Although quickly expanding, our understanding of transient stress granule structures and their role in disease progression is at a primitive stage. Currently, stress granules are known to contribute to the pathogenesis of several disorders, including fragile-X syndrome, spinal muscular atrophy, and ischemia-reperfusion injury (5
); like TDP-43 and FUS/TLS (fused in sarcoma/translated in liposarcoma) proteinopathies, these disorders are caused by altered RNA-binding proteins (4
). In this study, we show that sorbitol is a novel stressor mediating TDP-43 and hnRNP A1 localization to TIAR+
stress granules; this was observed in both somatic (Hek293T cells) and nervous system (primary cultured glia) cell types. The finding that colocalization of hnRNP A1 with TDP-43+
granules was conserved between somatic cell lines and primary cultured glia (Fig. and ) indicates that the readily accessible and common Hek293T cell line can be used to model the stress response initiated in cortex-derived glia. We used this information to develop a cell culture model to quantitatively test the stress response mounted by wild-type TDP-43 and the pathological TDP-43 G294A, A315T, G348C, and N390S mutants, in conjunction with several truncated TDP-43 mutants.
Our analysis indicates that a 57-residue region (residues 268 to 324) spanning the first one-third of the glycine-rich region (GRR) is necessary for the association of TDP-43 with stress granules (Fig. ). Moreover, removal of the distal 90 residues of GRR (residues 325 to 414) leads TDP-43 to behave like a pathological mutant in terms of its ability to increase stress granule size and facilitate assembly. While the current report was under review, another study using arsenite as a stressor similarly found that pathological TDP-43 mutants enhance their stress-induced localization to stress granules (53
). These observations are consistent with the model that the GRR harbors intrinsic determinants of stress granule formation: residues 268 to 324 are necessary for the association of TDP-43 with stress granules, while residues 325 to 414 are necessary, but not sufficient, for regulating this association. In this model, the pathological missense mutations—both inside and outside the distal GRR—negate this regulatory capability, perhaps by changing the structure of TDP-43 in a manner similar to distal truncation of TDP-43. However, we cannot rule out alternative explanations. For example, mutations in TDP-43 may alter the interaction of the GRR with structural components of the stress granules to regulate their size. Yet another explanation would be that these mutations result in the acquisition of an as-yet-unknown gain of function, such as higher toxicity in general, which in turn may result in larger stress granules. Future experiments will address each of these models.
Our work reinforces the notion that the GRR plays a crucial regulatory role in proper TDP-43 function. The GRR has been implicated in the regulation of splicing through hnRNP protein-protein interactions (47
) and association with stress granules (23
). This domain also harbors the vast majority of known pathological mutations in TDP-43 (and similarly in the pathologically related protein FUS/TLS) (51
). The osmotic-stress-responsive sequence (residues 268 to 324 [Fig. ]) identified in our study overlaps with the oxidative-stress-responsive sequence (residues 216 to 315) previously identified by Colombrita et al. (23
). We note, however, that the osmotic-stress-responsive sequence identified here is only half the size of the oxidative-stress-responsive sequence. It is conceivable that the overlapping GRR (residues 268 to 315) of these two stress-responsive sequences in TDP-43 represents a conserved stress response sequence that mediates its association with stress granules in different cell types (both somatic and nervous system specific) and with different stressors (oxidative and osmotic stressors). This region (residues 268 to 315) is particularly glycine rich (Fig. ), with multiple putative GXXXG motifs; these motifs are believed to mediate hydrophobic helical protein-protein interactions and have also been found in other proteins involved in neurodegenerative disorders (13
). We should also note that while sorbitol can be an oxidative stressor (Fig. ), the high concentrations of sorbitol used in this study are typically associated with hyperosmotic stress. Future studies will define the stress-responsive sequence motif in the GRR domain of TDP-43, which may include these GXXXG motifs, and will test whether such a motif is conserved in other GRR-containing proteins, in different cells, and with different stressors.
