|Home | About | Journals | Submit | Contact Us | Français|
The essential splicing factors U2AF65 and SF1 cooperatively bind consensus sequences at the 3′ end of introns. Phosphorylation of SF1 on a highly conserved ‘SPSP’ motif enhances its interaction with U2AF65 and the pre-mRNA. Here we reveal that phosphorylation-induces essential conformational changes in SF1 and in the SF1/U2AF65/3′ splice site complex. Crystal structures of the phosphorylated (P)SF1 domain bound to the C-terminal domain of U2AF65 at 2.29 Å resolution, and of the unphosphorylated SF1 domain at 2.48 Å resolution, demonstrate that phosphorylation induces a disorder-to-order transition within a novel SF1/U2AF65 interface. We find by small-angle X-ray scattering that the local folding of the SPSP motif transduces into global conformational changes in the nearly full length (P)SF1/U2AF65/3′ splice site assembly. We further determine that SPSP phosphorylation and the novel SF1/U2AF65 interface are essential in vivo. These results offer a structural prototype for phosphorylation-dependent control of pre-mRNA splicing factors.
Protein phosphorylation is required for pre-mRNA splicing (reviewed in (Stamm, 2008)), yet the structural changes induced by splicing factor phosphorylation currently are unknown. The serine/arginine-rich (SR) protein superfamily is a prototype for phosphorylation-dependent control of pre-mRNA splicing, in which phosphorylation of arginine-serine repeats regulates the localization and interactions among SR proteins and the pre-mRNA (reviewed in (Ghosh and Adams, 2011)). Phosphorylation of other classes of splicing factors is only beginning to be understood at the molecular level. One exemplary function of non-SR phosphorylation is to regulate interactions with U2AF65, which is an essential pre-mRNA splicing factor (Golling et al., 2002; Ruskin et al., 1988). A complex between U2AF65 and a second splicing factor, SF1, cooperatively recognizes consensus sequences at the 3′ splice sites of pre-mRNAs (Berglund et al., 1998; Zamore et al., 1992). There, U2AF65 recruits the U2 small nuclear ribonucleoprotein particle (snRNP) to the assembling spliceosome (Ruskin et al., 1988). Phosphorylation of SF1 by cGMP-dependent protein kinase-I (PKG-I) inhibits association with U2AF65 (Wang et al., 1999). Similarly, the SF3b155 subunit of the U2 snRNP is phosphorylated adjacent its U2AF65-interaction sites (Boudrez et al., 2002; Seghezzi et al., 1998), which suggests that SF3b155 phosphorylation could promote U2AF65 dissociation from the activated spliceosome (Bessonov et al., 2010; Shi et al., 2006; Wang et al., 1998).
A highly conserved SPSP motif of human SF1 (residues 80-83) was found to be predominately in the phosphorylated state in proliferating human embryonic kidney cells (Manceau et al., 2006). Subsequent phospho-proteome analyses of HeLa (Beausoleil et al., 2004), lymphoma (Shu et al., 2004), and prostate cancer cells (Myung and Sadar, 2012) confirm the prevalence of SF1 SPSP phosphorylation and suggest a role in cancer initiation or progression. A kinase called Kinase Interacting with Stathmin (KIS), or UHMK1, specifically phosphorylates SF1 serines at position 80 and 82 within the SPSP motif (Manceau et al., 2008; Manceau et al., 2006). Phosphorylation by KIS enhances SF1/U2AF65 interactions and promotes assembly of the ternary complex among SF1, U2AF65, and the 3′ splice site (Manceau et al., 2006). The phosphorylated SPSP motif resides in the most highly conserved region of SF1, which is located between a KH-QUA2 domain for pre-mRNA recognition and an N-terminal U2AF Ligand Motif (ULM) for U2AF65 interactions. Whereas most of the SF1 or U2AF65 domains engage in well-characterized protein or RNA interactions during pre-mRNA splicing and belong to documented families of protein folds (Figure 1A), the SPSP-containing region of SF1 lacks known structural homologues. Apart from phosphorylation, the function of the SF1 SPSP-containing domain is poorly understood.
