Abstract

   β-propeller protein-associated neurodegeneration (BPAN) is a rare
   X-linked neurodegenerative disorder caused by mutations in the WDR45
   gene, yet its molecular mechanisms remain poorly understood. Here, we
   identify a role for WDR45 in stress granule (SG) disassembly, mediated
   through its phase separation with Caprin-1. We demonstrate that WDR45
   forms gel-like condensates via its WD5 domain, which competitively
   displaces G3BP1 from Caprin-1 to promote SG disassembly.
   BPAN-associated WDR45 mutations impair condensate formation and
   Caprin-1 interaction, leading to delayed SG disassembly, which
   correlates with earlier disease onset. WDR45 depletion also exacerbates
   amyotrophic lateral sclerosis-associated pathological SGs, highlighting
   its broader relevance to neurodegenerative diseases. Using iPSC-derived
   midbrain neurons from a BPAN patient, we demonstrate delayed SG
   recovery, directly linking WDR45 dysfunction to neurodegeneration.
   These findings establish WDR45 as a critical regulator of SG dynamics,
   uncover a potential molecular basis of BPAN pathogenesis, and identify
   therapeutic targets for neurodegenerative diseases associated with SG
   dysregulation.

   Subject terms: Mechanisms of disease, RNA metabolism, Neurodegenerative
   diseases, Intrinsically disordered proteins
     __________________________________________________________________

   BPAN is a rare neurodegenerative disease caused by WDR45 mutations.
   Here, the authors discover that WDR45 can competitively displace G3BP1
   from Caprin-1 to promote stress granule disassembly, a function that is
   disrupted by BPAN-associated WDR45 mutations.

Introduction

   β-propeller protein-associated neurodegeneration (BPAN) is a rare
   X-linked neurodegenerative disorder characterized by developmental
   delays in childhood and progressive neurodegeneration in
   adulthood^[56]1,[57]2. Neuroimaging studies of BPAN patients have shown
   abnormal iron deposition in the basal ganglia, particularly in the
   globus pallidus, putamen, and substantia nigra^[58]3. In BPAN,
   degeneration of dopaminergic neurons and iron deposition in the
   substantia nigra pars compacta are implicated in the development of
   bradykinesia, a hallmark feature resembling Parkinson’s
   disease^[59]3,[60]4. Mutations in the gene encoding the WD repeat
   domain 45 (WDR45) protein have been identified as a primary cause of
   BPAN^[61]1,[62]2. WDR45, also known as WIPI4, is known to play
   important roles together with WDR45B in the process of autophagy, a
   critical cellular degradation and recycling pathway^[63]5–[64]8.
   Conventionally, BPAN pathogenesis has been attributed to autophagy
   dysfunctional caused by WDR45 mutations^[65]5,[66]6. However, emerging
   evidences suggest that WDR45 may also play critical roles beyond
   autophagy.

   Recent studies have revealed a more complex picture of BPAN
   pathogenesis. For instance, certain BPAN-related missense mutations
   produce neurodegenerative symptoms despite exhibiting near-normal
   autophagic activity, implying non-autophagic WDR45 functions are
   critical^[67]9. Furthermore, WDR45 has been implicated in preventing
   ferroptosis, a process distinct from autophagy, offering alternative
   explanations for BPAN pathogenesis^[68]10,[69]11. Additionally, WDR45
   deficiency has been shown to disrupt ER homeostasis and phospholipid
   metabolism^[70]12,[71]13. These findings suggest that the
   neurodegenerative mechanisms underlying BPAN might involve
   non-autophagic functions of WDR45, highlighting the need for further
   investigation into its diverse cellular roles.

   Stress granules (SGs), dynamic cytoplasmic condensates formed via
   liquid-liquid phase separation (LLPS), are essential for regulating
   mRNA translation and stability during cellular stress^[72]14–[73]16.
   However, persistent SGs, which can form due to chronic stress or
   disease-related mutations, are linked to the pathogenesis of several
   neurodegenerative diseases, including amyotrophic lateral sclerosis
   (ALS), frontotemporal dementia (FTD), inclusion body myopathy (IBM),
   and Charcot-Marie-Tooth disease (CMT)^[74]17–[75]28. While short-lived
   SGs are rapidly disassembled, persistent SGs, are typically removed
   through autophagy-dependent degradation^[76]23,[77]29,[78]30. Despite
   their importance, the precise mechanisms governing SG disassembly
   remain poorly understood, hindering the development of targeted
   therapies for SG-related neurodegenerative diseases.

   Caprin-1 (Cell Cycle Associated Protein 1), a core RNA-binding protein
   in SGs, regulates mRNA stability and translation^[79]31. In addition to
   its involvement in SG biology, Caprin-1 is essential for
   neurodevelopment^[80]32–[81]34. While it promotes SG assembly through
   interactions with Ras-GTPase-activating protein-binding protein (G3BP),
   it is not essential for SG formation^[82]15,[83]35,[84]36. However, the
   precise role of Caprin-1 in regulating SG dynamics remains unclear.

   Given WDR45’s potential involvement in processes beyond autophagy and
   the critical role of SG in neurodegeneration, we investigated the
   interplay between WDR45 and SG dynamics. Here we demonstrate that WDR45
   is recruited to SGs in response to various stress stimuli and is
   essential for SG disassembly through its interaction with Caprin-1. We
   show that WDR45 undergoes LLPS with Caprin-1 and possesses intrinsic
   gel-like properties that are crucial for SG disassembly. Furthermore,
   BPAN-associated mutations in WDR45 disrupt condensates formation and
   impair interaction with Caprin-1, leading to defective SG disassembly
   and pathological SGs accumulation. These findings provide mechanistic
   insights into BPAN pathogenesis, highlighting SG dysregulation as a
   potential therapeutic target in neurodegenerative diseases. Our study
   not only expands the understanding of WDR45 function but also reveals a
   layer of complexity in SG regulation, with broad implications for
   neurodegenerative disease research.

Results

WDR45 is incorporated into SGs and regulates SGs disassembly across different
cell lines and under various stress conditions

   The dynamics of SGs are known to be disrupted in various
   neurodegenerative diseases. To explore the potential involvement of SG
   in BPAN, we examined whether WDR45 is recruited into SGs under stress
   conditions. Immunostaining with antibodies against endogenous WDR45
   revealed a significant accumulation of WDR45 in cytoplasmic granules
   upon exposure to oxidative stress, heat shock (HS), or osmotic stress
   (Fig. [85]1a). These granules co-localized with established SGs markers
   G3BP1 in U2OS cells (Fig. [86]1a). Similarly, WDR45 was incorporated
   into SGs in human iPSC-induced neurons upon exposure to oxidative
   stress or heat shock (Fig. [87]1b and Supplementary Fig. [88]1a) and in
   SH-SY5Y upon oxidative stress (Supplementary Fig. [89]1b). These
   findings suggest that WDR45 recruitment into SGs is likely neither
   stress-specific nor cell type-dependent.

Fig. 1. WDR45 regulates SG disassembly under various stress conditions.

   [90]Fig. 1
   [91]Open in a new tab

   a Confocal images of U2OS cells treated with different stressors and
   stained with anti-G3BP1 (Red) and anti-WDR45 (Green) antibodies and
   DAPI (Blue) are shown. Stressors include sodium arsenite (SA, 500 μM,
   1 h), heat shock (HS, 44 °C, 2 h), and osmotic stress (Sorbitol,
   400 mM, 2 h). White boxes indicate the enlarged ROI and are displayed
   in separate channels (Zoom). Line graphs in the right panels show WDR45
   and G3BP1 signals along the indicated white lines in the insets. Scale
   bar, 10 μm. b Representative confocal microscopy images of neural
   progenitor cells induced from induced pluripotent stem cells stained
   with anti-PAX6 (Red) (Left). Following the differentiation and
   maturation process, neural progenitor cells were induced into midbrain
   neurons which were stained with anti-TUJ1 (Red) (Right). Scale bar, 10
   μm, and representative confocal images of G3BP1 (Red) and WDR45 (Green)
   in midbrain neurons treated with SA (SA, 500 μΜ, 45 min) and heat shock
   (HS, 1 h). Scale bar, 10 μm. The inset indicates the magnifying region
   of the white box. Inset scale bar, 1 μm. c WT, WDR45, WDR45B knockout
   cell lines and WDR45&WDR45B double knockout cell lines were untreated,
   treated with SA for 1 h, or recovered (Re) for 1.5 h, 2 h after SA
   removal. Confocal images of U2OS cells stained with anti-G3BP1 (Red)
   and DAPI (Blue) are shown. Scale bar, 10 μm. d Quantification of Cell
   with SGs percent (n = 3 independent experiments, every treatment cell
   number>200) under SA and at the recovery stage from stress as
   represented in (c). Data are shown as mean ± SD. ns, not significant,
   ****p < 0.0001 (two-way ANOVA with Tukey’s test). e WT, WDR45, WDR45B
   knockout cell lines and WDR45 WDR45B double knockout cell lines were
   untreated, treated with heat shock for 1 h, or recovered for 1.5 h, 2 h
   after removal of heat shock. Confocal images of U2OS cells treated with
   SGs induction stressors and stained with anti-G3BP1 (Red) and DAPI
   (Blue) are shown. Scale bar, 10 μm. f Quantification of Cell with SGs
   percent (n = 3 independent experiments, every treatment cell
   number>200) under heat shock and at the recovery stage from stress as
   represented in (e). Data were pooled from three independent
   experiments. Data are shown as mean ± SD. ns, not significant,
   ****p < 0.0001 (two-way ANOVA with Tukey’s test). Source data are
   provided as a Source Data file.

