ALS/FTD-associated FUS activates GSK-3β to disrupt the VAPB-PTPIP51 interaction and ER-mitochondria associations

Abstract

Defective FUS metabolism is strongly associated with amyotrophic lateral sclerosis and frontotemporal dementia (ALS/FTD), but the mechanisms linking FUS to disease are not properly understood. However, many of the functions disrupted in ALS/FTD are regulated by signalling between the endoplasmic reticulum (ER) and mitochondria. This signalling is facilitated by close physical associations between the two organelles that are mediated by binding of the integral ER protein VAPB to the outer mitochondrial membrane protein PTPIP51, which act as molecular scaffolds to tether the two organelles. Here, we show that FUS disrupts the VAPB-PTPIP51 interaction and ER-mitochondria associations. These disruptions are accompanied by perturbation of Ca(2+) uptake by mitochondria following its release from ER stores, which is a physiological read-out of ER-mitochondria contacts. We also demonstrate that mitochondrial ATP production is impaired in FUS-expressing cells; mitochondrial ATP production is linked to Ca(2+) levels. Finally, we demonstrate that the FUS-induced reductions to ER-mitochondria associations and are linked to activation of glycogen synthase kinase-3β (GSK-3β), a kinase already strongly associated with ALS/FTD.

Keywords: amyotrophic lateral sclerosis; frontotemporal dementia; glycogen synthase kinase‐3β; protein tyrosine phosphatase interacting protein 51; vesicle‐associated membrane protein‐associated protein B.

Figure EV1. Expression of EGFP‐FUS reduces the expression of endogenous FUS

HEK293 cells were transfected with control EGFP, EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K and 72 h post‐transfection, the samples were probed on immunoblots for FUS (using FUS antibody) and tubulin as a loading control.

Figure 1. Expression of wild‐type and ALS/FTD‐mutant FUS reduces ER–mitochondria associations in NSC34 cells

1. A

   Expression of FUS does not alter expression of VAPB, PTPIP51 or mitofusin‐2 (MFN2) in transfected NSC34 cells. Immunoblots of NSC34 cells transfected with EGFP as a control (CTRL), or wild‐type or mutant EGFP‐FUS. Transfected cells were purified via EGFP using a cell sorter and the samples probed on immunoblots as indicated. On the FUS immunoblot, samples were probed with FUS antibody to show endogenous and transfected proteins; tubulin is shown as a loading control.

2. B

   Representative electron micrographs of ER–mitochondria associations in NSC34 cells transfected with control EGFP vector (CTRL), EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K as indicated; arrowheads with loops show regions of association. Scale bar = 200 nm. Bar chart shows % of the mitochondrial surface closely apposed to ER in the different samples. Data were analysed by one‐way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. N = 27–30 cells and 247–424 mitochondria, error bars are s.e.m.; ***P < 0.001.

3. C, D

   siRNA loss of FUS does not affect ER–mitochondria associations or alter expression of VAPB, PTPIP51 or mitofusin‐2 (MFN2) in NSC34 cells. (C) Immunoblots of cells either mock transfected or treated with control (CTRL) or FUS siRNAs; GAPDH is shown as a loading control. (D) Representative electron micrographs of ER–mitochondria associations in control (CTRL) and FUS siRNA‐treated cells. Arrowheads with loops show regions of association. Scale bar = 200 nm. Data analysed by unpaired t‐test. N = 27–28 cells and 193–202 mitochondria, error bars are s.e.m.

Figure 2. Expression of wild‐type and mutant FUS reduces ER–mitochondria associations and the VAPB–PTPIP51 interaction in NSC34 cells

1. FU‐induced reductions in ER–mitochondria associations can be detected using SIM. NSC34 cells were transfected with either EGFP control vector, EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K and immunostained for TOM20 and PDI to label mitochondria (Mito) and ER, respectively; FUS was detected via their EGFP tags. Merge (ZOOM) shows zoomed images of boxed regions, and co‐localization shows co‐localized pixels. Scale bar = 2 μm. Bar chart shows ER–mitochondria co‐localization (Manders coefficient) normalized to control in the different samples. Data were analysed by one‐way ANOVA with Tukey's post hoc test. A total of 10–14 cells were analysed per condition from three independent experiments; error bars are s.e.m., **P < 0.01 and ***P < 0.001.