Notably, the SF1 SPSP motif typifies the phosphorylation sites of U2AF65 interaction partners, which often are found adjacent to ULM sequence motifs composed of a key tryptophan residue preceded by arginine and lysine residues. These ULMs bind RNA recognition motif (RRM) relatives that have acquired specialized features for ULM recognition, called U2AF Homology Motifs (UHMs). The prototypical UHM of U2AF65 sequentially engages the ULMs of SF1 and SF3b155 during the splicing process (Cass and Berglund, 2006; Selenko et al., 2003; Spadaccini et al., 2006; Thickman et al., 2006). Subsequent sequence analyses and a number of structures have identified UHMs among diverse nuclear proteins, including the KIS kinase, nuclear lamins, and other pre-mRNA splicing factors (Kielkopf et al., 2004). The atomic resolution structures of U2AF65, U2AF35, and SPF45 complexes (Corsini et al., 2007; Kielkopf et al., 2001; Selenko et al., 2003) reveal shared features of UHM-ULM recognition. The ULM tryptophan is buried within a hydrophobic pocket and positively-charged residues interact with negatively-charged residues of the UHM. The short length of this epitope raises the question of how ULMs specify distinct UHMs, or whether cross-talk occurs regularly among ULMs and diverse UHM-containing proteins. For example, the UHMs of SPF45, PUF60, and KIS kinase bind a range of ULMs, including those of U2AF65, SF3b155, and SF1 (Corsini et al., 2007; Corsini et al., 2009; Manceau et al., 2008). In contrast, an extended, proline-rich region of the U2AF65 ULM engages in a specific ‘tongue-in-groove’ interaction with the U2AF35 UHM (Kielkopf et al., 2001), indicating that subtle architectural changes can confer ULM/UHM specificity. In addition to sequence extensions of the minimal ULM, post-translational modifications such as phosphorylation also offer potential handles to confer specificity in the ULM-UHM network.
To advance our understanding of the role of phosphorylation in the SF1/U2AF65 interaction, we determined the structure of the doubly phosphorylated SPSP domain in the SF1/U2AF65 complex in comparison with unphosphorylated SF1, and further established that SPSP phosphorylation and U2AF65 interactions are required for mammalian cell proliferation. Crystal structures reveal that the phosphorylated domain of human SF1 engages the U2AF65 UHM in an extensive interface beyond the minimal ULM. The structure of the unphosphorylated SF1 domain, coupled with small-angle X-ray scattering of SF1/U2AF65 complexes in the different phosphorylated and RNA-bound states, further demonstrated that phosphorylation-induced folding of an arginine-rich loop around the phosphorylated serines amplifies into overall conformational changes in the nearly full length, phosphorylated SF1/U2AF65/3′ splice site complex. This tightly bent conformation of the phosphorylated splicing factor complex is compatible with coupling of the 5′ and 3′ ends of the splice site for the catalytic steps of splicing.
The importance of SF1/U2AF65 interactions and SPSP phosphorylation for mammalian cell proliferation was unknown. It was previously demonstrated that SF1 knockdown in HeLa cells by siRNA transfection substantially reduced cell proliferation (Tanackovic and Kramer, 2005). We leveraged this requirement as the basis for testing the importance of SF1 phosphorylation and U2AF65-interactions (Figures 1 and S1). Proliferation of mouse NIH3T3 fibroblasts was impaired severely following transfection with three different siRNAs targeting the known splice isoforms of mouse SF1 (Figures 1B and S1A-B). Three days post-transfection, SF1 protein was no longer evident and the cell numbers were substantially diminished compared to two control siRNAs (Figure 1B). We next tested the abilities of wild-type or myc-tagged SF1 containing mutated residues within the N-terminus (Figure 1A) to rescue the proliferation of cells treated continuously with siRNA. The transfected expression plasmids were derived from the human SF1 cDNA, which encodes an identical protein as mouse SF1, but has the advantage of at least two mismatches with our siRNA constructs. Accordingly, human SF1 and its mutant forms were readily detected at similar levels by immunoblotting 24 h post-transfection (Figure S1C). The transiently expressed proteins were phosphorylated efficiently in vivo, with the exception of the serine mutations that prevented SPSP phosphorylation (Figure S1C). Five days after plasmid transfection, cell proliferation was measured by cell counting or assessed by an MTT assay. Wild-type SF1 restored the proliferation potential and apparent normal morphology of the cells (Figure 1C).
