Structural basis for m₃G-cap-mediated nuclear import of spliceosomal UsnRNPs by snurportin1

Structure of the snurportin1–m₃G-cap complex

The crystal structure of the m₃G-cap-binding domain of human snurportin1 was determined by means of MIRAS using several heavy-atom derivatives and refined at a resolution of 2.4 Å (Table I). The overall structure is composed of five α-helices and 10 β-strands that form two almost coplanar β-sheets linked by two crossing β-strands (Figure 1). The m₃G-cap-binding pocket is located between the two β-sheets; several residues of strands β1, β3, β10 and adjacent loop regions interact with the entire m₃G-cap dinucleotide (Figure 2A, see Supplementary Figure 1). Both bases of the bound m₃GpppG are in a nearly coplanar orientation 3.5–3.8 Å apart, but slightly displaced with respect to a perfect base stacking (Figure 2B). The six-membered ring of the nonmethylated guanine is almost perfectly centered on the dimethylated N2 atom (Figure 2B). The base-stacking of the dinucleotide is continued by the side chain of Trp276 on the side of the trimethylated guanine, and is flanked by Leu104 on the side of the nonmethylated guanine (Figure 2A). Additionally, the m₃G-base is in hydrophobic contact with the side chains of Glu106 and Trp107, the latter in an almost perpendicular orientation to the stack with a distance of 4.3 Å. Remarkably, the protein forms only two hydrogen bonds with the trimethylated guanine base. The hydroxyl group of Ser105 donates a hydrogen to O6 and accepts one from N1 (Figure 2A). The nonmethylated guanine also forms hydrogen bonds with Ser105 as its NH₂ group at position 2, N1 is hydrogen-bonded to the main-chain carbonyl oxygen, and O6 accepts the hydrogen of the main-chain amide. The interaction is further stabilised by hydrogen bonds between several basic residues, the triphosphate and the riboses. The side chain of Lys128 and the main-chain amide of Arg129 interact with phosphate oxygens, and Lys144 is in contact with the cyclic oxygen of the ribose of the first snRNA nucleotide, as well as with phosphate oxygens (Figure 2A, see Supplementary Figure 1). A database search revealed that all residues involved in the interaction with the m₃GpppG dinucleotide are highly conserved among all known snurportin1 sequences. The only exception was found in Arabidopsis thaliana, where Ser105 is replaced by a proline residue. This replacement is structurally possible, but appears very unlikely due to the loss of hydrogen bonds involved in ligand binding. Since only a single base exchange is necessary to change serine into proline, an error of the database entry could also be plausible.

The binding mode of the m₃G-cap by snurportin1 differs significantly from that observed for the m⁷G-cap-binding proteins CBP20, eIF4E and VP39, as these proteins always intercalate the m⁷G-base between two aromatic side chains and keep the bound di- or oligonucleotides in an extended conformation (Figure 3) (Hodel et al, 1997, 1998; Calero et al, 2002; Mazza et al, 2002; Niedzwiecka et al, 2002; Tomoo et al, 2003).

So far only crystals of the m₃G-cap-binding domain with bound m₃GpppG have been obtained. All attempts to crystallise full-length snurportin1 have failed, probably due to a high degree of conformational flexibility between the N-terminal IBB domain and the m₃G-cap-binding domain. Also, no crystals of the m₃G-cap-binding domain could be obtained in the absence of m₃GpppG, indicating an effect of the dinucleotide on the structural integrity of the protein. Furthermore, it was not possible to crystallise a complex using a larger oligonucleotide, like synthetic m₃GpppA_mpU_mpA, which resembles the UsnRNAs more closely. The inspection of the crystal packing revealed that a larger cap oligonucleotide could not be accommodated in this crystal form of the snurportin1–m₃GpppG complex.

