated the mechanistic role of TDP-43’s RNA binding in this pathway. In contrast to wild-type TDP-43 (Fig. 2m), the RNA-binding deficient mutant TDP-43 (5FL) containing mutations in its RNA binding domain^24,25, or the acetylation mutant TDP-43 (2KQ), which inhibits its RNA binding²⁶, were both unable to localize to cytoplasmic RNA granules upon mitochondrial OXPHOS inhibition (Supplementary Fig. 4d, e). Thus, TDP-43 binding to RNA is required for its localization to cytoplasmic RNA granules upon mitochondrial OXPHOS inhibition.

Finally, we further examined the role of mitochondrial OXPHOS dysfunction in this pathway. We found that direct treatment with ROS (H₂O₂) or other inhibitors of mitochondrial OXPHOS (Rotenone, Oligomycin A, CCCP) was sufficient to promote TDP-43 localization to cytoplasmic RNA granules in OXPHOS conditions (Supplementary Fig. 4f, g), but not non-mitochondrial stressors such as sodium arsenite (NaAsO₂), nutrient deprivation (HBSS), or heat shock (43 °C) (Supplementary Fig. 4h–k). Together, these findings demonstrate that mitochondrial oxidative stress preferentially regulates TDP-43 dynamic localization to cytoplasmic RNA granules.

Mitochondrial OXPHOS regulation of TDP-43 localization to cytoplasmic RNA granules is mediated via TDP-43 oxidation and misregulated by disease mutant TDP-43

We then examined how TDP-43 RNA granule dynamics might mechanistically respond to mtROS. mtROS acts as a key signal for regulating cellular pathways via protein oxidation², and TDP-43 can be oxidized at specific cysteine residues²⁷. To determine if oxidation of TDP-43 mediated its ability to respond to mtROS in mitochondrial OXPHOS conditions, we generated a TDP-43 mutant (Ox) insensitive to changes in oxidation by mutating two key cysteine residues in TDP-43 previously shown to undergo oxidation (Cys173/Cys175)¹² (Fig. 3a, b). TDP-43 (Ox) mutant primarily localized to the nucleus and the cytoplasm in OXPHOS conditions (Fig. 3c, d and Supplementary Fig. 5a, c) but was surprisingly not altered upon mtROS generation from mitochondrial OXPHOS Complex III inhibition (AA) (Fig. 3e–g and Supplementary Fig. 5b, d, e). Moreover, TDP-43 (Ox) did not dynamically form cytoplasmic RNA granules over time upon mtROS generation (Supplementary Fig. 5f–h; Supplementary Video 4). Thus, the ability of mtROS generated by mitochondrial OXPHOS inhibition to drive TDP-43 accumulation in cytoplasmic RNA granules directly depends on cysteine oxidation of TDP-43.

Fig. 3. Mitochondrial OXPHOS regulation of TDP-43 dynamics is mediated by TDP-43 oxidation and misregulated by TDP-43 disease-associated mutants.

a, b Schematic of TDP-43 Oxidation mutant (TDP-43 (Ox)), which is mutated at two key oxidation residues (Cys173/Cys175) and is unable to be oxidized, and disease-associated mutant TDP-43 (M337V). Model of TDP-43 mutants in response to ROS. c–g Super-Resolution Airyscan2 live microscopy example of TDP-43 Oxidation mutant (TDP-43 (Ox)-Halo; purple) and RNA granules (mCherry-G3BP1; blue) in OXPHOS media, with corresponding linescans. TDP-43 (Ox) remains localized to the nucleus while RNA granules form in the cytoplasm without TDP-43 (Ox) at both baseline (yellow arrows; t = 0 min, Pre-treatment), and after AA treatment (yellow arrows; t = 45 min, Antimycin A) (n = 64 cells (Pre AA), n = 60 cells (Post AA)). h–l Super-Resolution Airyscan2 live microscopy example of TDP-43 disease mutant (TDP-43 (M337V)-Halo; purple) and RNA granules (mCherry-G3BP1; blue) in OXPHOS media, with corresponding linescans. TDP-43 (M337V) can already localize to cytoplasmic RNA granules at baseline (white arrows; t = 0 min, Pre-treatment), as well as after AA treatment (white arrows; t = 45 min, Antimycin A) (n = 65 cells (Pre AA), n = 62 cells (Post AA)). m TDP-43 (Ox) mutant does not localize to cytoplasmic RNA granules at baseline and is not altered by AA treatment (N = 3 independent experiments; n = 64 cells (Pre AA), n = 60 cells (Post AA)). Paired two-tailed t-test, ns not significant. n TDP-43 (M337V) mutant can already localize to cytoplasmic RNA granules at baseline and is not altered by AA treatment (N = 3 independent experiments; n = 65 cells (Pre AA), n = 62 cells (Post AA)). Paired two-tailed t-test, ns not significant. o Model of TDP-43 oxidation regulating TDP-43 localization to cytoplasmic RNA granules in response to ROS generation by mitochondrial OXPHOS, which is mediated by TDP-43 oxidation. This pathway is disrupted by the TDP-43 (Ox) oxidation mutant and misregulated by the TDP-43 (M337V) disease mutant. Data are means ± s.e.m. Scale bars: 5 µm (insets: 2 µm). Source data are provided as a Source data file.

We further confirmed that changes in TDP-43 (Ox) mutant response were not due to alterations in its expression level versus wild-type TDP-43 (Supplementary Fig. 5i), and showed that puromycin treatment which promotes TDP-43 cytoplasmic localization²⁸ is not mediated by TDP-43 oxidation, as it could still increase TDP-43 (Ox) mutant localization to cytoplasmic RNA granules (Supplementary Fig. 5j). Importantly, additional TDP-43 cysteine oxidation mutants at the two oxidation sites above (C173S/C175S) (TDP-43 2CS) or all four oxidation sites (C173S/C175S/C198S/C244S) (TDP-43 4CS) were also unable to localize to cytoplasmic RNA granules upon mtROS generation (Supplementary Fig. 5k, l), further supporting a key role for TDP-43 oxidation in mediating its localization to cytoplasmic RNA granules upon mitochondrial oxidative stress.

Next, as TDP-43 disease-associated mutations result in ALS and FTD^11,29–31, we investigated how this pathway might be modulated by the disease-associated mutant TDP-43 (M337V)¹¹ (Fig. 3a). Surprisingly, TDP-43 (M337V) could form cytoplasmic RNA granules at baseline (Fig. 3h, i and Supplementary Fig. 6a, c), which were not altered upon mtROS generation by mitochondrial OXPHOS Complex III inhibition (AA) (Fig. 3j–l and Supplementary Fig. 6b, d, e, i). Moreover, TDP-43 (M337V) RNA granule dynamics did not significantly change over time upon mitochondrial OXPHOS Complex III inhibition (Supplementary Fig. 6f–h; Supplementary Video 5), suggesting that the TDP-43 (M337V) disease-associated mutant acts independently of changes in mtROS.

Together, these findings demonstrate that both the TDP-43 oxidation mutant (Ox) and the TDP-43 (M337V) disease-associated mutant are insensitive to changes in mtROS from mitochondrial OXPHOS (Fig. 3m, n). However, while the TDP-43 (Ox) mutant is unable to respond to mtROS signaling to localize to cytoplasmic RNA granules, the TDP-43 (M337V) mutant already localizes to cytoplasmic RNA granules. Thus, a mechanistic pathway exists for mtROS generated by mitochondrial OXPHOS to dynamically recruit RNA granules in a manner dependent on the oxidation status of TDP-43 (Fig. 3o).