To assess the cellular consequences of abrogating the SF1/U2AF65 complex, we first tested a tryptophan to alanine (W22A) mutation in the SF1 ULM that is known to disrupt U2AF65 binding in vitro (Selenko et al., 2003). The W22A mutation severely reduced the ability of SF1 to rescue siRNA-treated cells, which is in agreement with the individual requirements for SF1 and U2AF65 in most eukaryotes tested (for example, (Golling et al., 2002; Kanaar et al., 1993; Mazroui et al., 1999; Potashkin et al., 1993; Shitashige et al., 2007; Zorio and Blumenthal, 1999)). The deleterious effect of the W22A mutation in SF1 demonstrated that formation of the SF1/U2AF65 complex is required for mammalian cell proliferation (Figure 1C). We next tested the in vivo consequences of mutating the serines in the SF1 SPSP motif (S80 and S82) that are normally phosphorylated in proliferating cells. Preventing phosphorylation by mutating the serines to alanines within the SPSP motif (S80/82A) reduced the ability of SF1 to rescue cell proliferation to a level comparable to the W22A mutation, whereas phosphomimetic mutations in the motif (S80/82E) that introduced a negative charge were significantly more effective at promoting proliferation than the S80/82A mutant (Figure 1C). The effects of the SF1 mutations indicated that both formation of the SF1/U2AF65 complex and phosphorylation of the SPSP motif are essential for mammalian cell proliferation.
Based on the observed in vivo functional importance of SF1 phosphorylation, we correlated the rescuing potential with the ability of the SF1 proteins to bind U2AF65 (Figure 1D). Nearly full-length, GST-fused SF1 fragments (comprising the ULM, SPSP, and KH-QUA2 domains) were purified from E. coli and incubated in pull-down assays with a HEK293T RNase-treated cell extract as a source of U2AF65, since endogenous U2AF65 in these extracts was less prone to aggregation than the recombinant protein. Each GST-fusion protein was tested after incubation in a phosphorylation reaction with or without purified KIS. SDS-PAGE and phosphoprotein staining confirmed that the proteins were phosphorylated efficiently using these conditions (Figure 1D). As expected based on previous observations (Manceau et al., 2006; Selenko et al., 2003), the critical W22A mutation nearly abolished association with U2AF65, whereas phosphorylation of SF1 by KIS increased association with U2AF65. The phosphorylation-impaired S80/82A mutant behaved similarly to unphosphorylated wild-type SF1, whereas the phospho-mimetic S80/82E mutant was comparable to phosphorylated SF1. These in vivo and interaction studies demonstrated that SPSP phosphorylation is correlated with SF1/U2AF65 complex formation and is required for mammalian cell proliferation.
Given an evident role for the SPSP domain in forming the SF1/U2AF65 complex, we proceeded to investigate the size of the SF1/U2AF65 interface in solution. Since heat capacity changes (ΔCp) are often related to interface size (reviewed in (Prabhu and Sharp, 2005)), we used isothermal titration calorimetry at four different temperatures to determine the ΔCp for U2AF65 UHM binding to a nearly full-length SF1 protein containing the conserved ULM, SPSP, and KH-QUA2 domains (residues 1-255) (Figures 2A and S2A-C). The ΔCp for SF1 binding to the U2AF65 UHM (−462 ± 55 cal mol−1 K−1) corresponded to ~1900 Å2 of additional buried surface area beyond the previously established SF1 ULM/U2AF65 UHM interface (Selenko et al., 2003). This relatively large, unknown interface between SF1 and the U2AF65 UHM corresponded in size to a typical heterodimer interface (Janin et al., 2008).
