Introduction
Amyotrophic lateral sclerosis (ALS or Lou Gehrig’s disease) is a progressive neurodegenerative disease with no cure available [
37,
49,
71]. A better understanding of the molecular etiology of the disease is needed to develop effective preventive measures or cures. Approximately 10–15 % of ALS cases are familial and studies of the ALS genes whose mutations cause familial ALS have provided valuable insights into the disease mechanism. The first identified ALS gene encodes copper/zinc superoxide dismutase (SOD1) [
14,
51]. The ALS mutations in SOD1 cause toxicity that is foreign to the wild-type (WT) protein (termed as “gain-of-function”) [
9], but the nature of such toxicity is still not fully understood. Mutations in a group of genes involved in RNA metabolism have been found in recent years, including TDP-43 [
45], FUS [
35,
67], ataxin-2 [
19], hnRNPA1 [
33] and Matrin-3 [
29]. The seemingly unrelated pathogenic mechanisms elicited by the mutations in SOD1 and in RNA metabolism regulators have not been reconciled yet.
The Ras GTPase-activating protein-binding protein G3BP1 [
48] is an important regulator of RNA metabolism [
5,
24,
66], translation [
1,
47] and stress granule (SG) dynamics [
2,
65,
70]. G3BP1 was reported to play a critical role in the secondary aggregation step of SG formation [
2], and has been used as a reliable marker of SGs [
31]. The misregulation of SG dynamics has been reported in many forms of ALS [
36]. G3BP1 is critical for neuronal survival since G3BP1 null mice demonstrate widespread neuronal cell death in the central nervous system [
73]. G3BP1 is also critical for synaptic plasticity and calcium homeostasis [
43].
In this study, we initially tested whether aggregated ALS mutant SOD1-containing inclusions were in any way related to G3BP1-positive stress granules. Interestingly, mutant SOD1 inclusions were co-localized with G3BP1-positive granules in spinal cord motor neurons of G93A SOD1 transgenic mice as well as in cultured cells. ALS-related mutants of SOD1, unlike wild-type SOD1, interacted with G3BP1 and the interaction was preserved in the presence of RNase, suggesting a direct protein–protein interaction. Domain deletion mutations, molecular modeling and point mutagenesis showed that the RNA-binding RRM domain of G3BP1 interacted with mutant SOD1 via residues critical to the RNA binding. Furthermore, the expression of mutant SOD1 perturbed SG dynamics and resulted in a delayed formation of SG in response to hyperosmolar stress and arsenite treatment. Our results suggest that G3BP1 represents a potential link between pathogenic SOD1 mutations and RNA metabolism alterations in ALS.
Materials and methods
Plasmids
The WT and A4V mutant SOD1-EGFP [
74], SOD1-3xHA [
21], 3xFLAG-FUS [
22], FLAG-TDP-43 (a gift from Dr. Francisco Baralle) [
3], FLAG-MATR3 (a gift from Dr. Yossi Shiloh, Addgene plasmid # 32880) [
52], and FLAG-hnRNPA1 (a gift from Dr. J. Paul Taylor) [
33] expression constructs were previously reported. The human G3BP1 expression constructs used in this study were based on the FLAG-G3BP1 plasmid [
34], a generous gift from Dr. Zhi-Min Yuan (University of Texas Health Science Center at San Antonio). The 3xFLAG-tagged G3BP1 constructs were generated from p3xFLAG-CMV10 (Sigma) using standard cloning techniques. The F380L/F382L G3BP1 mutation and the W32S SOD1 mutation were introduced with the QuikChange II Site-Directed Mutagenesis Kit (Agilent). The A4V/W32S double mutant SOD1-EGFP and the EGFP-G3BP1 expression constructs were generated by subcloning the respective fragments to pEGFP-N3 and pEGFP-C3 (Clontech), respectively. The mCherry-WT and F380L/F382L double mutant G3BP1 constructs were made by subcloning the respective G3BP1 fragments to pmCherry-C1 (Clontech). All plasmid constructs were verified with sequencing.