RNA granule-mitochondria contact site tethering is regulated by TDP-43 oxidation status and GADD34

Inter-organelle contact sites are tightly regulated by protein tethering complexes which act to tether organelles together^3–5. We next sought to identify whether RNA granule-mitochondria contact site tethering might be regulated via TDP-43 oxidation in mitochondrial OXPHOS conditions. Super-Resolution live imaging revealed that RNA granules (G3BP1) form dynamic contacts with mitochondria (TOM20) under wild-type TDP-43 (WT) conditions (Fig. 4a–d; Supplementary Fig. 7a; Supplementary Video 6).

Fig. 4. RNA granule-mitochondria contact site tethering is regulated by TDP-43 oxidation status.

a–c Super-Resolution Airyscan2 live microscopy example of RNA granule-mitochondria contact site tethering in wild-type TDP-43 condition (TDP-43 WT-Halo), showing RNA granules (GFP-G3BP1; orange) tethered to mitochondria (mApple-TOM20; blue) in OXPHOS media. Corresponding time-lapse and linescans of contact site tethering at t = 0 s (white arrows), which are untethered by t = 35 s (yellow arrow) (n = 48 contacts from 12 cells). d Corresponding histogram of the minimum contact duration (n = 48 contacts from 12 cells). e–g Super-Resolution Airyscan2 live microscopy example of RNA granule-mitochondria contact site tethering in TDP-43 oxidation mutant condition (TDP-43 (Ox)-Halo), showing RNA granules (GFP-G3BP1; orange) tethered to mitochondria (mApple-TOM20; blue) in OXPHOS media. Corresponding time-lapse and linescans of contact site tethering at t = 0 s (white arrows), which remain prolonged and still tethered at t = 49 s (yellow arrows) (n = 51 contacts from 13 cells). h Corresponding histogram of the minimum contact duration (n = 51 contacts from 13 cells). i–k Super-Resolution Airyscan2 live microscopy example of RNA granule-mitochondria contact site tethering in TDP-43 disease mutant condition (TDP-43 (M337V)-Halo), showing RNA granules (GFP-G3BP1; orange) tethered to mitochondria (mApple-TOM20; blue) in OXPHOS media. Corresponding time-lapse and linescans of contact site tethering at t = 0 s (white arrows), which are untethered by t = 42 s (yellow arrow) (n = 51 contacts from 13 cells). l Corresponding histogram of the minimum contact duration (n = 51 contacts from 13 cells). m The percentage of RNA granules that form RNA granule-mitochondria contacts is not disrupted by TDP-43 (Ox) oxidation mutant or TDP-43 (M337V) disease mutant (M337V) (N = 13 cells (TDP-43 WT); N = 13 cells (TDP-43 (Ox)); N = 13 cells (TDP-43 (M337V)). One-way ANOVA with Tukey’s post hoc test, ns not significant. n TDP-43 (Ox) oxidation mutant significantly prolongs RNA granule-mitochondria contact site tethering duration compared to the TDP-43 (M337V) disease mutant (n = 48 contacts from 13 cells (TDP-43 WT); n = 51 contacts from 13 cells (TDP-43 (Ox)); n = 51 contacts from 13 cells (TDP-43 (M337V)). One-way ANOVA with Tukey’s post hoc test, *p = 0.021. o Model of RNA granule-mitochondria contact site tethering and subsequent untethering event mediated by oxidation of TDP-43. TDP-43 (Ox) oxidation mutant inhibits contact untethering, while TDP-43 (M337V) disease mutant promotes contact untethering. Data are means ± s.e.m. Scale bars: 2 µm (time-lapse: 1 µm). Source data are provided as a Source data file.

Surprisingly, the TDP-43 (Ox) mutant, insensitive to changes in oxidation, was sufficient to prolong RNA granule-mitochondria contact sites (white arrows) (Fig. 4e–h; Supplementary Fig. 7b; Supplementary Video 7). Thus, altering the oxidation status of TDP-43 may be a key regulator of contact site untethering events, which is inhibited by the TDP-43 (Ox) mutant, leading to prolonged contact sites.

In contrast, the TDP-43 (M337V) disease-associated mutant led to decreased RNA granule-mitochondria contact tethering durations (Fig. 4i–l; Supplementary Fig. 7c; Supplementary Video 8). These findings suggest that the TDP-43 (M337V) disease-associated mutant may already prematurely drive RNA granule-mitochondria contact untethering, independently of changes in mtROS. Importantly, while the oxidation status of TDP-43 did not alter RNA granule-mitochondria contact formation (Fig. 4m), contact tethering duration was significantly increased by TDP-43 (Ox) mutant compared to TDP-43 (M337V) disease mutant (Fig. 4n). Of note, ROS (H₂O₂) or Antimycin A treatment, but not puromycin, significantly increased the percentage of RNA granules in contact with mitochondria (Supplementary Fig. 7d, e), further supporting a role for oxidation in regulating these contacts. Together, these findings highlight TDP-43 as a key regulator of RNA granule-mitochondria contact tethering, whose untethering is regulated by TDP-43 oxidation status (Fig. 4o).

Next, we investigated how TDP-43 might tether RNA granules to mitochondria in an oxidation-dependent manner. TDP-43 binds GADD34 upon increased ROS levels, and is stabilized by a TDP-43 mutant insensitive to changes in oxidation¹⁴. Moreover, GADD34 can be recruited to the OMM¹³, but the functional roles of these findings, and whether GADD34 might tether TDP-43 to mitochondria has never been examined. Therefore, to investigate whether RNA granule-mitochondria contacts were regulated via GADD34, we conducted Super-Resolution live microscopy of TDP-43 and GADD34 with the mitochondrial matrix marker (mito-BFP). Importantly, GADD34 expression was sufficient to induce the striking recruitment of TDP-43 oxidation mutant (TDP-43 (Ox)) around mitochondria (linescans; Fig. 5a), but not wild-type TDP-43 (WT) or TDP-43 (M337V) disease-associated mutant around mitochondria, resulting in their limited localization with GADD34 (Fig. 5b; Supplementary Fig. 7f–h). Thus, GADD34 can preferentially promote TDP-43 (Ox) recruitment around mitochondria.

Fig. 5. Mitochondrial GADD34 binding to TDP-43 regulates RNA granule-mitochondria contact site tethering and is dependent on TDP-43 oxidation status.

a, b Super-Resolution Airyscan2 live microscopy example with corresponding linescan showing GADD34 (mEmerald-GADD34; green) recruitment of TDP-43 oxidation mutant (TDP-43 (Ox)-Halo; blue) around the mitochondrial matrix (mito-BFP; purple) in OXPHOS media. TDP-43 (Ox) oxidation mutant localizes with GADD34 in contrast to TDP-43 WT or TDP-43 (M337V) (N = 3 independent experiments, from n = 57 cells (TDP-43 WT), n = 57 cells (TDP-43 (Ox)); n = 62 cells (TDP-43 (M337V)). One-way ANOVA with Tukey’s post hoc test, ***p = .0009 (WT vs Ox), **p = 0.0016 (Ox vs M337V). c Representative co-immunoprecipitation (co-IP) of GADD34 binding TDP-43. Western blot of input and IP (FLAG), showing IP of TDP-43-FLAG (WT, Ox, and M337V) and co-IP of GADD34 (Myc) in OXPHOS media. GADD34 preferentially binds TDP-43 (Ox), in comparison to TDP-43 (WT) or TDP-43 (M337V) (see arrow) (Representative image from N = 3 independent experiments). d Representative western blot of GADD34 siRNA knockdown in OXPHOS media, showing decreased GADD34 protein levels, compared to GAPDH (loading control) (n = 3 experiments per condition). e The percentage of RNA granules which form RNA granule-mitochondria contacts (GFP-G3BP1 and mApple-TOM20) is not disrupted by knockdown of GADD34 in cells expressing TDP-43 WT-Halo in OXPHOS media (N = 11 cells (siScr); N = 13 cells (siGADD34). Unpaired two-tailed t-test, ns not significant. f–h Corresponding histograms and quantification of contact site duration showing knockdown of GADD34 significantly decreases RNA granule-mitochondria contact site tethering duration compared to controls in OXPHOS media (n = 16 contacts from 11 cells (siScr); n = 21 contacts from 13 cells (siGADD34). Unpaired two-tailed t-test, **p = 0.0042. i Model of RNA granule-mitochondria contact site tethering mediated by TDP-43 on RNA granules tethered to GADD34 on the outer mitochondrial membrane. Subsequent untethering is mediated by oxidation of TDP-43, which releases its tethering to GADD34 on mitochondria. Data are means ± s.e.m. Scale bars: 2 µm (a). Source data are provided as a Source data file.

