Structural determinants of the cellular localization and shuttling of TDP-43

The TAR DNA-binding protein (TARDBP, hereafter referred to as TDP-43) is a highly conserved heterogeneous nuclear ribonucleoprotein (hnRNP) (Krecic and Swanson, 1999) that controls the transcription, splicing and RNA stability of specific genes (for a review, see Buratti and Baralle, 2008). The protein associates with single-stranded RNA and DNA sequences, and exhibits remarkable specificity for UG/TG dinucleotide repeats (Ayala et al., 2005; Buratti and Baralle, 2001). Previously, we have shown that TDP-43 strongly inhibits exon splicing through the specific recruitment to 3′ splice sites (for a review see Buratti and Baralle, 2008). In addition, our recent findings have indicated that TDP-43 downregulates cyclin-dependent kinase 6 (CDK6) transcript and protein levels in human cells. This interaction is possibly mediated by the 3′ untranslated region (3′ UTR) of the CDK6 mRNA (Ayala et al., 2008). Regulation of the human low-molecular-weight neurofilament (hNFL) by TDP-43 has also been reported to occur through 3′ UTR recruitment (Strong et al., 2007). Evidence that TDP-43 is the major protein in inclusions from patients suffering from frontotemporal lobar degeneration (FTLD) with ubiquitin-positive inclusions and amyotrophic lateral sclerosis (ALS) has prompted intense investigation into the determinants of the subcellular localization of this protein (Neumann et al., 2006; Arai et al., 2006). TDP-43 is predominantly nuclear, but in cases of neurodegenerative TDP-43 proteinopathies it is present as cytoplasmic aggregates, presumably containing post-translational modifications.

TDP-43 comprises two RNA recognition motifs (RRMs) followed by a glycine-rich C-terminal domain. The presence of the RRMs is a distinguishing feature of hnRNP proteins and, in general, these regions are known to mediate RNA recognition as well as protein-protein interactions. The first RRM (RRM-1) is necessary and sufficient to bind specific RNA or DNA sequences (Buratti and Baralle, 2001). The C-terminal tail of TDP-43 does not bind RNA but is necessary to modulate the splicing of CFTR exon 9, probably through the recruitment of an hnRNP complex. In fact, TDP-43 associates with hnRNP A isoforms (i.e. hnRNP A1 and hnRNP A2/B1) through this region specifically. The functional importance of this domain in additional biological properties is highlighted by studies on the regulation of the spermatid-specific SP-10 gene (also known as ACRV1) (Abhyankar et al., 2007). Transcriptional inhibition of SP-10 by TDP-43 depends on the presence of the carboxyl domain and RRM-1 of the protein. Moreover, the presence of C-terminal fragments is a characteristic feature of the affected tissues and/or neurons in ALS and FTLD. The mechanisms that result in TDP-43 degradation are still unknown, although it has been recently demonstrated that, in FTLD cases associated with mutations in the progranulin gene, these fragments may be generated by caspase activation (Zhang et al., 2007). Finally, the potential existence of a close link between this region and disease is underlined by increasing reports of mutations within this domain that are associated with familiar and sporadic cases of ALS and FTLD (Kabashi et al., 2008; Sreedharan et al., 2008; Van Deerlin et al., 2008; Yokoseki et al., 2008).

Elucidating the mechanism(s) that control cellular TDP-43 localization is of paramount importance to describe TDP-43 function and its role in pathogenesis. The sequences and regions of hnRNPs that control their cellular localization differ greatly among members of the family. For instance, distribution of hnRNP A1 is determined by the M9 sequence near the C-terminus (Michael et al., 1995; Siomi and Dreyfuss, 1995); a nuclear retention signal within the auxiliary domain of hnRNP C1 restricts the protein to the nucleus (Nakielny and Dreyfuss, 1996); whereas hnRNP K contains a classic bipartite nuclear localization signal (NLS) and a separate unique sequence that promotes bidirectional nuclear transport (Michael et al., 1997). A bipartite classic NLS has recently been identified at the N-terminus of TDP-43 composed of K82RK84 (NLS1) and K95VKR98 (NLS2) (Winton et al., 2008). Replacement of the basic residues in either or both NLSs with alanine were observed to promote cytoplasmic TDP-43 localization. Contemporarily, human wild-type (wt) and mutant forms of TDP-43 were expressed in a yeast model system to replicate the protein aggregation and abnormal cellular localization observed in disease (Johnson et al., 2008).

