Nucleo-cytoplasmic transport of TDP-43 studied in real time: impaired microglia function leads to axonal spreading of TDP-43 in degenerating motor neurons

Pathological deposition of protein aggregates characterises a large group of neurodegenerative diseases in which the key event is considered to be either the malformation of a key protein, or the failing of a crucial metabolic cascade, leading to the generation of a pathological proteinaceous agent that can self-propagate and result in gain- or loss-of-function progressing toward cell death [38].

An aggregating proteinopathy involving the TAR DNA-binding protein (TDP-43) provides a molecular link between two progressive neurodegenerative diseases: amyotrophic lateral sclerosis (ALS) of the motor cortex, brainstem and spinal cord, and frontotemporal lobar degeneration (FTLD) of the frontal, insular, and temporal cortex (non-SOD1 ALS and non-tau FTLD, termed ALS-FTLD; [3, 13, 15, 22, 43]).

In biochemical and cell culture models, TDP-43 has been identified as spontaneously aggregation prone [1, 27]. This propensity has been localised to a low-complexity domain in the C-terminal region that may play a role in RNA binding, granule formation, and/or liquid–liquid phase separation to subserve transcription dynamics and mRNA transport [6, 12, 32, 34, 55, 56]. It is hypothesised that inherited or spontaneous mutation of the encoding gene, TARDBP, or of an interacting partner, upsets the balance between aggregation and dissolution of TDP-43 [14, 71], which may be differentially regulated under changing conditions, including during the cellular stress/death response [7, 30].

TDP-43 also appears to display self-propagating behaviour. Feiler et al. [21] found that naive cultured mouse neurons will internalise tagged, oligomerized TDP-43 from the culture medium of transfected neurons. This was mediated by the packaging of TDP-43 into microvesicles or exosomes, which were more effective at transmission than free protein. Linking this phenomenon to clinical observations, when cell culture models were seeded with phosphorylated TDP-43^+ve insoluble fractions from ALS and FTLD patient brain or spinal cord, there was a significant increase in intracellular endogenous TDP-43 aggregation and subsequent cell death [44, 57]. To follow on from this we examined how motor neuron injury and subsequent degeneration mobilises TDP-43 in vivo and may contribute to pathogenicity in ALS.

We have developed an in vivo transgenic model to visualise subcellular compartments of individual motor neurons in zebrafish spinal cord. Using this new model, we have expressed fluorophore-tagged human TDP-43 within individual spinal motor neurons to observe the intraneuronal localisation of TDP-43 over time. After confirming that the subcellular localisation of TDP-43 is the same in zebrafish and human, we discovered in vivo that TDP-43 concentrated in highly dynamic nuclear granules with droplet-like properties. By application of a cell-specific UV injury, we were able to induce neuronal degeneration, within which the fate of TDP-43 could be precisely followed. We found that motor neurons underwent a stereotyped neurodegenerative process with structural changes closely mirroring those of degenerating neurons in human tissue. We further report upon live-imaging evidence for different pathways of TDP-43 release and distribution; as aggregated fragments, diffuse in cytoplasm, as a component of pyknotic or karryhorexic nuclear fragmentation, anterograde axonal diffusion or as a component of microglial uptake.

Our understanding of the molecular pathogenesis of ALS and FTLD is limited by our incomplete understanding of the sequence of events underpinning the pathology. Many of the insights we currently have on the hallmarks of ALS/FTLD were derived from postmortem histological studies. In almost all ALS patients and more than half of FTLD patients, the key molecular histopathological hallmark is the presence of abnormal amounts of intracellularly deposited TDP-43. It typically forms cytoplasmic-insoluble aggregates within neurons [3, 43]. Numerous in vitro models of TDP-43 overexpression (wildtype and mutant) have been investigated to understand how these TDP-43 aggregates are formed. These studies collectively suggest that TDP-43 is a highly aggregation-prone protein [27] and that a variety of molecular pathways can lead to cytoplasmic mislocalisation, aggregation and inclusion formation [14, 20, 30, 56, 71]. Moreover, although a true phenocopy of ALS/FTLD has been elusive in TDP-43 animal models [26], the overexpression of mutant forms of TDP-43 in mice does result in the development of motor impairment and lethality [29, 58, 63]. A key challenge with these animal models is that similar to the clinical postmortem situation, the disease histopathology only provides snapshots at different time points during disease progression. In contrast, in the present study, we were able to observe the fate and dynamics of TDP-43 cytoplasmic deposition in real-time at high resolution within the spinal cord in vivo. Notably, we were able to identify distinct stages of motor neuron degeneration and the mislocalisation and spread of TDP-43. In Fig. 8, we summarise the temporal sequence of neurodegeneration and TDP-43 mislocalisation. While this process of UV-mediated injury can reflect an accelerated cell-death pathway, we occasionally observed the formation of TDP-43 aggregates in the cytoplasm of spinal motor neurons. It will be important to explore if these aggregates are of similar structure and function to the pathological aggregates (e.g. phosphorylation status of TDP-43) that are observed in diseased human postmortem tissue and in animal models overexpressing mutant TDP-43.

