Two FTD-ALS genes converge on the endosomal pathway to induce TDP-43 pathology and degeneration

Abstract

Frontotemporal dementia and amyotrophic lateral sclerosis (FTD-ALS) are associated with both a repeat expansion in the C9orf72 gene and mutations in the TANK-binding kinase 1 (TBK1) gene. We found that TBK1 is phosphorylated in response to C9orf72 poly(Gly-Ala) [poly(GA)] aggregation and sequestered into inclusions, which leads to a loss of TBK1 activity and contributes to neurodegeneration. When we reduced TBK1 activity using a TBK1-R228H (Arg²²⁸→His) mutation in mice, poly(GA)-induced phenotypes were exacerbated. These phenotypes included an increase in TAR DNA binding protein 43 (TDP-43) pathology and the accumulation of defective endosomes in poly(GA)-positive neurons. Inhibiting the endosomal pathway induced TDP-43 aggregation, which highlights the importance of this pathway and TBK1 activity in pathogenesis. This interplay between C9orf72, TBK1, and TDP-43 connects three different facets of FTD-ALS into one coherent pathway.

One-Sentence Summary:

Two FTD-ALS-associated mutations converge on the endosomal pathway, the disruption of which induces TDP-43 aggregation.

Frontotemporal dementia and amyotrophic lateral sclerosis (FTD-ALS) are characterized by the formation of protein inclusions, notably the aggregation of TAR DNA-binding protein 43 (TDP-43). Aggregates imply impaired protein clearance, and several FTD-ALS-associated genes influence proteostasis, including TANK-binding kinase 1 (TBK1) (1). Loss-of-function mutations in TBK1 are risk factors for FTD-ALS (2) and studies in SOD1 mouse models have suggested that they require a so-called second hit to become pathogenic (3–5). Accordingly, some TBK1 carriers also harbor a mutation in another disease-associated gene (6, 7). Rare patients who co-harbor a TBK1 mutation and a G₄C₂ repeat expansion in chromosome 9 open reading frame 72 (C9orf72) (8, 9) present with an earlier age of onset and a more rapid disease course (7). Thus, TBK1 could modulate C9orf72-associated FTD-ALS (c9FTD-ALS).

We used an adeno-associated virus (AAV)-based C9orf72-(G₄C₂)₁₄₉ mouse model of c9FTD-ALS (10) to determine whether the repeat expansion influenced TBK1 phosphorylation at S172, a modification associated with its kinase-active form. Although total TBK1 levels were comparable in cortical brain lysates from C9orf72-(G₄C₂)₁₄₉ mice and C9orf72-(G₄C₂)₂ controls, mice expressing the expanded repeat showed an increase in the ratio of phosphorylated TBK1 (pTBK1) to total TBK1 (Fig. 1A, B). In the brains of C9orf72-(G₄C₂)₁₄₉ mice, pTBK1 localized to puncta (Fig. 1C, D). We investigated whether these puncta corresponded to dipeptide repeat (DPR) protein aggregates. Five DPRs are produced from the repeat associated non-ATG translation of the G₄C₂ repeat; all five aggregate in the brains of c9FTD-ALS patients and mice (10–14). Three DPRs co-localized with pTBK1 in C9orf72-(G₄C₂)₁₄₉ mice: poly(GA), poly(GR), and poly(PR) (Fig. 1E and Fig. S1A). These three DPRs are toxic in vivo (15–17).

Fig. 1. pTBK1 accumulates into DPR aggregates in c9FTD-ALS.

(A and B) Immunoblot (A) and quantification (B) of 6-month-old mouse cortical lysates (N = 4 mice). GAPDH, glyceraldehyde-3-phosphate dehydrogenase; A.U., arbitrary units. (C and D) Images (C) and quantification (D) of pTBK1 in 6-month-old mouse cortex (N = 9). Scale bar, 20 μm. (E) Immunofluorescence in 6-month-old mouse cortices. Scale bar, 10 μm. (F and G) Immunoblot (F) and quantification (G) of HEK293T lysates (N = 4 independent experiments). GFP monomer: ~27k Da; GFP-DPRs: high molecular weight (HMW) smears. (H) Immunostaining in the cortex of 3.5-month-old GFP and (GA)₁₀₀-V5 mice and 2-week-old GFP-(GR)₂₀₀ mice. Scale bars, 20 μm. (I) Immunostaining in healthy and C9-FTD frontal cortex (images labeled #1 and #2; scale bar, 10 μm) and hippocampus (images labeled #3; scale bar, 50 μm). (J) Immunofluorescence in healthy and C9-FTD frontal cortex. Scale bar, 10 μm. DAPI, 4′,6-diamidino-2-phenylindole. Data are means ± standard deviations (SDs). *P ≤ 0.05; ***P ≤ 0.001; ****P ≤ 0.0001; ns, nonsignificant. Statistics by Student’s t tests in (B) and (D) and one-way analysis of variance (ANOVA) in (G). Single-letter abbreviations for the amino acid residues are as follows: A, Ala; G, Gly; P, Pro; R, Arg.