This study shows that the predominantly nuclear TDP-43 constantly travels between the nucleus and the cytoplasm. At the same time, we investigate the effect that each functionally defined region of TDP-43 has upon subcellular distribution, by combining biochemical cellular fractionation and immunofluorescence methods. We find that disruption of the nucleic-acid-binding ability of TDP-43 or removal of its C-terminal tail interfere with the correct nuclear and cellular distribution of the protein.

The distribution of TDP-43 was investigated in HeLa and U2OS cell lines using two independent methods: confocal microscopy and biochemical fractionation of the cytoplasmic and nuclear fractions of transfected cells. Both methods showed that endogenous TDP-43 is predominantly nuclear, although lower levels of the protein are also present in the cytoplasm (Fig. 1B-C). These results are in keeping with previous observations regarding the distribution of endogenous TDP-43 in QBI-293 cells and primary hippocampal neurons (Winton et al., 2008). Transient transfection of a FLAG-tagged wild-type TDP-43 protein (wt TDP-43) did not significantly change cellular distribution. The overexpression of wt TDP-43 is probably responsible for the slight increase of TDP-43 in the cytoplasmic fraction (Fig. 1B). To rule out eventual nuclear and/or cytoplasmic contamination during cellular fractionation, we included cytoplasm-specific (tubulin) and nucleus-specific controls in the immunoblots of all samples (Fig. 1B, bottom). The yeast HPR1 homologue protein p84 (also known as THOC1) was used as nuclear marker (Durfee et al., 1994). We also assayed two mutants carrying mutations in the newly found nuclear localization sequences (NLSs) between amino acid residues 82 and 98 of TDP-43 (Fig. 1A) (Winton et al., 2008). As expected, disruption of either NLS1 or NLS2 of TDP-43 resulted in the accumulation of the protein in the cytoplasm, as seen by immunofluorescence (Fig. 1C) and biochemical fractionation (Fig. 1B). The constructs helped us to validate this system as a suitable model to investigate TDP-43 distribution between nuclear and cytoplasmic compartments. Nevertheless, our immunofluorescence experiments consistently showed the presence of the mutants in the nucleus, although at substantially lower amounts when compared with wt TDP-43. Disruption of both nuclear localization signals, NLS1 and NLS2, caused a substantial reduction in mutant TDP-43 in the nucleus, suggesting that the nuclear targeting signals are additive (supplementary material Fig. S1). However, the nuclei of cells expressing the double NLS1 mutant were not completely devoid of FLAG-tagged protein; we still observed the nuclear foci typical of wt TDP-43. These results indicate that additional factors are involved in the nuclear import of TDP-43. Alternatively, expression of tagged TDP-43 (wt or mutant) in the presence of endogenous TDP-43 might result in protein interactions, such as oligomerization, that affect the targeting of either form. TDP-43 oligomerization was already observed in vitro (Ayala et al., 2005). Recent data suggest that the process may also occur in cells as sequestration of endogenous TDP-43 or exogenous wt TDP-43 by NLS mutants in the cytoplasm was seen 72 hours post transfection (Winton et al., 2008).

The biochemical and immunofluorescence methods were used in parallel to obtain a comprehensive indication of TDP-43 distribution both at the single-cell and overall cell population levels. In some cases, such as with the NLS mutants shown in Fig. 1, the results obtained using both techniques were entirely consistent. However, as described below, accurate interpretation of experimental data obtained with other mutants benefited from these complementary techniques.

