The mammalian signal recognition particle (SRP) is an 11S cytoplasmic ribonucleoprotein that plays an essential role in protein sorting. SRP recognizes the signal sequence of the nascent polypeptide chain emerging from the ribosome, and targets the ribosome–nascent chain–SRP complex to the rough endoplasmic reticulum. The SRP consists of six polypeptides (SRP9, SRP14, SRP19, SRP54, SRP68 and SRP72) and a single 300 nucleotide RNA molecule. SRP9 and SRP14 proteins form a heterodimer that binds to the Alu domain of SRP RNA which is responsible for translation arrest. We report the first crystal structure of a mammalian SRP protein, that of the mouse SRP9/14 heterodimer, determined at 2.5 Å resolution. SRP9 and SRP14 are found to be structurally homologous, containing the same α‐β‐β‐β‐α fold. This we designate the Alu binding module (Alu bm), an additional member of the family of small α/β RNA binding domains. The heterodimer has pseudo 2‐fold symmetry and is saddle like, comprising a strongly curved six‐stranded amphipathic β‐sheet with the four helices packed on the convex side and the exposed concave surface being lined with positively charged residues.
The mammalian Signal Recognition Particle (SRP) plays an essential role in targeting of secretory and membrane proteins to the rough endoplasmic reticulum (RER) membrane (for reviews, see Walter and Johnson, 1994; Lütcke, 1995). Targeting occurs co‐translationally and translocation across the RER membrane begins before polypeptide synthesis is complete. The SRP acts in three distinct ways: (i) it binds the signal sequence of the nascent polypeptide to be translocated, which is exposed on the surface of the translating ribosome; (ii) it temporarily retards the nascent polypeptide from further elongation; and (iii) it mediates docking of the SRP–ribosome–nascent polypeptide chain complex to the RER membrane via the heterodimeric SRP‐receptor (SR). With the engagement of this machinery, the SRP is detached from the complex and recycled, and co‐translational translocation proceeds. GTP hydrolysis plays an important role in the SRP cycle, one SRP protein and both subunits of the SR contain G‐domains. The GTPase cycle of SRP, modulated by the ribosome, provides the regulatory link between translation and translocation machineries (Bacher et al., 1996).
In vitro studies show that mammalian SRP9 and SRP14 proteins form a tight heterodimer SRP9/14 in the absence of SRP RNA and bind specifically to a region of the SRP RNA which includes both the 3′ and 5′ ends (Figure 1) (Strub et al., 1991). The binding of SRP9/14 to the SRP Alu RNA is stoichiometric (Walter and Blobel, 1983; Bovia et al., 1994), of high affinity (<0.1 nM; Janiak et al., 1992) and independent of other SRP proteins (Strub and Walter, 1990). The part of SRP comprising SRP9/14 complexed with RNA forms a distinct structural domain known as the Alu domain due to the homology of the RNA sequences with the Alu family of repetitive DNA sequences and the small cytoplasmic Alu RNAs (scAlus) (Weiner, 1980; Chang et al., 1996). The Alu domain of SRP mediates the specific pause(s) in the synthesis of nascent ER‐targeted proteins whose signal sequence has been bound by SRP54 (Siegel and Walter, 1988). Both the mechanism and the functional rationale for the elongation arrest activity are unknown.
For structural and functional studies a fusion protein, denoted SRPΦ14‐9, has been constructed using SRP9 and SRP14 from Mus musculus, which can functionally replace the SRP9/14 heterodimeric subunit in the SRP (Bovia et al., 1994). SRPΦ14‐9 binds SRP Alu RNA and functions, indistinguishably from wild‐type, in elongation arrest and release of elongation arrest (Bovia et al., 1994). Here, we describe the crystal structure of the SRPΦ14‐9 fusion protein at 2.5 Å resolution and discuss features of the structure which may be relevant to the RNA binding and heterodimerization properties of SRP9/14.
Results and discussion
Solving the SRP9/14 structure by MIR methods
SRPΦ14‐9 was purified and crystallized as described elsewhere (Birse et al., 1996). The structure of the SRP9/14 heterodimer was determined by multiple isomorphous replacement (MIR) methods using mercury, platinum and selenium as heavy‐atom derivatives (Table I and Materials and methods). All data were collected on flash‐frozen crystals at 100 K. Solvent‐flattened electron density maps calculated to 2.8 Å resolution using phases from the three heavy‐atom derivatives were suitable to trace the SRP9/14 polypeptide chain. The structure has been refined to 2.5 Å resolution giving a model with a final Rcryst = 24.8% and Rfree = 29.9% with excellent stereochemistry (see Materials and methods).
