It is common when studying proteins associated with diseases to consider them mainly under the aspect of how they are linked to pathology. This is particularly the case for polyQ proteins: much has been said about their relationship to aggregation and cellular toxicity but still little is known about their cellular functions when they are in their non-pathologic state. In this study, we have followed a different approach. We focused on the functional role of non-pathologic Atx1 and set out to identify sequence motifs which could provide direct information about the mechanisms involved in recognition.
AAlthough a role in transcriptional regulation, often mediated by interaction of the AXH domain with several transcriptional co-regulators 
has also been suggested, independent evidence indicates that Atx1 is an RNA binding protein with a role in processing and/or exporting specific mRNAs to the cytoplasm: Atx1 recognizes RNA homo-polymers 
; it is linked directly through its interactome to other RNA binding proteins, some of which are splicing factors 
; Atx1 was shown to recruit the mRNA export factor TAP/NXF1 
We have shown in this study that Atx1 interacts with U2AF65, thus adding a new member to the Atx1 interactome. Atx1 does not co-localize with nuclear speckles, the subnuclear structures in which splicing factors accumulate when not actively forming spliceosomes (reviewed in 
). A splicing-independent association in extraspliceosomal complexes has recently been described for the splicing factors SF1 and U2AF 
Our splicing assays suggest an enhancing effect of Atx1 over-expression on U2AF65-mediated splicing. While the exact implications for the function of Atx1 in the context of splicing remain to be investigated in detail, it is worth noting that only non-expanded Atx1 seems to have a significant effect on U2AF65-mediated splicing suggesting that polyQ expansion could interfere with molecular recognition. Addionally, recruitment or trapping of U2AF65 in the single large nuclear aggregates of SCA1 affected neurons could have a detrimental effect on splicing.
Even more relevant for understanding the functions of native Atx1 is the identification of the motif that mediates the interaction: the presence of a ULM in the C-terminus of Atx1 strongly supports an involvement of the protein in pre-mRNA splicing and a direct involvement in the complex and highly dynamic macromolecular machinery represented by the spliceosome 
. This is in line with the reported presence of non-expanded Atx1 in a large protein complex associated in an RNA dependent fashion with the regulatory splicing factor SPF45 (this complex is most likely the early stage complex known as pre-spliceosome although no attempt to identify it was made) 
. Since many of the splicing factors contain ULM, UHM or both motifs, our results point directly to a participation of Atx1 to the network of multiple and competing interactions linking the splicing factors.
Further complexity to the overall picture is added by the presence of overlapping functional sites in Atx1. Two motifs overlap with ULM, the NLS and the 14-3-3 phospho-S776 ligand motif, known to be essential requirements for the functions of the protein and for development of disease. This prompted us to check the effect of phosphorylation on interaction and to compare our results with a previous study which describes binding of Atx1 with SPF45 by two-hybrid screen and co-immunoprecipitation from mammalian cells 
. Because of the techniques used, the authors made the a priori
reasonable assumption that the effect of phosphorylation could be mimicked or abolished by mutation of S776 to an aspartate and to an alanine respectively. They demonstrated that both wild-type and S776D Atx1 interact with SPF45 and that the interaction is strengthened for S776D but abolished when using a S776A mutant. These data are in some agreement with our in vitro
binding assays but as compared to the non-phosphorylated Atx1 ULM, the phosphorylated peptide binds with lower affinity. This indicates that aspartate does not have the correct properties to mimic a phosphate group, likely because it is smaller, unbranched and with a lower charge density, as further confirmed by observing that the S776D mutant peptide is not recognised by 14-3-3ζ.
S776 is therefore the molecular switch between alternative interactions: when non-phosphorylated, non-expanded Atx1 will be involved through its ULM in a vast network of interactions with splicing factors and contribute to the spliceosome. Phosphorylated Atx1 will bind instead 14-3-3. Phosphorylated Atx1 also recognises other cellular partners as shown by the evidence that the interactions with the transcriptional repressor Capicua (CIC) which targets the AXH domain of Atx1 
and other proteins involved in transcriptional regulation, such as the MEF2-HDAC4 complex 
, are weakened in the S776A mutant.
How do these findings tell us something about disease and how do interactions and signal overlap affect SCA1? It has been suggested that this pathology could be the result of the balance between a gain-of-function due to enhancement of interaction with SPF45 and a partial loss-of-function due to weakening of interaction with CIC. A different, non-mutually exclusive possibility is that phosphorylation causes Atx1 to interact with a substantially different network of nuclear proteins than non-phosphorylated Atx1 
. Our findings that Atx1 binding to UHMs is modulated in a posphorylation-dependent manner due to the intervention of 14-3-3 open a scenario that would merge the two hypotheses.
We have demonstrated that the S776A mutant peptide is able to recognize U2AF65 with affinity comparable to that of the native ligand sequence. This indicates that non-phosphorylated expanded Atx1, which does not manifest a SCA1 phenotype as we know from the S776A mutant 
, still recognizes UHM-containing nuclear factors despite being unable to bind 14-3-3. We suggest () that participation of Atx1 in these interactions plays a protective role against self-association of the protein induced by polyQ expansion: the spliceosome is a large and dynamic machinery composed of five major small nuclear ribonucleoprotein particles including U2 of which U2AF is an auxiliary factor, and more than three hundred non-snRNP protein splicing factors 
. Engagement in such an extended protein complex could easily ‘dilute out’ expanded Atx1 and prevent or reduce its self-association to an extent that allows efficient clearance by the machinery of the proteasome pathway. Conversely, when Atx1 is S776-phosphorylated and recruited by 14-3-3, the interaction with the spliceosome complex is hampered. As a consequence, expanded Atx1 becomes available both to self association through regions such as the polyQ stretches and/or the AXH domain 
, and to recognition of other partners. In support of this hypothesis is also the presence of the 14-3-3ζ isoform in protein aggregate deposits observed in different neurological diseases, such as Parkinson and Alzheimer diseases, Huntington and SCA1 itself 
. The effect is particularly evident in SCA1 where the 14-3-3ζ isoform changes localization from the cytoplasm to the nucleus of affected Purkinje cells 
Schematic model of the cellular interactions formed by Atx1 and the role that phosphorylation plays in pathology.
An important point that becomes clear from our studies is the importance of investigating the non pathologic molecular interactions as a strategy to identify mechanisms that prevent aggregation. Also, it is clear that it is far too simplistic to approach SCA1 or other polyQ diseases in terms of a gain or loss-of-function. These concepts are particularly inadequate when function is described in terms of enhanced or reduced interactions and strongly depend on which of the several interactions we may refer to. As remarked in a recent review 
and as indicated by our results, polyQ proteins take part in a complex and vast network of interactions, to which polyQ aggregation adds another competing pathway.
For complex regulatory systems, the difficult balance of multiple equilibria can only be appropriately described by gaining an overall picture which places each interaction into the bigger frame of all what is known about the individual components 
. It will be interesting to investigate further the role of different regions of Atx1 in modulating interactions since SPF45 and CIC, although binding to different sites, i.e. ULM and AXH, compete with each other