Recent findings that TDP-43 interacts with the 3′UTR of specific genes and the fact that it forms cytoplasmic inclusions in pathologies such as FTLD and ALS, indicated a function of TDP-43 through its localization in the cytoplasm (Ayala et al., 2008; Strong et al., 2007). Related proteins of the hnRNP family, such as hnRNP A1 and HuR, whose functions require nuclear as well as cytoplasmic localization have been shown to travel between the two compartments (Fan and Steitz, 1998; Piñol-Roma and Dreyfuss, 1992). The presence of TDP-43 in the cytoplasm under normal conditions, or its movement in and out of the nucleus has not been shown previously. We performed interspecies heterokaryon assays, commonly used to study hnRNP transport, to investigate TDP-43 nuclear-cytoplasmic shuttling. Under conditions of protein synthesis inhibition, transiently transfected human cells were fused with NIH-3T3 cells whose nuclei are readily distinguished by intense foci of DNA staining. As controls, we used two proteins whose shuttling – or lack of shuttling activity – is well established. The first hnRNP that was shown to shuttle, by using similar heterokaryon assays, was hnRNP A1, whereas hnRNP C1/C2 was not seen to translocate under the same experimental conditions (Piñol-Roma and Dreyfuss, 1992) (Fig. 2C,D). We observed that FLAG-tagged TDP-43 was present in mouse nuclei of fused cells, indicating that TDP-43 continuously travels in and out of the nucleus (Fig. 2, arrows indicating mouse nuclei). Like TDP-43, hnRNP A1 – but not hnRNP C1/C2 – was found in the nuclei of NIH-3T3 cells, confirming the validity of our experimental procedure (Fig. 2C,D). As further control, FLAG-tagged hnRNP C1/C2 was co-transfected with GFP-tagged TDP-43 to visualize specific nuclear translocation of both proteins in the same cells. Fig. 2D shows that GFP-TDP-43 is present in the mouse nuclei whereas hnRNP C1/C2 is completely absent. TDP-43 shuttling was seen with GFP- or FLAG-tagged protein in HeLa and U2OS cell lines. These results confirmed that TDP-43 travels rapidly and continuously in and out of the nucleus.

Not all shuttling hnRNPs have been observed to accumulate in the nuclei in a transcription-dependent manner (Sarkar et al., 2003). To test whether ongoing transcription is required for TDP-43 import to the nucleus, we treated cells with actinomycin D at concentrations that mainly inhibit RNA polymerase II activity. Actinomycin D treatment of HeLa or U2OS cells led to a substantial increase in cytoplasmic endogenous TDP-43 as compared with control cells (supplementary material Fig. S2). The same was true in the case of FLAG-tagged TDP-43 (supplementary material Fig. S2B). Again, we used FLAG-tagged-hnRNP A1 and hnRNP C1/C2 as controls. As previously reported (Piñol-Roma and Dreyfuss, 1992), hnRNP A1 accumulated in the cytoplasm upon actinomycin D treatment, whereas distribution of hnRNP C1/C2 was not affected (supplementary material Fig. S2C,D).

We next tested the role of the functional regions of TDP-43 in cellular localization. The two most structurally defined domains of TDP-43 are RRM-1 and RRM-2. Previous studies highlighted the fundamental role of RRM-1 regarding the tight and highly specific binding of TDP-43 to UG repeats (Ayala et al., 2005; Buratti and Baralle, 2001). Two mutants were constructed to disrupt TDP-43 association with its target sequence. The first mutant carries two single-amino-acid substitutions F147L and F149L (F147L/F149L) that have previously been shown to abolish binding to UG-repeat sequences (Buratti and Baralle, 2001); the second mutant ΔRRM-1 lacks the entire RRM-1 domain (including amino acid residues 106 to 175). Neither of the mutants resulted in their selective export to the cytoplasm, suggesting that diminished RNA/DNA binding does not affect nuclear targeting. Accordingly, biochemical fractionation did not reveal any significant differences between wt TDP-43 and ΔRRM-1 distribution (Fig. 3B). The fractionation experiments, however, reflected a moderate increase of F147L/F149L in the cytoplasmic fraction when compared to wt. The same was not observed by immunofluorescence suggesting that factors inherent to this particular mutant may affect its distribution in the biochemical fractions. During the immunofluorescence analyses of these mutants we immediately noticed that deletion of the entire RRM-1 region caused the formation of nuclear bodies that were substantially larger and more numerous when compared to endogenous or FLAG-tagged wt TDP-43 (compare Fig. 3C and Fig. 3D). At the same time, ∼20% of the F147L/F149L-transfected cells showed nuclear bodies similar to those seen with ΔRRM-1 (Fig. 3C-D).

