Mapping the self-association domains of ataxin-1: identification of novel non overlapping motifs

Plasmids, cell culture, transfections and imaging

Non-expanded (Q30) ataxin-1 and truncated ataxin-1 fusion proteins were constructed in pEYFP or pECFP vectors using the standard PCR and mutagenesis methods previously established in our laboratory (Menon et al., 2012). COS cells were grown in chamber slides in Dulbecco’s modified Eagle medium supplemented with 10% foetal bovine serum and 100 U/ml penicillin-streptomycin (Invitrogen Life Technologies). Cells were transfected with appropriate plasmid DNA using GeneCellin tranfection reagent (BioCellChallenge). Cells were fixed using 4% paraformaldehyde 54 h post-transfection and slides were mounted using CitiFluor (Agar Scientific). Cells were observed and recorded using a laser scanning confocal microscope (de Chiara et al., 2009).

Analytical size exclusion chromatography

Size exclusion chromatography was performed using a prepacked Superdex-75™ 10/300 GL column (Pharmacia) equilibrated with a 20 mM Tris–HCl pH 7, 150 mM NaCl, 1 mM TCEP buffer solution. Aliquots of 200 µl of 150 µM AXH and TLND2AXH incubated for 24 h at 37 °C in 20 mM pH 7, 150 mM NaCl, 1 mM TCEP were injected separately and eluted using a 0.8 ml/min flow rate. Albumin (67 kDa), Ovalbumin (43 kDa), Carbonic anhydrase (29 kDa), and Ribonuclease A (13.7 kDa) were used as standards for the molecular mass, whereas the Blue Dextran 2000 was used for the determination of the void volume of the column.

FRET microscopy

Samples for FRET were imaged on a Zeiss LSM 510 confocal microscope using a 63× 1.4NA Plan NeoFluar oil immersion objective and FRET analysis was carried out as previously described in detail (Menon et al., 2012). Pre- and post-bleach CFP and YFP images were imported into Mathematica 7.0 for processing as described (Matthews et al., 2008). Briefly, images were smoothed using a 3 × 3 box mean filter, background subtracted, and post-bleach images fade compensated. ${\displaystyle E=\frac{CF{P}_{\mathrm{postbleach}}-CF{P}_{\mathrm{prebleach}}}{CF{P}_{\mathrm{postbleach}}}}$ FRET efficiencies were then extracted from pixels falling inside the bleach region and plotted against the bleach efficiency on a pixel-by-pixel basis. FRET efficiency showed a linear correlation with bleach efficiency enabling determination of FRET efficiency at 100% bleach efficiency by extrapolation. Data from images were used only if YFP bleaching efficiency was greater than 50%. Finally, the FRET efficiency was converted in to the inter-fluorophore radius using: ${\displaystyle r=Ro\sqrt[6]{\frac{1}{E}-1}}$ where Ro is the Förster radius for CFP and YFP, which is 4.95 nm.

Ataxin-1 foci formation is independent of polyQ and mediated by the C-terminus

We first explored the relationship between foci formation and polyQ in non-expanded ataxin-1. We used non-expanded ataxin-1 to be able to detect the intrinsic properties of the functional protein that could then be transferred to the expanded form. We created several deletion constructs for expression in mammalian cells in which the protein was N-terminally attached to the yellow fluorescent protein (YFP) (Fig. 1). The first of these mutants, hereafter termed NT, contains the N-terminus of ataxin-1 up to the start of the polyQ tract. Its behaviour was compared with that of a mutant (termed Atx1ΔNT) that excludes the region preceding the polyQ (30Qs) tract. The results from the whole analysis are summarized in Table 1.