2. ER–mitochondria associations and the VAPB–PTPIP51 interaction are disrupted by wild‐type and ALS/FTD‐mutant FUS. NSC34 cells were transfected with EGFP control vector (CTRL), EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K and proximity ligation assays performed using VAPB and PTPIP51 antibodies. FUS were detected via their EGFP tags. Scale bar = 10 μm. Bar chart shows relative number of proximity ligation assay signals/cell. Data were analysed by one‐way ANOVA and Tukey's post hoc test; n = 47–53 cells from five experiments. Error bars are s.e.m.; ***P < 0.001.

3. Overexpression of FUS reduces the binding of VAPB to PTPIP51 in transfected cells. Cells were transfected as indicated with either control empty vector (CTRL), HA‐PTPIP51 + CTRL, or HA‐PTPIP51 + either EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K. PTPIP51 was immunoprecipitated using the HA tag and the amounts of endogenous bound VAPB detected by immunoblotting. Both inputs and immunoprecipitations (IP) are shown and no immunoprecipitating VAPB signals were obtained in the absence of HA‐PTPIP51. Bar chart shows relative levels of VAPB bound to PTPIP51 in the immunoprecipitations following quantification of signals from immunoblots. VAPB signals were normalized to immunoprecipitated PTPIP51‐HA signals. Data were analysed by one‐way ANOVA and Tukey's post hoc test; N = 4. Error bars are s.e.m.; ***P < 0.001.

Figure EV2. Control experiments involving omission of primary antibodies demonstrate the specificity of the VAPB‐PTPIP51 proximity ligation assays

1. A, B

   Panel (A) shows NSC34 cells; panel (B) shows mice spinal cords. Samples were probed with no primary antibodies (no Abs), VAPB only, PTPIP51 only or VAPB + PTPIP51 antibodies. In (A) samples are counterstained with DAPI to show nuclei. Scale bar = 10 μm (A) and 30 μm (B). Bar charts show proximity signals/cell. Data were analysed by one‐way ANOVA and Tukey's post hoc test. N = 16 cells (A) and 12 cells (B), error bars are s.e.m.; ***P < 0.001.

Figure 3. Overexpression of FUS reduces ER–mitochondria associations and the VAPB–PTPIP51 interaction in spinal cords of FUS transgenic mice

1. Overexpression of FUS does not affect expression of VAPB, PTPIP51 or mitofusin‐2. Immunoblots of spinal cord proteins from three 10‐week‐old FUS transgenic mice and three age‐matched littermates are shown; FUS was detected via its HA tag. Tubulin was used as a loading control.

2. Representative electron micrographs of ER–mitochondria associations in lumbar spinal cord motor neurons of FUS transgenic mice and their non‐transgenic littermates; arrowheads with loops show regions of association. Scale bar = 200 nm. Bar chart shows % of mitochondrial surface closely apposed to ER in the two samples. Data were analysed by unpaired t‐test. N = 67–88 cells and 438–749 mitochondria, error bars are s.e.m.; ***P < 0.001.

3. ER–mitochondria associations and the VAPB–PTPIP51 interaction are disrupted in lumbar motor neurons in spinal cords of FUS transgenic mice. Representative images of proximity ligation signals in 11‐week‐old FUS and non‐transgenic (NTg) littermate mice. Data were analysed by unpaired t‐test; N = 40 cells from 3 FUS and 3 non‐transgenic littermates (age 10 weeks). Error bars are s.e.m.; ***P < 0.001.