We then compared differences in the 1H-15N HSQC spectra of 15N-labeled U2AF65 UHM bound to three different SF1 constructs (Figures 2B-C and S2D): the minimal SF1 ULM peptide, SF114-132 composed of the ULM and the adjacent SPSP domain, or nearly full-length SF1. The 1H-15N HSQC spectrum of the (15N)U2AF65 UHM/SF1 ULM complex was assigned based on prior assignments (courtesy of M. Sattler (Selenko et al., 2003)) and extrapolated to the SF1 and SF114-132 complexes with the assistance of 3D NOESY-NHSQC and TOCSY-NHSQC spectra. The HSQC spectra of SF1/(15N)U2AF65 UHM and SF114-132 /(15N)U2AF65 UHM are nearly identical, in contrast to the HSQC spectra of the above compared with the SF1 ULM/U2AF65 UHM complex, which differs significantly at UHM residues surrounding the bound ULM C-terminus. These results implicated the SPSP domain of SF1, which follows the SF1 ULM C-terminus, as an additional region that interacts with U2AF65. Such an extended SF1/U2AF65 interface explains prior observations that a ULM deletion mutant of SF1 continues to support spliceosome assembly at a low but detectable level (Rain et al., 1998).
To understand the role for the phosphorylated SF1 domain in complex formation with U2AF65, we determined the 2.29 Å resolution crystal structure of the U2AF65 UHM bound to phosphorylated (P)SF114-132 protein comprising the N-terminal ULM and SPSP domains (Figure 3, Table 1, Supplemental Experimental Procedures). The SPSP domain consists of two α-helices arranged in an anti-parallel coiled-coil, thereby representing a novel fold family. A DALI search (Holm and Sander, 1993) of the protein data bank (PDB) identified the coiled-coil of the endosomal sorting complex subunit Vps23 (PDB ID 2F6M (Kostelansky et al., 2006)) as the closest match among known structures, which only shares 12% sequence identity, a modest Z-score of 5.7, and 3.8 Å RMSD between 60 matching Cα-atoms. The ULM is linked to the SPSP domain by residues running anti-parallel to the first helix of the coiled-coil, and the phosphorylated SPSP motif is located at the junction of the two α-helices (Figure 3A-B). Apart from slight revisions of the secondary structure elements (Figures 3C and S3), the core SF1 ULM/U2AF65 UHM structure generally agrees with the NMR structure of the minimal SF1 ULM/U2AF65 UHM complex (Selenko et al., 2003) (RMSD 1.4 Å for 103 matching Cα atoms of PDB ID 1O0P). Additionally, the SF114-132/U2AF65 UHM crystal structure identified a hydrogen bond that is poorly defined by NMR (Liepinsh et al., 1992), between a conserved serine in the SF1 ULM (S20) and a conserved aspartate (D401) in the acidic U2AF65 UHM α-helix (α2) (Figure 4A). Unfavorable electrostatic repulsion explains the ability of S20 phosphorylation by PKG-I to dissociate the SF1/U2AF65 complex and inhibit spliceosome assembly (Wang et al., 1999).
The (P)SF114-132/U2AF65 UHM structure revealed that the phosphorylated SPSP domain embraces the U2AF65 UHM in an α-helical extension of the ULM (Figures 3 and and4).4). The C-terminal edge of the U2AF65 UHM β-sheet engages in a novel interface with the first α1-helix, preceding loop, and to a minor extent the α2-helix of SF1. The novel SF1/ U2AF65 interface is corroborated by the ΔCp and chemical shift changes for complex formation in solution. The buried surface area of the SF114-132/U2AF65 UHM structure accounts for the observed ΔCp (1400 Å2; 2200 Å2 including the ULM) (Figure S2C). Likewise, the interactions in the structure match substantial differences in the 1H-15N HSQC spectra of 15N-labeled U2AF65 UHM bound to either SF1 or SF114-132 compared with the minimal (15N)U2AF65 UHM/SF1 ULM complex (Figure S2D).
Consistent with the large negative ΔCp (Prabhu and Sharp, 2005), the interface of the SF1 SPSP domain with U2AF65 is primarily hydrophobic. In the loop preceding the coiled-coil, SF1 V39 and I40 contact U2AF65 M446 and M381/V460 residues, respectively (Figure 4C). In the SF1 α1-helix, I53 and L56 pack against U2AF65 M381 and V460, whereas SF1 L56 contacts U2AF65 V458. Otherwise, the SF1 D60 in the α1-helix forms a salt bridge with U2AF65 K453, and SF1 E108, which is the sole interacting residue in the SF1 α2-helix, engages U2AF65 R457 (Figure 4B). The SF1 R46 guanidinium group further encloses the C-terminal U2AF65 W475 in a cation-π interaction (Figure 4D). This SF1 R46 interaction induces U2AF65 W475 and the preceding F474 residue to mask U2AF65 F433, a singular example of a conserved UHM residue that would typically stack against nucleobases bound to canonical RRMs (Figure S3) (Maris et al., 2005).