Mode of interaction between snurportin1 and m₃G-cap

During UsnRNP biogenesis, the hypermethylation of the m⁷G-cap occurs after assembly of the Sm-core UsnRNP complex and initiates the import of this complex into the nucleus. Therefore, the specificity of m₃G-cap binding by snurportin1 is of particular interest with respect to the discrimination of m⁷G-cap-bearing UsnRNAs and mRNAs, preventing their accidental reimport into the nucleus. The physiological cargoes of human snurportin1 share the common sequence m₃GpppA_m at the 5′ end of the UsnRNA. Previous studies have already demonstrated that an m₃GpppG dinucleotide is sufficient to be bound by snurportin1, and that the binding of m⁷GpppG to snurportin1 is two to three orders of magnitude less efficient than that of m₃GpppG (Huber et al, 1998). In order to quantify the interaction of snurportin1 with different 5′-caps, we determined the binding constants of various 5′-oligonucleotides by means of fluorescence spectroscopy (Table II). For these experiments the full-length snurportin1 was used, as the crystallised 26 kDa domain formed in situ during the co-crystallisation of the 37 kDa fragment and m₃GpppG (Strasser et al, 2004). Attempts to obtain the 26 kDa fragment either by preparative proteolysis or as recombinant protein failed, since this fragment precipitates due to unfolding or incorrect folding, respectively, in the absence of m₃GpppG and is therefore not suitable for biochemical experiments. The K_d-value for the m₃GpppG dinucleotide (which is present in the crystal structure) was determined to be 1.0 μM, while it is 12.1 μM for the m₃GpppA dinucleotide. The crystal structure of the complex containing the m₃GpppG dinucleotide reveals that the adenosine of m₃GpppA could be harboured in the binding pocket as well, since there is no strong discrimination between the purines commonly caused by polar interactions. The higher K_d-value of the m₃GpppA appears to be related to the lack of two hydrogen bonds with Ser105. It turned out to be impossible to determine the exact K_d-value of the m⁷GpppA dinucleotide by fluorescence spectroscopy, owing to an extremely large inner filter effect caused by the high dinucleotide concentrations required, but it can be estimated that the K_d-value for m⁷GpppA is greater than 170 μM. Obviously, the difference in K_d-values between m₃GpppA and m⁷GpppA must be related to the two methyl groups on N2 of the m₃G-base. These two methyl groups are in van der Waals (VDW) contact with Trp107, but the difference in affinities between m₃GpppA/G and m⁷GpppA cannot be caused exclusively by the contribution of these two VDW contacts to the free binding energy. Hence, the difference in affinity should be related to the binding process itself. Upon binding to snurportin1 the cap is dehydrated, and hydrogen bonds to water molecules are replaced by hydrogen bonds to the protein. Obviously, the m⁷G- and m₃G-caps differ in solution with respect to their hydration shells, as the m⁷G-cap is capable to bind two water molecules via its NH₂ group. These two water molecules have to be released for the binding of the m⁷G-cap requiring a substantial amount of energy, as these hydrogen bonds are not replaced by hydrogen bonds to the protein. The dimethylated N2 of the m₃G-cap has no bound water molecules; hence its binding to snurportin1 and the hydrophobic interaction with the Trp107 side chain is much more favourable. These differences could already cause the observed differences in K_d, but other molecular properties might additionally contribute to the discrimination.

The discrimination between m⁷G- and m₃G-caps by snurportin1 might also be related to the strength of the cation–π interaction. With respect to the pK_a-value of 7.46 of m₃GpppG (Wieczorek et al, 1995), the m₃G-base is expected to be positively charged under crystallisation conditions (pH 5.5). As suggested for m⁷G-cap-binding proteins (Quiocho et al, 2000), a cation–π interaction between the positively charged trimethylated guanine and the coplanar aromatic side chain of Trp276 is expected. The presence of the two methyl groups bound to N2 probably changes the distribution of the positive charge within m₃G in comparison to m⁷G and therefore significantly affects the cation–π interaction. Interestingly, Trp107 is only 4.5 Å distant to the N2 atom of the m₃G-base (Figure 2C); hence, a cation–π interaction appears reasonable assuming a significant positive partial charge on N2 stabilised by the two methyl groups.