A recent report has shown that TDP-43, acting as a transcriptional insulator at the promoter of the spermatid-specific SP-10 gene, is predominantly present in the nuclear matrix fraction (Abhyankar et al., 2007). We wanted to test whether removal of TDP-43 disrupted chromatin organization. HeLa cells were depleted of TDP-43 by RNA interference (supplementary material Fig. S3B) as previously described (Arrisi-Mercado et al., 2004; Ayala et al., 2006). Both mock-depleted and TDP-43-depleted HeLa cells were treated with micrococcal nuclease and subsequent extractions were performed according to standard protocols (Martic et al., 2005). We obtained three sequential fractions: S1 (chromatin soluble in divalent ions, depleted of H1), S2 (EDTA soluble, more compact chromatin), and P (chromatin enriched in actively transcribed genes and nuclear matrix components) (Fig. 4). As shown by the DNA analysis reported in supplementary material Fig. S3, the nucleosomal profiles of the S1 and S2 fractions were similar in both mock-treated and TDP-43-depleted cells, suggesting that removal of TDP-43 does not have a drastic effect on nucleosomal ladder formation.

We then looked for changes in the distribution of ΔRRM-1 and F147L/F149L upon chromatin fractionation with respect to wt TDP-43 because these mutants caused alterations in the nuclear localization pattern. In agreement with previous studies (Abhyankar et al., 2007) we found that endogenous TDP-43 accumulated in the nucleoplasm and nuclear matrix portions (Fig. 4, upper panel). Western blot analysis showed that, similar to the endogenous protein, FLAG-tagged TDP-43 expressed in HeLa cells was evenly distributed between the S1 and P fractions, and almost absent in the S2 fraction. As control we looked at the distribution of histone H1 which, as expected, was enriched in the S2 and P fractions (Fig. 4, bottom panel) (Martic et al., 2005). Strikingly, the distribution of ΔRRM-1 and F147L/F149L was strictly confined to the P fraction (Fig. 4, center panels). These results indicate that the distribution of TDP-43 between the soluble nucleoplasm and the structure-bound portion of the nucleus changes upon disruption of RNA and/or DNA binding. The incremental accumulation of ΔRRM-1 and F147L/F149L in the chromatin and/or nuclear matrix fraction may well have the same biological effect as a loss-of-function mutation. At the same time, these observations suggest that the cellular mobility of TDP-43 largely depends on RNA and/or DNA association or on protein interactions mediated by RRM-1.