Cell culture and transfection
N2A and HEK293T (293T) cells were cultured in DMEM (Sigma, D5796) supplemented with 10 % fetal bovine serum and penicillin–streptomycin at 37 °C in 5 % CO2/95 % air with humidification. N2A and 293T cells were transfected with Lipofectamine 2000 (Life Technologies) and Polyethylenimine “Max” (Polysciences, Inc.), respectively. The G3BP1-null 293T cells were generated by employing CRISPR technology (G3BP1 Double Nickase Plasmids, Santa Cruz, sc-400745-NIC) following the manufacturer’s protocol.
Animals
Transgenic mouse strains B6.Cg-Tg(SOD1)2Gur/J and B6.Cg-Tg(SOD1-G93A)1Gur/J [
26] were bred and maintained as hemizygotes at the University of Kentucky animal facility. Transgenic mice were identified using PCR. The mice were killed at age 60, 90 and 125 ± 5 days. Mice were anesthetized with an intraperitoneal injection of 0.1 ml Pentobarbital (50 mg/ml, Abbott Laboratories) and transcardially perfused with 0.1 M phosphate buffered saline (PBS), pH 7.5 before spinal cords were dissected. All animal procedures were approved by the university IACUC committee.
Clinical materials
Human skin fibroblast cell cultures were established as previously described [
16]. Briefly, punch skin biopsy (3 mm) was obtained after informed consent from a 63-year-old male with symptomatic ALS with a documented L144F SOD1 mutation (Athena Diagnostics). The control skin biopsy was obtained from a 64-year-old healthy male who was free of neurological disease or any known ALS gene mutation. Skin biopsies were washed with PBS and cut into small pieces. Fibroblast growth medium [MEM (Sigma, M5650) supplemented with 20 % FBS, 2 mM
l-glutamine, 100 unit/ml penicillin, and 100 µg/ml streptomycin] was added to the minced biopsy tissue and transferred along with tissue fragments into tissue culture plates. Cultures were maintained in a humidified atmosphere of 5 % CO
2/95 % air at 37 °C to allow fibroblast cells grow from tissue fragments. Fibroblast cells were then maintained under the same conditions as above. The study was approved by the Institutional Review Board of the University of Kentucky.
Fluorescence microscopy
N2A or 293T cells were seeded on gelatin-treated glass coverslips and transfected with SOD1-EGFP constructs. Twenty-four hours later, cells were rinsed with 1× PBS, fixed with 4 % formaldehyde in 1× PBS, and permeabilized with 1× PBS supplemented with 0.25 % Triton-X100. Primary fibroblast cells were cultured, fixed and permeabilized similarly as above. Mouse spinal cords were dissected, post-fixed in 4 % formaldehyde in 1× PBS for 3 h, cryopreserved in 30 % sucrose overnight, embedded in Tissue-Tek OCT compound (Sakura). Sections were cut at 12 μm and permeabilized with 1× PBS supplemented with 0.1 % Triton-X100. The primary antibodies were sheep anti-human SOD1 (The Binding Site, PC077), mouse anti-G3BP1 (BD Biosciences, 611126), rabbit anti-G3BP1 (Proteintech, 13057-2-AP), goat anti-TIA1 (Santa Cruz, sc-1751), mouse anti-eIF3 p110 (Santa Cruz, sc-74507) and mouse anti-GE-1/p70 S6 kinase alpha (Santa Cruz, sc-8418). The secondary antibodies were Alexa Fluor 488 donkey anti-sheep (Life Technologies, A11015), Alexa Fluor 568 donkey anti-mouse (Life Technologies, A10037), Alexa Fluor 647 donkey anti-mouse (Life Technologies, A-31571), Alexa Fluor 568 donkey anti-rabbit (Life Technologies, A-10042), Alexa Fluor 647 donkey anti-rabbit (Life Technologies, A-31573), and Alexa Fluor 568 donkey anti-goat (Life Technologies, A-11057). The samples were mounted by applying Vectashield Mounting Medium (Vector Laboratories) and visualized using a Nikon A1 or a Leica SP5 confocal microscope with a 60× objective.