Figure 4. Expression of FUS disrupts cellular Ca²⁺ homoeostasis and mitochondrial ATP production

1. A, B

   FUS disrupts Ca²⁺ homoeostasis. HEK293 cells were transfected with M3R and either control vector (CTRL), FUS, FUSR521C or FUSR518K as indicated. Release of ER Ca²⁺ was induced by treatment of cells with OxoM. Panel (A) shows cytosolic Ca²⁺ levels with representative Fluo4 fluorescence traces on the left and normalized peak values on the right. Fluo4 fluorescence shows a transient increase in cytosolic Ca²⁺ levels upon OxoM treatment but compared to control, wild‐type and mutant FUS all increase peak cytosolic Ca²⁺ levels. Panel (B) shows mitochondrial Ca²⁺ levels with representative Rhod2 fluorescence traces on the left and normalized peak values on the right. Data were analysed by one‐way ANOVA and Tukey's post hoc test. (A) N = 49–52 cells from three experiments; (B) N = 50–52 cells from five experiments, error bars are s.e.m.; *P < 0.05, **P < 0.01, ***P < 0.001.

2. C

   FUS reduces mitochondrial ATP production. ATP levels were measured in NSC34 cells transfected with the ATP indicator AT1.03 and either control vector (CTRL), HA‐FUS, HA‐FUSR521C or HA‐FUSR518K. Cells were imaged in time‐lapse prior to and after KCN treatment to inhibit oxidative phosphorylation. Representative traces of YFP/CFP ratios are shown for the different samples; initial YFP/CFP ratios prior to KCN treatment and those after KCN treatment are indicated. The fall in YFP/CFP ratios correlates with ATP produced by oxidative phosphorylation. Bar chart shows relative ATP levels produced by oxidative phosphorylation (OXPHOS) in the different samples. Data were analysed by one‐way ANOVA and Tukey's post hoc test. N = 29–54 cells from five experiments, error bars are s.e.m.; *P < 0.05, ***P < 0.001.

Figure EV3. FUS does not bind VAPB or PTPIP51 in immunoprecipitation assays from transfected HEK293 cells

1. Cells were transfected as indicated with control vector (CTRL), HA‐FUS + CTRL, myc‐VAPB + CTRL, or myc‐VAPB + either HA‐FUS, HA‐FUSR521C or HA‐FUSR518K. VAPB was immunoprecipitated via the myc‐tag and the samples probed on immunoblots for VAPB using rabbit VAPB antibody and for co‐immunoprecipitating FUS via the HA tag. Input VAPB and FUS were detected using myc and HA antibodies.

2. Cells were transfected as indicated with control vector (CTRL), HA‐FUS + CTRL, HA‐PTPIP51 + CTRL or HA‐PTPIP51 + either HA‐FUS, HA‐FUSR521C or HA‐FUSR518K. PTPIP51 was immunoprecipitated using rat anti‐PTPIP51 and the samples probed for PTPIP51 using rabbit anti‐HA antibody and for co‐immunoprecipitating FUS using rabbit FUS antibody. Input PTPIP51 and FUS were detected using PTPIP51 and EGFP antibodies, respectively.

Figure 5. FUS activates GSK‐3β in transfected cells and transgenic mice

1. Cells were transfected with either control vector (CTRL), HA‐FUS, HA‐FUSR521C or HA FUSR518K and the samples probed on immunoblots for GSK‐3β phosphorylated on serine 9 (GSK‐3β‐S9), total GSK‐3β, FUS (using FUS antibody) and tubulin as a loading control. Phosphorylation of GSK‐3β serine 9 is the principal mechanism for regulating its activity; serine 9 phosphorylation inhibits GSK‐3β activity. Bar chart shows relative levels of GSK‐3β serine 9 phosphorylation following quantification of signals from immunoblots and normalization to total GSK‐3β signals. Data were analysed by one‐way ANOVA and Tukey's post hoc test. N = 4, error bars are s.e.m.; *P < 0.05.

2. Immunoblots of total and serine 9 phosphorylated GSK‐3β in spinal cords from three 10‐week‐old FUS transgenic mice and their non‐transgenic littermates. Samples were probed for tubulin as a loading control. Bar chart shows relative levels of GSK‐3β serine 9 phosphorylation following quantification of signals from immunoblots and normalization to total GSK‐3β signals. Data were analysed by unpaired t‐test, error bars are s.e.m.; *P < 0.05.