As expected, wild-type ataxin-1 fused to YFP readily formed nuclear foci (Figs. 2A and 2B). The YFP fusions of ataxin-1 constructs were similar in expression pattern to equivalent constructs that lacked YFP fusion and were stained with antibodies (data not shown), suggesting that the presence of YFP does not significantly influence the data. The NT construct showed a diffused pattern of expression, demonstrating that this region is not involved in foci formation (Fig. 2C). Addition of the polyQ tract to the N-terminal region (construct named NTQ) did not alter the behaviour of the cells (Fig. 2D). We observed that constructs Atx1ΔNTQ and Atx1ΔNT were both able to form foci (Figs. 2E and 2F).

The observation that Atx1ΔNTQ, which lacks the polyQ tract, is able to form foci demonstrates that polyQ region is not per se a major foci forming factor. These results thus indicate that the C-terminus of ataxin-1 alone is involved in foci formation which does not involve the polyQ tract.

The AXH domain is insufficient to form foci but needs N-terminal extension

In order to further map the region required for nuclear foci formation, we expressed deletion mutants from the C-terminal region of ataxin-1 fused with YFP. We first expressed the region starting at the AXH domain and ending at the last residue of ataxin-1 (AXH2End, residues 562–816), which also contains the endogenous Nuclear Localisation Signal (NLS). As expected, this protein was mostly nuclear, but expressed itself in a diffused pattern (Figs. 3A and 3B). A diffused pattern of nucleo-cytoplasmic expression was also observed for a construct with the AXH domain alone starting at amino acids ASPAA and comprising residues 562–689, termed AXH (Figs. 1, 3C and 3D). Foci formation was observed instead when the AXH2End construct was N-terminally extended (residues 410–816, termed TLND2End, where TLND represents the N-terminal amino acid sequence of the construct) (Figs. 3E and 3F).

These results thus suggest that the aggregation-prone AXH domain is insufficient for foci formation by its own.

Evidence for the self-association prone CT2AXH motif

Since TLND2End formed foci while AXH2End did not, we next expressed the TLND2AXH region (residues 410–561) as YFP fusion to analyse if this is independently capable of foci formation. We observed that this construct formed foci that were, however, smaller in size as well as in number (Figs. 4A–4C). This construct appeared to express mainly in a diffused form, but compared to other diffusedly expressing constructs such as the AXH, we could clearly observe small foci. Therefore this construct appeared to have a behaviour intermediate between those forming foci and those that do not. N-terminal extension of TLND2AXH (i.e., residues 227–561, construct CT2AXH) showed that the CT2AXH construct enhances foci formation ability (Figs. 4D and 4E). The foci in both instances were mostly extranuclear, which is not surprising as this region is not known to possess a functional NLS. C-terminal addition of SV40 NLS to CT2AXH resulted in exclusively nuclear foci formation (Figs. 4F and 4G). Lastly, deletion of the CT2AXH region from full-length YFP ataxin-1 (construct Atx1ΔCT2AXH) abolished foci formation (Figs. 4H and 4I).

These results show that foci formation takes place independently from the AXH domain and that the nuclear localization signal has an influence on foci localisation.

Identification of a self-association motif outside the polyQ region and the AXH domain

The isolated AXH domain is known to dimerize in vitro (de Chiara et al., 2003; Chen et al., 2004; de Chiara et al., 2005; de Chiara et al., 2013a). Dimerization rather than other forms of self-association could then be the seeding event for foci formation. Since nothing is known about the region N-terminally upstream to the AXH domain, we further characterized the TLND2AXH motif (residues 410–561) by analytical size exclusion chromatography (SEC) techniques to better understand the relationship between foci formation and the properties of the individual domains (Fig. 5). Our original intention was to do the experiment with the CT2AXH construct. However, we ran into problems of expression as the protein went into inclusion bodies. For this reason we switched to TLND2AXH, which expressed well and in soluble form.

Analytical gel-filtration chromatograms indicate that, in analogy with the AXH domain, the construct TLND2AXH is in equilibrium between two species which have an elution volume of 11.0 and 12.0 ml, corresponding to roughly 36 kDa and 30 kDa respectively if it were a globular protein (from the standards). Lower retention volumes are observed for TLND2AXH as compared to AXH despite their similar molecular weight. This would be justified by the elongated shape of the putatively unfolded TLND2AXH construct (de Chiara & Pastore, 2011) as compared to the compact fold of the AXH domain (Chen et al., 2004).