   Prompted by these findings, we next examined whether WDR45 regulates
   the dynamics of SGs. By silencing WDR45 using siRNA or employing
   CRISPR-mediated genetic knockout (KO) methods, we analyzed the dynamics
   of SGs induced by sodium arsenite (SA) or heat shock (HS)
   (Fig. [92]1c–f, Supplementary Figs. [93]1c–f and [94]2a–c). WDR45 KO in
   U2OS was validated through sequencing and immunostaining (Supplementary
   Fig. [95]1c, d). Intriguingly, genetic interference of WDR45 led to a
   significant delay in SG disassembly, while SG formation remained
   largely unaffected in the absence of WDR45 both across different cell
   lines, including Neuro-2a (N2a) and U2OS cells, and under alternative
   stress conditions, such as oxidative stress and heat shock
   (Fig. [96]1c–f, Supplementary Figs. [97]1e, f and [98]2a–c). Live-cell
   imaging of HeLa cells with G3BP1-GFP knock in under arsenite stress
   following WDR45 knock down also revealed the delayed SGs disassembly
   and fewer SGs per cell, though the size of the SGs remained unchanged
   (Supplementary Fig. [99]1g–j, Supplementary Movies [100]1–[101]4). To
   gain deeper insight into the role of WDR45 in SGs dynamics, we
   overexpressed WDR45 in U2OS cells. The results showed that
   overexpression of WDR45 did not influence SGs assembly upon arsenite
   stress (Supplementary Fig. [102]1k).

   We next investigated whether the other WIPI family proteins, namely
   WIPI1, WIPI2, and WDR45B play a similar role in SG disassembly as
   WDR45. We first examined their colocalization with SGs. GFP-tagged
   WDR45 but not WIPI1, WIPI2, or WDR45B colocalized with SGs
   (Supplementary Fig. [103]2d). Next, we performed siRNA-mediated
   silencing of WIPI1, WIPI2, WDR45B or ATG2A/2B. Among these, fonly the
   knockdown of WDR45B, a paralogue of WDR45, affected SG disassembly
   (Supplementary Fig. [104]2a–c).

   To further evaluate the role of WDR45B, we used CRISPR-mediated KO of
   WDR45B to deplete WDR45B or both WDR45 and WDR45B and evaluated the
   impact on SG dynamics in U2OS and N2a cells (Fig. [105]1c–f,
   Supplementary Fig. [106]1e, f and Supplementary Fig. [107]2a–c). While
   knockdown or KO of either WDR45 or WDR45B individually resulted in a
   substantial delay in SG disassembly, simultaneous silencing of both
   genes did not decelerate the delay (Fig. [108]1c–f, Supplementary
   Figs. [109]1e, f and [110]2a–c). Importantly, these effects were
   consistent across different cell lines and stress conditions
   (Fig. [111]1c–f, Supplementary Figs. [112]1e, f and [113]2a–c).
   Together, these results indicate that WDR45 and WDR45B have distinct,
   non-redundant roles in SG disassembly.

   Consistent with previous study by Ji et al. (2021)^[114]5, WDR45 and
   WDR45B single KO cells showed no significant autophagy defects
   (Supplementary Fig. [115]2e–g). These observations indicate that WDR45
   regulates SG disassembly likely through autophagy-independent pathway.
   This finding aligns with prior studies that SGs undergo disassembly
   process independently of autophagy mechanisms^[116]29. We also
   supplemented the experiment of SGs dynamics under heat shock, while
   cells were pretreated with 20 µM bafilomycin (an autophagy inhibitor)
   for 18 h. Results demonstrated that bafilomycin treatment had no
   influence on SGs dynamics under this condition (Supplementary
   Fig. [117]2h, i). Given that WDR45 depletion may promote ER
   stress^[118]12, we confirmed that CRISPR-mediated knockout of WDR45 did
   not significantly alter the expression of HSPA5, indicating that WDR45
   depletion does not permanently affect ER stress levels (Supplementary
   Fig. [119]2j, k). In contrast, siRNA-mediated knockdown of WDR45
   transiently increased HSPA5 expression, potentially leading to ER
   stress (Supplementary Fig. [120]2l). Notably, defects in SG disassembly
   were observed in both WDR45 KO and KD cells (Fig. [121]1c–f,
   Supplementary Fig. [122]2b, c). Together, our results suggest that
   WDR45-dependent SG disassembly is mediated by distinct mechanisms
   separate from its autophagy-related or ER stress-related functions,
   highlighting a role for WDR45 in cellular stress responses.

Caprin-1 forms condensates with WDR45 and facilitates its recruitment into
SGs

   To gain insight into the mechanisms underlying the incorporation of
   WDR45 into SGs and its role in modulating disassembly, we analyzed the
   interactomes of WDR45 proteins in human cells under both physiological
   and stress conditions. WDR45-GFP was expressed in human U2OS cells,
   immunoprecipitated with anti-GFP nanobody, and co-purified proteins
   were analyzed by liquid chromatography-tandem mass spectrometry
   (LC-MS/MS) (Fig. [123]2a, Supplementary Table [124]1).

Fig. 2. WDR45 interacts with Caprin-1 and undergoes LLPS with Caprin-1.

   [125]Fig. 2
   [126]Open in a new tab

   a Comparative proteomic analysis of GFP (control group) and WDR45-GFP
   (experimental group) under SA treatment and untreated conditions. Cells
   expressing either GFP or WDR45-GFP were treated with SA for 1 hour or
   left untreated as controls. Each experimental condition was performed
   in triplicate to ensure reproducibility and statistical robustness
   (Created with Biorender.com). b Venn diagrams comparing the proteomes
   of CON-WDR45 and SA-WDR45 with the core components of SGs. The green
   box highlights the shared proteins between CON-WDR45 proteome and the
   SG core components. The orange box highlights the shared proteins
   between SA-WDR45 proteome and the SG core components. c Western blot
   analysis of immunoprecipitation (IP) experiments using GFP (control),
   WDR45-GFP, and WDR45B-GFP. IP was performed using antibodies against
   GFP to assess the interaction with Caprin-1. d Representative confocal
   images of U2OS cells under control and SA treatment (500 μM 45 min)
   stained with anti-Caprin-1 (Red) and anti-WDR45 (Green). DAPI is shown
   in blue. Scale bar, 10 μm. e Representative confocal images of U2OS WT
   and CAPRIN1 KO cells under SA treatment (500 μM 45 min) stained with
   anti-G3BP1 (Red) and anti-WDR45 (Green). DAPI is shown in blue. Scale
   bar, 10 μm, Zoom’s Scale bar, 1 μm. f Representative SIM (Structure
   Illumination Microscopy) images of G3BP1-positive SGs (Red) and WDR45
   (Green) in WT and CAPRIN1 KO U2OS cells. Scale bar, 1 μm, Zoom’s Scale
   bar, 0.5 μm. g Scatterplot representing the co-localization correlation
   between G3BP1 and WDR45 signal (WT and CAPRIN1 KO cells) during the SA
   stage as represented in (f). Data are shown as mean ± SD. Statistical
   significance was determined using an unpaired and two-sided t-test
   comparison. Each cell represents the Pearson coefficient analysis of
   cell ROI. ****P < 0.0001. n = 20 from 3 independent experiments. h
   Relative fluorescence intensity of WDR45-GFP before and after
   photobleaching over time. Data are represented as mean ± SEM for n = 7.
   i Dynamics of WDR45 foci indicated by the yellow circle in U2OS cells
   stably transfected with WDR45-GFP were analyzed using FRAP.
   Representative images of the same foci before and after photobleaching
   are shown. Scale bar, 1 μm. j Phase diagram of WDR45 in vitro. k
   1,6-hexanediol (1,6-HD) destroys the WDR45 condensates in vitro. Scale
   bar, 10 μm. l LLPS of purified recombinant G3BP1 (Cyan), Caprin-1
   (Red), and WDR45 (Green) in vitro. Recombinant proteins were mixed
   pairwise at a 1:1 molar ratio (Caprin-1 + WDR45, G3BP1 + Caprin-1,
   G3BP1 + WDR45) to observe LLPS under physiological conditions with
   150 mM NaCl. Scale bar, 10 μm. m Dynamics of WDR45 (or +Caprin-1) foci
   (Green) indicated by yellow circle in vitro were analyzed using FRAP.
   Representative images of the same foci before and after photobleaching
   are shown. Scale bar, 5 μm. n Relative fluorescence intensity of WDR45
   (or +Caprin-1) before and after photobleaching over time. Data are
   represented as mean ± SEM. ****p < 0.0001 by paired and two-sided
   t-test. n = 17 for WDR45+Caprin-1, n = 7 for WDR45 alone. Source data
   are provided as a Source Data file.

   Our results indicated significant changes in WDR45-interacting proteins
   under stress conditions. While 300 interactors were identified under
   normal conditions, this number was reduced to 79 following arsenite
   treatment (Fig. [127]2b). This suggests that WDR45 is responsive to
   cellular stress. Subsequently, we performed pathway enrichment analysis
   using the online tool ShinyGo^[128]37. The analysis revealed that
   pathways associated with neurodegenerative diseases—such as Amyotrophic
   Lateral Sclerosis, Alzheimer’s disease, Huntington’s disease, and
   Parkinson’s disease—were enriched across both conditions (Supplementary
   Fig. [129]3a). These findings underscore the potential role of WDR45 in
   neurodegenerative diseases.