The main chains of SRP9 and SRP14 proteins both fold into three β‐strands and two α‐helices with a α1‐β1‐β2‐β3‐α2 connectivity. As a heterodimer, the two proteins form a six‐stranded anti‐parallel β‐sheet stacked against the four α‐helices with pseudo 2‐fold symmetry (Figure 2A). The structure is compact with a central hydrophobic core sandwiched between the β‐sheet and the four α‐helices. Both the β‐sheet and the four α‐helices are amphipathic, having predominantly non‐polar side chains interacting in the hydrophobic core and polar residues in solvent‐exposed regions. The four α‐helices are inclined by ∼27° relative to the β‐strands. The anti‐parallel strands (β1, β2 and β3) are connected via short hairpin loops except for the larger loop (22 residues) between β1 and β2 in SRP14, which is partially disordered in the crystal structure. The 10 C‐terminal residues of SRP9 extend outward from the heterodimer, having well‐defined electron density, forming a long arm that makes contact with the β‐sheet of a symmetry‐related molecule. The four N‐terminal residues of SRP14 form a short β‐strand (14βN) at one end of the β‐sheet, parallel to 14β3. At the other extremity of the β‐sheet, clear electron density is observed for an eighth strand anti‐parallel to 9β3, but with no visible connections to any other part of the molecule. The shape of the electron density for the side chains and stereochemical considerations suggest that this most likely corresponds to five residues of the linker peptide in the SRPΦ14‐9 fusion construct (QGGEQK). The four N‐terminal residues of SRP9 are disordered. They may have been displaced by the linker peptide and thereby prevented from forming a parallel strand analogous to 14βN. There is no electron density for the lysine‐rich C‐terminus of SRP14, nor the artificial N‐terminal Φ extension of the fusion construct.
The heterodimer interface is formed by the anti‐parallel strands 9β1 and 14β1 and the two anti‐parallel helices 9α2 and 14α2. Interdigitation of hydrophobic residues from these strands and helices and six main chain hydrogen bonds between 9β1 and 14β1 (residues 27–31 of SRP9 and 26–30 of SRP14) are the major contributions to the dimer interface stability. In addition, there is a hydrogen bond between the side chains of SRP9‐His66 and SRP14‐Tyr83 buried within the hydrophobic core. Outside the hydrophobic core, additional SRP9–SRP14 interactions are made between the peptide 91–95 near the C‐terminus of SRP14 which wraps along the edge of SRP9 in the vicinity of the loops between 9β1 and 9β2 and 9β3 and 9α2. SRP14‐Lys95 makes two main chain hydrogen bonds to SRP9‐Asp54 and SRP9‐Ala56 and a salt bridge with SRP9‐Asp54, and the main chain amide of SRP14‐Gly93 is hydrogen bonded to the hydroxyl group of SRP9‐Tyr31. Also, SRP14‐Leu94 is inserted into the hydrophobic core. The α‐helices provide additional dimer stability through the polar side chain interactions of SRP9‐Glu63 with SRP14‐Arg88, and SRP14‐Gln80 which hydrogen bonds with SRP9‐Ser67 and SRP9‐Arg71. Finally, the side chains of SRP9‐Arg32 and SRP14‐Ser25 also form a hydrogen bond.