We next explored which DPR or DPRs were responsible for these effects on TBK1 phosphorylation and localization. In HEK293T cells, green fluorescent protein (GFP)-(GA)₁₀₀ increased the pTBK1-to-total TBK1 ratio and coimmunoprecipitated with TBK1, whereas GFP-(GR)₁₀₀ and GFP-(PR)₁₀₀ did not (Fig. 1F, G and Fig. S1B). Accordingly, pTBK1 formed puncta in the brains of mice expressing poly(GA)₁₀₀-V5, but not in the brains of mice expressing GFP-(GR)₂₀₀ or GFP alone (Fig. 1H). Thus, poly(GA) was responsible for triggering the phosphorylation and aggregation of TBK1. Additionally, we observed pTBK1 puncta in the frontal cortex and hippocampus of C9-FTD patients but not in healthy controls (Fig. 1I and Table S1); these puncta colocalized with poly(GA) (Fig. 1J and Fig. S1C). Puncta were also detected by a total TBK1 antibody in C9-FTD (Fig. S1D).

TBK1 is activated when the oligomerization of its interactors triggers its clustering and trans-autophosphorylation (18–20): The oligomerization of poly(GA) could induce TBK1 phosphorylation by a similar mechanism. In cells, the proline-interrupted poly(GA₅P) – which is unable to form compact aggregates (15) – failed to induce the phosphorylation of TBK1 (Fig. 2A, B), did not colocalize with pTBK1 (Fig. 2C), and did not coimmunoprecipitate with TBK1 (Fig. 2F). Moreover, pTBK1 shifted to the insoluble fraction of cell lysates when poly(GA) was aggregated (Fig. 2D, E and Fig. S2A). The interaction of TBK1 with poly(GA) was independent of its phosphorylation, as coimmunoprecipitation still occurred when the S172 locus was mutated (Fig. S2B). Thus, TBK1 autophosphorylates after clustering into poly(GA) aggregates.

Fig. 2. Poly(GA) aggregation sequesters TBK1 into inclusions, which limits its activity.

(A and B) Immunoblot (A) and quantification (B) of HEK293T lysates (N = 3 independent experiments). The number symbol in (A) indicates nonspecific bands. (C) Immunofluorescence of HEK293T cells. Scale bar, 10 μm. (D and E) Immunoblot (D) and quantification € of HEK293T lysate fractions (N = 3). S, soluble; Ins, insoluble. The asterisk in (D) indicates higher exposure. (F) Coimmunoprecipitation analysis of HEK293T lysates. P, immunoprecipitation. (G and H) Immunoblot (G) and quantification (H) of HEK293T lysates (N = 3). In all blots, GFP monomer: ~27 kDa; GFP-DPRs: ~50 kDa unaggregated, HMW oligomerized. Data are means ± SDs. *P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.0001; ns, nonsignificant. Statistics by one-way ANOVA.

We hypothesize that as long as poly(GA) inclusions persist in diseased neurons, pTBK1 clusters cannot disassemble. Because TBK1 regulates the autophagic clearance of aggregates (18), its sequestration could increase poly(GA) inclusion levels. Replenishing TBK1 could help alleviate this burden. When we overexpressed wildtype TBK1 in cells expressing GFP-(GA)₁₀₀, we saw a significant increase in the phosphorylation of the autophagy adaptor p62, a known target of TBK1 (18), and a decrease in poly(GA) aggregation in immunoblots (Fig. 2G, H). This decrease was not observed when we overexpressed TBK1-R228H (Arg²²⁸→His) (Fig. 2G, H), an ALS-associated partial loss-of-function mutant with reduced autophosphorylation activity (5). As expected, overexpression of wildtype TBK1 increased the ratio of pTBK1 to total TBK1 and the ratio of phosphorylated p62 to total p62, while TBK1-R228H only slightly increased these ratios (Fig. 2G, H). The TBK1-R228H mutation did not hinder its interaction with poly(GA) (Fig. 2F).