Consistent with a previous report (Huber et al, 1998), the binding of the synthetic tetranucleotide m₃GpppA_mpU_mpA, which resembles more closely UsnRNAs in vivo, to full-length snurportin1 is even stronger (K_d=0.23 μM) than that of m₃GpppG, which should be related to the presence of additional nucleotides and/or due to the 2′O-methylated riboses. Inspection of the electrostatic surface potential of the m₃G-cap-binding domain gives no hint on binding sites for the additional nucleotides (data not shown). However, the crystallised domain lacks the C-terminal 61 residues, which could extend the RNA-binding surface. To clarify the role of these C-terminal residues, the determination of the K_d-value for the crystallised snurportin1 fragment would be necessary, but the 26 kDa fragment aggregates rapidly in the absence of m₃GpppG (see above) and could therefore not be used for biochemical experiments.

The crystallised domain is also lacking the N-terminal 78 residues, raising the question whether the IBB domain could affect the m₃G-cap-binding analogous to importin-α. This appears possible since the IBB domain of importin-α is known to compete with the NLS of a cargo for binding to importin-α (Kobe, 1999). To examine this potential autoinhibitory effect of the snurportin1 IBB domain on m₃G-cap binding, the affinity of the m₃GpppG was evaluated in the presence of increasing amounts of IBB domain (residues 1–65). Remarkably, even at 100-fold excess of IBB domain over snurportin1 (and a five-fold excess of m₃GpppG over snurportin1) the affinity remains unaltered (see Supplementary Table Ia). In a second approach, an N-terminally truncated snurportin1 comprising residues 66–360 was used. The K_d-value for m₃GpppG turned out to be 1.24 μM (±0.3 μM), which is close to the value of 1.0 μM (±0.03 μM) observed for the full-length protein. Both experiments demonstrate that the IBB domain of snurportin1 does not exhibit an autoinhibitory effect on cargo binding as reported for importin-α.

Role of tryptophan residues Trp276 and Trp107 in m₃G-cap binding

An obvious feature of the m₃G-cap-binding pocket is the presence of the two tryptophan residues Trp107 and Trp276 (Figure 2). In order to understand their functional significance, these tryptophan residues were substituted by alanine, either singly or together. The resulting mutated proteins were purified and characterised with regard to m₃G-cap-binding affinity using fluorescence spectroscopy and an in vitro nuclear import assay. However, an exact determination of K_d-values for the mutated proteins was not possible, as the strongly reduced affinities (see below) would require high concentrations of m₃GpppG, leading to a very large internal filter effect. Therefore, the change in affinity was estimated comparing the changes in fluorescence upon addition of m₃GpppG.

Binding of m₃GpppG to wild-type snurportin1 leads to a decrease in fluorescence of 62.7% (±3.9%), indicating a strong interaction. Taking the changed fluorescence properties of each mutant into account, the binding of m₃GpppG decreases the fluorescence for the W276A and W107A mutants only by 20.0% (±6.1%) and 25.0% (±6.3%), respectively, and even less (15.8% (±4.2%) for the double mutant, corresponding to a reduced cap-binding activity. The effect of the W276A substitution is expected, as this tryptophan stacks on the m₃G-base and forms a strong cation–π interaction. More interestingly, the W107A mutation shows almost the same reduction in fluorescence, suggesting an important role in m₃G-cap binding. Since the two methyl groups of N2 are pointing towards Trp107, this tryptophan should have a substantial impact on the discrimination of m⁷G-cap-bearing RNAs. Indeed, the binding affinity of m⁷GpppA to the W107A mutant is identical to that of m₃GpppA, as determined by the decrease in tryptophan flourescence (see Supplementary Table Ib). Hence, the fully solvent-exposed Trp107 plays an important role in the process of cap-binding and discrimination of m⁷G-caps.

Furthermore, the β-strand (β1) harbouring Trp107 exhibits a highly twisted conformation and might adopt a different, more relaxed conformation in the absence of the cap, with the side chain of Trp107 located inside the cap-binding pocket. However, there is currently no experimental evidence for such a conformational flexibility of the β-strand β1 or for an induced fold upon cap binding as observed for CBP20 (Calero et al, 2002; Mazza et al, 2002).