Based on the structure, we investigated the importance of the novel SF1/U2AF65 interface in vivo and in vitro starting with mutation of the highly-conserved, hydrophobic I40 and I53 residues in the SPSP domain to arginines (I40/53R mutant) (Figure 1C-D). Although the I40/53R mutation did not impair SF1 as severely as W22A, the I40/53R mutation still substantially inhibited cell proliferation when transfected in siRNA-treated cells and reduced SF1 association with U2AF65. We separately replaced the R46 residue that caps the U2AF65 tryptophan with aspartate, and found that this R46D mutation substantially reduced the interaction with U2AF65 relative to wild-type SF1 (Figure 1D). The inhibitory effects of the I40/53R and R46D mutants demonstrate the functional importance of the extended SF1/U2AF65 interface. Furthermore, these mutations prevented phosphorylation from affecting the association of SF1 with U2AF65, which suggests that the effects of phosphorylation depend on the integrity and coordination among different structural elements in the N-terminus of SF1.
The phosphorylated SPSP sequence is located in a loop connecting the two anti-parallel α-helices of the SF1 coiled-coil (Figure 2A, Figure 5A). Three conserved arginines, located across from the phosphates either in the loop preceding the α2-helix (R93, R97) or in the first turn of this helix (R100), enclose the phosphorylated (P)S80 and (P)S82 serines of the SPSP motif in a positively-charged, molecular ‘claw’. In addition, the positively-charged side chains of K104 and R79 are localized near the phosphates. Interactions with the (P)S80 phosphoryl group stabilize the R79 conformation, such that the hydrophobic portion of the R79 side chain packs against the well-defined R97 side chain. The absence of direct contacts between the U2AF65 UHM and the phosphorylated serines suggested that phosphorylation instead could serve as a molecular buttress to stabilize the fold of the SF1 coiled coil. To test this possibility, we compared the thermostability of the unmodified and phosphorylated SF1 domain by monitoring changes in the circular dichroism spectra as a function of temperature (Figure S4C). We found that SPSP phosphorylation significantly increases the thermostability of the SF1 domain, consistent with phosphorylation-dependent stabilization of the SF1 coiled-coil structure for U2AF65 UHM interactions.
To understand the local conformational changes induced by SPSP phosphorylation, we determined the complementary structure of the unphosphorylated SF1 domain (residues 26-132, SF126-132) at 2.48 Å resolution (Table 1). Whereas the coiled-coil region remains essentially the same (RMSD 0.50 – 0.68 Å between Cα atoms of residues 46-66/97-115 of the unmodified versus phosphorylated structures, range given among crystallographically independent copies), the conformation of the SPSP loop differs dramatically (Figure 5B). In contrast to phosphorylated SF114-132 (Figure 5C), residues immediately preceding the unmodified SPSP motif (residues 74-81) exhibit little detectable electron density in the absence of phosphorylation (Figure 5D). Whereas S80 is absent in all four SF126-132 copies in the crystallographic asymmetric unit, the S82 residue is absent in two of the copies and exposed to solvent in the other two molecules. Analysis of the SF126-132 crystals by SDS-PAGE confirmed that the crystallized protein is intact (Figure S4B). Although the structural disorder potentially could be due to the lack of an interaction partner rather than phosphorylation, the SPSP-containing loop does not directly participate in the interface with the U2AF65 UHM. Further, SF1 residues 74-81, R93, R97, or R100 are free of crystal packing contacts among the structures. As such, the unphosphorylated SPSP motif appears to be highly flexible. The release of R97 from interactions with the (P)S80 phosphoryl group leads to disorder in the arginine side chain, as well as surrounding residues such as the R79 side chain. Taken together, phosphorylation appears to induce local folding of residues preceding and partially including the SPSP motif.