The C-terminal tail of TDP-43 promotes interactions with hnRNPs – as seen in vitro (Buratti et al., 2005). Moreover, this domain is required for the regulation of alternative splicing, as seen both in vitro and in transfected human cells (Ayala et al., 2005; Buratti et al., 2005) (and our unpublished results). Mutants that gradually shorten the C-terminal tail of TDP-43 (amino acid sequences 1-366, 1-315, and ΔC truncated at residue 261) were engineered to investigate the importance of the C-terminus in cellular targeting. We were also curious to test whether the complex formed by TDP-43 and the other hnRNPs has a role in cellular targeting (Fig. 5A). Biochemical fractionation of cells expressing the different mutants showed that the ΔC and 1-315 truncations caused a progressive shift toward cytoplasmic localization (Fig. 5B). The immunofluorescence experiments performed in U2OS cells, however, revealed a picture in which the 1-315 and ΔC localization patterns were variable, ranging from nuclear wt-like distribution to cytoplasmic localization (Fig. 6). Moreover, we observed that a fraction of ΔC-transfected cells contained large nuclear inclusions that corresponded to hollow spaces in DAPI staining (Fig. 6B, central panels), suggesting dense and insoluble bodies. The number of cells showing inclusion bodies increased proportionally with the amount of protein expressed. Transfection of similar amounts of wt protein, however, did not result in similar changes (data not shown). In the case of HeLa cells, higher amounts of the truncated ΔC protein were required to obtain cytoplasmic localization and inclusion bodies, as observed by immunofluorescence microscopy.

Once the subcellular distribution of the mutant TDP-43 forms was investigated, we tested their ability to shuttle between nuclear and cytoplasmic compartments. As expected, disruption of the bipartite nuclear localization signal inhibited the ability of NLS1 and NLS2 to be imported into the mouse nuclei. Most of the detectable TDP-43 mutants remained in the cytoplasm of the heterologous nuclei, in agreement with the requirement of this amino-terminal sequence to allow nuclear entry (supplementary material Fig. S4A,B). Mutants ΔRRM-1 and F147L/F149 showed shuttling activity, but at reduced levels when compared with wt TDP-43 (data not shown). In the case of the mutant lacking the C-terminus of TDP-43, we observed shuttling in cells where the human donor nuclei presented a wt pattern (supplementary material Fig. S4C). No translocation of ΔC into the mouse nuclei was observed in the presence of cytoplasmic localization or inclusion bodies (supplementary material Fig. S4D). These last observations are in agreement with a decrease in TDP-43 mobility upon removal of the C-terminal region.

Until recently, TDP-43 has been mainly studied in connection to the abnormal splicing of the cystic fibrosis transmembrane conductance regulator (CFTR) exon 9, an event that is closely associated with the occurrence of cystic fibrosis (Delaney et al., 1993; Strong et al., 1993). Subsequently, aberrantly localized and post-translationally modified forms of TDP-43 were found to represent the main accumulating protein in neuronal cytoplasmic inclusions in patients affected by two neurodegenerative diseases, frontotemporal dementias (FTDs) and ALS (Neumann et al., 2006; Arai et al., 2006). The pathological role of TDP-43 in these neurodegenerative diseases is still not understood. As with other neurodegenerative pathologies characterized by intracellular protein aggregation, the cause and effect relationship between TDP-43 inclusions and disease is unknown. In particular, it remains to be established whether TDP-43 inclusions are inherently toxic, or whether the sequesteration of TDP-43 during aggregation is the pathogenic event. Formation of TDP-43 inclusions in the affected cells is accompanied by depletion of the protein from the nucleus, suggesting that the various regulatory activities carried out by TDP-43, namely splicing, transcription and the more recently described role in cell-cycle control, are disrupted by aberrant TDP-43 aggregation (Ayala et al., 2008; Buratti and Baralle, 2008). Further details on the determinants for correct localization and the connection between cellular targeting and protein activity are required to better understand TDP-43 function and its role in neurodegenerative diseases.

Our present work shows that, although predominantly nuclear, TDP-43 continuously travels between the nucleus and cytoplasm. The nuclear-cytoplasmic shuttling activity of other proteins, such as hnRNP A1 and SFRS1 (also known as SF2 or ASF), is consistent with their various functions in both cellular compartments. Likewise, TDP-43 would require similar mobility to carry out functions including interactions with introns, 3′ UTR, and the processing of microRNA (Ayala et al., 2008) (and our unpublished work). In the case of almost all shuttling hnRNPs and SR proteins, transcription inhibition has been used to decrease nuclear import as an alternative means of visualizing protein permanence in the cytoplasm. According to our results, TDP-43 also accumulates in the cytoplasm upon inhibition of RNA polymerase II following treatment with actinomycin D. As in the case of the other shuttling hnRNPs, whose cellular distribution is affected by actinomycin D, the mechanism by which transcription inhibition affects the nuclear recruitment of TDP-43 is unknown (see Piñol-Roma and Dreyfuss, 1993). We can only speculate that continuous mRNA synthesis is required to signal TDP-43 import via RNA or protein-protein interactions.

