Sus1, Cdc31, and the Sac3 CID Region Form a Conserved Interaction Platform that Promotes Nuclear Pore Association and mRNA Export

Summary The yeast Sac3:Cdc31:Sus1:Thp1 (TREX-2) complex facilitates the repositioning and association of actively transcribing genes with nuclear pores (NPCs)—“gene gating”—that is central to integrating transcription, processing, and mRNA nuclear export. We present here the crystal structure of Sus1 and Cdc31 bound to a central region of Sac3 (the CID domain) that is crucial for its function. Sac3CID forms a long, gently undulating α helix around which one Cdc31 and two Sus1 chains are wrapped. Sus1 has an articulated helical hairpin fold that facilitates its wrapping around Sac3. In vivo studies using engineered mutations that selectively disrupted binding of individual chains to Sac3 indicated that Sus1 and Cdc31 function synergistically to promote NPC association of TREX-2 and mRNA nuclear export. These data indicate Sac3CID provides a scaffold within TREX-2 to integrate interactions between protein complexes to facilitate the coupling of transcription and mRNA export during gene expression.

Initial crystals of complex containing Sac3 residues 753-813 diffracted very poorly. Therefore, truncation mutants were used to define more precisely the region of Sac3 that binds Sus1 and Cdc31. A series of Sac3 fragments were made that progressively removed residues from the C-terminus of the centrin-binding site. In this way, a fragment comprising residues 753-805 was found that retained the ability to bind both Sus1 and Cdc31 ( Figure S3). Crystals grown from this minimal Sac3 fragment yielded high-resolution diffraction data. A Sac3 construct spanning residues 723-805, incorporating both Sus1 binding sites and also ending in residue 805, yielded high resolution data readily.

Crystal structure determination
An initial MR solution of the P2 1 crystal of the Sac3 753-805 :Sus1:Cdc31 complex (crystal form 2, Table 1) was obtained using Cdc31 (PDB 2GV5) as model. Although the phasing was not adequate to trace the Sac3 and Sus1 chains reliably, it enabled identification of the Se sites present in Sac3 (2 sites) and Cdc31 (4 sites). The resultant SAD map obtained after processing using SHARP (Bricogne et al. 2003) followed by solvent flipping (Abrahams & Leslie, 1996) enabled most of the Sac3 chain to be built, after which solvent-flattened maps began to show a series of α-helices from which a putative Sus1 model was built. Molecular replacement was then used to place this model into P2 1 2 1 2 1 crystals formed from unlabelled Sac3 and Cdc31 and Se-Met labelled Sus1. Identification of the two Se sites on the Sus1 chain allowed unequivocal assignment of the Sus1 sequence and some rebuilding of the Sus1 chain. This structure was finally placed into the 2.5Å resolution P2 1 native data set (crystal form 1, Table 1) and, after iterative cycles of refinement (using REFMAC5 with TLS (CCP4, 1994) and CNS (Brunger, 2007) together with local rebuilding, resulted in a model with an R-factor of 18.8% (Rfree = 23.5%) and excellent geometry ( Table 1). This structure had a MolProbity (Davis et al. 2007) score of 1,85 (97th percentile). We also obtained the 3.1 Å resolution molecular replacement structure (R-factor=19.9%; Rfree=24.6%) of the P2 1 2 1 2 1 crystal form that showed the same arrangement of chains in the complex.
The structure of the Sac3 753-805 complex was then used to obtain the structure of the Sac3 723-805 complex by molecular replacement using MOLREP (CCP4, 1994). These crystals diffracted to 2.7 Å resolution ( Table 1), although the diffraction data showed some asymmetry, which was corrected using the anisoscale web site (http://www.doembi.ucla.edu/~sawaya/anisoscale). The self-Patterson function showed that there was a clear pseudo-translation of (0.5, 0, 0.5) which complicated the determination, although eventually a model that contained four copies of the Sac3 753-805 complex was obtained. Sac3 residues 723-752 and an additional Sus1 chain were then built into the density and the structure refined, initially using CNS (Brunger, 2007) followed by REFMAC (CCP4, 1994) and finally PHENIX (Terwilliger et al. 2008) using TLS refinement using the groups indicated by the TLSMD website (http://www.skuld.bmsc.washington.edu/~tlsmd/index.html - Painter and Merritt, 2007) to yield a final structure that contained four Cdc31 chains, four Sac3 chains, eight Sus1 chains, 4 sulphate ions, and 108 water molecules and which had an R-factor of 21.0% (Rfree = 26.4%) and excellent geometry ( Table 1). This structure had a MolProbity (Davis et al. 2007) score of 2.13 (92nd percentile).