   Comparing the WDR45 interactome under normal conditions and stress
   conditions with the core SGs network published by Yang^[130]15, we
   identified Caprin-1 as the only overlapping protein (Fig. [131]2b). We
   confirmed the interaction between WDR45 and Caprin-1 using
   co-immunoprecipitation and Glutathione S-transferase (GST) pull-down
   (Fig. [132]2c and Supplementary Fig. [133]7c). Furthermore, WDR45
   co-localized with Caprin-1 under oxidative stress in U2OS cells,
   iPSC-derived neurons and SH-SY5Y (Fig. [134]2d and Supplementary
   Fig. [135]3b, c). In addition, WDR45B also interacted with Caprin-1
   (Fig. [136]2c). To study the WDR45 interactors more comprehensively, we
   also confirmed the interaction between WDR45 and NUP98 and G3BP1, with
   no detectable binding observed under standard assay conditions.
   (Supplementary Fig. [137]3d). Given the critical role of Caprin-1 in
   the core network of stress granules, Caprin-1 was selected for
   subsequent analysis.

   To determine whether WDR45 incorporation into SGs is dependent on
   Caprin-1, we generated a CAPRIN1 KO cell line using the CRISPR-Cas9,
   which showed complete loss of Caprin-1 expression (Supplementary
   Fig. [138]3e). The incorporation of WDR45 into SGs was significantly
   reduced in CAPRIN1 KO cells (Fig. [139]2e). The result demonstrated
   that Caprin-1 is essential for WDR45 recruitment to SGs.
   Super-resolution structured illumination microscopy (SIM) further
   confirmed this finding. In WT cells, WDR45 showed a high degree of
   co-localization with SGs, as indicated by the high Pearson correlation
   coefficient (Fig. [140]2f, g). However, in CAPRIN1 KO cells, the
   Pearson correlation coefficient between WDR45 and SGs was significantly
   reduced, suggesting that the recruitment of WDR45 to SGs is impaired in
   the absence of Caprin-1 (Fig. [141]2f, g).

   The enrichment of WDR45 in SGs promoted us to investigate the dynamics
   of WDR45 puncta in the cytosol using fluorescence recovery after
   photobleaching (FRAP). WDR45-GFP puncta in cytosol were highly dynamic,
   recovering and plateauing within 60 s post-photobleaching
   (Fig. [142]2h, i). Caprin-1 has been shown to undergo LLPS with RNA,
   G3BP1, or FMRP, but not individually^[143]14,[144]33,[145]36,[146]38.
   We hypothesized that WDR45 might form droplets in vitro or modulate the
   LLPS properties of Caprin-1.

   To test this, we purified MBP-HIS-WDR45 (hereafter WDR45) and
   MBP-HIS-mCherry-Caprin-1 (hereafter Caprin-1) and removed the MBP-HIS
   tag using 3 C protease for in vitro phase separation assay. The
   cellular concentration of WDR45 is 34 nM^[147]39. At physiological
   concentrations, WDR45 formed condensates in vitro without requiring any
   crowding agent (Fig. [148]2j, Supplementary Fig. [149]3f). This
   observation is consistent with WDR45 being one of the top-scored phase
   separation candidates according to the DosPS predictor^[150]40, despite
   its relatively low percentage of disordered regions. However, in the
   presence of RNA or PEG, WDR45 protein produced amorphous, fiber-like
   aggregates (Supplementary Fig. [151]3g). We further employed
   1,6-hexanediol (1,6-HD), which inhibits phase separation by disrupting
   multivalent hydrophobic protein-protein interactions^[152]17. These
   WDR45 condensates were no longer observed in the presence of 1,6-HD,
   indicating that the hydrophobic interactions drive WDR45 condensates
   formation (Fig. [153]2k).

   Consistent with previous reports, Caprin-1 did not form droplets
   individually but did undergo LLPS with G3BP1, RNA, or in the presence
   of 10% PEG (Fig. [154]2l and Supplementary Fig. [155]3g). When mixed
   with purified full-length WDR45, Caprin-1 formed condensates with
   WDR45, while WDR45 did not form condensates with G3BP1 (Fig. [156]2l
   and Supplementary Fig. [157]4a). With increasing WDR45 concentration,
   WDR45/Caprin-1 condensate correspondently grew in size, demonstrating
   the concentration-dependent nature of WDR45/Caprin-1 condensate
   (Supplementary Fig. [158]4b, c). Notably, purified WDR45B did not form
   condensate, even with Caprin-1, which is consistent with the
   observation that WDR45B did not form puncta under stress conditions
   (Supplementary Fig. [159]2d, [160]4a).

   Next, using FRAP on in vitro WDR45 condensates with or without
   Caprin-1, we found that the FRAP ratio of WDR45 condensates without
   Caprin-1 was approximately 10%, significantly lower than that with
   Caprin-1 (∼60%) (Fig. [161]2m, n). This difference in FRAP ratios
   indicates that WDR45 homotypic interaction-mediated gel-like phase
   separation, while heterotypic interactions with Caprin-1 promote a
   transition to a more dynamic, liquid-like state. Consistently,
   WDR45/Caprin-1 droplet fused completely within 12 seconds, whereas
   WDR45 alone exhibited slower fusion rates (Supplementary Fig. [162]4d,
   e). These results consistent with the observation that WDR45 interacted
   with Caprin-1 and formed dynamic puncta in cell culture (Fig. [163]2c,
   d, h, i). These findings demonstrate that Caprin-1 plays a critical
   role in modulating the dynamic properties of WDR45 condensates,
   facilitating a transition from a static, gel-like state to a dynamic,
   liquid-like phase. This transition likely enables WDR45 to rapidly
   respond to cellular stress, highlighting its functional versatility in
   SG dynamics and stress responses.

WDR45 competes with G3BP1 for Caprin-1 binding to facilitate SG disassembly

   To investigate whether the interaction between WDR45 and Caprin-1 is
   critical for SG disassembly, we examined their binding dynamics during
   stress and recovery. Cells stably expressing WDR45-GFP were subjected
   to stress and monitored during recovery at 1.5 and 2 hours. The
   interaction between WDR45-GFP and Caprin-1 increased significantly upon
   stress, peaking at 1.5 hours of recovery (Fig. [164]3a, b). This
   suggests that WDR45 binds more tightly to Caprin-1 during stress and
   recovery phases, potentially facilitating SG disassembly.

Fig. 3. WDR45 competes with G3BP1 for Caprin-1 binding to facilitate SG
disassembly.

   [165]Fig. 3
   [166]Open in a new tab

   a Western blot analysis of immunoprecipitation (IP) experiments
   examining the Caprin-1 enrichment capability in U2OS cells expressing
   WDR45-GFP under different conditions: control (untreated), SA treatment
   for 1 hour (SA 1 h), and recovery periods of 1.5 hours and 2 hours
   post-treatment. (n = 3 independent experiments). b Quantification of
   Caprin-1 enrichment ability with WDR45-GFP under SA treatment and
   during recovery stages, as represented in (a). Data were pooled from
   three independent experiments and normalized to the control condition.
   Results are presented as mean ± SD, with statistical significance
   indicated by *p < 0.05 and **p < 0.01, determined by one-way ANOVA with
   Tukey’s test. c WT, WDR45, CAPRIN1 knockout cell lines and
   WDR45&CAPRIN1 double knockout cell lines were untreated, treated with
   SA for 1 h, or recovered for 1.5 h, 2 h after removal of SA. Confocal
   images of U2OS cells treated with SGs induction stressors and stained
   with anti-G3BP1 (Green) and DAPI (Blue) are shown. Scale bar, 10 μm. d
   Quantification of Cell with SGs percent under SA (n = 3 independent
   experiments) and at the recovery stage from stress as represented in
   (c). Data were pooled from three independent experiments. Data are
   shown as mean ± SD. ns, not significant, ****p < 0.0001 (two-way ANOVA
   with Tukey’s test). e Western blot analysis of endogenous
   immunoprecipitation (IP) performed with G3BP1 to assess the interaction
   with Caprin-1 in wild-type (WT) and WDR45 knockout (KO) cells. The
   enrichment ability of G3BP1 for Caprin-1 was assessed under two
   conditions: control (untreated) and SA treatment (SA). (n = 3
   independent experiments). f Quantification of G3BP1’s ability to enrich
   Caprin-1 under control and SA treatment conditions, as shown in (e).
   Data were pooled from three independent experiments and normalized to
   Caprin-1 enrichment in WT cells under the control condition. Results
   are presented as mean ± SD, with statistical significance indicated by
   *p < 0.05, determined by one-way ANOVA with Tukey’s test. g Scatterplot
   representing the co-localization correlation between G3BP1 and Caprin-1
   signal (WT and WDR45 KO cells) during the SA stage. Data are shown as
   mean ± SD. Statistical significance was determined using an unpaired
   and two-sided t-test. ***p < 0.001. n = 20 from 3 independent
   experiments. h Representative SIM (Structure Illumination Microscopy)
   images of G3BP1-positive SGs (Red) and Caprin-1 (Green) in WT and WDR45
   KO U2OS cells. Scale bar, 1 μm, Zoom’s Scale bar, 0.5 μm. (i), LLPS of
   purified recombinant G3BP1 (Cyan), Caprin-1 (Red), and WDR45 (Green) in
   vitro. Keep the concentrations of G3BP1 and Caprin-1 constant while
   gradually increasing the concentration of WDR45 to observe LLPS under
   physiological conditions with 150 mM NaCl. LLPS was visualized using
   confocal microscopy. Scale bar, 10 μm. j LLPS of purified recombinant
   G3BP1 (Cyan), Caprin-1 (Red), and WDR45 (Green) in vitro. Keep the
   concentrations of WDR45 and Caprin-1 constant while gradually
   increasing the concentration of G3BP1 to observe LLPS under
   physiological conditions with 150 mM NaCl. Scale bar, 10 μm. Source
   data are provided as a Source Data file.

