Many TDP-43 fragments are prone to aggregate when expressed in mammalian cells
To understand the properties of fragmented TDP-43, we made a series of constructs that synthesize the N-terminal and C-terminal truncated TDP-43 fragments (). We tagged the constructs with EGFP at their carboxyl termini so that we can easily follow the behavior of the proteins synthesized from these constructs. After transfection into motor neuron-like NSC-34 cells 
, we observed intracellular aggregates from several TDP-43 fragments (). We quantified the percentage of aggregate-containing cells among the GFP-fluorescent cells and also the percentage of cells with aggregates in the nucleus or the cytoplasm (). We did not observe aggregates from the full-length TDP-43-EGFP fusion protein (). However, removal of the N-terminal end of TDP-43 up to the middle of the RRM2 (constructs 2, 3 and 4) resulted in aggregates (; ). Further removal of the C-terminal half of the RRM2 (construct 5) diminished aggregation (, ). Similarly, removal of the C-terminal end of TDP-43 up to the entire glycine-rich domain (constructs 6, 7 and 8) led to aggregation (; ). Further removal of RRM2 (constructs 9 and 10) abolished aggregate formation, so did removal of both N- and C-termini of TDP-43 (construct 11; ). These observations were fully replicated in HeLa cells (data not shown) and suggest that C-terminal half of the RRM2 domain is necessary but not sufficient for TDP-43 aggregation.
TDP-43 aggregation and distribution of non-aggregated TDP-43-EGFP fragments in NSC-34 cells.
TDP-43 aggregation in NSC34 cells.
In general, cells transfected with the C-terminal fragments 2, 3 and 4 had a high percentage of aggregate-containing cells than the cells transfected with the N-terminal fragments 6, 7 and 8 (). In cells where there were no aggregates, all the C-terminal fragments were predominantly cytoplasmic while all the N-terminal fragments were predominantly nuclear (, ). This is expected based on the literature that the nuclear localization signal is located in the N-terminal region 
. Surprisingly, we notice that the full length construct 1 did not show a pattern of nuclear enrichment by visualizing EGFP fluorescence (). To determine whether the EGFP signal truly reflects the distribution of TDP-43, we examined the protein integrity in the cell extracts using Western blot. When probed by an anti-GFP antibody, all constructs showed a fusion protein band (). However, construct 1 also showed a strong EGFP band at ~25 KD, indicating the EGFP signal in the cells transfected with construct 1 largely reflects the distribution of EGFP, which probably obscured the distribution of TDP-43-EGFP fusion protein. How the EGFP was produced from the construct is not clear but by immunostaining using an anti-TDP-43 antibody, TDP-43 was predominantly nuclear in cells transfected with construct 1 (), thus indicating that the TDP-43-EGFP fusion protein is predominantly distributed in nucleus, consistent with the literature 
Expression of the various TDP-43-EGFP constructs in NSC-34 cells.
TDP-43 fragments impair neurite growth in an aggregation-dependent or independent manner
To investigate the functional consequence of TDP-43 aggregation, we examined the cell death/viability. By several assays, including trypan blue staining, MTS assay, counting EGFP-fluorescent cells at 1, 2 and 3 days after transfection, TUNEL, activated caspase 3 staining and LDH release, we did not find a significant difference among the cells transfected with the constructs expressing EGFP and various TDP-43 fragments ( and data not shown), suggesting that the TDP-43 fragments, including those that form aggregates, do not significantly affect the viability of the cells in our culture.
TDP-43 fragments did not increase cell death.
To determine whether the fragmented TDP-43 impairs neuronal characteristics, we transfected the constructs into NSC-34 cells and then induced neuronal differentiation. The cells transfected with full-length TDP-43-EGFP grew long neurites that were not different from the cells transfected with EGFP (, EGFP, construct 1). In contrast, cells transfected with the C-terminal TDP-43 fragments grew short or no neurites (, constructs 2–5), indicating that the C-terminal fragments impair neurite growth. Furthermore, this impairing activity was not dependent on TDP-43 aggregation. Both cells with aggregates (, construct 2–4 inserts) or without (, constructs 2–4) showed short or no neurites (, white and black bars). Furthermore, construct 5, which had almost no aggregates () and did not form any observable aggregates under the differentiation condition, strongly impaired neurite growth (). These observations indicate that aggregation is not a prerequisite for the C-terminal fragments to impair neurite growth.
Expression of TDP-43 fragments inhibits neurite growth in motor neuron-like NSC-34 cells.
Knockdown of TDP43 inhibits neurite growth during neuronal differentiation of NSC-34 cells.
In contrast to the C-terminal fragments, the majority of cells transfected with the N-terminal fragments or the middle fragment grew nearly normal neurites (, constructs 6-11; , white bars). However, a small number of cells that were transfected with constructs 6, 7 and 8 displayed aggregates (, constructs 6–8 inserts). In these cells, neurite growth was impaired (, black bars), suggesting that for these N-terminal fragments the impairing activity for neurite growth depends on their aggregation.