The Cdc31:Sac3 interface
The fold of Cdc31 is homologous to that of calmodulin and is based on two domains, each of which contains two putative Ca-binding EF-hands. The N-domain (residues 18-91) had the "closed" conformation (see Yap et al. 1999) and the C-terminal domain (residues 95-158) in the "open" conformation, similar to that observed for Cdc31 bound to Sfi1 (Li et al. 2006). As observed in Cdc31:Sfi1 crystals (Li et al. 2006), there was little clear electron density for Cdc31 residues 1-13 in any of the different crystal forms we investigated. These residues vary between species and are not required for function (Li et al. 2006) and were probably disordered. The Sac3:Cdc31 interaction interface involved primarily the C-terminal domain (EF hands III and IV, residues 95-158) of Cdc31, with fewer contacts observed between Sac3 and the N-terminal domain of Cdc31 than were seen with the Sfi1:Cdc31 interaction (Li et al. 2006). The residues in the C-terminal domain of Cdc31 that form the large hydrophobic interface with Sac3 are similar to those involved both with the interaction with Sfi1 (Li et al. 2006) and the analogous interaction between Clamydomonas centrin and Kar1 (Hu & Chazin, 2003), albeit the Sac3 helix is oriented in the opposite direction. Trp802 of Sac3 appears to play a central role in the interaction and becomes buried in a hydrophobic cavity in the C-terminal domain of Cdc31 formed by the side chains of Phe105, Met137, Ile138, Phe141, Ile149 and Ile157 ( Figure 2C). Additional hydrophobic contacts are also made by Phe799 Sac3 and Phe798 Sac3 that contact smaller hydrophobic patches on Cdc31 formed by Phe105, Leu118, Leu125 and Glu97, Ile98, Ala101, respectively. There are also putative salt bridges formed between Lys795 Sac3 -Glu124 Cdc31 and Arg790 Sac3 -Glu97 Cdc31 . By comparison, the interaction interface between Sac3 and the Cdc31 N-terminal domain is much less substantial and comprises hydrophobic contacts between Met782 Sac3 -Val47 Cdc31 and Thr793 Sac3 -Glu24 Cdc31 together with putative salt bridges between Lys789 Sac3 -Glu27 Cdc31 and Lys797 Sac3 -Glu20 Cdc31 . The ~24-residue Cdc31-binding region of Sac3 is somewhat shorter than the ~33 residue Cdc31-binding motifs present in Sfi1 (Li et al. 2006). Although the Cdc31 binding region of Sac3 is not directly homologous to the consensus Cdc31 binding motif seen in Sfi1 (Li et al. 2006), it clearly shares several common features, especially the roles played by Phe798, Phe799 and Trp802 that dominate the interaction interface with the Cdc31 C-terminal domain. The primary difference between the Cdc31 binding motifs in Sfi1 and Sac3 is the absence in Sac3 of the residues involved in the interaction with the Cdc31 N-terminal domain.

Affinity of Sus1 for Sac3 CID
To measure the binding affinity of S-tagged Sus1 to the Sac3:Cdc31 complex Fluotrac-600 96well black plates (Greiner Bio-one) or 96-well black glutathione coated plates (Pierce) were coated with 1 to 5 nM per well of GST-Sac3 723-805 :Cdc31 complex or GST-alone for 16 hours at room temperature. Solid phase binding assays were then performed as described (Bayliss et al. 2002), except that binding was carried out for 16 hours at 4 °C. The bound S-tagged Sus1 was detected by incubation with horseradish peroxidase conjugated S-tag antibody (Abcam, Cambridge, UK) and subsequent addition of the horseradish peroxidase substrate SuperRed (Virolabs, Virginia, USA). The fluorescent signal was determined using a Tecan Safire II plate reader at excitation and emission wavelengths of 530 nm and 590 nm respectively. Binding data ( Figure S6) were analyzed with GraphPad Prism 5.0a for Mac OS X (GraphPad Software, San Diego, California, USA) using nonlinear regression assuming one site binding.  a ;ade2, his3, leu2, trp1, ura3, nup133::HIS3 (Doye et al., 1994) Figure S1. Sequence alignment of the Sac3 CID region with homologous proteins. Sac3 homologs were identified by BLAST searches. Sequences downstream of the conserved PFAM Sac3/GANP domain were submitted to Jpred (http://www.compbio.dundee.ac.uk/~www-jpred) for secondary structure prediction. Regions predicted to form long α-helices were then further aligned using ClustalW2 (http://www.ebi.ac.uk/Tools/clustalw2). Amino acids are coloured based on chemical properties and aligned conservation. Numbers above the alignment correspond to human GANP and below to yeast Sac3 amino acid residues.  Figure S3. A minimal fragment of the Sac3 CID (residues 753-805) retains binding to Sus1 and Cdc31. Co-expressed GST-Sac3 fragments and Cdc31 together with singly expressed Sus1, both from bacterial cell lysates, were incubated with glutathione sepharose for 1hr, the resin washed thoroughly and analysed by SDS-PAGE. Figure S4. Human Sac3/GANP binds human Sus1/ENY2 in vitro.  contains the Sus1A site whereas GST-Sac3(758-805) contains the Sus1B and Cdc31 sites. GST-GANP(1162-1204 contains the putative Sus1A site, whereas GANP(1162-1240) contains both the putative binding Sus1 sites identified by sequence analysis (Figure S1). Recombinant fragments, GST-GANP or GST-Sac3, or GST alone were immobilized on glutathione sepharose and incubated with purified recombinant His-tagged ENY2 or His-tagged Sus1. The sepharose beads were washed to remove non-specifically bound protein and analyzed by SDS-PAGE and Coomassie staining. The complex between GST-tagged Sac3 723-805 , Cdc31 and Sus1 was immobilised on glutathione sepharose from clarified bacterial lysate in the presence of 5mM EGTA or 5mM CaCl 2 . After 1hr incubation, resins were thoroughly washed in buffer containing either 5mM EGTA or 5mM CaCl 2 and analysed by SDS-PAGE. Figure S6. Affinity of Sus1 for Sac3 CID :Cdc31. The affinity of Sus1 for the Sac3 CID :Cdc31 complex was measured using solid phase binding assays (Bayliss et al. 2002). The apparent dissociation constant was calculated from the average of ten normalized data sets using nonlinear regression and is expressed ± the standard error.