   To explore whether WDR45 modulates SG dissolution through a
   Caprin-1-dependent pathway, we generated CAPRIN1 single KO cells and
   WDR45&CAPRIN1 dKO cells, then monitored the dynamics of SGs induced by
   SA or HS. CAPRIN1 KO rescued the dissolution defect observed in WDR45
   KO cells (Fig. [167]3c, d and Supplementary Fig. [168]5a, b).
   Interestingly, genetic interference with CAPRIN1 individually led to
   faster dissolution of SGs, whereas SGs formation was largely unaffected
   in the absence of Caprin-1 under both stress conditions (Fig. [169]3c,
   d, Supplementary Fig. [170]5a, b). We also conducted the FRAP assay in
   WT and WDR45 KO cells to exam whether WDR45 regulates SGs mobility. The
   mobility of GFP-G3BP1 puncta remained unchanged following WDR45
   knockout, suggesting that WDR45 does not regulate SG dissolution by
   altering SG mobility (Supplementary Fig. [171]5f, g). In addition,
   WDR45 depletion did not affect the protein levels of G3BP1 or Caprin-1
   (Supplementary Fig. [172]5c–e). These findings indicate that WDR45
   modulates SG dissolution through a Caprin-1-dependent pathway.

   SGs assemble through a core protein-RNA interaction network, with G3BP
   serving as the central node of this network^[173]15. Based on our
   observations, we hypothesized that WDR45 competes with G3BP1 for
   Caprin-1 binding, thereby modulating the SG core network and dynamics.
   To test this hypothesis, we first validated that the interaction
   between G3BP1 and Caprin-1 increased during stress and decreased during
   recovery, using a HeLa G3BP1-GFP knock-in cell line (Supplementary
   Fig. [174]5h). In contrast, the interaction between WDR45-GFP and
   Caprin-1 increased during recovery, peaking at 1.5 h post-stress
   (Fig. [175]3a, b). To further investigate this competition, we
   performed Co-IP in WDR45 KO cells and G3BP1/2 dKO cells. Interestingly,
   the interaction between G3BP1 and Caprin-1 significantly enhanced in
   WDR45 KO cells under both control and stress conditions (Fig. [176]3e,
   f). Conversely, the interaction between WDR45 and Caprin-1
   significantly enhanced in G3BP1/2 dKO cells under both conditions
   (Supplementary Fig. [177]5i). Given that previous studies have shown
   that the interaction between G3BP1 and Caprin-1 is
   RNA-dependent^[178]36, we next examined the interaction between WDR45
   and Caprin-1 following RNase treatment. Following RNase treatment, we
   observed a significant enhancement in the interaction between WDR45 and
   Caprin-1 (Supplementary Fig. [179]5j). Furthermore, we mapped the
   interaction domain of WDR45 and Caprin-1 to the NTD-GIM domain of
   Caprin-1, which is also responsible for its interaction with
   G3BP1^[180]36 (Supplementary Fig. [181]5k, l). Together, these findings
   demonstrate that WDR45 competes with G3BP1 for binding to Caprin-1,
   providing a mechanistic basis for its role in modulating SG dynamics.

   To further confirm this finding, we conducted SIM-based
   super-resolution imaging. WDR45 KO cells exhibited an increased Pearson
   correlation coefficient between G3BP1 and Caprin-1 (Fig. [182]3g, h).
   Moreover, SIM-based super-resolution imaging revealed that the
   molecular density of G3BP1 in SGs significantly increased in WDR45 KO
   cells, whereas it was reduced in CAPRIN1 KO cells (Supplementary
   Fig. [183]5m, n). These results also suggest that WDR45 and Caprin-1
   have opposing roles in regulating the SG core, with WDR45 KO leading to
   a strengthened and more rigid SG core, while CAPRIN1 KO results in a
   looser and less stable core.

   We next asked whether WDR45 and G3BP1 compete directly for Caprin-1
   binding in vitro. In GST pull-down assays, purified GST–Caprin-1
   efficiently bound WDR45, but the addition of purified G3BP1 markedly
   reduced this interaction (Supplementary Fig. [184]5o).

   To further determine whether the competition is dose-dependent, we
   conducted in vitro phase separation experiments using three purified
   proteins. WDR45 could undergo phase separation with G3BP1 and Caprin-1
   together to form droplets. As the concentration of WDR45 increased,
   G3BP1 was excluded from the droplets (Fig. [185]3i). Conversely,
   increasing the concentration of G3BP1 excluded WDR45 from the droplets
   (Fig. [186]3j). Notably, high concentration of WDR45 led to irregularly
   shaped droplets, indicating potentially gel-like condensates
   (Fig. [187]3i). FRAP assay showed that increasing concentrations of
   WDR45 reduced droplet mobility, consistent with the formation of
   gel-like condensates (Supplementary Fig. [188]5p, q).

   Previous studies have shown that stress granules contain a stable core
   structure^[189]30 and SG assembly is driven by a core protein-RNA
   network^[190]15. Our findings reveal a critical regulatory mechanism
   underlying SG dynamics: WDR45 competes with G3BP1 for Caprin-1 binding,
   loosening the SG core and facilitating SG disassembly during the
   recovery process. This competition highlights the essential interplay
   between WDR45, Caprin-1, and G3BP1 in destabilizing the SG core network
   and regulating SG dynamics.

The WD5 domain of WDR45 mediates condensates formation and SG disassembly

   WDR45, a member of the WIPI protein family, has a relatively simple
   structure composed of seven WD (tryptophan-aspartate) domains with
   significantly varying amino acid compositions (Fig. [191]4a). To
   identify the domains responsible for WDR45’s function in SGs, we
   systematically deleted segments of its WD domain (Fig. [192]4a and
   Supplementary Fig. [193]6a).

Fig. 4. WD5 domain in WDR45 mediates LLPS with Caprin-1.

   [194]Fig. 4
   [195]Open in a new tab

   a Schematic diagram of WDR45 domain structure and amino acid sequence
   of each domain. Truncations of ΔWD1-4, ΔWD5-7, ΔWD5-6, ΔWD6-7, ΔWD7,
   ΔWD6, ΔWD5 are shown. b Western blot analysis of immunoprecipitation
   (IP) experiments examining the Caprin-1 enrichment capability in U2OS
   cells expressing GFP, WDR45-GFP and its truncations under control
   (untreated). c LLPS of purified recombinant Caprin-1 (Red), WDR45 and
   its truncations (Green) in vitro. Recombinant proteins were mixed
   pairwise at 1:1 molar ratio to observe LLPS under physiological
   conditions with 150 mM NaCl. Scale bar, 10 μm. d Confocal images of
   U2OS cells (expressing GFP, WDR45-GFP and its truncations) that were
   treated with SA and stained with anti-G3BP1 (Red) antibody and DAPI
   (Blue) are shown. Scale bar, 10 μm. e Confocal images of WDR45 KO U2OS
   cells (expressing GFP, WDR45-GFP and its truncations) that were treated
   with SA and recovery 2 h stained with anti-G3BP1 (Red) antibody and
   DAPI (Blue) are shown (n = 3 independent experiments, every treatment
   cell number>200). Scale bar, 10 μm. f Quantification of Cell with SGs
   percent under SA, at the recovery stage from stress as represented in
   (e). Data were pooled from three independent experiments. Data are
   shown as mean ± SD. ns, not significant, **p < 0.01, ****p < 0.0001
   (one-way ANOVA with Tukey’s test). Source data are provided as a Source
   Data file.

   Initial analysis of purified truncated proteins revealed that the
   ΔWD1-4 segment retained condensation properties, whereas ΔWD5-7 did
   not, indicating that the WD5-7 region is critical for WDR45 phase
   separation (Supplementary Fig. [196]6a). Further truncation experiments
   with ΔWD7, ΔWD6-7, ΔWD5-6, ΔWD5, and ΔWD6 variants demonstrated that
   the WD5 and WD6 domains are essential for mediating WDR45 phase
   separation (Supplementary Fig. [197]6a). Specifically, deletion of WD6
   partially inhibited phase separation, whereas deletion of WD5
   completely abolished it (Supplementary Fig. [198]6a). Notably, the
   isolated WD5 domain formed aggregates in vitro, suggesting that WD5
   mediates the gel-like properties of WDR45, while other domains may
   stabilize or enhance WD5’s phase separation capability (Supplementary
   Fig. [199]6b).

   To further investigate the role of WD5 in WDR45’s interaction with
   Caprin-1, we performed Co-IP assays. Truncated variants lacking WD5
   (ΔWD5 and ΔWD5-6) lost the ability to interact with Caprin-1, whereas
   variants lacking only WD6 (ΔWD6) retained this interaction both in cell
   culture and in vitro (Fig. [200]4b, Supplementary Fig. [201]6c).
   Consistently, ΔWD5 variants failed to form droplets with Caprin-1 in
   vitro, highlighting the importance of WDR45 gel-like condensates in
   facilitating Caprin-1 LLPS (Fig. [202]4c).

   Given that WDR45 forms condensates with Caprin-1 via the WD5 domain, we
   next examined whether WDR45’s incorporation into SGs also depends on
   WD5. Using U2OS cell lines expressing WDR45-GFP with deletions in WD1-4
   (ΔWD1-4) or WD5-7 (ΔWD5-7), we observed that under stress conditions,
   WDR45-GFP(ΔWD5-7) failed to localize to SGs, whereas WDR45-GFP(ΔWD1-4),
   which retains phase separation properties, successfully incorporated
   into SGs (Fig. [203]4d). Further analysis of ΔWD5 and ΔWD6 variants
   confirmed that deletion of WD5, but not WD6, abolished SG localization
   (Fig. [204]4d).