Co-immunoprecipitation assays
The lysates were prepared in 1× RIPA buffer (Millipore) supplemented with protease inhibitor cocktail (Sigma, P-8340, 1:500) and 1 mM sodium orthovanadate. Mouse spinal cord extracts were prepared using a dounce homogenizer. Transfected cell lysates were homogenized by passing through a 23-gauge needle several times. Immunoprecipitations were performed at 4 °C for 2 h. The endogenous G3BP1 immunoprecipitations were done from spinal cord extracts with rabbit anti-G3BP1 Antibody (Millipore, 07-1801) and Protein A UltraLink Resin (Thermo Scientific Pierce, 53139). The FLAG immunoprecipitations were performed using EZview Red Anti-FLAG M2 Affinity Gel (Sigma, F2426) and the bound proteins were eluted with 3xFLAG peptide (Sigma, F4799). Where indicated, RNase Cocktail (Ambion Life Technologies, AM2286) was added to the immunoprecipitation mixtures at 1:100 dilution.
The in vitro SOD1–G3BP1 binding assays were performed using 2 μg WT or G93A mutant human SOD1 purified from insect cells as described [
27] and 1.5 μg of 6xHis-tagged human G3BP1 purified from
E. coli (Fitzgerald Industries, 80R-1601) in 500 μl 1× RIPA buffer. The mixtures were incubated at 37 °C for 2 h, cooled to 4 °C and G3BP1 was immunoprecipitated with mouse anti-G3BP1 (Millipore, 05-1938) and Protein G UltraLink Resin (Thermo Scientific Pierce, 53126).
Western blotting
The nitrocellulose membranes were blocked and antibodies were applied in 5 % milk in TBST (100 mM TRIS–HCl, pH7.5, 0.9 % NaCl, 0.1 % Tween-20). The antibodies used were rabbit anti-G3BP1 (Millipore, 07-1801), rabbit anti-SOD1 (Santa Cruz, sc-11407), mouse anti-FLAG M2-HRP (Sigma, A8592), rabbit anti-HA (Santa Cruz, sc-805), mouse anti-PABP1 (Santa Cruz, sc-32318) and goat anti-actin (Santa Cruz, sc-1616). All immunoblotting images were acquired using a BioRad ChemiDoc MP system.
In silico docking
The homology model of the RRM domain of G3BP1 was obtained using a leading homology modeling server Raptor X [
30]. The model was docked to the SOD1 A4V mutant dimer (PDB ID: 3GZQ, [
23]) using the protein–protein interaction modeling server HADDOCK [
17]. The docking model was further refined using Rosetta Docking [
12] as implemented in the ROSIE server [
40], which evaluates protein–protein complexes by using rigid body perturbations of the protein chains.
Stress granule induction and analysis
N2A cells grown on gelatin-treated glass coverslips were transfected with WT or A4V mutant SOD1–EGFP constructs. Twenty-four hours later, the cells were treated with 0.5 M
d-sorbitol (Sigma, S1876) or 0.5 mM sodium arsenite dissolved in fresh medium for the indicated times at 37 °C to induce stress granules, with or without recovery in fresh medium as indicated. Control cells were treated with fresh medium without sorbitol or arsenite. The cells were fixed and G3BP1 immunofluorescence experiments were performed as above. Z-stack images of random view-fields were acquired with identical imaging parameters. Maximum intensity projections of the Z-stacks were analyzed for stress granule formation as published [
10] using ImageJ (
http://imagej.nih.gov/ij). ANOVA with post hoc Tukey HSD (honest significant difference) test was used to determine
p values for multiple pair-wise comparisons. Student’s
t test (two-tailed distribution, two-sample unequal variance) was used to determine
p values for simple pair-wise comparison.