These results provide solid evidence of the existence of a newly identified self-association region within ataxin-1 that is independent from the AXH domain.

The two distinct dimerization motifs are independently capable of self-association in cells

To complement these in vitro studies of dimerization, we next tested the self-association regions identified in cultured cells and explored if they are capable of direct interaction leading to self-association also in cell. To verify that the full-length ataxin-1 protein is indeed capable of self-association, we expressed full-length non-expanded ataxin-1 as CFP and YFP fusions and carried out FRET analysis. Self-association of ataxin-1 was evident from positive FRET signal (Figs. 6A–6F). We observed comparable FRET signals both when full-length ataxin-1 was expressed as C-terminal fusion to CFP (CFP-C-Atx1, Fig. 6A) and YFP (YFP-C-Atx1, Fig. 6B) and when ataxin-1 was C-terminal to CFP (CFP-C-Atx1, Fig. 6D) and N-terminal to YFP (YFP-N-Atx1, Fig. 6E). This suggests that the position of the fluorophore in ataxin-1 does not influence the FRET signal significantly.

We then expressed the newly identified dimerization motif TLND2AXH as CFP and YFP fusions. Self-association was evident in FRET experiments using the TLND2AXH construct (Figs. 6P–6R). As expected, the N-terminal extended form of TLND2AXH (CT2AXH) was also found to interact directly in cells (Figs. 6J–6L). Similarly, we carried out FRET analysis with CFP and YFP fusions of the AXH domain. The results confirmed direct interaction between the YFP and CFP fusion proteins (Figs. 6M–6O).

Testing the effects of destabilization of AXH dimerization in cell

Finally, we used the FRET based approach to test the hypothesis that stabilization of the AXH domain into its monomeric form could result in reduction of protein aggregation. We used FRET analysis on co-expressed ataxin-1 YFP vs ataxin-1 CFP where ataxin-1 YFP plasmid also expressed a Capicua (CIC) peptide (residues 34–48) of the protein that we have recently been shown to interact with the AXH domain with high affinity and to force it in a monomeric form (de Chiara et al., 2013b). Unfortunately, we were unable to verify the effect of the peptide since the antibodies raised against it did not recognize the peptide (data not shown) preventing confirmation of peptide expression.

We resorted to a different strategy as a proof of principle. We have recently reported that mutation of a glycine at position 567 keeps the AXH domain in a predominantly monomeric form (de Chiara et al., 2013a). We thus tested if self-association in cells is affected by this mutation. We reasoned that a direct comparison is possible as this is only a point mutation which is unlikely to influence FRET data. We carried out a FRET analysis using the CFP ataxin-1 567 mutant and YFP-wild-type ataxin-1. Interestingly, we observed a significant reduction in the FRET signal (CFP 567 Atx1 versus YFP wt Atx1, Figs. 6G–6I and 7). This result suggests that identifying and manipulating vulnerable regions of ataxin-1 could be a tool to reduce self-association and possibly also have an effect on ataxin-1 aggregation. The corrected FRET efficiencies obtained with the different protein pairs and combined from different photo-bleaching experiments were calculated and are summarised in Fig. 7. We found that the FRET signal of CT2AXH was much higher as compared to full-length ataxin-1. This may be due to the fact that the CT2AXH fragments are smaller than full-length ataxin-1, the fluorophores are likely to be closer to each other and the radial distance of the resonance energy transfer is accordingly smaller, leading to stronger FRET. We cannot also rule out the possibility that the truncated fragments tend to bind more strongly to each other. Reduced FRET of TLND2AXH, compared to CT2AXH is expected as the smaller TLND2AH is obviously forming smaller foci—pointing towards the possibility of a weaker association.

We can thus conclude that both dimerization domains of ataxin-1 contribute to the self-interaction of the protein in cells.