   To assess the functional relevance of WD5 in SG disassembly dynamics,
   we examined WDR45 KO cells. WDR45-GFP(ΔWD5) failed to rescue SG
   disassembly, whereas both WDR45-GFP(ΔWD6) and wild-type WDR45
   successfully restored normal SG dynamics (Fig. [205]4e, f).

   To further explore the role of WD5, we analyzed a truncated version of
   WDR45 (Δ171–191, termed “del peak”) predicted using the FuzDrop online
   tool ([206]https://fuzdrop.bio.unipd.it/predictor)^[207]41
   (Supplementary Fig. [208]6d). The del peak protein was unable to form
   condensates, suggesting that this region is critical for phase
   separation (Supplementary Fig. [209]6e). Previous studies have shown
   that substituting serine (S) with proline (P) can significantly alter
   the functional properties of various protein domains, particularly in
   highly conserved serine residues within WD repeat domains, leading to
   structural changes^[210]42. Based on this, we focused on serine 183
   (S183), located within the overlapping region of the WD5 domain and the
   peak. By generating S183P and S183A mutants, we discovered that
   mutations at S183 abolished WDR45’s ability to form condensates,
   weakened its interaction with Caprin-1, and failed to rescue SGs
   disassembly defects in WDR45 KO cells (Supplementary Fig. [211]6e–h).

   These findings demonstrate that the WD5 domain, particularly S183, is
   critical for condensate formation, mediating the interaction between
   WDR45 and Caprin-1, and regulating SG disassembly. This highlights the
   importance of the WD5 region in maintaining proper SG dynamics and
   provides insights into the molecular mechanisms underlying WDR45
   function.

BPAN-associated mutations in WDR45 impair phase separation and SG disassembly

   Seven female patients developed neurodegeneration symptoms and there
   was no clear correlation between the autophagic activity and the age of
   symptoms onset^[212]9. To assess whether BPAN-associated mutations in
   WDR45 directly affect its phase separation properties, we constructed
   and purified proteins for following eight BPAN-related WDR45 mutations
   associated with neurodegeneration symptoms at known ages: R13P, N61K,
   L98P, N202K, G204D, A209D, S210P and S251del (Fig. [213]5a). These
   mutant proteins exhibited a marked sensitivity to both salt and protein
   concentration for condensate formation (Supplementary Fig. [214]7a, b).
   Compared to wild-type WDR45, all BPAN-related mutations significantly
   reduced condensate formation capacity, with mutations in the WD5 domain
   (N202K, G204D, A209D, and S210P) reducing condensate formation capacity
   to only 10–20% of wild-type levels. (Fig. [215]5a, b, Supplementary
   Fig. [216]7a, b).

Fig. 5. BPAN-associated mutations in WDR45 impair its function in SG
disassembly.

   [217]Fig. 5
   [218]Open in a new tab

   a Schematic diagram of β-propeller protein-associated neurodegeneration
   (BPAN) related WDR45 mutations’ domain structure. b The phase
   separation ability of respective WDR45 mutants was graphed alongside
   the age of onset for neurodegenerative symptoms. The age of onset for
   these symptoms is provided in parentheses after each mutation. The
   Pearson’s correlation coefficient (R) reflecting the correlation
   between the extent of condensates formation ability defects and the age
   of onset for these symptoms is presented. c Western blot analysis of
   Co-IP experiments using GFP-trap to assess the interaction between
   WDR45-GFP, and its mutations with Caprin-1. d LLPS of purified
   recombinant Caprin-1 (Red), WDR45 and its mutations (Green) in vitro.
   Recombinant proteins were mixed pairwise at 1:1 molar ratio to observe
   LLPS under physiological conditions with 150 mM NaCl. Scale bar, 10 μm.
   e Quantification of droplets area for Caprin-1 and WDR45 and its
   mutations as represented in (d). Data are shown as mean ± SD. ns, not
   significant, ****p < 0.0001 (one-way ANOVA with Tukey’s test). WT:
   n = 72; R13P: n = 63; L98P: n = 43; N202K: n = 45; G204D: n = 49;
   A209D: n = 51; S210P: n = 52; S251 del: n = 56 from 3 independent
   experiments. f Confocal images of WT U2OS cells (expressing GFP,
   WDR45-GFP and its mutations) that were treated with SA and stained with
   anti-G3BP1 (Red) antibody and DAPI (Blue) are shown. Scale bar, 10 μm.
   g Summary of WDR45 and its mutants in relation to SG localization. h
   Confocal images of WDR45 KO U2OS cells (expressing GFP, WDR45-GFP and
   its mutations) that were treated with SA and recovery 2 h stained with
   anti-G3BP1 (Red) antibody and DAPI (Blue) are shown (n = 3 independent
   experiments, every treatment cell number>200). Arrowheads indicate the
   successfully transfected cells. Scale bar, 10 μm. i Quantification of
   Cell with SGs percent under SA, at the recovery stage from stress as
   represented in (h). Data were pooled from three independent
   experiments. Data are shown as mean ± SD. ns, not significant,
   ***p < 0.001, ****p < 0.0001 (one-way ANOVA with Tukey’s test). Source
   data are provided as a Source Data file.

   These findings directly demonstrate that BPAN-associated WDR45
   mutations impair the protein’s ability to form condensate, likely
   disrupting its physiological role. Given that BPAN is characterized by
   severe neurodegeneration, we explored whether the degree of
   condensation impairment correlates with disease progression. Linear
   regression analysis revealed a strong positive correlation between the
   age of BPAN onset and the extent of phase separation impairment
   (Pearson correlation coefficient R = 0.598) (Fig. [219]5b). Mutations
   associated with earlier disease onset exhibited the most severe defects
   in phase separation capacity (Fig. [220]5b). These results suggest that
   WDR45’s condensate formation capability is closely tied to BPAN
   progression, potentially through interference with its role in SG
   dynamics.

   Further analysis confirmed that these mutant WDR45 proteins exhibited
   reduced condensate formation properties and failed to interact with
   Caprin-1 both in cell culture and in vitro (Fig. [221]5c, Supplementary
   Fig. [222]7c). Additionally, these mutations showed significantly
   reduced ability to promote Caprin-1 phase separation (Fig. [223]5d, e).

   To investigate whether BPAN-associated WDR45 mutations affect SG
   dynamics, we introduced these mutations into WDR45-GFP constructs.
   Under stress conditions, mutations within the WD5 domain (N202K, G204D,
   A209D, S210P, and S251del) failed to localize to SGs and were unable to
   rescue disassembly defects in WDR45 KO cells (Fig. [224]5f–i,
   Supplementary Fig. [225]7d). Notably, these mutations even exacerbated
   the disassembly defects (Fig. [226]5h, i). In contrast, mutations R13P
   and L98P, which retained phase separation properties, successfully
   incorporated into SGs and rescued SG disassembly defects
   (Fig. [227]5f–i and Supplementary Fig. [228]7d).

   These findings directly link WDR45’s role in SG disassembly to BPAN
   pathology, underscoring the importance of its interaction with Caprin-1
   in SG dynamics. The inability of BPAN-associated mutations to mediate
   phase separation and promote SG disassembly highlights a critical
   mechanism underlying disease progression.

WDR45 deletion exacerbates accumulation of pathological SGs

   Pathological SGs, induced by disease-associated factors, are
   characterized by their impaired dynamics, reduced disassembly, and
   harmful effects on cellular function^[229]28. Factors driving the
   formation of these pathological SGs include mutant proteins such as FUS
   and dipeptide-repeat proteins (DPRs). To assess the broader
   implications of WDR45 in neurodegenerative diseases, we firstly
   examined its role in the accumulation of pathological SGs containing
   ALS-associated mutant FUS-R521C protein. The FUS-R521C mutant is known
   to translocate into SGs and contributes to disease
   pathology^[230]43–[231]45.

   Plasmids encoding either wild-type FUS (FUS WT) or the FUS-R521C were
   transfected into WT and WDR45 KO cells. Forty-eight hours
   post-transfection, we observed a significant increase of
   FUS-R521C-positive SGs in WDR45 KO cells compared to WT cells following
   transfection (Fig. [232]6a–d). Moreover, the FUS-R521C-positive SGs in
   WDR45 KO cells were approximately three times larger than those in WT
   cells, indicating a pronounced accumulation of pathological SGs in the
   absence of WDR45 (Fig. [233]6a–d).

Fig. 6. The accumulation of ALS-related pathological SGs in WDR45 KO cells.

   [234]Fig. 6
   [235]Open in a new tab

   a WT and WDR45 KO cells under normal conditions were transfected with
   plasmids encoding HA-tagged wild-type FUS (FUS WT) and FUS-R521C (FUS
   R521C) and stained for HA and TIAR (a SG marker). Scale bar, 10 μm. b
   Quantification of Cell with FUS-positive SGs percent under control as
   represented in (a). Data are shown as mean ± SD. ***p < 0.001 (unpaired
   and two-sided t-test) n = 3 independent experiments. c WT and WDR45 KO
   cells under normal conditions were transfected with plasmids encoding
   HA-tagged wild-type FUS (FUS WT) and FUS-R521C (FUS R521C) and stained
   for HA and TIAR. Scale bar, 10 μm. d Quantification of area for
   FUS-positive SGs. Scatter plot showing the cytoplasmic FUS area (μm^2)
   in WT (mean=0.705 μm^2, n = 147) and WDR45 KO (mean=2.797 μm^2,
   n = 143) cell lines. ****p < 0.0001 (unpaired and two-sided t-test)
   n = 3 independent experiment. e WT and WDR45 KO cells under normal
   condition were transfected with plasmid encoding GFP-tagged GR50 and
   stained for G3BP1. White arrowheads indicate the GR-positive SGs. Scale
   bar, 10 μm. Zoom’s scale bar, 10 μm. f Quantification of Cell with
   GR50-SGs percent under control as represented in (e). Data are shown as
   mean ± SD. ***p < 0.001 (unpaired and two-sided t-test) n = 3
   independent experiment. Source data are provided as a Source Data file.