Discussion
SOD1 was the first gene whose mutations were identified to cause familial ALS [
14,
51]. The cytoplasmic SOD1-positive protein inclusions in affected motor neurons are a hallmark of mutant SOD1-mediated ALS [
50,
68]. In other ALS cases, the pathological inclusions are often immune-positive of the RNA-binding protein TDP-43 [
41,
60]. In recent years, alterations in the dynamics of stress granules have emerged as a common theme in a wide range of ALS cases [
18,
36,
72]. ALS-related mutations in a range of genes including TDP-43, FUS, hnRNPA1, profilin-1, angiogenin and C9ORF72 were found to result in abnormalities in stress granule dynamics [
2,
4,
7,
13,
20,
22,
33,
38,
59,
61,
63,
69]. Ataxin-2, an ALS risk factor with polyglutamine expansions, can also regulate stress granules [
19,
46]. These observations suggest that mutant SOD1-mediated ALS might represent an outlier in the ALS disease mechanism. At the same time, the nature of the gain-of-toxicity caused by mutant SOD1 [
9] is still not fully understood. This study was initiated to determine whether mutant SOD1 also causes perturbed stress granule dynamics.
We have demonstrated here that cytoplasmic inclusions of mutant SOD1 were immune-positive for two reliable stress granule markers G3BP1 [
31,
32] and TIA1 in the G93A mutant SOD1 transgenic mouse spinal cord (Fig.
1a, Supplemental Fig. 1), fibroblast cells derived from human ALS patient (Fig.
1c, Supplemental Fig. 3), and cultured cells (Fig.
1d, Supplemental Figs. 4, 5). A third stress granule marker eIF3 was also co-localized with mutant SOD1 and G3BP1 in cultured N2A cells (Supplemental Fig. 4). Moreover, the G3BP1-positive mutant SOD1 inclusions were found closely juxtaposed to P-bodies (Fig.
1f), supporting that they constitute stress granules [
32]. It was previously reported that RNA stabilizer Hu Antigen R (HuR) and TIA1-related protein (TIAR) were found along with mutant SOD1 in detergent-insoluble aggregates [
39]. However, that study did not test whether mutant SOD1 was co-localized in stress granules. Our results provide the evidence that mutant SOD1 is co-localized with stress granules and consequently can potentially influence the dynamics of stress granules.
We found that the molecular basis for the co-localization of mutant SOD1 inclusions and G3BP1 is that multiple mutants of SOD1, but not WT SOD1, interact with G3BP1 (Fig.
2). This interaction was independent of RNA (Fig.
2c) and was reconstituted with purified mutant SOD1 and G3BP1 proteins (Fig.
2d), showing that mutant SOD1 and G3BP1 directly interact with each other. In the previous study showing the presence of HuR and TIAR in mutant SOD1 aggregates, the amount of HuR and TIAR was significantly reduced by RNase treatment [
39], suggesting an indirect interaction between mutant SOD1 and HuR/TIAR. This is in contrast to the RNA-independent direct interaction between mutant SOD1 and G3BP1 found in this study. Moreover, the interaction appeared to be specific for G3BP1, while four other RNA-binding proteins implicated in ALS, hnRNPA1, FUS, TDP-43 or Matrin-3 did not interact with mutant SOD1 (Fig.
3).
We further demonstrated that the RRM domain of G3BP1 is essential and sufficient for the interaction with mutant SOD1 (Fig.
4b, c). Molecular modeling and in silico docking results suggest that the F380 and F382 residues of G3BP1 and the W32 residue of mutant SOD1 play a role in this interaction (Fig.
4d). Site-directed mutagenesis results showed that mutating F380 and F382 in G3BP1 and W32 in SOD1 indeed impaired the interaction (Fig.
4e). The W32 residue of human SOD1 resides in the middle of the third β-strand [
25] and is exposed on the protein surface, thus it is reasonable that it is involved in the interaction with G3BP1. It has been reported that W32 potentiated the aggregation and cytotoxicity of ALS mutant SOD1 [
62]. In addition, other ALS-related mutants of SOD1 were reported to display aberrantly increased hydrophobicity [
64], which can enable the interaction with G3BP1. Consistent with the interaction results, the W32S SOD1 mutation and the F380L/F382L G3BP1 mutation significantly impaired the co-localization of A4V SOD1 with G3BP1 (Fig.
5).
We found that expression of mutant SOD1 delayed the formation of G3BP1 stress granules in response to hyperosmolar stress and arsenite treatment (Fig.