Our results on the cellular distribution of the carboxy-domain mutants reveal that disruptions in this region affect normal targeting of TDP-43 to the nucleus and cytoplasm. The observed heterogeneity of cellular targeting of the ΔC-truncated mutant, however, suggests that lack of the C-terminal domain does not affect a single localization signal for nuclear import, export or nuclear retention. This mutant shows cytoplasmic localization despite the presence of the recently described NLS at the N-terminus (Winton et al., 2008). The fact that ΔC TDP-43 forms cytoplasmic and nuclear inclusion bodies strongly points to a role of the C-terminal tail in TDP-43 solubility. Decreased mobility of the mutant might cause aberrant nuclear or cytoplasmic retention. In fact, heterokaryon assays show greatly decreased nuclear import of ΔC TDP-43 in fused cells where the mutant is outside the nucleus and/or forms inclusion bodies (supplementary material Fig. S4C,D). From a functional point of view, the C-terminal region of TDP-43 is necessary to promote splicing and transcriptional inhibition of TDP-43 target genes (Abhyankar et al., 2007; Ayala et al., 2005; Buratti et al., 2005). This region has been shown to associate with hnRNP A/B isoforms in RNA-bound and -unbound forms (Buratti et al., 2005). Collectively, these results suggest that TDP-43 solubility depends on the association with these additional factors through its C-terminal tail region. Regarding the potential effects on the cellular environment, high expression of ΔC isoforms are likely to reduce cell viability – as observed in the case of polyglutamine protein aggregation (Welch and Diamond, 2001). In contrast to our observations, most of the C-terminal-truncated mutants expressed in the yeast model are reported to be present in the nucleus and to cause no aggregation (Johnson et al., 2008). We do not have an explanation for this discrepancy yet, although it is possible that TDP-43 expression in yeast undergoes different transport dynamics than in human cells.

The RRM-1 region of TDP-43 has been well characterized in terms of protein function and nucleic acid interactions. In addition to disrupting RNA binding (Buratti and Baralle, 2001), RRM-1 mutants fail to regulate alternative splicing (our unpublished observations). We showed that although disruption of RNA and/or DNA binding does not change the predominantly nuclear targeting of TDP-43, the deletion of the N-terminal RRM-1 (ΔRRM-1) domain or point mutations therein (F147L/F149L) seem to alter TDP-43 dynamics in the nucleus. In fact, the ΔRRM-1 bodies observed by immunofluorescence are clearly distinct from the foci formed by wt TDP-43; they are larger and more numerous. The chromatin fractionation studies confirm that the nuclear distribution of the mutants is different from wt TDP-43. The fact that expression of both RRM-1-deletion mutants and the more conservative point mutants result in similar nuclear localization and chromatin fractionation patterns argues against protein insolubility of these mutants. As summarized in Fig. 7, endogenous and FLAG-tagged TDP-43 are present in both soluble and chromatin/nuclear matrix bound fractions, whereas ΔRRM-1 and F147L/F149L are almost entirely associated with the nuclear matrix. These results suggest that functions associated with RRM-1, such as (UG)_n RNA binding, confer TDP-43 mobility in the nucleus. The function of RRM-2 remains elusive because it is dispensable for targeted RNA binding. The strong nuclear-matrix–scaffold association displayed by mutants that lack functional RRM-1 may imply a role for RRM-2 in chromatin organization.