   Additionally, we examined the impact of poly(GR) repeats associated
   with the ALS/FTD gene C9orf72, which are known to disrupt SG
   dynamics^[236]18,[237]46. When GR50 was transfected into cells, U2OS
   WDR45 KO exhibited a higher percentage of GR-positive SGs compared to
   WT cells (Fig. [238]6e, f).

   These findings highlight WDR45’s protective role in regulating SG
   dynamics and mitigating the formation of pathological SGs associated
   with neurodegenerative diseases. The loss of WDR45 appears to
   exacerbate pathological protein aggregation, suggesting a potential
   link between WDR45 dysfunction and the progression of neurodegenerative
   disorders.

iPSC-derived midbrain neurons from WDR45-mutant patient exhibited delayed SG
recovery

   To establish a disease model for investigating the molecular mechanisms
   underlying WDR45-related disorders, we previously derived iPSCs from a
   female BPAN patient carrying a non-canonical splice site mutation in
   the WDR45 gene (c.827+1 G > A), which encodes a truncated version of
   WDR45 termed p.(Ala276AlafsTer1)
   ([239]https://hpscreg.eu/cell-line/CIPi005-A)^[240]47 (Supplementary
   Fig. [241]8a). No obvious defects in autophagy flux were detected in
   the patient-derived iPSCs, which indicates that this mutation may cause
   some defect in a non-autophagic function (Supplementary Fig. [242]8b,
   c).

   To explore the role of WDR45 in SG dynamics in neuronal model, the
   derived iPSCs were differentiated into neural progenitor cells and
   subsequently into midbrain neurons (Supplementary Figs. [243]1a,
   [244]8d, [245]e). The induced neurons displayed normal autophagy
   levels, as indicated by comparable SQSTM1 and LC3B expression between
   patient and control neurons (Supplementary Fig. [246]8f). Additionally,
   there was no significant difference in endoplasmic reticulum (ER)
   stress between the two groups, as shown by similar HSPA5 expression
   levels (Supplementary Fig. [247]8g). Interestingly, midbrain neurons
   derived from the patient exhibited significantly slower SG recovery
   following arsenite-induced stress compared to healthy control neurons,
   highlighting the critical role of WDR45 in SG disassembly
   (Fig. [248]7a, b). To further investigate the molecular basis of this
   defect, we purified the truncated p.(Ala276AlafsTer1) protein and found
   that it was unable to form condensates with Caprin-1 and also failed to
   associate with Caprin-1, as shown by Co-IP and GST pull-down assays.
   (Fig. [249]7c, Supplementary Fig. [250]8h, i). Furthermore, plasmids
   encoding the p.(Ala276AlafsTer1) could not colocalize with SGs
   (Supplementary Fig. [251]8j) and failed to rescue the SG disassembly
   defect in WDR45 KO cell line (Fig. [252]7d, e). This finding supports
   the notion that BPAN-associated mutations compromise WDR45’s function
   in SG dynamics by disrupting its phase separation property and
   interaction with Caprin-1.

Fig. 7. iPSC-derived midbrain neurons from WDR45-mutant patient show delayed
SG recovery.

   [253]Fig. 7
   [254]Open in a new tab

   a Representative confocal images of neurons from patient and
   age-matched control. Neurons were subjected to SA stress for 30 min and
   recovered for 120 min and 150 min, and then stained with anti-Caprin-1
   (Red) and anti-TUJ1 (Green). Nuclei were stained with DAPI. Scale bar,
   10 μm. Zoom’s scale bar, 10 μm. b Statistical results of cells with SGs
   under different treatments as represented in (b). Data were performed
   with three biological repeats. Data are presented as mean ± SD. ns, not
   significant, **p < 0.01 by two-way ANOVA. c LLPS of purified
   recombinant Caprin-1 (Red), WDR45 Patient (Green) in vitro. Recombinant
   proteins were mixed pairwise at a 1:1 molar ratio to observe LLPS
   physiological conditions with 150 mM NaCl. Scale bar, 10 μm. d Confocal
   images of WDR45 KO U2OS cells (expressing GFP, p(Ala276AlafsTer1)) that
   treated with SA following recovered for 120 min, and stained with
   anti-G3BP1 (Red) antibody and DAPI (Blue) are shown, n = 3 independent
   experiment. Scale bar, 10 μm. e Quantification of Cell with SGs percent
   under SA, at the recovery stage from stress as represented in (d). Data
   were pooled from three independent experiments. Data are shown as
   mean ± SD. ns, not significant (unpaired and two-sided t-test). f A
   proposed model of WDR45-mediated SG disassembly. Functional WDR45 helps
   SGs disassembly through competing with G3BP1 for Caprin-1, thus
   weakening the SGs core network. While mutated WDR45 (WDR45 KO or
   BPAN-related WDR45 mutations) failed to interact with Caprin-1, and
   causing SGs disassembly defect (Created with Biorender.com). Source
   data are provided as a Source Data file.

Discussion

   This study identifies a function for WDR45 in SG disassembly and
   advances our understanding of SG dynamics regulation and BPAN
   pathogenesis (Fig. [255]7f). We demonstrate that WDR45, via its WD5
   domain, forms gel-like condensates and undergoes LLPS with Caprin-1
   (Fig. [256]2l). This interaction is essential for efficient SG
   disassembly. Critically, BPAN-associated mutations, particularly those
   within WD5, disrupt WDR45’s phase separation capacity and its
   interaction with Caprin-1, leading to defective SG disassembly
   (Fig. [257]5d–i). This mechanistic defect correlates with earlier
   disease onset in BPAN patients, directly implicating impaired SG
   dynamics in disease pathogenesis (Fig. [258]5b).

   Furthermore, the distinct roles of WDR45’s WD5 (SG disassembly) and WD3
   (ATG2B binding for autophagy) domains^[259]48 highlight the functional
   diversity of this protein and underscore the specificity of our
   findings. We emphasize that various disease-related mutations in WDR45
   contribute to BPAN through divergent molecular mechanisms. Some
   mutations primarily affect autophagy, such as R13P and L98P^[260]5,
   while others compromise WDR45’s role in SG disassembly or selectively
   affect WDR45’s ability to undergo phase separation (Fig. [261]5h, i,
   Supplementary Fig. [262]7a, b).

   Our findings reveal that WDR45 competes with G3BP1 for Caprin-1
   binding, thereby modulating the core SG network. This competition
   weakens the G3BP1-Caprin-1 interaction, which in turn promotes SG
   disassembly (Fig. [263]3c–j). While previous research has primarily
   focused on the role of the core SG network in SG assembly, our work
   shifts the perspective by revealing its equally crucial function in SG
   disassembly. The role of Caprin-1 in transitioning WDR45 condensates
   from a static gel-like state to a more dynamic liquid-like phase is
   essential for WDR45 localization to SGs and subsequent disassembly
   (Fig. [264]4c–f). This dynamic interplay between WDR45 and Caprin-1
   represents a layer of regulation for SG dynamics. Our discovery
   underscores the importance of maintaining a dynamic balance between SG
   assembly and disassembly to ensure effective cellular stress responses.

   The protective role of WDR45 against pathological SG formation
   containing mutant FUS (Fig. [265]6a–d), implicated in ALS and FTD,
   underscores the broader relevance of our findings beyond BPAN. WDR45
   depletion exacerbates the accumulation of these aberrant SGs,
   suggesting a potential therapeutic avenue for modulating SG dynamics in
   these related neurodegenerative disorders. The delayed SG recovery
   observed in iPSC-derived midbrain neurons from a BPAN patient further
   strengthens the clinical relevance of our findings and validates the in
   vitro observations (Fig. [266]7a, b). This patient-derived midbrain
   neurons model provides a powerful tool for future investigations into
   the molecular mechanisms linking WDR45 dysfunction to
   neurodegeneration. Additional studies are required to directly link
   these mechanisms to neuronal degeneration. Future research will focus
   on assessing whether restoring normal SG dynamics can mitigate neuronal
   degeneration in BPAN patient-derived models and mouse models.

   BPAN is characterized by neurodevelopmental defects in the early stage
   of the disease and neurodegeneration in the late stage^[267]3. However,
   the mechanisms underlying the transition from neurodevelopmental
   impairments to neurodegeneration remain unclear. Caprin-1 is essential
   for normal brain function, has been implicated in a neurodevelopmental
   disorder caused by loss-of-function variants^[268]49. Recent studies
   have also highlighted a strong association between SG-essential genes
   and neurodevelopmental disorders, further underscoring the importance
   of SG dynamics in brain health^[269]34. Building on these findings and
   our own results, we hypothesize that WDR45 functions together with
   Caprin-1 to regulate key processes in brain function (Fig. [270]7f),
   with this interaction potentially altering with aging. This dynamic
   interplay may provide a crucial mechanistic link between the
   neurodevelopmental and neurodegenerative features of BPAN, offering
   valuable insights into the disease progression and pathophysiology.

   Overall, our research sheds light on the molecular mechanisms
   underlying BPAN and potentially other neurodegenerative disorders
   associated with SG dysfunction. By elucidating the role of WDR45 in SG
   disassembly and its interaction with Caprin-1, we propose that
   targeting WDR45-Caprin-1 interactions or modulating SG dynamics could
   offer promising therapeutic strategies for BPAN and related
   neurodegenerative diseases. This approach could pave the way for
   developing interventions aimed at restoring normal SG dynamics and
   mitigating disease progression.