To further confirm the binding of TDP-43 to GADD34, we conducted biochemical co-immunoprecipitation (co-IP) studies of TDP-43 with GADD34. As expected¹⁴, GADD34 (Myc) showed increased binding to TDP-43 oxidation mutant (TDP-43 (Ox)) but not TDP-43 (WT) or TDP-43 (M337V) disease-associated mutant (Fig. 5c; arrow), showing that GADD34 binding to TDP-43 is regulated by TDP-43 oxidation.

Finally, to demonstrate GADD34’s role in RNA granule-mitochondria contact tethering, we knocked down GADD34 (Fig. 5d) and examined RNA granule-mitochondria contact site dynamics. Importantly, while GADD34 knockdown did not alter contact formation (Fig. 5e), it significantly decreased RNA granule-mitochondria contact tethering duration (Fig. 5f–h), further supporting a role for GADD34 in tethering RNA granules to mitochondria. Together, these findings suggest that RNA granule-mitochondria contact tethering in OXPHOS conditions is mechanistically regulated by TDP-43’s binding to GADD34, with subsequent contact untethering driven by altering TDP-43 oxidation status, which releases its binding from GADD34 (Fig. 5i).

Mitochondrial GADD34-PP1 complex regulates TDP-43 RNA granule dynamics

One key role for inter-organelle contact sites is to allow for proteins on both organelles to directly crosstalk and regulate one another^3–5. Thus, we asked whether important functional crosstalk might occur between GADD34 on mitochondria and TDP-43 RNA granules. GADD34 is a key co-factor of the phosphatase PP1, which it binds to form a GADD34-PP1 complex to allow PP1 to dephosphorylate specific substrates^15–17. While GADD34 can localize to the OMM¹³, whether GADD34 can recruit PP1 to mitochondria has not been investigated. We found using Super-Resolution live microscopy of the mitochondrial matrix that GADD34 was able to recruit a subpopulation of PP1 around mitochondria (Fig. 6a, Supplementary Fig. 8a; linescans). We further confirmed this by directly imaging the OMM using Super-Resolution live microscopy, showing GADD34 recruitment of a subpopulation of PP1 to the OMM labeled with TOMM20 (Supplementary Fig. 8b; linescans).

Fig. 6. GADD34 recruits PP1 to mitochondria, and a GADD34-PP1 complex regulates TDP-43 dynamics.

a Super-Resolution Airyscan2 live microscopy example with corresponding linescans showing GADD34 WT (mEmerald-GADD34 WT; green) recruitment of a subpopulation of PP1 (mScarlet-PP1; purple) around the mitochondria (Mitochondrial matrix marker; PDH1-SNAP; blue) in OXPHOS media (n = 25 cells). b Model of GADD34 recruitment of PP1 to the outer mitochondrial membrane, and inhibition by a GADD34 (KARA) mutation in its PP1 binding domain. c Quantification showing GADD34 and PP1 colocalization is inhibited by the GADD34 (KARA) mutation in OXPHOS media (N = 3 independent experiments, from n = 76 cells (GADD34 WT); n = 68 cell (GADD34 (KARA)). Unpaired two-tailed t-test, ****p = 0.000048. d, e Quantification showing both GADD34 and PP1 can localize to mitochondria in GADD34 WT condition (N = 3 independent experiments, from n = 72 cells (GADD34 WT); n = 76 cells (PP1) but only GADD34 but not PP1 is localized to mitochondria in GADD34 (KARA) condition in OXPHOS media (N = 3 independent experiments, from n = 65 cells (GADD34 (KARA)); n = 68 cells (PP1). Unpaired two-tailed t-test, ns not significant, **p = 0.0097. f, g Super-Resolution Airyscan2 live microscopy example showing TDP-43 WT-Halo (purple) localizes to the nucleus and not to RNA granules (GFP-G3BP1; blue) in GADD34 wild-type condition (TagBFP2-GADD34 WT) in OXPHOS media, with corresponding linescan (n = 53 cells). h, i Super-Resolution Airyscan2 live microscopy example showing TDP-43 WT-Halo (purple) is localized to cytoplasmic RNA granules (GFP-G3BP1; blue) (arrows) in GADD34 (KARA) mutant condition, which is unable to bind PP1 (TagBFP2-GADD34 (KARA)) in OXPHOS media, with corresponding linescan (n = 65 cells). j GADD34 (KARA) mutant which is unable to recruit PP1 disrupts TDP-43 dynamics resulting in significantly increased TDP-43 localization at cytoplasmic RNA granules (N = 3 independent experiments, from n = 53 cells (GADD34 WT); n = 65 cells (GADD34 (KARA)). Unpaired two-tailed t-test, *p = 0.035. Data are means ± s.e.m. Scale bars: 1 µm (a); 5 µm (insets: 2 µm) (f, h). Source data are provided as a Source data file.

To further elucidate the role of GADD34 in recruiting PP1 to the OMM, we expressed GADD34 with a mutation in its PP1 binding domain (GADD34 (KARA))³² and investigated PP1 recruitment to mitochondria (Fig. 6b, Supplementary Fig. 8c, d). We first confirmed that GADD34 (WT) and GADD34 (KARA) indeed showed similar expression (Supplementary Fig. 8e). As expected, mutant GADD34 (KARA) unable to bind PP1 showed significantly decreased colocalization with PP1, as compared to GADD34 (WT) (Fig. 6c). Moreover, while both wild-type GADD34 (WT) and mutant GADD34 (KARA) localized to the OMM (Supplementary Fig. 8f), only GADD34 (WT) recruited a subpopulation of PP1 to mitochondria (Fig. 6d). In contrast, PP1 localization to the mitochondria was significantly decreased in mutant GADD34 (KARA) conditions (Fig. 6e), suggesting that GADD34 may recruit a subpopulation of PP1 to the OMM via its PP1 binding domain (Fig. 6b). Additionally, we conducted biochemical mitochondrial subcellular fractionations to investigate GADD34 and PP1 localization and found that both GADD34 and PP1 were indeed present in the mitochondrial fraction (Supplementary Fig. 8g). Together, these findings highlight a role for a GADD43-PP1 complex on the mitochondria mediated by GADD34’s recruitment of PP1.

We next investigated whether the GADD34-PP1 complex might be important for directly regulating TDP-43 RNA granule dynamics. Using Super-Resolution live microscopy, we compared the role of GADD34 (WT) versus mutant GADD34 (KARA). Interestingly, while GADD34 (WT) did not alter TDP-43’s localization to the nucleus (Fig. 6f, g and Supplementary Fig. 9a, c), the PP1 binding-deficient mutant GADD34 (KARA) significantly increased TDP-43 cytoplasmic RNA granules (Fig. 6h–j and Supplementary Fig. 9b, d). Thus, the GADD34-PP1 complex is able to regulate TDP-43 RNA granule dynamics.