In conclusion, here we show that the different nuclear and cytoplasmic functions of TDP-43 can easily perturb its cellular distribution and solubility. Thus, insight into the possible links between TDP-43 function and certain diseases will probably derive from studies that focus on how the disease-related TDP-43 mutations affect TDP-43 function.

FLAG-tagged TDP-43 was generated by PCR amplification from the previously described vector pGEX3X-TDP43 (Buratti et al., 2001) using primers 1 and 2 (see below) to clone the sequence between the HindIII-BamHI sites of pFLAG-CMV-2 (Sigma). TDP-43 was cloned into eGFP-N1 vector (Clontech) between HindIII-BamHI sites. NLS1 and NLS2 were expressed from a Myc-tagged vector pcDNA3.1(+)/Myc-His A (Invitrogen). The TDP-43 sequence was cloned using BamHI-XbaI sites. pF-ΔRRM1 was amplified by PCR from pGEX3X-ΔRRM1 (Buratti and Baralle, 2001) using primers 1 and 3 to clone between HindIII-KpnI. Point mutations F147L/F149L were introduced by PCR using primers 4 and 5, and cloned between HindIII-KpnI sites. FLAG-tagged C-terminal truncation mutants were constructed by PCR amplification and cloned between HindIII-KpnI sites using the following oligonucleotides: pF-ΔC (1 and 6), 1-315 (1 and 7), and 1-366 (1 and 8). FLAG-tagged hnRNP A1 and hnRNP C1/C2 were cloned by PCR amplification using primers 9 and 10, and 11 and 12, respectively, between KpnI-BamHI sites. cDNA from HeLa cells was used as template. Primer sequences are: (1) pFHindwtFW_cccaagctttctgaatatattcgggtaaccg; (2) pFBamwtRV_cgcggatccctacattccccagccagaagacttag; (3) pFKpnwtRV_cggggtaccctacattccccagccagaagacttag; (4) F147L/F149LFW_aaggggttgggcttggttcgtttt; (5) F147L/F149LFW_aaaacgaaccaagcccaacccctt; (6) pFKpnDCRV_cggggtaccctaggcattggatatatgaacgctg; (7) pFKpn315RV_ggggtacctcacgcaccaaagttcatcccaccacc; (8) pFKpn366RV_ggggtacctcaggcctggtttggctccctctg; (9) KpnA1FW_cggggtaccatctaagtcagagtctcctaaag; (10) BamHA1RV_cgcggatccttaaaatcttctgccactgcc; (11) KpnC1FW_cggggtaccagccagcaacgttaccaacaag; (12) BamHC1RV_cgcggatccttaagagtcatcctcgccattg.

Human HeLa and U2OS, and mouse NIH-3T3 cell lines were grown in DMEM-Glutamax-I (GIBCO) supplemented with 10% fetal bovine serum (EuroClone) and antibiotic-antimycotic-stabilized suspension (Sigma). Cells were grown overnight and transfected with Effectene transfection reagent (Qiagen). Cell electroporation of U2OS cells was carried out using 0.5×10⁶ cells in 0.7 ml of PBS at room temperature in a Gene Pulser/MicroPulser, using a 0.4-cm gap cuvette (Bio-Rad) at 250 V, 975 μF with a Gene Pulser II (Bio-Rad). Depletion of TDP-43 by RNA interference for the chromatin-fractionation studies was carried out as previously described (Arrisi-Mercado et al., 2004; Ayala et al., 2006). Transcription-inhibition experiments were performed by treating cells with 5 μg/ml of actinomycin D for 3 hours.

Assays were carried out as previously described (Caceres et al., 1998; Piñol-Roma and Dreyfuss, 1992), fixing cells for 1 hour after cell fusion that was carried out 24 hours post transfection.