Methods

Cell culture and treatment

   HeLa, U2OS, HEK293T cell lines were cultured in Dulbecco’s modified
   Eagle medium with high glucose (HyClone, SH30022.01), and N2a was
   cultured in DME/F-12 (HyClone, SH30022.01), both supplementing with 10%
   fetal bovine serum (HUANKE, HK-CH500) and 1% penicillin–streptomycin
   (Vivacell, C3420-0100). Cells were maintained at 37 °C incubator
   supplemented with 5% CO[2]. The HEK293T (CRL-1573), HeLa (CCL-2), and
   Neuro-2a (CCL-131) cell lines were obtained from the American Type
   Culture Collection (ATCC). While U2OS cells (SCSP-5030) were purchased
   from Cell Bank of Chinese Academy of Sciences. iPSCs healthy control
   were kindly gifted from JiaYu Chen lab. All cells are regularly tested
   and confirmed to be free of mycoplasma contamination.

   For oxidative stress, cells were cultured in medium with 500 μM sodium
   arsenite (Millipore, 106277); for osmotic stress, cells were cultured
   in medium with 400 mM sorbitol (Sigma-Aldrich, S1876); for heat stress,
   cells were exposed to 44 °C in a humidified incubator (ESCO CelMate
   CO[2] INCUBATOR) with 5% CO[2].

Antibodies

   For Western blot, the following antibodies were used: Caprin-1
   (Proteintech, 15112-1-AP, dilution: 1:500), GFP (Santa Cruz, sc-9996,
   dilution: 1:1000), GST (Proteintech, 10000-0-AP, dilution: 1:1000),
   6*His (Proteintech, 66005-1-Ig, dilution: 1:1000), G3BP1 (Santa Cru,
   sc-365338, dilution: 1:1000), WDR45 (Proteintech, 19194-1-AP, dilution:
   1:1000), GAPDH (Proteintech, 19194-1-AP, dilution: 1:10000), SQSTM1
   (Proteintech, 18420-1-AP, dilution: 1:5000), LC3 (HUABIO, ET1701-65,
   dilution: 1:1000), NUP98 (HUABIO, HA722523). For immunostaining, the
   following antibodies were used: HA (Invitrogen, 26183, dilution:
   1:500), TIAR (CST, 8509S, dilution: 1:500), Caprin-1 (Proteintech,
   15112-1-AP, dilution: 1:500), Caprin-1 (Proteintech, 66352-1-Ig,
   dilution: 1:500), GFP (Santa Cruz, sc-9996, dilution: 1:500), G3BP1
   (Proteintech, 13057-2-AP, dilution: 1:500), G3BP1 (Santa Cru,
   sc-365338, dilution: 1:500), WDR45 (Proteintech, 19194-1-AP, dilution:
   1:400). TUJ1 (Proteintech, 66375-1-Ig, dilution: 1:400). For
   endo-immunoprecipitation, the following antibodies were used, G3BP1
   (Santa Cru, sc-365338, dilution: 1 μg/400 μl in lysis buffer).

Plasmid constructs and transfection

   The full length of Homo sapiens WDR45, CAPRIN1, and G3BP1 were
   amplified using PCR from human cDNA. The pET-MBP-WDR45 construct was
   generated from pET-MBP.3 C and WDR45 fragment using seamless cloning
   (LABLEAD, D0204P). The pLVX-WDR45-GFP construct was generated from
   pLVX.puro with an GFP tag inserted after gene fragment and WDR45
   fragment. Constructs of GFP tag inserting before WDR45 fragment were
   also generated. Different truncations and mutations of WDR45 were
   obtained through PCR applying site-directed mutagenesis. Constructs of
   CAPRIN1 including pET-MBP-CAPRIN1, GST-CAPRIN1, pLVX-GFP-CAPRIN1 were
   generated using seamless cloning. Constructs of G3BP1 including
   pET-MBP-G3BP1, and pLVX-GFP-G3BP1 were generated using seamless
   cloning.

   Plasmids were transfected into corresponding cell lines using
   Lipofectamine 3000 (Invitrogen, L3000015) according to the
   manufacturer’s instructions.

CRISPR–Cas9-mediated knockout cell lines

   Cells were plated on one well of six-well plate at a density of around
   20% and transfected with 2.5 μg PX459 plasmids with corresponding guide
   RNA (gRNA) sequences^[271]50. gRNA sequences were designed using
   [272]CRISPOR (Concordet and Haeussler, 2018). Knockout plasmids were
   generated through T4 ligation (LABLEAD, T5205) with PX459 digested by
   BbsI-HF® (NEB, R3539S) and annealed oligonucleotides. Two days later,
   1 μg ml^−1 puromycin (LABLEAD, P067) was added to the medium to select
   positive transfected cells for 3-5 days. After selection, survived
   cells were plated into 96 well plates at a single cell concentration.
   Monoclonal knockout cell lines were validated by sequencing analyses,
   western blot and immunostaining. The 20-nt guide sequences for the
   target genes are as follows:

   WDR45:

   gRNA: 5’-TGACTCAACAGCCACTTCG-3’

   WDR45B:

   gRNA1: 5’-CATCAGTTGCACGTCTTCGA-3’

   gRNA2: 5’-GAGCTGCATTGCACTCAACC-3’

   CAPRIN1:

   gRNA1: 5’-GGAGTGAGGCGGCCGCGGGA-3’

   gRNA2: 5’-TCGGGTTCCTCCGGGAGTGA-3’

Creation of stable expression cell lines

   Stable expression cell lines were constructed through
   lentivirus-mediated transfection. Briefly, 2 μg corresponding
   overexpression plasmids, 1 μg psPAX2 (Addgene plasmid #12260) and 1 μg
   pMD2.G (Addgene plasmid #12259) plasmids were used to co-transfect
   293 T cells in a 3.5 cm dish firstly. Viruses were collected at 48 h
   post transfection and used to infect U2OS or HeLa cells for another
   48 h, following selection by adding 1 μg ml^−1 puromycin to the medium
   for 3-5 days. Western blot was performed to validate the cell lines.

siRNA transfection and RT-qPCR

   siRNA was transfected into cells using Lipofectamine 3000 (Invitrogen,
   L3000015) according to the manufacturer’s instructions. Total RNA was
   extracted using MagZol Reagent (Magen, R4801-01) and then
   reverse-transcribed (YESEN, 11141ES10). Real-time quantitative PCR was
   conducted with SYBR qPCR Mix (Magen, MD70101) using QuantStudio3
   (Applied Bioscience). Quantitation of all target gene expression was
   normalized to the control gene GAPDH. qPCR primers are listed in
   Supplementary Table [273]2. siRNA sense sequences are listed below.

   si-WIPI2:

   5’-GACAGUCCUUUAGCGGCA-3’

   si-ATG2a:

   5’- GCAUUCCCAGUUGUUGGAGUUCCUA-3’

   si-ATG2b:

   5’- AGGUCUCUCUUGUCUGGCAUCUUUA-3’

   si-WIPI1-1:

   5’- GCUGCUGGUUUGCCUAGAA-3’

   si- WIPI1-2:

   5’- CACCAACAGUCGACCAGAA-3’

   si- WIPI1-3:

   5’- GGAGAGUGUGUCUUAAUCA-3’

   si- WDR45:

   5’- GCAUCUAUGUGUACUCCUU-3’

   si- WDR45B:

   5’- GGGAUGACCUGAAGAAGAA-3’

Protein purification

   WDR45 (full length and mutants) and G3BP1 were subcloned to pET-MBP.3 C
   for protein purification. Plasmids were transformed into E. Coli BL21
   competent cells. After 2 liters of E. coli were grown to OD[600] of
   0.8, 0.5 M IPTG (LABLEAD, 0487) was added to induce protein expression
   at 16 °C for 16 hours. Pellets were collected through centrifugation
   and resuspended in lysis buffer (500 mM NaCl, 50 mM HEPES 7.5, 1 mM
   DTT, 30 mM imidazole). After sonication (60% power, 2 s on, 8 s off,
   working time 10 min), lysates were pelleted at 18000 g for 50 min.
   Supernatants were applied to the His-tag column with 5 mL beads
   (GenScript, L00666-100) overnight. Proteins were eluted with 300 mM
   imidazole in lysis buffer. The proteins were then purified by Cytiva
   AKTA go, collecting different fractions which were subjected to
   SDS-PAGE. According to the results of SDS-PAGE, certain fractions were
   pooled, concentrated, flash frozen in liquid nitrogen and stored at
   -80 °C.

Western blot

   Cells were scraped and lysed in cold RIPA buffer (LABLEAD, R1090) for
   30 min supplemented with 1% protease cocktail (LABLEAD, C0101),
   pipetting up and down every ten minutes, following centrifugation at
   12000 g for 30 min. The pellets were discarded and supernatants were
   boiled in the loading buffer at 95 °C for 15 min. Protein samples were
   separated on an SDS-PAGE gel and transferred to a PVDF membrane
   (Millipore, IPVH00010), followed by incubation with primary antibodies
   overnight. Membrane was then washed with TBST three times, and
   horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson
   Immuno, 111-035-003) were added for visualization using ECL (LABLEAD,
   E1050). Protein quantification analysis was performed using NIH ImageJ
   (Fiji) software.