The essential role of these two tryptophan residues, Trp276 and Trp107, was also observed in the in vitro nuclear import assay using intact U1snRNPs as transport cargo. The decrease in import observed for the three mutants correlates quantitatively with the reduced cap-binding affinities evaluated by fluorescence spectroscopy (Figure 4). Therefore, it appears unlikely that snurportin1 forms other strong contacts, like protein–protein interactions, with the U1snRNP. However, the nuclear import of UsnRNPs in vivo was reported to occur by means of a larger complex containing snurportin1, importin-β, the SMN protein and presumably other components of the SMN complex (Narayanan et al, 2002, 2004).

Structural homology

Searching the PDB for structurally related proteins revealed that the overall fold of the m₃G-cap-binding domain exhibits high structural similarity to the GTP-binding domain of the mRNA-guanylyltransferase indicated by an r.m.s.d. of 2.6 Å for 204 common C_α atoms. Interestingly, this enzyme is involved in the formation of the 5′-cap as it transfers GMP to the 5′end of mRNAs and UsnRNAs via a 5′–5′-triphosphate bridge (Hakansson et al, 1997). This domain is also found in DNA ligases, where it is involved in ATP binding (Subramanya et al, 1996). Superposing the structures of snurportin1 and mRNA-guanylyltransferase, which shares the highest structural similarity to snurportin1, reveals that the binding sites for GTP and the m₃G-cap are in a similar area of the protein surface, but the positions of the nucleotides and the residues forming the binding pocket, and therefore all interactions, are completely different (Figure 5A/B). Furthermore, the GTP-binding pocket of the mRNA-guanylyltransferase penetrates much more deeply between the β-sheets than the m₃G-cap-binding pocket of snurportin1.

The structural homology between the m₃G-cap-binding domain and the mRNA-guanylyltransferase was unexpected, since all searches for sequences homologous to the m₃G-cap-binding domain gave no significant hint to any other protein (Huber et al, 1998). Indeed, the amino-acid sequence identity between the human m₃G-cap-binding domain and the human mRNA-guanylyltransferase is only 12.7% (Figure 5C). A more elaborate database search recently identified the m₃G-cap-binding domain of snurportin1 as an inactive paralogue of the mRNA-guanylyltransferase (Mans et al, 2004). The structural homology suggests that the m₃G-cap-binding domain shares a common ancestor with the GTP-binding domain of the mRNA-guanylyltransferase that was linked to an IBB domain by shuffling and fusion of the corresponding exons.

Structure determination

Purification, m₃G-cap-binding affinity test and crystallisation strategy are described elsewhere (Strasser et al, 2004). Shortly, crystals of a 26 kDa m₃G-cap-binding domain of snurportin1 were only obtained in the presence of the synthetic m₃GpppG-cap dinucleotide. Single crystals for X-ray data collection were grown in sitting drops using 8–10% (w/v) PEG 10 k and 200 mM sodium citrate (pH 5.5) as precipitant. Diffraction data of a native crystal were collected at the PSF beamline BL1 (BESSY, Berlin) to a resolution of 2.4 Å. Three different heavy-atom derivative data sets were collected on a Micromax 007 rotating anode generator (Rigaku/MSC, USA) operating at 40 kV and 20 mA, equipped with Osmic focusing mirrors and a Mar345dtb detector system (Xray-Research, Germany) in order to solve the phase problem. Data were processed using Denzo/Scalepack (HKL-research) and CCP4 programs (Bailey, 1994). Heavy-atom sites were found and initial phases were improved by AutoSHARP (Globalphasing Ltd, UK). The graphics program O (Jones et al, 1991) was used for model-building and the structure was refined with CNS (Brünger et al, 1998). A 37 kDa fragment of snurportin1 was used for crystallisation, but the crystals contain only a 26 kDa fragment. Inspection of possible GluC-protease cleavage sites and preliminary mass spectroscopic data indicates that this fragment consists of residues 79–300. The refined model comprises residues 97–300. The missing residues are not defined in the electron densisty map due to conformational flexibility. The m₃GpppG-cap dinucleotide was well defined in the electron density map.