Finally, as GADD34 acts as a co-factor for PP1 to dephosphorylate substrates^15–17, and as TDP-43 is a known substrate for PP1 dephosphorylation³³, we asked if GADD34 might be a co-factor for PP1’s dephosphorylation of TDP-43. Indeed, in contrast to GADD34 (WT), the GADD34 (KARA) mutant resulted in increased levels of phosphorylated TDP-43 (pTDP-43), which localized to TDP-43 cytoplasmic granules (Supplementary Fig. 9e, f; arrows). These data show that GADD34 binding to PP1 is an important driver of TDP-43’s dephosphorylation and suggest that GADD34 may additionally act as a co-factor for PP1’s dephosphorylation of TDP-43. Altogether, these findings highlight the GADD34-PP1 complex as a key regulator of TDP-43 RNA granule dynamics.

TDP-43 oxidation modulates PP1 phase separation dynamics

We then asked whether TDP-43 might conversely regulate PP1 granule dynamics. Proteins undergo phase separation in response to a variety of factors³⁴, and both TDP-43³⁵ and PP1³⁶ have been found to phase separate, but their ability to regulate one another’s dynamics has never been studied. As a result, we sought to investigate the dynamic relationship between TDP-43 and PP1 and the role of TDP-43 oxidation using Super-Resolution live microscopy in OXPHOS conditions with PP1 overexpression. As expected, wild-type TDP-43 (WT) did not promote phase separation of TDP-43 with PP1 into cytoplasmic granules (Fig. 7a–c). Indeed, in the presence of wild-type TDP-43 (WT), PP1 rarely formed granules, and further analysis showed that these were independent of TDP-43 (Supplementary Fig. 10a, d).

Fig. 7. TDP-43 oxidation status modulates PP1 phase separation dynamics.

a–c Super-Resolution Airyscan2 live microscopy example of TDP-43 WT-Halo (purple) and expressed mScarlet-PP1 (green) in OXPHOS media. Corresponding linescan and model showing TDP-43 WT is localized to the nucleus, and PP1 does not form cytoplasmic granules (b, c) (n = 65 cells). d, e Super-Resolution Airyscan2 live microscopy example of TDP-43 (Ox)-Halo oxidation mutant (purple) and expressed mScarlet-PP1 (green) co-recruited to the same granules in the cytoplasm (white arrows) in OXPHOS media. Corresponding linescan depicts their colocalization (black arrow) (n = 60 cells). f Live time-lapse imaging of dynamic TDP-43 (Ox) and PP1 granules over 56 s (n = 6 cells). g Model of TDP-43 (Ox) and PP1 localized to the same phase-separated granule. h, i Super-Resolution Airyscan2 live microscopy example of TDP-43 (M337V) Halo disease mutant (purple) and expressed mScarlet-PP1 (green) in OXPHOS media showing PP1 forms its own granules in the cytoplasm, which lack TDP-43 (yellow arrows). Corresponding linescan depicts PP1 only (yellow arrow) (n = 65 cells). j Live time-lapse imaging of dynamic PP1 granule over 56 s (n = 6 cells). k Model of PP1 phase-separated granule in the TDP-43 (M337V) condition. l TDP-43 (Ox) oxidation mutant promotes the formation of PP1 granules with contain TDP-43, in contrast to TDP-43 WT or TDP-43 (M337V) disease mutant (N = 3 independent experiments, from n = 65 cells (TDP-43 WT); n = 60 cells (TDP-43 (Ox)); n = 65 cells (TDP-43 (M337V)). One-way ANOVA with Tukey’s post hoc test, **p = 0.0091 (WT vs Ox), *p = 0.0179 (Ox vs M337V). m TDP-43 (M337V) disease mutant significantly promotes the formation of PP1 granules which lack TDP-43 (N = 3 independent experiments, from n = 65 cells (TDP-43 WT); n = 60 cells (TDP-43 (Ox)); n = 65 cells (TDP-43 (M337V)). One-way ANOVA with Tukey’s post hoc test, ** = 0.0393 (Ox vs M337V). n Model of TDP-43 oxidation status regulating PP1 phase separation dynamics. Data are means ± s.e.m. Scale bars: 5 µm (i: 2 µm; ii and time-lapse: 0.5 µm). Source data are provided as a Source data file.

In contrast, the TDP-43 mutant (Ox), which is insensitive to changes in oxidation, resulted in the striking formation of PP1 cytoplasmic granules together with TDP-43 (Fig. 7d–g). Surprisingly, as TDP-43 (Ox) alone did not form cytoplasmic granules in OXPHOS media (Fig. 3c–g, m), we found that the expression of PP1 here was sufficient to promote TDP-43 (Ox) recruitment to cytoplasmic granules (Fig. 7d–g), suggesting that both PP1 and TDP-43 (Ox) are able to promote each other’s phase separation. Indeed, PP1 and TDP-43 (Ox) phase separated together and dynamically interacted over time as revealed by time-lapse Super-Resolution imaging (Fig. 7f; Supplementary Video 9). Additional analysis showed that PP1 formed granules that were almost exclusively with TDP-43 (Ox) (Supplementary Fig. 10b, e).

Finally, we found that the TDP-43 (M337V) disease-associated mutant also led to the formation of cytoplasmic PP1 granules (Fig. 7h–k), but these surprisingly did not contain TDP-43 (Fig. 7h–k; Supplementary Video 10). Indeed, PP1 formed granules which always lacked TDP-43 (Supplementary Fig. 10c). In addition, further analysis of TDP-43 (M337V) granules confirmed their localization with RNA granules (G3BP1) but never with PP1 (Supplementary Fig. 10f). Moreover, quantification of PP1 and TDP-43 granule dynamics revealed that TDP-43 (Ox) was sufficient to stall PP1 and TDP-43 together in a phase-separated state in cytoplasmic granules (Fig. 7l), while disease-associated mutant TDP-43 (M337V) led to the striking formation of PP1 granules which lacked TDP-43 (Fig. 7m).

We further examined whether granules containing TDP-43 (Ox) and PP1 were in contact with mitochondria. Using Super-Resolution live microscopy, we found that most of these granules were in contact with mitochondria (Supplementary Fig. 10g, white arrows). Additionally, we knocked down GADD34 (Fig. 5d) and found that PP1 granules could still colocalize with TDP-43 (Ox) in cytoplasmic granules (Supplementary Fig. 10h) and that the proportion of these granules in contact with mitochondria was not altered (Supplementary Fig. 10i). Moreover, we investigated the role of GADD34 in regulating TDP-43 (Ox)’s interaction with PP1. To test this, we performed co-IP studies of TDP-43 (Ox) with PP1 in GADD34 knockdown cells and found that TDP-43 (Ox) could still bind to PP1 even with decreased GADD34 in GADD34 knockdown (Supplementary Fig. 10k, l). Together, these data suggest that GADD34 may not be required for PP1’s interaction with TDP-43 (Ox). Furthermore, the TDP-43 mutant (Ox) may override the role of GADD34 in this pathway, leading to the phase separation of TDP-43 (Ox) with PP1 granules. Finally, to confirm the binding between TDP-43 and PP1, we conducted co-IP studies of TDP-43’s interaction with PP1. Consistent with our imaging studies, PP1 (V5) preferentially bound to TDP-43 oxidation mutant (TDP-43 (Ox)), in comparison to TDP-43 (WT) or TDP-43 (M337V) disease-associated mutant (Supplementary Fig. 10j; arrow), further demonstrating PP1’s binding to TDP-43 is regulated by TDP-43 oxidation.