FIG. 8. Model summarizing distinct stress granule species formed by wild-type and mutant (G348C) TDP-43. (a) The GRR harbors putative structural determinants for the association of TDP-43 with stress granules (box) and for the control of stress granule size. (more ...)
Since all missense mutations tested in our study mount similar stress responses, we analyzed one of these mutations (G348C) in more detail under progressively longer stress treatments. We found, in comparing cells expressing G348C mutant versus wild-type TDP-43, that a higher percentage of G348C mutant-expressing cells directed the protein into stress granules, that the number of granules formed peaked after a short stress exposure, and that the stress granule size progressively increased at the expense of the number of granules. Yet wild-type TDP-43-expressing cells coped with stress in the converse manner: instead of progressively forming larger granules over time, the wild-type protein progressively formed a greater number of smaller granules of relatively unchanged size (Fig. ). Our results indicate that the pathological mutant handles stress in a manner markedly different from that of the wild type; these differences will be explored in future studies in order to understand how pathological TDP-43 mutants contribute to neurodegeneration and cell death.
It has been suggested that the formation of stress granules delays apoptosis by sequestering proapoptotic factors (such as RACK1 and ROCK1) (8
). Our observation that cells recover after an initial stress response is compatible with this model. However, it is also expected that prolonged stress (beyond the cell's capability to handle such stress) would trigger cell death. Indeed, TDP-43, like PARP, is a target of caspase-3; multiple sites in TDP-43 are reportedly targeted by caspase-3 (DEND10-13
, and DVMD216-219
), producing 25- and 35-kDa species (30
). In our study, we observed that prolonged sorbitol treatment reduces cell viability, which coincides with the formation of TDP-43+
stress granules in native Hek293T cells (Fig. ). In spite of the drastically increased stress granule size in cells overexpressing pathological mutants (compared to that for the wild-type protein), we did not observe premature induction of apoptosis in these cells (as indicated by PARP cleavage [data not shown]). We interpret this observation as demonstrating that stress granule size is not a primary determinant of the initiation of apoptosis and that mutants may actually delay the onset of cell death before the host cells commit suicide.
In interpreting our data, we consider that the effect and time course of stress granule formation in cells overexpressing a pathological mutant should best be compared to those in cells overexpressing a wild-type protein. This is particularly relevant when one considers that overexpression of wild-type TDP-43 protein (or several other proteins implicated in neurodegenerative disorders) in vivo
also promotes TDP-43 pathology and neuronal death over time. In this context, while the general effects are similar, there are subtle differences between the stress response of overexpressed wild-type protein and that of endogenous TDP-43. Another related consideration is that TDP-43 protein levels are likely to be stringently regulated, such that overexpression of TDP-43 alone could cause cell death in some cell types (9
). However, at our level of overexpression, we did not observe cell death in unstressed stable cell lines expressing wild-type and mutant TDP-43 (Fig. ).
Both neurons and glia display TDP-43+
cytoplasmic aggregates in a spectrum of ALS and FTLD-U neurodegenerative disorders, as well as in a subset of Alzheimer's and Parkinson's diseases (21
). Degeneration and the eventual death of neurons in these diseases likely reflect a combined outcome of impaired neurons and their supporting glial cells. It is plausible that the neuronal and glial cytoplasmic aggregates in ALS or FTLD-U are end products of stress granules or derivatives of these structures generated after an unsuccessful response to stress. TDP-43 has previously been shown to localize to T-cell-restricted intracellular antigen-1 (TIA-1)-positive stress granules in axotomized C57BL/6 mouse motor neurons (56
). While previous studies failed to detect colocalization of TDP-43 with known stress granule markers in patient brain (57
) and spinal cord (23
) samples, improved techniques now allow the detection of TDP-43 in stress granules in brain samples from ALS and FTLD-U patients (53
). Further investigation is needed to analyze the relationship of stress granule formation and TDP-43 proteinopathies in cellular and animal models and in human patients in order to understand the molecular and cellular mechanisms by which TDP-43 and RNA-binding proteins respond to stress in health and disease.