We next tested the in vivo and in vitro importance (P)S80 and (P)S82 interactions with the arginine ‘claw’ by mutating the R93, R97, and R100 arginines of SF1 to glutamate (R93/97/100E mutant) (Figure 1). The R93/97/100E mutation of SF1 reduced cell proliferation to a comparable level as preventing SPSP phosphorylation with the S80/82A mutation. In the absence of phosphorylation, the U2AF65 interactions of the R93/97/100E mutant remained similar to the wild-type SF1. Instead, the R93/97/100E mutation interfered with the ability of SPSP phosphorylation to enhance U2AF65 interactions, which were reduced drastically when comparing the phosphorylated (P)R93/97/100E and wild-type (P)SF1 proteins. The combined results from these structural and functional experiments demonstrate that the arginine ‘claw’ is imperative for sensing and transmitting the phosphorylation state of the SPSP motif for formation of the SF1/U2AF65 complex.
To investigate whether phosphorylation alters the overall conformation of the early 3′ splice site complex, we used small-angle X-ray scattering (SAXS) to characterize the solution shapes of nearly full length phosphorylated and unphosphorylated SF1 (residues 1-255) and U2AF65 (residues 148-475) either as protein complexes or bound to a prototypical 3′ splice site from the adenovirus major late (AdML) pre-mRNA (Figures 6 and S5-S6, Table S1). The phosphorylation state, stoichiometry, and monodisperse nature of the samples were verified as described in the Supplementary Experimental Procedures and shown in Figures S5-S6. Since the bound RNAs contribute approximately 10% of the total scattering mass, the scattering data primarily reflects the conformations of the protein components.
Phosphorylation of the SF1 SPSP motif induced striking conformational changes in the (P)SF1/U2AF65/RNA complex. Whereas phosphorylation of the SPSP motif already causes the maximum dimensions (Dmax) determined in pairwise distance distribution [P(r)] functions to decrease slightly (from 150 Å for SF1/U2AF65 to 135 Å for (P)SF1/U2AF65 complexes, phosphorylation significantly compresses the (P)SF1/U2AF65/RNA complex by 25 Å to a maximum dimension of 110 Å (Figure 6C, Table S1). We further investigated the nature of the conformational changes in three-dimensions by comparing averaged and filtered ab initio restorations of the scattering data (Figures 6D and S6). From a relatively ellipsoidal arrangement of tandem domains in the SF1/U2AF65 or (P)SF1/U2AF65 complexes, the (P)SF1/U2AF65/RNA complex converts to a nearly uniform ‘C’-shape following SPSP phosphorylation. These differences demonstrated that SPSP phosphorylation exacerbates global conformational changes in the ternary (P)SF1/U2AF65/RNA complex.
In this study we have established that a novel coiled-coil structure within the SF1 SPSP domain provides an extended interface to bind U2AF65, which is required for mammalian cell proliferation. Phosphorylation of the SPSP motif in the coiled-coil domain induces local folding and induces global rearrangements in the SF1/U2AF65/RNA complex. Taken together, these structural and functional results support a model whereby phosphorylation indirectly facilitates SF1/U2AF65 interaction by ordering the SPSP motif, which in turn favors the coordinated contribution of the N-terminal ULM peptide and coiled-coil SF1 domains at the U2AF65 interface. This phosphorylated extension of the SF1/U2AF65 UHM interface beyond the short ULM epitope raises the possibility that domains adjoining the minimal U2AF65, SF1, and SF3b155, or other, as yet uncharacterized ULMs, could serve common roles in regulating the UHM network, or more broadly pre-mRNA splicing factor activity.
The local disorder-order transition and ensuing global conformational changes of the phosphorylated SPSP motif of SF1 bear striking resemblance to the well-characterized example of protein kinases (reviewed in (Taylor and Kornev, 2011)). Phosphorylation of protein kinase activation loops is a rare structural view of protein conformational changes induced through this modification. As exemplified by the prototype of cAMP-dependent protein kinase, phosphorylation shifts a disordered or variable loop to adopt a well-defined structure that is conserved among diverse families of protein kinases. Localized folding of the kinase activation loop depends on electrostatic interactions between the phosphorylated side chain and a key arginine residue. In turn, the phosphorylation-dependent relocation of the arginine shifts neighboring residues and promotes inter-domain closure to achieve active conformation of the kinase. As revealed by our structures of the phosphorylated SF1/U2AF65 UHM complex compared to unphosphorylated SF1, the activation of protein kinases is analogous to SPSP phosphorylation in the SF1/U2AF65 complex. We have shown that phosphorylation of SF1 induces (i) local folding of the SPSP motif, (ii) arginine side chain relocation, and (iii) global domain closure in an analogous manner to protein kinase activation.