Thus, these findings demonstrate a pathway in which TDP-43 and PP1 dynamically phase separate together, dependent on TDP-43 oxidation status, but are misregulated by TDP-43 (M337V) disease-associated mutant (Fig. 7n) and further highlight an important bidirectional crosstalk between TDP-43 and PP1 dynamics.

Discussion

This study identifies a cellular pathway for mitochondrial OXPHOS to regulate TDP-43 localization to cytoplasmic RNA granules, leading to RNA granule-mitochondria contact site tethering via TDP-43 and GADD34 (Fig. 8a: Model). Mechanistically, this pathway is regulated by mtROS generated by mitochondrial OXPHOS dysfunction, which promotes TDP-43 oxidation at Cys173/Cys175, leading to increased GADD34 binding¹⁴ around mitochondria, while subsequent oxidation changes release TDP-43 from binding GADD34, leading to RNA granule-mitochondria contact site untethering and the release of TDP-43 from PP1 granules.

Fig. 8. Model: Mitochondrial OXPHOS regulates TDP-43 oxidation and crosstalk with GADD34/PP1 via RNA granule-mitochondria contact sites.

a–c Model of the role of mitochondrial OXPHOS and TDP-43 oxidation in regulating TDP-43 RNA granule dynamics and crosstalk with GADD34-PP1 complex on the mitochondria via RNA granule-mitochondria contact sites. This pathway is properly regulated by ROS in the TDP-43 WT condition but is inhibited by TDP-43 Oxidation mutant (Ox) and is misregulated by TDP-43 (M337V) disease-associated mutant. See “Discussion” for details.

In contrast, a TDP-43 oxidation mutant (Ox), insensitive to changes in oxidation, stalls this pathway at multiple steps regulated by mtROS. TDP-43 (Ox) is unable to be recruited to cytoplasmic RNA granules upon elevated mtROS levels and does not respond to subsequent oxidation changes, leading to increased GADD34 binding, prolonged RNA granule-mitochondria contact tethering, and stalled TDP-43 (Ox) and PP1 in phase-separated granules (Fig. 8b: Model). Finally, the TDP-43 disease-associated mutant (M337V) overrides this pathway by not properly responding to changes in mtROS levels. The TDP-43 disease mutant localizes to cytoplasmic RNA granules independently of mtROS and drives the formation of PP1 granules without TDP-43 (Fig. 8c: Model). Functionally, RNA granule-mitochondria contacts may act as an important hub to regulate the dynamic bidirectional crosstalk between the GADD34-PP1 complex and TDP-43.

Inter-organelle contacts form between multiple organelles to regulate cellular homeostasis^3,4. RNA granules contact the ER³⁷, early endosomes³⁸, late endosomes³⁹, and lysosomes⁴⁰, while mitochondria form contact sites with many organelles^3–5. RNA granule-mitochondria contact sites^40,41 may additionally play important roles, such as FMRP RNA granules, which were recently found to contact mitochondria to regulate local protein synthesis and mitochondrial fission⁴¹. Our work further demonstrates the formation of TDP-43 RNA granule-mitochondria contact sites and elucidates its regulation by mitochondrial OXPHOS.

We found that mitochondrial OXPHOS can regulate mtROS signaling to promote TDP-43 localization to cytoplasmic RNA granules⁴², which was dependent on TDP-43’s RNA binding ability and regulated by its acetylation^24–26. Importantly, TDP-43 oxidation plays key roles^12,43,44 and we showed that oxidation of TDP-43 at Cys173/Cys175 is a critical mechanism through which TDP-43 recognizes mtROS signaling in this pathway. As mtROS from mitochondrial OXPHOS generates H₂O₂, which can cross the nuclear membrane², this may allow for mitochondrial OXPHOS to functionally recruit TDP-43 from the nucleus to the cytoplasm via TDP-43 oxidation. In addition, TDP-43 undergoes other post-translational modifications, including phosphorylation and oxidation at methionine residues^45–47, which may further modulate TDP-43’s dynamics in this pathway. Moreover, as mitochondrial OXPHOS dysfunction can lead to changes in ATP levels, changes in ATP levels upon OXPHOS misregulation may additionally contribute to the changes in TDP-43 cytosolic granule formation we observed, as ATP has previously been shown to be a hydrotrope for RNA-binding proteins⁴⁸, and also regulates TDP-43’s oligomerization⁴⁹.

GADD34 is a key regulatory subunit for PP1, which requires co-factors for its phosphatase activity^15–17, and GADD34 can shuttle between the OMM and the endoplasmic reticulum¹³. Here, we found that GADD34 is able to recruit a subpopulation of PP1 to the OMM, and the GADD34-PP1 complex regulates TDP-43 localization to RNA granules and phosphorylation. Thus, GADD34 may play an additional role as PP1’s co-factor for TDP-43 dephosphorylation^14,33, potentially allowing GADD34-PP1 on mitochondria to modulate TDP-43’s phosphorylation status at RNA granule-mitochondria contact sites. Of note, while we found that GADD34 expression was sufficient to recruit both TDP-43 (Ox) and PP1 directly to the OMM, there may be additional regulators that promote PP1’s recruitment to mitochondria and PP1’s dephosphorylation of TDP-43.

Interestingly, the GADD34-PP1 complex plays a critical role in activating the integrated stress response pathway¹⁵, where it dephosphorylates EIF2α to promote protein synthesis from RNA at ribosomes. Thus, the GADD34-PP1 complex at mitochondria may act as a master regulator for protein translation of nuclear-encoded mitochondrial proteins, by modulating the phosphorylation status of both TDP-43 and EIF2α, to coordinate RNA delivery by TDP-43⁸ with subsequent protein translation by EIF2α¹⁵. Indeed, multiple studies have identified a key role for TDP-43 in delivering nuclear-encoded RNA from the nucleus to the mitochondria and regulating the translation of mitochondrial proteins, which support mitochondrial OXPHOS function^42,50–54. This could, in turn, alter mitochondrial oxidation levels to modify TDP-43’s oxidation, allowing TDP-43 to subsequently return to the nucleus (Fig. 8a: Model, gray arrows). Moreover, disruption of this pathway may contribute to the alterations in mitochondrial dynamics, function, and metabolism previously observed in TDP-43 models^55–65.

TDP-43 is known to form phase-separated granules^35,66,67 regulated by RNA binding⁶⁸ and additional proteins^69–73. We show that TDP-43 can phase separate into cytoplasmic granules with the phosphatase PP1. PP1 phase separates in yeast following oxidative stress³⁶, but its live cell dynamics and regulation in mammalian cells have never been investigated. We identified TDP-43 oxidation as a key regulator of PP1’s phase separation dynamics, whereby mutant TDP-43, insensitive to changes in oxidation, stalls PP1 with TDP-43 (Ox) mutant in cytoplasmic granules. Furthermore, our findings showed that while TDP-43 (Ox) alone did not form cytoplasmic granules in OXPHOS media, PP1 expression was sufficient to promote TDP-43 (Ox) localization to cytoplasmic granules with PP1, demonstrating that both PP1 and TDP-43 (Ox) are able to promote each other’s phase separation. As PP1 also undergoes oxidation, which regulates its activity⁷⁴, PP1 oxidation may additionally modulate this pathway. Moreover, we found PP1 and TDP-43 (Ox) could still interact and phase separate together into cytoplasmic granules even upon GADD34 knockdown, suggesting that GADD34 may not be necessary for PP1 and TDP-43 (Ox)’s interaction. Thus, further studies investigating the interactions between TDP-43, GADD34, and PP1 and additional regulators of this pathway will shed important light on how this pathway is dynamically regulated.