If TBK1 is sequestered by poly(GA), further impairing TBK1 activity would be detrimental to cells. Accordingly, mutations that reduce TBK1 activity would exacerbate defects in poly(GA)-positive neurons – explaining why patients with mutations in both C9orf72 and TBK1 show a more severe disease course (7). We modeled this double-mutant condition by using an established AAV-based protocol (10, 15) to express GFP or (GA)₁₀₀-V5 in TBK1-R228H knockin (5) and wild-type mice. Mice expressing (GA)₁₀₀-V5 developed a progressive hindlimb clasping defect (Fig. 3A) and showed impaired performance in the rotarod (Fig. S3A), hanging wire (Fig. 3B), and fear conditioning tests (Fig. S3B). These phenotypes were enhanced in TBK1-R228H mice (Fig. 3A, B and Fig. S3). Mice expressing GFP behaved normally (Fig. 3A, B and Fig. S3), consistent with the TBK1-R228H model being phenotypically normal (5).

Fig. 3. Reducing TBK1 activity exacerbates the behavioral, neurodegenerative, and pathological effects of poly(GA) in mice.

(A) Hindlimb clasping scores (N = 9, 10, 16, and 14 mice at each time point). WT, wild-type. (B) Hanging wire test in 3-month-old mice (N = 9, 10, 16, and 14). (C) Whole mouse brain weight (N = 12, 15, 16, and 14). (D and E) Images (D) and quantification (E) of NeuN in mouse cortex (N = 11, 12, 13, and 12). Scale bar, 200 µm. (F and H) Images (F) and quantification (H) of pTBK1 in mouse cortex (N = 14 each). Scale bar, 20 µm. (G and I) Images (G) and quantification (I) of poly(GA) in mouse cortex (N = 14 and 16). Scale bar, 20 µm. (J to L) Immunofluorescence (J) and quantification [(K) and (L)] in mouse cortex (N = 6 each). Scale bar, 10 μm. NT, non-transduced or poly(GA)-negative. Except in (A) and (B), mice were 3.5 months old. Data are means ± SDs. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001; ns, nonsignificant. Statistics by two-way ANOVA in (A), Student’s t test in (H) to (I), and one-way ANOVA in remaining panels.

Mice expressing (GA)₁₀₀-V5 showed reduced total brain weight (Fig. 3C) and cortical neuronal loss (Fig. 3D, E). Both phenotypes were enhanced on the TBK1-R228H background (Fig. 3D, E), where the cortex was noticeably thinner (Fig. 3D). Neuroinflammation levels often correlate with neurodegeneration in mice, and as such, expression of (GA)₁₀₀-V5 increased the expression of the microglial markers Iba1 (Fig. S4A) and CD68 (Fig. S4B) and the astroglial marker GFAP (Fig. S4C). This was enhanced on the TBK1-R228H background (Fig. S4A-C).

As in SOD1 models (3, 5), our “double mutant” study is consistent with the idea that reducing TBK1 activity in neurons cell-autonomously exacerbates FTD-ALS phenotypes. Loss of glial TBK1 may be beneficial (3, 5), however, because altered immune responses could modulate c9FTD-ALS (21). We investigated whether TBK1-R228H affected TBK1-regulated immune response pathways (18) in our mice. Poly(GA) increased the expression of two interferon-stimulated genes, IFIT1 (Fig. S4D) and MX1 (Fig. S4E), but their levels were comparable in wild-type and TBK1-R228H mice. Thus, the effects of TBK1-R228H on poly(GA)-induced defects may not involve the interferon pathway, but further experiments are required to confirm this hypothesis.

We observed pTBK1 puncta in the cortex of both wild-type and TBK1-R228H mice expressing (GA)₁₀₀-V5, although pTBK1 puncta were less abundant in the TBK1-R228H mice due to the mutant’s reduced autophosphorylation activity (Fig. 3F, H). Poly(GA) inclusions were increased in TBK1-R228H mice compared to wild-type mice (Fig. 3G, I). These inclusions were intracellular and neuronal (Fig. S4F). There was also a corresponding increase in the number of puncta detected by a total TBK1 antibody in the TBK1-R228H mice (Fig. S4G, H). Total TBK1 and pTBK1 colocalized with poly(GA) aggregates in mice expressing (GA)₁₀₀-V5 (Fig. S4I and Fig. 3J-L). Thus, the R228H mutation did not prevent TBK1 sequestration but did reduce its activity, impairing its ability to induce aggregate clearance and increasing poly(GA) pathology.

