Introduction
The major classes of frontotemporal lobar degeneration (FTLD) are those characterized by the presence of neuronal and glial inclusions composed of either tau protein (FTLD-tau) or TAR DNA-binding protein of 43 kDa (TDP-43; FTLD-TDP) [
27,
40]. Familial forms of FTLD-tau are associated with mutations in the
MAPT gene that encodes the tau protein, whilst mutations in
Granulin (
GRN),
Valosin Containing Protein (
VCP), or
C9ORF72 genes can cause FTLD-TDP or amyotrophic lateral sclerosis [
3,
10,
22,
54]. Neurodegenerative conditions such as Alzheimer’s disease (AD), Huntington disease (HD), as well as Parkinson disease (PD) and dementia with Lewy bodies (DLB) are proposed to be “secondary” TDP-43 proteinopathies in which TDP-43 pathology occurs in the context of the distinctive hallmark pathology of each of these disorders [
1,
21,
37,
47,
51]. Furthermore, TDP-43 pathology has been reported in the tauopathies argyrophilic grain disease [
14] and corticobasal degeneration [
51], but it is sparse in progressive supranuclear palsy [
59]. The mechanistic connection between primary and secondary TDP-43 proteinopathies is unclear, but it is possibly related to unknown environmental or genetic factors.
One common feature in most human TDP-43 proteinopathies is the presence of cytoplasmic phosphorylated TDP-43 (pTDP-43), while normally TDP-43 is readily detected in the nucleus. Several studies have shown that antibodies specific for phosphorylated S403/404 and S409/410 TDP-43 recognize TDP-43 proteinopathies in humans [
17,
38] and in transgenic mice overexpressing TDP-43 [
7,
24,
57]. We sought to address the possible association between TDP-43 aggregation and other proteinopathies through the neuropathological analysis of mouse models of amyloidosis, tauopathy, α-synucleinopathy, and HD. This approach attempts to isolate the effect of each model’s defining genetic trigger and proteinopathy on TDP-43 aggregation, thereby eliminating parallel mechanisms that may cause TDP-43 pathology in humans (i.e., unrelated genetic or environmental factors). We discovered significant age-dependent accumulation of cytoplasmic, phosphorylated TDP-43 in two independent mouse models of tauopathy, but not in mouse models of amyloidosis, α-synucleinopathy, or (HD). As such, we demonstrate that tau-driven mechanisms can drive abnormal TDP-43 pathology in tau transgenic in vivo models.
Discussion
TDP-43 aggregation, cytoplasmic redistribution, phosphorylation and misprocessing characterize the pathology found in FTLD-TDP, associated with mutations in the
GRN,
VCP, and
C9ORF72 genes [
3,
10,
54]. In contrast, a second form of frontotemporal lobar degeneration, FTLD-tau, is characterized by hyperphosphorylated, aggregated tau pathology, and many of the familial forms are caused by mutations in the
MAPT gene that encodes tau protein [
22]. Recently, Bieniek et al. [
4] reported tau pathology in brains of individuals with FTLD associated with the
C9ORF72 expansion mutation, but similar elevation of tauopathy were not observed in FTLD associated with
GRN mutations suggesting that an overlap of FTLD-TDP and FTLD-Tau may occur in the context of
C9ORF72. Interestingly, King et al. [
29] reported an individual with an A239T sequence variant in the
MAPT gene as well as the
C9ORF72 expansion. This individual presented with dominant Pick-like tau pathology as well as the TDP-43 and p62 pathology that characterizes
C9ORF72 carriers; however, her siblings lacked this tau variant and developed typical pathology associated with the
C9ORF72 hexanucleotide repeat. Surprisingly, we were unable to find publications in which the authors clearly screened FTLD-tau cases with known pathogenic
MAPT mutations for the level and/or distribution of phosphorylated TDP-43. TDP-43 pathology has been identified in a subset of different proteinopathies including tauopathies that occur in the absence of
MAPT mutations [
2,
19,
37]. The importance of this pathological overlap has been unclear. In the current study, we sought to utilize mouse models of Aβ amyloidosis, tauopathy, α-synucleinopathy, and a polyglutamine disorder (HD) to determine if any aspect of TDP-43 pathology can be driven in vivo by an independent primary pathological aggregate (e.g., tau) caused by a defined genetic event (e.g., mutant tau).
In healthy neurons, TDP-43 is primarily localized within the nucleus, and the redistribution and aggregation of TDP-43 within the cytoplasm are thought to be critical events in TDP-43-proteinopathies [
40]. Much of the TDP-43 found within these aggregates is phosphorylated at serine residues (409/410) [
38]. In the current study, we demonstrated the association of pTDP-43 within the cytoplasm of neurons burdened with pathological tau aggregates—a tauopathy triggered solely by the expression of mutant human tau in transgenic mice. These findings are consistent with a recent report showing the partial cytoplasmic redistribution of TDP-43 in JNPL3 mice during the course of tau inclusion formation using a non-phospho-specific TDP-43 antibody, although they did not show that this was directly associated with tau pathology [
52].