3. Cytosolic FUS activates GSK‐3β more potently than wild‐type FUS. Cells were transfected with EGFP control, EGFP‐FUS or EGFP‐FUS lacking its C‐terminal nuclear localization signal (FUSΔC). Samples were probed on immunoblots for total GSK‐3β, GSK‐3β phosphorylated on serine 9, FUS (using EGFP antibody) and tubulin as a loading control. Bar chart shows relative levels of GSK‐3β serine 9 phosphorylation following quantification of signals from immunoblots and normalization to total GSK‐3β signals. Data were analysed by one‐way ANOVA and Tukey's post hoc test. N = 3, error bars are s.e.m.; **P < 0.01, ***P < 0.001.

4. Cytosolic FUS reduces the VAPB–PTPIP51 interaction more potently than wild‐type FUS. Cells were transfected with PTPIP51‐HA, and either EGFP control, EGFP‐FUS or EGFP‐FUSΔC. PTPIP51 was immunoprecipitated using the HA tag and bound endogenous VAPB detected by immunoblotting. Both inputs and immunoprecipitations (IP) are shown. FUS was detected using EGFP antibody. Bar chart shows relative levels of VAPB bound to PTPIP51 in the immunoprecipitations following quantification of signals from immunoblots. VAPB signals were normalized to immunoprecipitated PTPIP51‐HA signals. Data were analysed by one‐way ANOVA and Tukey's post hoc test; N = 3. Error bars are s.e.m.; **P < 0.01, ***P < 0.001.

Figure 6. Inhibition of GSK‐3β increases ER–mitochondria associations and the VAPB–PTPIP51 interaction

1. Representative electron micrographs of ER–mitochondria associations in NSC34 cells treated with either vehicle or GSK‐3β inhibitors AR‐A014418 (1 μM) or CT99021 (100 nM) for 16 h. Arrowheads with loops show regions of association; scale bar = 200 nm. Bar charts shows % of the mitochondrial surface closely apposed to ER in the different samples. Data were analysed by one‐way ANOVA followed by Tukey's multiple comparison test. N = 32 cells and 181–210 mitochondria. Error bars are s.e.m.; *P < 0.05.

2. VAPB‐PTPIP51 proximity ligation assays of NSC34 cells treated with either vehicle, 1 μM AR‐A014418 or 100 nM CT99021 for 16 h. Cells were also stained for nuclei with DAPI. Bar chart shows relative number of proximity ligation assay signals/cell. Data were analysed by one‐way ANOVA and Tukey's post hoc test; n = 103–173 cells from five experiments. Error bars are s.e.m.; **P < 0.01, ***P < 0.001.

Figure 7. Inhibition of GSK‐3β rescues defective ER–mitochondria associations induced by FUS

Representative electron micrographs of ER–mitochondria associations in NSC34 cells transfected with EGFP‐FUS, EGFP‐FUSR521C or EGFP‐FUSR518K and treated with vehicle or 1 μM AR‐A014418 for 16 h. Arrowheads with loops show regions of association; scale bar = 200 nm. Bar charts shows % of the mitochondrial surface closely apposed to ER in the different samples. Data were analysed by one‐way ANOVA followed by Tukey's multiple comparison test. N = 26–27 cells and 201–236 mitochondria. Error bars are s.e.m.; *P < 0.05, ***P < 0.001.

Figure 8. Inhibition of GSK‐3β rescues FUS‐induced defects in mitochondrial Ca²⁺ uptake following its release from ER stores

HEK293 cells were transfected with M3R and either control vector (CTRL), FUS, FUSR521C or FUSR518K and then treated with vehicle (DMSO) or 1 μM AR‐A014418 for 16 h as indicated. Release of ER Ca²⁺ was induced by treatment of cells with OxoM. Representative Rhod2 fluorescence traces showing mitochondrial Ca²⁺ are shown along with bar chart displaying normalized peak values. Data were analysed by one‐way ANOVA and Tukey's post hoc test. N = 30–100 cells from four experiments. Error bars are s.e.m.; *P < 0.05, ***P < 0.001.