The ‘C’-shaped morphology of the phosphorylated SF1/U2AF65/RNA complex observed here is consistent with a bent conformational model of the 3′ splice site during the early stages of spliceosome assembly (Figure 6E). Our model accounts for known chemical interactions, as well as functional requirements, and places these features into a physical context. The fact that the 5′ and 3′ splice sites communicate in the early splicing factor complex is exemplified by the association of the U1 snRNP at the 5′ splice site, which facilitates association of U2AF65 (Michaud and Reed, 1993) and subsequently the U2 snRNP at the 3′ splice site (Barabino et al., 1990). Moreover, the close proximity of the branch-point sequence, the 5′ splice site, and the 3′ splice site during early spliceosome assembly has been postulated based on hydroxyl radical probing (Kent and MacMillan, 2002). Association of U2AF65 with the assembling spliceosome appears to provide sufficient driving force to promote interactions among sequences in the 3′ part of the intron (Kent et al., 2003). Accordingly, the RS domain of U2AF65 contacts pre-mRNA sequences located both upstream and downstream of the 3′ splice site (Kent et al., 2003; Shen and Green, 2004; Valcarcel et al., 1996). Significantly, the ’C’-shaped conformation of the (P)SF1/U2AF65/RNA complex would juxtapose the 5′ and 3′ boundaries of the splice site in preparation for nucleophilic attack by the branch point sequence in the first catalytic step of pre-mRNA splicing (Figure 6E). Our data demonstrate that phosphorylation of the SF1 SPSP motif works synergistically with other domains or RNA to promote the bent organization of the SF1/U2AF65/3′ splice site complex. An important future goal is to determine the molecular pathway whereby SPSP phosphorylation prepares the (P)SF1/U2AF65/3′ splice site conformation for the early stages of pre-mRNA splicing.
NIH3T3 cells were maintained in DMEM containing 10% fetal calf serum in a humidified incubator and 5% CO2 at 37 °C. The cells were seeded in 35 mm dishes at a density of 30,000 cells/dish with 2.5 mL of medium and transfected the next day (Day 1) by adding 30 pmol of siRNA and 4 μL of Lipofectamine™ RNAiMAX in 0.5 mL of Opti-MEM™ following the manufacturer’s instructions (Invitrogen). On Day 4, cells were trypsinized and counted after Trypan blue staining. Extracts were prepared from cell pellets and analyzed by SDS-PAGE and immunoblotting (Figure S1B). The siRNAs used were all Stealth™ siRNAs from Invitrogen: siSF1-1, 5′-GCTTCGAGAGTTGGCTCGCTTGAAT; siSF1-2 5′-GCTCAGGATAAAGCACGGATGGATA; siSF1-3, 5′-GAGAAGGAATGCAACGCCAAGATCA; siCTL1, GFP Reporter Control (Invitrogen); siCTL2, 5′-GCUAGAGUGGUCGCUCGUUGCUAAU.
NIH3T3 cells were seeded in 35 mm dishes at a density of 50,000 cells/dish and transfected with siRNA siSF1-1 (Day 1). On Day 4 the cells were split and reverse transfected with siSF1-1 at a density of 5,000 cells/100 μL of medium per well in 96 wells plates. On Day 5, cells were transfected with SF1 (Manceau et al., 2006) or mutant expression plasmids with Lipofectamine™ 2000. Cells were incubated with siRNA siSF1-1 for a third time on Day 8. On Day 11, cells were stained with CellTiter 96® AQueous One Solution Cell Proliferation Assay (Promega) for 1 h at 37 °C in a 5% CO2 humidified incubator and absorbance at 490 nm was measured on a Victor3 multilabel counter (Perkin Elmer).