Immunofluorescence and microscopy

   Cells were plated on glass coverslips at appropriate concentrations,
   followed by treatments or transfection. Cells were washed by PBS three
   times and fixed with 4% PFA (LABLEAD, P4500) for 15 min, followed by
   permeabilization with 0.2% Triton X-100 (Sigma-Aldrich, T8787) for
   15 min and blocking with 3% BSA (LABLEAD, 0332-1) for 60 min. Primary
   antibodies were added, and incubated overnight. Corresponding Alexa
   Fluor Cy3, 488 secondary antibodies were added for visualization.
   Slides were then mounted using Mounting Medium, and antifading (with
   DAPI) (Solarbio, S2110). The images were captured using a Zeiss LSM880
   with a ×63 oil objective.

Immunoprecipitation assays

   Cells were seeded in 10 cm dishes at a desired concentration, following
   transfection or continuing to culture to a density of 90%. Cells were
   then scraped and lysed in 300 μL cold IP lysis buffer (25 mM Tris-HCl
   pH 7.4, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 1% NP40) for 30 min,
   lysates were subjected to centrifugation at 12000 g for 30 min, 60 μL
   supernatants were boiled as input. For exogeneous immunoprecipitation,
   15 μL GFP-magnetic beads (LABLEAD, GNM-50-2000) were added to the rest
   supernatants, following incubation overnight. For endogenous
   immunoprecipitation, lgG (Applygen, C1754) and primary antibodies were
   added to equal amounts of supernatants, following incubation overnight.
   Protein A/G-Sepharose beads (Thermo Scientific, 88802) were added to
   incubate for another 3 hours. The beads were washed 3 times before
   adding the loading buffer. Samples were analyzed by SDS-PAGE for
   immunoblotting.

GST Pull-down

   Protein Glutathione S-transferase (GST) was conjugated to glutathione
   beads (GenScript, L00206-50) at 4 °C for 2 hours. GST-Caprin-1 was
   purified, as the process of protein purification using glutathione
   beads, without the step of elution. WDR45 full-length and its mutants
   were added to the complexes for pull-down. After incubating overnight,
   the beads were washed 3 times with PBS, and analyzed by SDS-PAGE using
   anti-GST and anti-HIS antibodies.

FRAP

   FRAP experiments were performed on a Leica TCS-SP8 microscope with a
   ×63 oil immersion objective. For intracellular experiments, cells were
   seeded into 35 mm glass bottom dishes and were transiently transfected
   with WDR45-GFP or GFP-G3BP1 plasmids for 2 days. Droplets were
   photobleached using a laser intensity of 80% at 488 nm. Recovery was
   recorded for the indicated time. For in vitro experiments, Droplets of
   GFP-WDR45 alone or along with mCherry-Caprin-1 were photobleached using
   a laser intensity of 80% at 488 nm and 561 nm and recovered. Data were
   processed using the Fiji plugin FRAP profiler v2 and plotted with
   GraphPad Prism.

Mass spectrometry analysis

   Cells stably expressed WDR45-GFP or GFP were treated with 500 μM sodium
   arsenite for an hour or left untreated as a control.
   Immunoprecipitation using anti-GFP magnetic beads was performed to
   track interacting proteins of WDR45. Protein samples were subjected to
   SDS-PAGE, and the gel blocks were cut off when samples were about to
   enter stacking gel. The gels were dehydrated with 200 μL of
   acetonitrile, air-dried for 5-10 minutes, then rehydrated and digested
   with trypsin (10 ng/μL). An additional 50 μL of extraction solution
   (60% acetonitrile, 1% TFA) was added to the gel pieces, which were
   sonicated for 10 minutes. The supernatant was dried by centrifugal
   evaporation. The resulting peptides were desalted using a Mono-spin C18
   column (GL Sciences) and dissolved into 0.1% formic acid and spiked
   with iRT peptides before DIA analysis. LC-MS/MS analyses were performed
   on the Vanquish Neo UHPLC (Thermo Fisher Scientific), the trypsin
   digested peptides were separated using mobile phase A (0.1% formic acid
   (Sigma) in water) and phase B (0.1% formic acid and 80% acetonitrile
   (Thermo Fisher Scientific)). The Thermo Orbitrap Ascend mass
   spectrometer (Thermo Fisher Scientific) was configured for
   data-independent acquisition (DIA) by combining two experiment
   elements, corresponding to a full MS experiment and an MS/MS
   experiment, and was operated in positive polarity mode. The precursor
   ions were isolated by Quadrupole and fragmented by High Energy
   Collision Dissociation (HCD) with normalized collision energy (%) at
   32. A direct DIA Analysis built in Spectronaut software (version 18.0)
   was used for data analysis. The MS/MS spectra were matched against the
   human UniProt database (20,230 entries, downloaded in August 2023). DIA
   files were processed in the default mode. The FDR for PSM and protein
   quantification were both set to 0.01. We quantified differences of
   protein abundance between groups using log2 fold change and applied the
   Student’s t-test to assess statistical significance to obtain p values.
   Proteins with log2 fold change > 0, and p value < 0.05 were retained.

In vitro phase separation assay

   For single component phase separation, WDR45 Full length or mutations
   were diluted to different salt or protein concentrations in Eppendorf
   tubes and transferred to a glass slide, covered with a coverslip, and
   imaged with OLYMPUS IX73 equipped with 40×objectives. For co-phase
   separations between WDR45, G3BP1 and mCherry-Caprin-1, WDR45 was
   labeled with XFD488 NHS Ester (AAT Bioquest, 1812) and G3BP1 was
   labeled with XFD647 NHS Ester (AAT Bioquest, 1833). Proteins were
   diluted to appropriate concentrations as indicated under physiological
   conditions with 150 mM NaCl, images were captured using a Zeiss LSM880
   with a ×63 oil objective.

Structure Illumination microscopy (SIM)

   U2OS cells were seeded and grown on coverslips and treated as
   indicated. The mounting technique was the same as immunofluorescence.
   The data were collected through the Polar-SIM system (Airy Technology
   Co., Ltd., China). A 488 nm and 561 nm laser were used to excitation
   with SIM. The SIM reconstruction process was conducted using the
   Airy-SIM software with pre-processing (Dark) or post processing
   (MRA/MLE). Fluorescence intensity and Pearson correlation coefficient
   were analyzed using Fiji.

Midbrain neurons differentiation

   The iPSCs were plated on matrigel (Corning® Matrigel®, 355247) and
   maintained in mTeSr Plus medium (STEMCELL technology, 100-0276).
   Midbrain neurons were induced using DAXING technology neuron induction
   kits according to the manufacturer’s recommendations. Briefly, 4×10^5
   iPSCs were plated on neuron coating matrix (DAXIANG, NK100107) and kept
   in mTeSr Plus. The next day, medium was changed to Neuron progenitor
   induction medium (DAXIANG, NK100101) for 14 days. Subculturing was
   necessary when cell confluency reached 90%. After 14 days, neuron
   progenitor cells were successfully induced and maintained in NPC
   expansion medium (DAXIANG, ND100102). 2×10^5 neuron progenitor cells
   were plated in Midbrain Differentiation medium (DAXIANG, NK100103) for
   14 days and Midbrain Maturation medium (DAXIANG, NK100107) for 7 days,
   changing medium daily.

   For oxidative treatment, cells were exposed to 500 μM sodium arsenite
   for 30 min, for heat shock, cells were exposed to 44 °C for 1 hour.

Living cell imaging

   HeLa knock in G3BP1-GFP cell line was used for living cell imaging.
   Briefly, cells growing on confocal dish (Cellvis, D35C4-20-1.5-N) were
   transfected with siRNA targeting WDR45 or control, then following
   incubation for 48 h. For arsenite (500 μM) stress phase, images were
   captured at 15 min, 40 min, 60 min. Using Nikon Elements software with
   definite focus engaged, multipoint images were taken every 20 s with
   the 488-nm laser at 10% power, average is 2, zoom is 1024*1024. For
   recovery phase (using fresh culture medium after arsenite stress
   (500 μM, 60 min)), images were captured at 0 min, 60 min, 120 min and
   150 min. Multipoint images were taken every 1 min with the 488-nm laser
   at 5% power, average is 2, zoom is 1024*1024. The living cell images
   were collected by Nikon AXR NSPARC and analyzed by Nikon Elements
   viewer and image J software. The LUTS was changed in the same
   situation. Cells were maintained at 37 °C incubator supplemented with
   5% CO[2].

Statistics & reproducibility

   For the immunostaining of treated cells, cells harboring SGs were
   quantified manually, at least three fields per experiment were
   analyzed; areas of FUS-positive granules were measured using Fiji, and
   all cells were considered for the quantification. Quantification of
   western strips was performed using Fiji software, and at least three
   biological repeats were measured. Colocalization analysis was performed
   using the Fiji plugin Coloc2. GraphPad Prism was used for statistical
   analysis. Statistical values are displayed as the means ± SD (standard
   deviation)/SEM (standard error of the mean). p > 0.05 was considered
   not significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by
   Student’s t-test, one-way ANOVA, or two-way ANOVA. The details are
   included in the corresponding figure legends. All experiments were
   repeated at least three times with similar results.

Reporting summary

   Further information on research design is available in the [274]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [275]Supplementary Information^ (10.6MB, pdf)
   [276]DOASF^ (45.4KB, pdf)
   [277]Supplementary Movie 1^ (9.7MB, mp4)
   [278]Supplementary Movie 2^ (9.7MB, mp4)
   [279]Supplementary Movie 3^ (78.8MB, mp4)
   [280]Supplementary Movie 4^ (86.9MB, mp4)
   [281]Reporting summary^ (2.8MB, pdf)
   [282]Transparent Peer Review file^ (959.1KB, pdf)

Source data

   [283]source data^ (19.8MB, xlsx)

Acknowledgements