In order to determine if the accumulation of pTDP-43 is directly related to the aggregation of hyperphosphorylated tau or simply the expression of high levels of human mutant tau protein, we examined brain and spinal cord from young rTg4510 and JNPL3 mice, respectively, with high levels of transgenic tau expression and minimal levels of hyperphosphorylated, aggregated tau protein. No change in TDP-43 phosphorylation or cytoplasmic distribution was observed in young tau mice. In contrast, cytoplasmic pTDP-43 was observed in both rTg4510 and JNPL3 mice after they developed overt tau pathology and, in the case of JNPL3, motor dysfunction. These results strongly support the idea that mutant tau expression alone does not induce pTDP-43 accumulation within the cytoplasm and that aggregation of tau is also critical.
The distribution of neurofibrillary tau pathology and neuronal loss in rTg4510 and JNPL3 mouse models of tauopathy closely correlated with the distribution of the neurons containing cytoplasmic accumulations of pTDP-43 protein. Hyperphosphorylated tau and pTDP-43 protein co-localized within many of the affected neurons; however, the overlap was incomplete since hyperphosphorylated tau aggregates could be found in some neurons in the absence of cytoplasmic pTDP-43. Rarely, the converse was observed. This data suggests that tau pathology precedes the redistribution of TDP-43 into the neuronal cytoplasm. Interestingly, pTDP-43 was present within the neurites of rTg4510 mice in the absence of AT8 immunopositive neuritic tau, suggesting that TDP-43 cytoplasmic accumulation may develop in other cellular domains after cytoplasmic redistribution is initiated by tau aggregation or other factors in the perikarya. Another interpretation is that neuritic pTDP-43 may co-localize with tau that is not phosphorylated at S202 or S202/T205, the phospho-tau epitopes recognized by CP13 and AT8 in the current study, respectively. Since most remaining cortical and hippocampal (CA1) neurons in the rTg4510 mice examined had tau pathology within the neuronal cell body, it seems likely that the pTDP-43 positive neurites originate from these affected cells; however, the methods utilized in this paper cannot exclude other origins of pTDP-43 localization within the neurites. Nevertheless, tau is a microtubule binding and stabilizing protein and it is possible that its aggregation perturbs normal microtubule function leading to the cytoplasmic accumulation of TDP-43 that might normally be transported to the nucleus or the presynaptic domains [
35]. The notion that perturbation of microtubule function can lead to this cytoplasmic redistribution of TDP-43 is consistent with the observation of TDP-43 pathology in Perry syndrome, a rare parkinsonian disorder [
56]. Perry syndrome is caused by mutations in
DCTN1, the large p150
glued subunit of the dynactin complex [
12] and cell culture studies show that the disruption of dynein-mediated microtubule transport can promote TDP-43 cytoplasmic aggregation [
44].
To determine if the association between pTDP-43 and hyperphosphorylated tau altered the biochemical profile of TDP-43, we performed protein fractionation from brains of rTg4510 tau transgenic and NT mice. There was no change in total levels of TDP-43 in the soluble tau fraction across genotypes; however, rTg4510 mice did have a significant increase in a higher molecular weight smear that was immunopositive for TDP-43. The nature of these high molecular weight species is unclear; however, similar high molecular weight species have been identified in affected human brains although these tend to be found in insoluble, not soluble, fractions [
1,
25,
39,
40]. It is unlikely that the TDP-43 in the HMW species is aggregated since it was localized in the soluble fraction, but it could reflect various post-translational modifications such as ubiquitination or oxidative modifications [
9,
40]. Given that tau in human tauopathy and in the tau transgenic mice utilized in this study becomes hyperphosphorylated and aggregated, thereby shifting into the detergent insoluble fraction, we sought to determine if TDP-43 similarly shifted into the sarkosyl-insoluble fraction in association with the tauopathy in rTg4510. Indeed, we saw a significant increase of full-length TDP-43 within the sarkosyl-insoluble fraction when compared to NT mice, supporting the close association between the tau pathology observed in these mice and the cytoplasmic accumulation of pTDP-43. Intriguingly, we also observed a significant increase in a low molecular weight species of TDP-43 which we termed TDP-35 for its migration of ~35 kDa. The exact nature of TDP-35 and its relevance to the tauopathy observed in our models is unclear. In humans, it has been suggested that a similar 35 kDa species observed in TDP-43 proteinopathies may be generated from alternative translational or splicing pathways or may be the result of cleavage by caspase activity [
41,
53,
61]. Indeed, caspase activation is a feature of the mouse models of tauopathy utilized in this study [
49,
60].