Overexpression of full-length TDP-43 rescues neurite growth
The TDP-43 fragments could impair neurite growth by a gain of toxicity that is independent of the normal TDP-43 function or by dominant negatively interfering with the normal TDP-43 function. If a gain of toxicity is true, overexpression of full length wild type TDP-43 should not affect this toxicity. On the other hand, if the dominant negative mechanism is true, overexpression of the wild type TDP-43 should compensate for the functional inhibition caused by the fragments and rescue neurite growth. To differentiate these two possibilities, we cotransfected the full-length TDP-43 with the TDP-43 fragments into the NSC-34 cells and quantified the number of transfected cells with neurites (, green and red bars). The full-length TDP-43 was able to rescue the neurite growth from the inhibition caused by TDP-43 fragments. These results are consistent with the possibility that the TDP-43 fragments impair neurite growth by dominant negatively interfering with the normal function of TDP-43.
TDP-43 function is required for neurite growth during neuronal differentiation
If interference of the normal function of TDP-43 caused impairment in neurite growth, then a loss of TDP-43 function is expected to have the same effect. To test this, we transfected NSC-34 cells with a microRNA (miRNA)-expression construct that silences TDP-43 expression. This construct employed the CAG promoter to drive expression of a primary miRNA, which encode a red fluorescent protein (dsRed2) and a pre-miRNA targeting mouse TDP-43 () 
. Therefore, these constructs expressed both dsRed2 and a miRNA, which allowed us to track the transfected cells. To control for possible non-specific effects, such as stress associated with transfection, non-specific silencing or non-specific cellular responses to particular miRNA sequences, we used constructs that express two different miRNAs (miR-TDP-43a and miR-TDP-43b) targeting different regions in mouse TDP-43 mRNA. In addition, a control construct that expresses miRNA with scrambled sequence (miR-SCR) was used. We first tested whether the miRNA knocked down TDP-43 by transfecting these constructs into the NSC-34 cells. After 48 hours, we selected red-fluorescent cells using FACS and carried out Western blot on the protein extracted from the fluorescent cells. TDP-43 was knocked down in cells transfected with either TDP-43-specific miRNA compared with those cells transfected with the scrambled miRNA (). Similar to the expression of TDP-43 fragments, TDP-43 knockdown did not have a significant effect on cell death/viability ( and data not shown) but significantly impaired the neurite growth ().
Confirmation of the observations in primary neurons
To confirm the neurite phenotype observed in NSC-34 cells, we transfected the constructs into the rat forebrain neural precursors and then differentiated them into neurons. The neurons transfected with full-length TDP-43-EGFP, the N-terminal fragments (construct 6-10) and the RRM2 fragment (construct 11) grew normal neurites that were not different from the cells transfected with EGFP alone (). In contrast, cells transfected with two C-terminal TDP-43 fragments (construct 2 and 3) grew short or no neurites (). Differing from the NSC-34 cells, the C-terminal fragments 4 and 5 and the N-terminal fragments 6, 7 and 8 did not impair neurite growth in primary neurons (). This was most likely due to the relatively low transfection rate and the low expression levels achieved in the primary neurons, where no constructs formed aggregates. However, the two C-terminal fragments 2 and 3 impaired the neurite growth, suggesting that these two fragments have relatively potent toxicity.
Expression of TDP-43 fragments inhibits neurite growth in primary rat forebrain neurons.
To determine whether TDP-43 is required for neurite growth in primary neurons, we also conducted the knockdown experiments. Because the transfection efficiency was low, we first tested whether our two miRNAs (which were designed to silence mouse TDP-43) could knockdown rat TDP-43 by transfecting our constructs into rat-2 cells, which can be transfected with above 90% efficiency. Although both miR-TDP43a and miR-TDP43b matched rat TDP-43 mRNA sequence well, miR-TDP-43a did not knockdown rat TDP-43 (data not shown), probably due to the sequence differences in the regions flanking the miRNA target between the rat and mouse TDP-43 mRNA. miR-TDP43b knocked down rat TDP-43 efficiently (). Therefore, we transfected miR-TDP-43b and miR-SCR into the rat primary neurons and found that neurite growth in cells transfected with miR-TDP-43b was also significantly impaired (). These results confirm that TDP-43 also is required for neuronal differentiation in primary neurons.
Knockdown of TDP43 inhibits neurite growth in differentiated primary rat fore brain neurons.