TDP-43 forms cytoplasmic aggregates in both sporadic and familial forms of ALS/FTD, but the mechanisms driving its mislocalization remain unclear^9,10. Our work supports a key role for defective mitochondrial OXPHOS as a driving factor in sporadic disease leading to TDP-43 aggregation⁶⁵, as directly inhibiting mitochondrial OXPHOS function was sufficient to induce TDP-43 cytoplasmic granules over time. Indeed, mitochondrial dysfunction is linked to ALS/FTD⁹, and TDP-43 undergoes increased aggregation under oxidative stress compared to non-respirating cells⁷⁵. In addition, familial disease-associated mutations in TDP-43 have been shown to impair TDP-43 RNA granule transport in axons and its viscosity^76,77. Our study further suggests that in familial disease, TDP-43 disease-associated mutations override their ability to respond to mtROS regulation. While our studies conducted Super-Resolution live microscopy of overexpressed fluorescently labeled TDP-43 to track its dynamics over time in conjunction with G3BP1 granule dynamics, future studies on the dynamics of endogenous TDP-43 will help shed additional light on this pathway. Interestingly, while modulating the GADD34-PP1 pathway may rescue ALS models^78–81, the mechanistic crosstalk between GADD34-PP1 and TDP-43 has remained unclear. Here, we showed that the GADD34-PP1 complex regulates TDP-43 dynamics, and conversely that the TDP-43 disease-associated mutant disrupts PP1 phase separation dynamics, resulting in PP1 cytoplasmic granules lacking TDP-43. Thus, further understanding the molecular roles of this dynamic crosstalk will be important for elucidating how this pathway is misregulated in disease. Overall, our study suggests that RNA granule-mitochondria contacts represent an important site for the mechanistic crosstalk between TDP-43 oxidation and the GADD34-PP1 complex, which has key consequences for TDP-43-associated pathogenesis in neurodegenerative diseases such as ALS and FTD.

Methods

Plasmids

The following plasmids were obtained from Addgene: pEGFP-C1-G3BP1-WT⁸² was a gift from Anthony Leung (Addgene plasmid # 135997; http://n2t.net/addgene:135997; RRID:Addgene_135997), mApple-TOMM20-N-10 was a gift from Michael Davidson (Addgene plasmid # 54955; http://n2t.net/addgene:54955; RRID:Addgene_54955), mito-BFP⁸³ was a gift from Gia Voeltz (Addgene plasmid # 49151; http://n2t.net/addgene:49151; RRID:Addgene_49151), pCS2+MLS-HyPer7¹⁹ was a gift from Vsevolod Belousov (Addgene plasmid # 136470; http://n2t.net/addgene:136470; RRID:Addgene_136470), PDHA1-N-10 was a gift from Michael Davidson (Addgene plasmid # 58195; http://n2t.net/addgene:58195; RRID:Addgene_58195), and Halo-TOMM20-N-10 was a gift from Kevin McGowan (Addgene plasmid # 123284; http://n2t.net/addgene:123284; RRID:Addgene_123284). mCherry-G3BP1 was obtained from GeneCopoeia (P0037-M55). The following plasmids were generated using VectorBuilder: TDP-43 Halo (WT, 5FL (F147L/F149L/F194L/F229L/F231L), 2KQ (K145Q/K192Q), Ox (C173G/C173V), M337V, 2CS (C173S/C175S), 4CS (C173S/C175S/C198S/C244S)), TagBFP2-GADD34 (WT and KARA (V556A/F558A)), mEmerald-GADD34 (WT and KARA (V556A/F558A)), and mScarlet-PPP1CA.

Cell culture and transfections

HeLa cells (ATCC CCL-2) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco; 11995-065) supplemented with 10% (vol/vol) FBS, 100 units per ml penicillin, and 100 μg/ml streptomycin for standard passaging and Control media conditions. Cells cultured in OXPHOS media were changed to the following media prior to transfection for 24 h before imaging (OXPHOS media^20,21: DMEM, no glucose (DMEM) (Gibco; 11966-025) supplemented with: 10 mM Galactose, 10% (vol/vol) FBS, 100 units per ml penicillin, and 100 μg/ml streptomycin). All cells were maintained at 37 °C in a 5% CO₂ incubator and were tested and found negative for mycoplasma contamination. Cells were transfected using Lipofectamine 2000 (Invitrogen, 11668019) 24 hrs prior to imaging or lysis, and 0.5 μg of DNA was used per plasmid. To confirm that expression levels of TDP-43 did not alter its response to Antimycin A treatment, titration experiments for TDP-43 expression were conducted using transfection of 0.5 μg (1×), 1 μg (2×), and 1.5 μg (3×) DNA plasmid for TDP-43 WT-Halo. GADD34 knockdown was performed using Lipofectamine RNAiMAX (Invitrogen, 13778075) 48 hrs prior to imaging or lysis for western blotting and coimmunoprecipitations, and cells were treated with 60 pmol of RNA per condition (Horizon Discovery; siScr (D-001810-10-20 (ON-TARGETplus Non-targeting Pool)); siGADD34 (L-004442-02-0005 (ON-TARGETplus Human PPP1R15A (23645) siRNA – SMARTpool))). For live imaging, cells were grown on glass-bottomed culture dishes (MatTek; P35G-1.5-14-C). All experiments were conducted in OXPHOS media except those listed in control media (Fig. 2h–l and Supplementary Fig. 3).

Drug treatments

For drug treatments, live cells were imaged while being treated with Antimycin A (10 μM; 45 min) (Sigma A8674), DMSO (2 μL; 45 min) (Sigma D2650), H₂O₂ (1 mM; 1 h) (Sigma H1009), Rotenone (2 μM; 1 h) (Sigma R8875), Oligomycin A (1 μM; 1 h) (Sigma 75351), CCCP (20 μM; 1 h) (Sigma C2759), Sodium arsenite (0.5 mM; 30 min) (Sigma 1062771000), or Puromycin (10 μg/ml; 1 h) (Thermo A1113803). For nutrient deprivation, cells were incubated in HBSS (Thermo 24020117) for 1 h prior to imaging. For heat shock experiments, cells were kept at 43 °C in a 5% CO₂ incubator for 1 h and during imaging. For antioxidant experiments, cells were treated with the non-thiol antioxidant Butylated hydroxyanisole (BHA) for 18 h (100 μM) (Sigma W218308) or the thiol antioxidant N-Acetyl-L-cysteine (NAC) for 1 h (10 mM) (Sigma A9165) prior to treatment with Antimycin A (45 min) or H₂O₂ (1 h). For nuclear export experiments, cells were treated with Leptomycin B (45 nM; 18–20 h) (Cell Signaling 9676).

Super-Resolution live Airyscan2 microscopy

Super-Resolution live Airyscan2 microscopy was conducted with a Zeiss LSM 980 confocal microscope with Airyscan2 detector using an α Plan-Apochromat 100x 1.46 Oil DIC 435 immersion objective (Zeiss) in a temperature-controlled chamber (37 °C) at 5% CO₂ using Zen Blue 3.3 (Zeiss). Airyscan2 images were processed in Zen Blue 3.3 using the following settings (2D SR processing; Auto Filter; Strength: standard). Time-lapse imaging movies were conducted at 1 frame every 7–10 s, 1–3 min for Figs. 1, 4, 5 and 7 and Supplementary Figs. 1 and 7; 1 frame every 17 s for 30–45 min for Figs. 2 and 3 and Supplementary Figs. 2, 3 and 5; 1 frame every 5 min for Supplementary Fig. 2i, j; and 1 frame every 22 s for 30–45 min for Supplementary Fig. 6.