The association between cytoplasmic pTDP-43 and tau appears specific since we saw no evidence of cytoplasmic relocalization of pTDP-43 in mouse models of Aβ amyloidosis, α-synucleinopathy or polyglutamine disease (HD), regardless of the broad spectrum of ages and stages of primary proteinopathy examined. A number of papers report TDP-43 pathology in AD with estimates ranging from 23 to 56 % [
1,
2,
19,
21,
30,
51]. Many of the inclusions in human brains display close overlap between tau and TDP-43 [
1,
2], similar to that observed in the tauopathy mice here. Lin and Dickson [
32] also previously reported that in human AD brains, TDP-43 also can associate with tau within neuronal inclusions at the ultrastructural level.
The amyloid models that we utilized in the current study do not develop tauopathy similar to that observed in AD; therefore, it is possible that amyloidosis and tauopathy act in concert in AD to produce TDP-43 pathology. Caccamo et al. [
6] reported that the 3XTg-AD amyloid model [
42], which express mutant amyloid precursor protein, mutant presenilin 1, and P301L mutant tau protein, have increased full-length and ~35 kDa TDP in the low salt fraction and cytosolic fraction at 6 months of age, but not at 2 months or 12 months. Caccamo et al. [
6] suggested that high levels of soluble amyloid beta oligomers positively correlated with TDP-43 changes, but they did not report an association with tau. Since the tauopathy in the 3XTg-AD model is much later and more modest than the amyloid pathology, it is not clear if the impact of tauopathy would have been observed by Caccamo et al. in the ages of mice examined. Herman et al. [
18] also reported increased TDP-43 expression, cleavage and aggregation in association with intracellular amyloid beta 1–42 using lentiviral expression of amyloid beta 1–42 in rat motor cortex. Neither study reported an association of TDP-43 with the extracellular plaques that we examined in this report.
TDP-43 pathology is frequently observed in the brains (18–60 %) of patients with DLB [
2,
19,
37], however, tau, α-synuclein and Aβ amyloid deposits often coexist in these brains making it difficult to assess which of these primary insults may trigger TDP-43 inclusion formation. In the current study, we used transgenic mice that primarily develop each specific type of these three inclusions to provide a useful indication of which one is more likely to contribute to TDP-43 cytoplasmic aggregation. The association between tau aggregation and pTDP-43 cytoplasmic aggregation in these tau transgenic mice could suggest that tau is the most critical factor driving TDP-43 aggregation in human DLB. Alternatively, that the co-occurrence of α-synuclein, tau and amyloid pathology in DLB could trigger an alternative mechanism which drives TDP-43 aggregation. Our currently available mouse models would not allow us to explore this scenario.
Our group has also shown that constitutive overexpression of wild-type and less so mutant TDP-43 can cause aggregation of hyperphosphorylated tau protein at S202, one of the two epitopes that is recognized by the AT8 antibody used in the current study [
58]. This data also suggested that activation of PKC in the TDP-43 mice led to the hyperphosphorylated tau providing another link between tau and TDP-43. More recently, Jinwal et al. [
26], reported that clearance of TDP-43 protein via a Cdc37/Hsp90 complex is impaired by the accumulation of tau. This recent finding also could underlie our in vivo findings in the tau transgenic mice that show robust aggregation of hyperphosphorylated tau protein.
Our results show a clear link between tau pathology and cytoplasmic accumulation of phosphorylated TDP-43 in the controlled in vivo systems of tau transgenic mice. It is currently unknown if this association between tau and TDP-43 can affect the disease course in either the mouse models or in human tauopathies. Our studies lay the groundwork for such investigations. Certainly, our data would suggest that groups with large cohorts of
MAPT mutation carriers should assess their autopsy tissue for overlapping tau and TDP-43 pathology; however, such screens are generally precluded by the availability of tissue from known
MAPT carriers. Functional studies which exploit the capabilities of the in vivo model systems utilized in this report could compliment these human studies. For example, the rTg4510 model of tauopathy can now be crossbred with conditional TDP-43 models created by our group and others [
7,
24] to determine if the two pathologies act in concert to accelerate the FTLD-like neurodegeneration of these models. Furthermore, we can now suppress tau expression in the rTg4510 mice and determine if the TDP-43 pathology is reversible and if any reversion of TDP-43 pathology tracks with the cognitive recovery observed in tau suppressed rTg4510 mice [
46]. In addition, seeding and spreading techniques [
23] that are proving informative for the tau field could be expanded into TDP-43 transgenic mice (and vice versa) to help determine the cross talk between tau and TDP-43. Clarification of the disease relevance between tau and TDP-43 will ultimately allow us to determine if therapeutic efforts aimed at one molecule may hold promise against diseases characterized by the other protein.