Human U2AF65 (residues 148-475 of NCBI RefSeq NP_009210), ΔRRM1 (residues 258-475), UHM (residues 375-475), SF1 (residues 1-255 of NCBI RefSeq NP_004621), SF114-132 (residues 14-132), and SF126-132 (residues 26-132) were expressed as recombinant GST-tagged proteins using Escherichia coli strain BL21 cells. The GST-tag was cleaved using PreScission Protease™ (GE Healthcare). The SF1 proteins were phosphorylated using KIS kinase in similar conditions as previously described (Maucuer et al., 2000). Phosphorylation was verified by shifted mobility on SDS-PAGE gels and by Pro-QR Diamond phosphoprotein gel stain (Invitrogen) (Figures S1C, S4A, and S5A). Details of recombinant protein production are provided in the Supplementary Experimental Procedures.
Before performing the pull-down assays, the GST-fusion proteins were incubated with or without purified KIS kinase in kinase assay conditions as previously described (Manceau et al., 2006). The phosphorylation efficiency was monitored by SDS-PAGE and Pro-QR Diamond phosphoprotein stain (Invitrogen) (Figure S1D). The GST-SF1 fusion proteins were immobilized on glutathione beads and incubated with 293T cell extracts for pull-down of endogenous U2AF65. Proteins were detected by immunoblotting. Details of the GST pull-down procedures are provided in the Supplementary Experimental Methods.
Procedures for heat capacity change determination and NMR analyses are provided in the Supplementary Experimental Methods.
The (P)SF114-132/U2AF65 UHM structure was determined by a three-wavelength Se-Met multiwavelength anomalous dispersion (MAD) experiment resulting in a figure-of-merit of 0.68 before density modification. A high-resolution data set from a separate crystal was used for structure refinement (Table 1). The SF126-132 structure was determined by molecular replacement using Phenix (Adams et al., 2010). Details of protein crystallization and structure determination are provided in the Supplementary Experimental Methods.
SAXS sample preparation and data collection at the Advanced Light Source followed procedures as described (Gupta et al., 2010). Samples were monodisperse and free of detectable aggregation based on scattering curves, Porod volumes, and linear Guinier plots (Figure S5D). A separate series of SAXS data were collected at the Cornell High Energy Synchrotron Source for independently prepared (P)SF1/U2AF65 and (P)SF1/U2AF65/RNA complexes, and produced similar results (Figure S6). Details of the SAXS procedures are provided in the Supplementary Experimental Methods. Molecular dimensions derived from the SAXS data and discrepancy values of the fitting procedure are reported in Table S1.
We are grateful to G. Hura, R. Gillilan, and J. Jenkins for advice with SAXS and M. Sattler for sharing NMR data. This work was supported by grants from the National Institutes of Health (NIH) R01 GM070503 (to C.L.K.) and R01 GM035490 (to M.R.G.), and by the Institut National de la Santé et de la Recherche Médicale (INSERM), the University Pierre et Marie Curie, and The Brain and Behavior Research Foundation (to A.M.). Crystallographic data were collected with support of NIH NCRR grant S10 RR026501 in-house (to J.E.W.) and U.S. DOE, NIH grant P41RR001209, and NIGMS for the Stanford Synchrotron Research Laboratory. SAXS data were collected at the SIBYLS beamline of the Advanced Light Source, Lawrence Berkeley National Laboratory, which is supported by NIH NCI Grant CA92584 and U.S. DOE Grant DE-AC02-05CH11231, and at the G1 beamline of the Cornell High Energy Synchrotron source, which is supported by the NSF grant DMR-0936384 and the NIH/NCRR grant RR-01646. W.W., A.G., K.R.T., A.M., and V.M. performed experiments. C.L.K., J.E.W., W.J.B., S.D.K., and A.M. analyzed experiments. C.L.K, A.M., and M.R.G. designed the experiments and wrote the manuscript with input from W.W, V.M, S.D.K., and J.E.W.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Present addresses: Dept. Biochemistry & Molecular Biophysics, University of Chicago, Chicago, IL 60637 (A.G.); Lane Center for Computational Biology, Carnegie Mellon University, Pittsburgh, PA 15213 (K.R.T.); Hauptman-Woodward Biomedical Research Institute, Buffalo, NY 14203 (W.J.B.)
Coordinates and structure factors have been deposited in the Protein Data Bank with accession codes 4FXW for (P)SF114-132/U2AF65 UHM and 4FXX for SF126-132. The SAXS data and models have been deposited in BIOISIS with the accession codes reported in Table S1.
Supplementary information includes six figures, one table and Supplementary Experimental Procedures and can be found with this article at doi:TBA