Super-Resolution live structured illumination microscopy (SIM)

Super-Resolution live Lattice SIM² (structured illumination microscopy) images of RNA granules (GFP-G3BP1) and mitochondria (mApple-TOMM20) were acquired on a Zeiss Elyra 7 system with a Plan-Apochromat 63x/1.4 oil-immersion DIC M27 objective lens using ZEN Black 3.0 SR (Zeiss) software in a temperature-controlled chamber (37 °C) at 5% CO2 for the entire duration of imaging. Individual cells expressing all plasmids of interest were identified at random using the Locate function in ZEN Black 3.0 (Zeiss) for Supplementary Fig. 1. Images were processed using Lattice SIM² with the following parameters: 16 iterations, 0.065 regularization weight, 2× processing sampling, 4× output sampling, median filter fit, and detrend.

Electron microscopy

HeLa cells were grown on Thermanox plastic coverslips (Electron Microscopy Services 77280) for electron microscopy (EM) in OXPHOS media for 24 h, prior to fixation in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer for 1 h at room temperature, followed by 2–24 h at 4 °C. Cells were then post-fixed with 1% osmium tetroxide and 3% uranyl acetate, dehydrated in an ethanol series, embedded in Epon resin, and polymerized for 48 h at 60 °C. Ultrathin sections were made with a UCT ultramicrotome (Leica Microsystems) and contrasted with 4% uranyl acetate and Reynold’s lead citrate. Samples were imaged with an Eagle 4k HR 200 kV CCD camera on a FEI Tecnai Spirit G2 transmission electron microscope operated at 80 kV. RNA granules were identified as membraneless organelles consistent with structures previously confirmed by prior CLEM (correlative light electron microscopy) studies to contain RNA-binding proteins^40,41.

Immunofluorescence

Cells were plated on coverslips and transfected with Halo-G3BP1 24 h prior to fixation. Cells were washed with PBS and fixed in 4% (vol/vol) paraformaldehyde for 15 min and blocked and permeabilized with 2% BSA and 0.1% saponin. Fixed cells were incubated in primary antibody for 1 h, washed three times for 5 min each with blocking solution, incubated in Alexa-conjugated secondary antibodies (Invitrogen A11077 and A11029 (1:300)) for 1 h, washed three times for 5 min each with blocking solution, and mounted on glass slides with mounting medium (Vector Laboratories; H140010). Images were obtained using a Zeiss LSM 980 confocal microscope with Airyscan2 detector using an α Plan-Apochromat 100x 1.46 Oil DIC 435 immersion objective (Zeiss) using Zen Blue 3.3 (Zeiss). The following primary antibodies were used: TDP-43 (Proteintech, 10782-2-AP, 1:200) and TDP-43 Phospho (Ser409/410) (Biolegend, 829901, 1:150).

Co-immunoprecipitation

To examine co-immunoprecipitation of TDP-43 with GADD34 and PP1, HeLa cells were cotransfected for 24 h with FLAG-tagged TDP-43 (WT, Ox, M337V), myc-tagged GADD34, and V5-tagged PP1 using Lipofectamine 2000 (Invitrogen). Cells were lysed on ice with NP-40 lysis buffer (Boston BioProducts, BP-119) with Roche cOmplete Protease Inhibitor Cocktail (Sigma, 11697498001). Lysates were immunoprecipitated using Protein G-coupled Dynabeads (Invitrogen 10003D), incubated in anti-FLAG antibody (Sigma F7425), rotating for 30 min. Beads were subsequently washed 5× in with NP-40 and eluted by heating in 50 μL 2x Laemmeli sample buffer (Bio-Rad 1610737) for 20 min at 55 °C. Equal volumes of lysates and eluates were analyzed by SDS-Page and Western blot according to standard protocols.

Mitochondrial fractionation

Cells were transfected with FLAG-tagged TDP-43 WT, myc-tagged GADD34, and V5-tagged PP1 for 24 h prior to mitochondrial isolation and lysed with PBS with protease inhibitors (Thermo 78425), and the Mitochondria Isolation Kit (Thermo 89874) was used according to the manufacturer’s protocols using the reagent-based method. The final mitochondrial pellet was resuspended in 2x Laemmli sample buffer (Bio-Rad 1610737) and boiled for 5 min at 95 °C. Cytosolic and mitochondrial fractions were then analyzed by SDS-PAGE and western blotting.

Western blotting

HeLa cells were grown in 6-well plates for 72 h and transfected with 0.5 μg of DNA plasmid for 24 h or 60 pmol siRNA for 48 h. Cells were transferred to an ice bucket, lysed in RIPA buffer (Sigma, R0278-500ML) with Roche cOmplete Protease Inhibitor Cocktail (Sigma, 11697498001), and sonicated. Protein concentrations were determined with Pierce BCA assays (Thermo Scientific, 23223/23224), diluted in 4X Laemmli sample buffer (Bio-Rad, 1610747), and boiled for 5 min at 95 °C. Samples were subsequently loaded onto a Novex 4–20% Tris-Glycine gel (Invitrogen, XP04202BOX), run at 135 V for 1 h in Tris-Glycine SDS Running Buffer (Boston BioProducts; BP-150), and transferred to 0.2 µm Nitrocellulose membranes using the Trans Blot Turbo Transfer system (Bio-Rad, 1704158). Successful transfer was detected by Ponceau staining (Thermo Scientific, A40000279). Membranes were blocked with 5% milk in 1× Tris-buffered saline (50 mM Tris, pH 7.4, 150 mM NaCl) with 0.1% Tween (TBST) for 1 h at room temperature and incubated with a primary antibody, 4 °C, overnight: TDP-43 (Proteintech, 12892-1-AP, 1:1000), GADD34 (Proteintech, 10449-1-AP, 1:500), FLAG (Cell Signaling, 8146, 1:1000), Myc (Cell Signaling, 2276, 1:1000), V5 (Cell Signaling, 13202, 1:1000), TOMM20 (Abcam, 56783, 1:500), and GAPDH (Sigma, MAB374, 1:10,000). The next day, membranes were washed 3× with 1x TBST and incubated with a secondary goat anti-mouse and goat anti-rabbit HRP antibodies (Cell Signaling 7074 and 7076 (1:3000)) for 1 h. HRP signaling was developed using Clarity or Femto chemiluminescence substrate (Bio-Rad 1705061; Thermo 34096) and imaged using a Bio-Rad image reader. To visualize expression of different proteins on the same membrane, the membrane was incubated in Restore™ PLUS Western Blot Stripping Buffer (Thermo, 46430) for 15 min. Following this, the initial blocking step and primary and secondary antibody treatment were performed again. Bio-Rad Image Lab Software was used for quantification using the “Volume Tools: Rectangle” function. For quantification of TDP-43 expression or GADD34 expression levels, protein levels were normalized against GAPDH and renormalized to the average of the wild-type condition (TDP-43 or GADD34).