The subcellular redistribution of TDP-43 from the nucleus to the cytoplasm in injured motor neurons (when microglia were depleted and, therefore, the cells were not rapidly phagocytosed) was especially striking, and this study represents the first real-time in vivo observation of the phenomenon. This redistribution of TDP-43 to the cytoplasm has been suggested to represent a landmark feature of the onset and progression of ALS/FTLD [23, 43], and in cell culture the disrupted trafficking of TDP-43 is a precursor to the formation of cytoplasmic TDP-43 inclusions [66]. Previously, Sato et al., [54] have reported the cytoplasmic mislocalisation of TDP-43 in response to neuronal stress in a mouse ligation model. Our observations provide in vivo support for the hypothesis that cytoplasmic accumulation of TDP-43 in motor neurons could occur as a direct downstream consequence of neuronal stress or injury. It is important to point out that in our experiments microglia did not directly prevent the cytoplasmic mislocalisation of TDP-43 in UV-injured motor neurons, rather this process was circumvented by the microglial phagocytosis of the degenerating neuron. This raises an intriguing question—why is TDP-43 pathology limited to ALS/FTLD and not other neurodegenerative diseases (such as Alzheimer’s, Parkinson’s and Huntington diseases)? It is possible that the TDP-43 redistribution that we have observed is elicited by specific types of neuronal stress/injury (such as UV that was used in this study). Alternatively, it is possible that microglia are specifically defective in their phagocytic function in ALS/FTLD (and not other diseases), therefore, allowing intracellular TDP-43 redistribution to occur. These are important directions for future follow-up. Interestingly, Spiller and colleagues recently demonstrated the selective uptake of TDP-43 by reactive microglia in mice, and that this process is likely to play an important role in clearing pathological (mislocalised) TDP-43 from neurons [59].

Significant recent interest in the field has focussed on the potential cell-to-cell transfer of TDP-43 and/or associated proteins, and the possibility that this might facilitate propagation of the disease process through the motor system. In vitro, the intercellular transfer of TDP-43 has been demonstrated [21, 47, 57, 67]. However, whether this also occurs in vivo remains an open question. In our study, we observed two pathways for transfer of TDP-43 from the originating dying neuron: uptake by microglia via direct phagocytosis and transfer into the surrounding neuropil.

UV injury-induced degeneration resulted in a phagocytic response by spinal microglia, which took up predominantly nuclear TDP-43. This represents the first in vivo demonstration of TDP-43 uptake by microglia as well as the prevention of TDP-43 spreading by normal microglia. Further study into the fate of TDP-43 and whether microglial function is altered following TDP-43 uptake in the zebrafish model may help expand upon the early evidence for microglial involvement in the progression of ALS/FTLD [8, 35, 46]. For example, phagocytosis of degradation-resistant TDP-43 aggregates [69] may disrupt the lysosomal pathway in microglia and, therefore, impair their ability for clearance of TDP-43. Interestingly, Paolicelli et al. have recently reported that TDP-43 depletion in Cx3Cr1-expressing cells (including microglia) led to an increase of phagocytosis and elevated transcription of lysosomal components [46]. This suggests that both endogenous and exogenously acquired TDP-43 may alter the functional behaviour of microglia, and contribute towards the progression of ALS/FTLD.

Depletion of microglia through PU.1 knockdown allowed us to visualise intracellular spreading of TDP-43. In fact, we observed at least three pathways for TDP-43 dispersal following motor neuron degeneration after injury: this included diffusion of TDP-43 from the nucleus to the cytoplasm after neuronal swelling, compartmentalised TDP-43 release in discrete quanta, and the breakdown of homeostatic TDP-43 exclusion from the axon. Concerning the interesting axonal spreading of TDP-43 it is important to note that we consistently did not detect any eGFP-TDP43^WT inside axons of healthy spinal motor neurons. Neuronal cell cultures and drosophila imaging have identified axonal TDP-43 as a component of transport vesicles in keeping with a role as a chaperone of RNA transport [2, 19]. We frequently observed the cytoplasmic mTagBFP and the CAAX-tagged membrane-bound mKO2 fluorophores incorporated into transport vesicles moving antero- as well as retro-grade along the axon (data not shown). However, we never observed this active transport for eGFP-TDP43^WT. This difference indicates the intriguing possibility that the incorporation of TDP-43 into transport vesicles may be differentially regulated between neuronal cell types, in an in vivo setting (i.e. cell culture vs vertebrate spinal cord) or in distinct environmental or developmental contexts.

This study provides the first live-imaging data to support axonal-mediated spreading of TDP-43 [9, 45]. In combination with observations obtained in cell cultures suggesting TDP-43 transfer via the axonal membrane [21], this provides evidence that axonal TDP-43 might be an important mediator of ALS/FTLD pathology. It is worth noting that alongside our observations of spinal microglia engulfment of motor neuron cell bodies in this study and in two previous studies [40, 41], we have never observed microglial phagocytosis of degenerating axons. Given that microglial neuronophagia preceded TDP-43 nucleo-cytoplasmic transport and that the anterograde movement of TDP-43 was a late process in neuron degeneration (that we could observe best when microglia were depleted), it is reasonable to conclude that rapid microglial phagocytosis of injured motor neuron somata is an effective mechanism for the prevention of TDP-43 spread.

Taken together, nucleo-cytoplasmic transport of TDP-43 was studied in real time in the living spinal cord, and revealed that depletion of functional microglia leads to axonal and extracellular spreading of the pathological protein. This provides new insight into the potential direct role of microglia as well as pathways for protein distribution in the cascade of cellular and molecular events that cause ALS and FTLD.