Fragmented TDP-43 co-aggregate with the wild type TDP-43 and reduce nuclear TDP-43
The above data are consistent with the hypothesis that TDP-43 fragments that contain RRM2 are prone to aggregation and impairing neurite growth by dominant-negatively interfering with the normal TDP-43 function. The functional impairment could be mediated through the interaction between the TDP-43 fragments and the full-length TDP-43. To test this possibility, we cotransfected the EGFP-tagged TDP-43 fragments with a full-length TDP-43 that was tagged with a red fluorescent protein tdTomato (RFP). Unlike the EGFP-tagged TDP-43, this construct expressed the full-length fusion protein and unfused tdTomato was detected (). When cotransfected with the mutants that do not form aggregates (, constructs 5 and 11), this wild type TDP-43 was predominantly in the nucleus with a diffused pattern. In contrast, when cotransfected with mutants that form aggregates, the wild type TDP-43 colocalized with the aggregates and the diffused TDP-43 signal in the nucleus was reduced (, compare the red signal of constructs 2, 3, 4, 6, 7 and 8 with that of constructs 5 and 11). This suggests that the TDP-43 fragments can co-aggregate with the wild type TDP-43 and consequently deplete the TDP-43 in the nucleus.
Coaggregation of wild type TDP-43 with its fragments.
To determine whether the aggregation reduces the endogenous nuclear TDP-43, we transfected the cells with the EGFP-tagged fragments. To avoid the interference from the transfected TDP-43 fragments, we observed the endogenous TDP-43 using antibodies against the N- or C-terminal peptide. In cells transfected with wild type TDP-43, there was an increase in the nuclear TDP-43 staining compared with the untransfected cells (, #1, compare the nuclei pointed with filled arrowheads with those with the open arrowheads) or EGFP-transfected cells (not shown), but the overall ratio between the nuclear and cytoplasmic TDP-43 remained the same as the EGFP-transfected cells (), indicating that both nuclear and cytoplasmic TDP-43 are increased to a similar extent. Similarly, in cells transfected with a fragment that does not form aggregates (construct 11), the TDP-43 staining intensity was the same as the untransfected cells (, #11) or EGFP-transfected cells (), indicating that this construct does not alter the nuclear and cytoplasmic distribution of endogenous TDP-43. In contrast, in cells transfected with the aggregation-prone fragments, the endogenous TDP-43 was detected in the aggregates (, #2, 3, 4, 6 and 7, arrows) but was reduced in the nucleus (, #2, 3, 4, 6 and 7, compare the nuclei pointed with filled arrowheads with those with open arrowheads). This observation was confirmed by quantification of the ratio between the nuclear and cytoplasmic staining density (). Interestingly, in cells transfected with the C-terminal fragments, the ratio between the nuclear and the cytoplasmic TDP-43 was significantly reduced regardless whether the cells had aggregates or not (). On the other hand, in cells transfected with the N-terminal fragments, this nuclear to cytoplasmic ratio was significantly reduced in cells with aggregates but remained unchanged in the cells without aggregates. Thus, the reduction of nuclear TDP-43 () correlates with the inhibition of neurite growth (). These results suggest that these TDP-43 fragments can co-aggregate with the wild type TDP-43 and reduce the functional TDP-43.
Expression of truncated TDP43 proteins reduces endogenous TDP-43 in the nucleus.
Untagged TDP-43 fragments behave similarly as the tagged fragments
To rule out the possibility that the above observations were caused by the tags, we constructed untagged constructs and transfected them into NSC-34 cells. Like the tagged constructs, constructs 1 (the full-length TDP-43) and 11 did not form aggregates whereas the aggregation-prone constructs (#2, 3 and 6) formed aggregates (). Furthermore, when cotransfected with the full-length TDP-43, the aggregation-prone fragments induced aggregation of the full-length TDP-43 and reduced the diffused TDP-43 staining in the nucleus (). Finally, expression of the untagged aggregation-prone constructs impaired neurite growth (). These results confirm the observations on the tagged TDP-43 fragments.
Untagged TDP-43 fragments form aggregates and coaggregate with the full-length TDP-43.
Untagged TDP-43 impaired neurite growth during neuronal differentiation.
Altering splicing is not a unifying property of the TDP-43 fragments that impaired neurite growth
A well-known function of TDP-43 is modulation of mRNA splicing. A best characterized example is that it promotes exon 9 skipping during the CFTR mRNA splicing 
. To test whether the TDP-43 fragments alter the spicing-modulation activity, we assayed the CFTR splicing after transfecting various TDP-43 fragments into cultured cells. As expected, full length TDP-43 (construct 1) promoted exon 9 skipping while knockdown of TDP-43 by an siRNA inhibited this activity (). Construct 3, 4, 5 and 11 did not noticeably affect the CFTR splicing, whereas construct 2 inhibited exon 9-skipping activity (similar to TDP-43 knockdown by siTDP43) and construct 6 promoted an aberrant splicing (). It is notable that both constructs that interfered with the splicing contained RRM1, suggesting that RRM1 is required for the TDP-43 fragments to affect splicing. These results indicate that some TDP-43 fragments are capable of altering the CFTR splicing activity. However, this alteration is insufficient to account for the neurite growth-impairing activity of the TDP-43 fragments.
Effects of TDP43 fragments on CFTR splicing.