Image analysis

RNA granule-mitochondria contacts imaged in living cells were categorized as those that showed RNA granules and mitochondria in contact for >8 s in time-lapse images. All contacts analyzed for the minimum duration of contacts were those that had already formed at the beginning of the video. The minimum duration of contact in HeLa cells was quantified as the time before contact termination and dissociation (RNA granules and mitochondria detaching from one another) over a 3-min (178 s) (1f/7) video. For contact analysis in puromycin treatments, contacts were analyzed over a 3-min (178 s) (1f/10) video. Any contacts that lasted throughout the entire 3-min video and were still in contact by the end of the video were categorized as 178 s in bar graphs and histograms for the minimum duration of RNA granule-mitochondria contacts. The percentage of RNA granules in contact with mitochondria was quantified as the number of RNA granules in contact with mitochondria (tethered >8 s) in a region of interest containing ten RNA granules. For analysis of RNA granule-mitochondria contact tethering duration in TDP-43 conditions in siScr versus GADD34 siRNA conditions, only contacts that were clearly tethered throughout the entire contact tethering duration and for which both RNA granule and mitochondria could be clearly tracked after the untethering event occurred were analyzed. Linescans were generated using ImageJ (NIH) and normalized per protein for Figs. 1, 4, 5 and 6a and Supplementary Figs. 1 and 8. For linescans in Figs. 2, 3, 6g, i and 7, proteins were normalized to the maximum intensity of the cell.

TDP-43 localization was analyzed from Super-Resolution live microscopy images and categorized per cell into five mutually exclusive categories: nuclear (N), cytoplasmic (C), nuclear and cytoplasmic (N&C), cytoplasmic granules (G), and nuclear and cytoplasmic granules (N&G) for quantification in Figs. 2 and 3 and Supplementary Figs. 2, 3, 5, 6 and 9 are shown per TDP-43 condition. Graphs c and d in Supplementary Figs. 2, 3, 5, 6 and 9 are shown without statistical analyses. TDP-43 analysis was performed at baseline (Pre AA) and after Antimycin A treatment (Post AA) for each cell from >15 cells per experiment (N = 3 independent experiments) for each TDP-43 condition. The same cells were analyzed at baseline (Pre AA) and after Antimycin A treatment (Post AA) for the formation of G3BP1 granules. TDP-43 localization analysis using antioxidants and other stressors was only analyzed at the final time point for Supplementary Fig. 4a, b and Supplementary Fig. 4f–k.

TDP-43 colocalization with GADD34 was analyzed from Super-Resolution live microscopy images and quantified as the proportion of cells that showed TDP-43 (wild-type, (Ox) mutant, or (M337V) mutant) colocalized with GADD34. GADD34 colocalization with PP1 was quantified as the proportion of cells that showed GADD34 (wild-type or KARA mutant) colocalized with PP1. GADD34 or PP1 localization to mitochondria was quantified as the proportion of cells that showed GADD34 or PP1 localization around mitochondria (mito-BFP) in either GADD34 (wild-type) or GADD34 (KARA) conditions.

PP1 and TDP-43 granule dynamics were analyzed in TDP-43 (WT), TDP-43 (Ox), and TDP-43 (M337V) conditions in OXPHOS media with PP1 overexpression. In Supplementary Fig. 10a–c, PP1 granule dynamics were analyzed from Super-Resolution live microscopy images, and the proportion of cells were analyzed for five mutually exclusive categories, based on the category in which the majority of PP1 localized to: Lack of PP1 localization to granules (none), localization to PP1 granules by itself (alone), colocalization with RNA granules (G3BP1) (with RG), colocalization with TDP-43 granules (with TDP-43), or colocalization with both RNA granules (G3BP1) and TDP-43 (with RG + TDP-43). Data in Supplementary Fig. 10a–c are shown per TDP-43 condition without statistical analyses. For Fig. 7l, the following two categories were added together from this analysis (Supplementary Fig. 10a–c) to determine “PP1 granules with TDP-43”: colocalization with TDP-43 granules (with TDP-43) and colocalization with both RNA granules (G3BP1) and TDP-43 (RG + TDP-43). For Fig. 7m, the following two categories were added together from this analysis (Supplementary Fig. 10a–c) to determine “PP1 granules without TDP-43”: PP1 localization to granules by itself (alone) and colocalization with RNA granules (G3BP1) (with RG). The same dataset was further analyzed for TDP-43 dynamics, as shown in Supplementary Fig. 10d–f, where the proportion of cells were analyzed for five mutually exclusive categories, based on the category in which the majority of TDP-43 localized to: Lack of TDP-43 localization to granules (none), localization to TDP-43 granules by itself (alone), colocalization with RNA granules (G3BP1) (with RG), colocalization with PP1 granules (with PP1), or colocalization with both RNA granules (G3BP1) and PP1 (with RG + PP1). Data in Supplementary Fig. 10d–f are shown per TDP-43 condition without statistical analyses.

For Hyper7 imaging and analysis, HeLa cells were transfected with mitochondrial-targeted Hyper7 (Addgene 136470). Hyper7 was excited at 405 nm and 480 nm, and emission was collected at 517 nm over 45 min with 5 min intervals using a Zeiss LSM 980 confocal microscope with Airyscan2 detector using an α Plan-Apochromat 100x 1.46 Oil DIC 435 immersion objective (Zeiss) in a temperature-controlled chamber (37 °C) at 5% CO₂ using Zen Blue 3.3 (Zeiss). Movies were analyzed using ImageJ (NIH) to obtain the Hyper7 ratio (F = max fluorescent intensity of emission signals excited by 493 nm/353 nm). ΔF = F_t−F₀ was calculated for each time point and plotted by subtracting the ΔF at each time point from the average DMSO value at each time point for 5 cells per experiment (N = 3 independent experiments) for each condition.

For TMRE experiments, cells were treated with 100 nM TMRE for 30 min (Thermo T669) and washed 3× prior to imaging. Images were quantified using ImageJ (NIH), where the maximum TMRE intensity was measured per cell for at least 18 cells per experiment for 3 independent experiments. TMRE intensity was normalized to the average of the untreated control prior to Antimycin A treatment for each experiment.

For ATP experiments, cells were treated with 5 μM ATP-red live cell dye for 15 min (Sigma SCT045) and washed 2× prior to imaging. Cells were imaged following 45 min of DMSO (Ctrl) or 10 μM AA treatment. Images were quantified using ImageJ (NIH), where the maximum ATP intensity was measured per cell for at least 21 cells per experiment for 3 independent experiments. ATP intensity was normalized to the average of the control treatment (DMSO) for each experiment.

PP1 and TDP-43 interactions were analyzed for TDP-43 (Ox) by counting the number of cells that had granules containing both PP1 and TDP-43 (Ox) in siScr and siGADD34 cells from >15 cells per experiment (N = 3 independent experiments) for each condition. From this, the percentage of granules (which contained both PP1 and TDP-43 (Ox)) in contact with mitochondria was quantified as the number of granules in contact with mitochondria divided by the total number of granules, analyzed from a region of interest which contained 5–10 granules from N = 3 independent experiments.

Statistical analysis, graphing, and figure assembly

Data were analyzed using an unpaired two-tailed Student t-test (for two datasets) or a one-way ANOVA with Tukey’s post hoc test (for multiple datasets). All statistical tests were justified as appropriate and were analyzed from N ≥ 3 independent experiments per condition (see text and figure legends for details). Statistics and graphing were performed using Prism 10.4.1 (GraphPad) software. All videos and images were exported and assembled using Zen 3.5 (Zen Blue edition). Models were made in PowerPoint (Microsoft), and all final figures were assembled in Illustrator (Adobe).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

H.E.B. conducted all experiments. A.C.W. prepared plasmids. H.E.B. and Y.C.W. designed the overall study, analyzed data, prepared figures, and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was funded by the National Institutes of Health (NIH/NIGMS) DP2 New Innovator Award Grant DP2GM146322 to Y.C.W.

Data availability

All data generated in this study are included in the manuscript or are available upon request by contacting the corresponding author. Source data are provided with this paper.

Materials availability

Correspondence and requests for materials should be addressed to Y.C.W. (yvette.wong@northwestern.edu).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74009-9.