Loss of TBK1 impairs endosome maturation in human motor neurons (22). We asked whether TBK1 sequestration also impaired this pathway. Mice expressing (GA)₁₀₀-V5 developed enlarged early endosomes in the cortex (Fig. 4A and Fig. S5A) that were more abundant on the TBK1-R228H background (Fig. 4A, B). We also observed aberrant Rab7-positive late endosomes in mice expressing (GA)₁₀₀-V5 (Fig. S5B-D). These endosomal defects occurred in cells that contained poly(GA) aggregates (Fig. 4C, D), suggesting a causal link.

Fig. 4. Poly(GA) induces endosomal defects that trigger TDP-43 proteinopathy, and a reduction in TBK1 activity intensifies this process.

(A and B) Images (A) and quantification (B) of early endosome antigen 1 (EEA1) in mouse cortex (N = 6, 6, 13, and 13 mice). Scale bar, 20 µm. (C and D) Immunofluorescence (C) and quantification (D) in mouse cortex (N = 6 each). Scale bar, 10 μm. (E and F) Images (E) and quantification (F) of pTDP-43 in mouse cortex (N = 6, 6, 14, and 13). Scale bar, 20 µm. (G and H) Immunofluorescence (G) and quantification (H) in mouse cortex (N = 6 each). Scale bar, 10 μm. (I to K) Immunofluorescence (I) and quantification [(J) and (K)] in HEK293T cells. Scale bar, 10 μm. (L to M) Immunofluorescence (L) and quantification [(M) and (N)] in HEK293T cells ± apilimod treatment (N = 3 independent experiments). Scale bar, 10 μm. All mice were 3.5 months old. Data are means ± SDs. *P ≤ 0.05; ***P ≤ 0.001; ****P ≤ 0.0001; ns, nonsignificant. Statistics by Student’s t test in (J), (K), (M), and (N) and by one-way ANOVA in remaining panels.

TBK1-mediated endosomal defects induce TDP-43 pathology in human motor neurons (22). Rare pTDP-43 inclusions are also observed in mice expressing poly(GA) (15). We asked whether our double-mutant mice would show more pronounced pTDP-43 pathology. In fact, pTDP-43 inclusions were more abundant in mice expressing (GA)₁₀₀-V5 on the TBK1-R228H background (Fig. 4E, F). These inclusions occurred in poly(GA)-positive cells (Fig. 4G, H), which suggests that poly(GA) aggregates, endosomal defects, and TDP-43 pathology are linked.

To determine whether endosomal defects are sufficient to induce TDP-43 pathology, we used a constitutively active Rab5-Q79L mutant to induce enlarged early endosomes in cells (Fig. 4I, J and Fig. S6A-C). In a subset of these cells, exogenously expressed TDP-43, but not GFP, showed cytoplasmic aggregation and reduced nuclear localization (Fig. 4I, K and Fig. S6A). We saw similar results when we used the phosphatidylinositol-3-phosphate 5-kinase inhibitor apilimod to disrupt endosome maturation (Fig. 4L-N). Rab5-Q79L also induced endogenous TDP-43 aggregation but not its nuclear depletion (Fig. S6B, D). Therefore, endosomal defects induce TDP-43 aggregation, but TDP-43 nuclear depletion may be downstream of initial cytoplasmic seeding. Defects in nucleocytoplasmic transport predate TDP-43 mislocalization in c9FTD-ALS (23), and TDP-43 aggregates sequester nuclear pore components and related factors (24).

We suggest a model in which poly(GA) inclusions sequester TBK1, leading to a reduction in TBK1 function that disrupts endosome maturation and induces TDP-43 aggregation. Although our work focuses on C9orf72 and TBK1, mutations in the multiple FTD-ALS genes affet endosomes or lysosomes in cells (25–28). Enlarged endosomes are also seen in an ALS-like mouse models and sporadic ALS tissue (27, 29). Furthermore, filaments of lysosomal transmembrane protein 106B (TMEM106B) are a pathological hallmark in multiple neurodegenerative diseases, including FTD-ALS (30–32), and variants in TMEM106B are associated with FTD (33). Several disease processes therefore converge on the endolysosomal pathway, the disruption of which could sensitize cells to pathogenic insults, like protein aggregation, and act as a primary driver of TDP-43 proteinopathy and neurodegeneration.