6). Hyperosmolar stress was chosen in our study since it was shown to induce FUS [
53] and TDP-43 [
15] stress granules. Arsenite treatment is widely used to induce stress granules as it induced oxidative stress [
56]. There are multiple potential mechanisms how the ALS mutant SOD1–G3BP1 interaction could perturb stress granule dynamics. The F380 and F382 residues are well conserved in the “3” and “5” positions of the RNP1 motif (Supplemental Fig. 6A), which are critical to RNA binding [
42]. Substitution of conserved phenylalanine residues in the RNP1 motif were reported to impair RNA binding [
8,
11,
54]. It is noted that F380L/F382L mutation significantly reduced the interaction between G3BP1 and PABP1 (Fig.
4e, lanes 6 and 8). Similarly, the G3BP1-PABP1 interaction was abolished in the presence of RNase (Fig.
2c, lane 6), suggesting that F380L/F382L mutation impaired the RNA binding of G3BP1. Furthermore, we performed the RNA immunoprecipitation experiment to measure the amount of c-myc RNA (one of the few known RNA-binding partners of G3BP1 [
66]) co-precipitated with G3BP1. The F380L/F382L double mutant G3BP1 co-precipitated significantly less endogenous c-myc mRNA than WT G3BP1 (
p = 2.7 × 10
−4), confirming that the F380 and F382 residues are important for G3BP1 RNA binding (Supplemental Fig. 6b). As discussed earlier, the interaction between mutant SOD1 and G3BP1 is not mediated by RNA, but rather direct protein–protein interaction (Figs.
2,
4). It is conceivable that, when F380 and F382 are engaged in the interaction with mutant SOD1, the RNA-binding ability of G3BP1 is impaired (a “competitive binding” model). Since G3BP1 plays a critical role in stress granule dynamics [
2,
65,
70], its impaired RNA binding can result in the delayed formation of stress granules.
An alternative mechanism is that mutant SOD1 can sequester G3BP1 into inclusions, thus interfering with the G3BP1-mediated stress granule formation and dynamics. It is interesting to note “rimming” of mutant SOD1 inclusions by G3BP1 in cultured cells (Supplemental Fig. 5). Rimming of pathological aggregate structures was reported in other neurodegenerative diseases, e.g., the rimming of huntingtin aggregates by p62 [
6] or HDAC6 [
28]. This observation suggests that G3BP1 could be secondarily sequestered to mutant SOD1 inclusions, thus contributing to impaired stress granule dynamics (a “sequestration” model). Future studies are needed to distinguish these two potential mechanisms. It is also noted that G3BP1-positive granules were relatively small in cells under hyperosmolar stress (Fig.
6b) and large in cells treated with arsenite (Fig.
6c). In addition to the number of granules in this study, future studies using live cell imaging techniques will examine sizes of individual granules and determine whether multiple smaller granules merge into large granules under different stress conditions.
A delayed formation of stress granules in the presence of mutant SOD1 (Fig.
6) are reminiscent of the reports that the expression of ALS mutants of TDP-43 and FUS also resulted in altered stress granules [
4,
15,
38]. These findings suggest that the misregulation of stress granule dynamics represents a common pathophysiological feature shared by ALS cases mediated by mutant SOD1 and those caused by mutations in proteins involved in RNA metabolism. Perturbation of the physiological function of G3BP1 is especially harmful to neurons since G3BP1 is a neuronal survival factor [
43,
73]. The full complement of RNAs regulated by G3BP1 and in general by stress granules in the central nervous system is unknown. In addition, it remains unclear whether RNA molecules reside in the mutant SOD1 and G3BP1 positive co-inclusions or what functional role RNAs play in the inclusions. Future studies will determine the range of RNAs whose stability and function is affected by altered stress granule dynamics in ALS.
Taken together, our results suggest that the aberrant interaction of ALS-related SOD1 mutants with G3BP1 and the resulting perturbation of stress granule dynamics are likely important components of the toxicity of SOD1 mutations. In addition, a number of stress conditions are known to induce G3BP1-positive stress granules that might in turn seed mutant SOD1 inclusions. This two-hit scenario might enhance the pathological aggregation of mutant SOD1. Our findings reconcile the seemingly disparate disease mechanisms caused by mutations in SOD1 and several other genes involved in RNA metabolism.