Two 100-mm confluent dishes were harvested and washed in PBS by repeated centrifugation. The protocol was carried out at 4°C. The pellet was resuspended in five volumes of buffer N [15 mM Tris-HCl pH 7.5, 60 mM KCl, 15 mM NaCl, 5 mM MgCl₂, 1 mM CaCl_2, 1 mM DTT, 2 mM Na₃VO₄, 1 mM PMSF, 0.25 M sucrose, Complete Protease Inhibitor Cocktail (Roche Applied Science)]. Cell lysis was obtained by adding an equal amount of buffer N plus 0.6% NP-40. Following 5 minutes of incubation nuclei were pelleted and gently resuspended in 1 ml of buffer N. The nuclei were again pelleted and lysed using an equal volume of solution 2 (10 mM HEPES pH 7.9, 10 mM KCl, 1.5 mM MgCl₂, 0.1 mM EGTA, 0.1 mM DTT, 0.5 mM PMSF, 5% glycerol, 0.4 M NaCl). The nuclear fraction was cleared by centrifugation following 30 minutes of incubation. Nuclear and cytoplasmic fractions were quantified (Bradford Protein Assay, Bio-Rad) and visualized by SDS-PAGE and western blotting on a nitrocellulose membrane.

S1, S2 and P chromatin fractions from HeLa-cell nuclei were prepared according to previously described procedures (Martic et al., 2005; Rose and Garrard, 1984), with minor modifications. HeLa cells were grown overnight to confluence in a 100-mm dish. The harvested cells were washed with PBS followed by a wash with hypotonic buffer A (20 mM HEPES pH 7.9, 20 mM NaCl, 5 mM MgCl₂, 1 mM ATP) and incubated on ice for 15 minutes in the same buffer. Lysis of the cytoplasmic membrane was achieved with 20 strokes in a 15-ml cell douncer (Wheaton) using the type B pestle. The nuclear pellet was obtained by centrifugation and was then resuspended in buffer B (20 mM HEPES pH 7.9, 0.15 M NaCl, 0.5 mM MgCl₂, 0.3 mM sucrose, 2 mM CaCl₂, 1 mM ATP, and 0.5% NP-40), incubated for 30 minutes at 4°C and centrifuged to obtain a total chromatin fraction and a soluble nuclear fraction. The chromatin fraction was resuspended in buffer B without NP-40 and incubated for 4 minutes at 16°C with 1 unit of micrococcal nuclease (Sigma). The solution was centrifuged to obtain the supernatant (S1). The pellet was resuspended in 8 mM EDTA, incubated at 4°C for 15 minutes, and centrifugated to obtain the chromatin fraction (P) and supernatant (S2). The proteins contained in each fraction were separated by 12% SDS-PAGE and detected by western blotting.

Indirect immunofluorescence was carried out as previously described (Ayala et al., 2008). Cells were observed on a Zeiss LSM 510 confocal microscope.

Endogenous TDP-43 was detected using rabbit anti-TDP43 polyclonal antibodies (Buratti et al., 2001) and ProteinTech (10782-2). Anti-FLAG M2 monoclonal antibody was from Sigma (F1804), mouse anti-Myc monoclonal antibody (9B11) from Cell Signaling, goat anti-Lamin A/C polyclonal antibody from Santa Cruz Biotechnology (sc-6215) and anti-LAP2β was kindly provided Tom Misteli (NCI NIH, Bethesda, MD). Antibodies used for the fractionation controls included mouse anti-tubulin monoclonal (Sigma; T5168), rabbit anti-histone H1 polyclonal (Santa Cruz Biotechnology; sc-10806), and mouse anti-p84 monoclonal (Abcam; ab487). HRP-conjugated secondary antibodies (Dako) were used for immunoblotting, and FITC- or Texas-Red-conjugated secondary antibodies (Jackson ImmunoResearch) were used for immunofluorescence detection.

This work was supported by Telethon Onlus Foundation (Italy) (GGP06147), FIRB (RBNE01W9PM), and by a European community grant (EURASNET-LSHG-CT-2005-518238). We also wish to thank Monica Caggiano for technical help and Maurizio Romano (ICGEB, Trieste, Italy) for the eGFP-TDP43 vector.