TDP-43 is a member of the hnRNP protein family and its 414 amino acid sequence consists of an N-terminal region, two RRM domains and a C-terminal region. Like many members of this family, the C-terminal region contains Gly-rich sequences that may be essential to recruit cellular factors that can modulate TDP-43 function. Currently, the only protein factors whose association to TDP-43 has been verified are several members of the hnRNP family (20
) and the survival of motor neuron (SMN) protein (37
). The C-terminal tail does not seem to play a role in the interaction with SMN. Other additional factors have also been suggested to interact with this protein on the basis of protein–protein association studies that use high-throughput methodologies (38
). In particular, two proteomics studies (39
) involving yeast two hybrid systems found some other potential TDP-43 binding partners, namely XRN2 and PM/Scl100, involved in mRNA decay, ZHX1, a transcriptional repressor, SETDB1, a chromatin remodeling regulator, and NSFL1C and ARF6, both involved in membrane trafficking. In this respect, it is important to note that TDP-43 has been described to be part of RNA granules responsible for trafficking, sequestering and degrading RNA species (41
) and has been observed to colocalize strongly with Staufen, moderately with TIA-1 and weakly with XRN1, an exoribonuclease involved in mRNA decay (42
). Finally, TDP-43 has also been found associated with both human and mouse microprocessor complexes (43
), suggesting that it may also be involved in the biosynthesis of microRNAs. However, many of these interactions are difficult to interpret in terms of the functional role of the complexes due to the lack of more stringent biochemical evidence. In this work, we have presented a new siRNA/minigene coupled system to assess the in vivo
splicing inhibitory activity of TDP-43 mutants and homolog proteins. Our results have fully confirmed previous in vitro
data regarding the functional interaction of TDP-43 with hnRNP proteins. In particular, we have focused on hnRNP A2, the major hnRNP protein recognized by TDP-43 according to pull-down assays (19
). We have now identified residues 321 to 366 of the TDP-43 C-terminal tail as the minimal binding region required to bind hnRNP A2. In addition, we also find that this region is necessary for TDP-43 splicing inhibitory activity in our in vivo
system. Taken together, these experiments strongly support the requirement for TDP-43 to form an hnRNP complex through its C-terminus to inhibit exon splicing. Interestingly, the size of the 321-366 region (although not its sequence) is strikingly similar to the 39 amino acid-long M9 regions described in the hnRNP A/B proteins that enables the bidirectional transport of these proteins across the nuclear envelope through binding to transportin, also known as Kapβ2 (45
). It has been recently shown that the M9 region of hnRNP A1 protein binds a concave surface of a C-terminal arch in Kapβ2 in an extended conformation (residues 263 to 289 of this protein) making an extensive network of polar and hydrophobic contacts (45
). A similar situation may occur in the TDP-43–hnRNPA2 interaction and could explain the fact that three missense mutations in the 321-366 region associated with neurodegeneration do not substantially affect binding efficiency and splicing function. Altogether, these data suggest that the mutations in TDP-43 found in patients do not contribute to a drastic protein loss of function. More likely, they may contribute to a predisposition to develop the disease (i.e. more readily form aggregates) through some still unidentified mechanisms. Given the late age of onset of both ALS and FTLD, another possibility is that the mutations affect TDP-43 function to a small degree (not detectable in our assays), but the slight disruption of its activity perpetuated over a long time might be the cause for the neurodegeneration. The same concept has been proposed to be the cause of neurodegeneration in Spinal Muscular Atrophy patients with a slight decrease in SMN protein (47
). Further work is currently in progress to better define the TDP-43–hnRNP A2 interaction through the identification of the hnRNP A2 residues involved. The TDP-43 interacting domain should be localized in the C-terminal tail of hnRNP A2 as previously reported (20
). These experiments will provide a better indication of the TDP-43 residues involved in the interaction and thus in a better position to judge the effect of eventual disease causing mutations.
From a functional point of view, characterization of the TDP-43–hnRNP interaction is essential to understand its splicing regulatory properties, especially in light of recent observations regarding the potential existence of human alternatively spliced variants that lack the C-terminal tail (49
). Although at the moment there is scant biochemical evidence regarding the relative abundance or distribution pattern of these isoforms in humans, it is clear that their existence and or production would carry distinct biological properties with respect to the WT protein but would still be able to compete for the same binding sites (UG repeats). Therefore, in addition to variations in the relative hnRNP A/B proteins present in different tissues or developmental stages, expression of these truncated isoforms may also prove to be a good way to modulate TDP-43 function(s) without necessarily altering TARDBP
basal expression levels.
More in general, our results represent a clear indication that interactions between splicing regulatory proteins belonging to distinct classes can be a powerful modifier of their functional properties. In fact, it was well known that functional biochemical interactions between hnRNPs of the same type can explain their effect on splicing. For example, it has been proposed that a variety of proteins, such as PTB, hnRNP H and hnRNP A/B can potentially multimerize to create ‘zones of silencing’ across exons or modulate the conformation of the pre-mRNA and thereby influence exon recognition (51–54
). Less often, protein networking has been described to occur between different factors. One such interaction has been described for the PTB–Raver1 interaction in the control of the tropomyosin gene (55
), a case where both proteins can synergistically repress exon recognition. More recently, the presence of a functionally relevant biochemical interaction between the hnRNP H/F proteins and Fox2 has been reported (57
). In this case, the interaction between hnRNP H and Fox2 is capable of altering the binding ability of the complex and thereby to influence the splicing inhibitory effect of the H/F proteins on FGFR2 exon IIIc.
Finally, all these observations of intricate biochemical connections between splicing factors of different classes will certainly add a layer of complexity to the well-established concept of combinatorial and context-dependent control in splicing (58–60
). In order to understand splicing outcomes in the future, it will not be enough to simply identify all the trans-acting factors that bind to the RNA sequence in the vicinity of an exon. More probably, it will also be necessary to consider the relative expression levels and regulation of cellular factors that do not directly contact the RNA under study, but which can modify the functional properties of the factors that do.