Topology and parameter files for the m₃GpppG-cap dinucleotide were created with the help of the Dundee PRODRG2 Server (van Aalten et al, 1996). Secondary structure elements were assigned by STRIDE (Frishman and Argos, 1995) and figures were generated with Pymol (www.pymol.org).

Expression constructs

For all expression constructs described, full-length snurportin1 in pGEX6P1 (Amersham Biosciences) was used as a PCR template (Strasser et al, 2004). The coding region of the deletion mutant of snurportin1 (residues 66–360) was amplified using primers with restriction sites for BamHI and XhoI and inserted into pGEX6P1. The IBB domain (residues 1–65) was cloned in pET28b (Novagen) via NcoI and XhoI restriction sites. Tryptophan mutants W276A, W107A and W276A/W107A in pGEX6P1 were generated using the QuickChange™ Site Directed Mutagenesis Kit (Stratagene).

Protein expression and purification

Expression and purification of snurportin1 wild-type, truncated snurportin1 (residues 66–360) and the tryptophan mutants were performed as described previously (Strasser et al, 2004). The IBB domain (residues 1–65) was expressed as a His-tagged protein at 30°C for 5 h using 1 mM IPTG and purified at first by Ni-NTA agarose. Therefore, the supernatant was applied to the resin in 300 mM NaCl, 20 mM Tris, pH 7.5, 2 mM 2-mercaptoethanol, and eluted with the loading buffer containing 300 mM imidazole. The fractions containing the IBB domain were pooled, concentrated and further purified by gel filtration on Superdex 75 media (100 mM NaCl, 50 mM HEPES, pH 7.5, 2 mM DTT).

Fluorescence spectroscopy

Fluorescence measurements were performed on a Fluoromax3™ spectrofluorimeter (Jobin Yvon) at 20°C in 100 mM NaCl, 50 mM HEPES/NaOH (pH 7.5), using cuvettes with 0.5 × 1.0 cm² section (Hellma). Emission at 315 or 340 nm, used for the calculation of dissociation constants, was recorded during excitation at 295 nm (1 nm bandwidth) for 60 s with time constants of 0.5 s. Equilibrium dissociation constants were obtained by fitting the solutions of a quadratic function, assuming a 1:1 stoichiometry and taking into account the fluorescence of the cap oligonucleotides (Graphit 3.1, Erithacus Software). Standard deviations were calculated from three independent experiments. To determine the cap-binding affinity of wild-type snurportin1 and of the mutants W107A, W276A and W107A/W276A, emission spectra (310–500 nm) were collected after excitation at 295 nm for the proteins alone and in the presence of m₃GpppG in a 20-fold, or in case of m⁷GpppA and m₃GpppA in a 50-fold, molar excess. Spectra were corrected for the fluorescence of the cap dinucleotides and buffer contributions. The decrease in fluorescence upon ligand binding to snurportin1 was determined in three independent measurements. For competition experiments with the IBB domain, the decrease in tryptophan fluorescence after addition of m₃GpppG in a five-fold molar excess to snurportin1 full length was compared to the decrease after addition of increasing amounts of the IBB domain (residues 1–65). The IBB domain was titrated up to a 100-fold molar excess to snurportin1 wild type.

Nuclear import assay

Nuclear import assays were performed as described by Huber et al (2002). Each transport reaction mixture contained 100 nM Cy3-labelled human U1snRNPs, and, except for the negative control, 500 nM importin-β, 300 nM of snurportin1 or the mutated snurportin1 proteins (W107A, W276A and W107A/W276/A). The transport was performed on permeabilised HeLa cells for 10 min at room temperature in a humidified chamber. After extensive washing with 1 × PBS, the cells were fixed with paraformaldehyde (5% in 1 × PBS) and mounted on coverslips. Images were taken at 400-fold magnification. Rates of U1snRNP import were calculated with the help of the program ImageJ (http://rsb.info.nih.gov/ij/), by averaging nuclear fluorescence intensities of 50 cells. In order to compare the results, the intensities of each experiment were normalised against wild-type snurportin1 intensities.