Structural homology from dissimilar sequences
SRP9 and SRP14 are structurally homologous but differences are observed in the loop between strands β1 and β2 (residues 33–54 in SRP14) and the offset angles between the two α‐helices (∼15° in SRP9 and ∼7° in SRP14). The root mean square (r.m.s.) distance of the α‐carbon positions between α1, β1, β2 and β3 (38 residues) of the two polypeptides is 1.0 Å. Using the structure, we have constructed a multiple sequence alignment for SRP9 and SRP14 proteins from various eukaryotes (Figure 3). Although the number of identical residues between SRP9 and SRP14 is very low, the pattern of hydrophobic residues which gives rise to the hydrophobic core is conserved and to a lesser extent, the positions of solvent‐exposed arginines and lysines. A section of the longer loop between strands β1 and β2 loop of SRP14 is apparently important for specific SRP Alu RNA binding (N.Bui et al., 1997) consistent with the fact that it is largely disordered in the crystal structure in the absence of RNA. Another significant difference between SRP9 and SRP14 is the interaction with SRP9 of the C‐terminal region following helix α2 of SRP14, as described above. These interactions could account for the much more detrimental effect on heterodimer stability of C‐terminal truncations of SRP14 beyond −10 residues compared with SRP9. The arginine‐ and lysine‐rich extreme C‐terminus of SRP14 (Figure 2A) is also a distinguishing feature, but is disordered in the crystal structure and apparently dispensable for RNA binding (Bovia et al., 1994; N.Bui et al., 1997).
The structural homology between SRP9 and SRP14 suggests that the SRP Alu RNA binding protein may have originally been a homodimer which evolved into the presently observed heterodimer by gene duplication. This could perhaps be explained by the need to accommodate asymmetry in the co‐evolving SRP Alu RNA. This argument could also account for the situation in Saccharomyces cerevisiae where the SRP14 homologue is larger than its higher eukaryotic counterparts (Hann and Walter, 1991) yet the 5′ end of the SRP RNA (scR1; Felici et al., 1989), as in the case of Schizosaccharomyces pombe (Strub et al., 1991), is much simpler. A search through the entire yeast genome has revealed no SRP9 homologue, nor has one been characterized experimentally (Hann and Walter, 1991; Brown et al., 1994). From these observations one can hypothesize that either the yeast SRP14 functions as a monomer, possibly with its longer loop replacing SRP9, or that it forms a homodimer.
Putative RNA binding surface
Electrostatic interactions between basic residues and the phosphates of nucleic acids are generally thought to be the important forces involved in the formation and stabilization of protein–RNA complexes. We have therefore examined the nature of the SRP9/14 heterodimer surface using electrostatic potential calculations (Nichollis et al., 1993). As shown in Figure 4, these calculations strikingly reveal that the β‐sheet possesses a highly positive charged concave surface due to the abundance of exposed basic residues (SRP9‐Arg26, SRP9‐Lys30, SRP9‐Arg32, SRP9‐Lys41, SRP9‐Arg52, SRP14‐Lys31, SRP14‐Lys55, SRP14‐Arg59, SRP14‐Lys66 and loop residues SRP9‐Lys24 and SRP14‐Lys74). In the crystal structure, a phosphate ion is observed bound to SRP9‐Arg26, and SRP14‐Lys31 and SRP14‐Arg59. The proposed binding site of SRP9/14 on SRP Alu RNA (Figure 1), as determined by chemical footprinting (Strub et al., 1991), covers a substantial region of RNA and at least 86 nucleotides of RNA are required to maintain highly specific binding (O.Weichenrieder, S.Cusack and K.Strub, unpublished results). Some of this RNA is likely to be double‐stranded (Zwieb et al., 1996) and Figure 2B shows that the heterodimer has indeed the right curvature to form a saddle on a double‐stranded domain of Alu RNA. Also protruding from the concave β‐sheet surface is a solvent‐exposed aromatic residue, SRP14‐Phe27, surrounded by positively charged residues. This is a candidate to interact with an RNA base as is often observed in protein–RNA complexes (Mattaj and Nagai, 1995).
The hypothesis that important RNA binding interactions occur at the concave β‐sheet surface is consistent with experiments using the specific sulphydryl alkylating reagent N‐ethylmaleimide (NEM; Siegel and Walter, 1988). These experiments showed that SRP9 contains two sulphydryl groups accessible to alkylation in the free protein, but which are protected from modification in the RNA complex. The conclusion was that these cysteines lie along an RNA binding region in the SRP9/14 heterodimer. The structure supports this observation by showing that the two highly conserved residues, SRP9‐Cys39 and SRP9‐Cys48, are indeed on the solvent‐exposed β‐sheet surface, whereas the two cysteines of SRP14 are buried. In the structure, the SRP9‐Cys39 is covalently bound to a β‐mercaptoethanol molecule arising from the purification procedure (Birse et al., 1996).
A number of deletion and point mutations have been made on both SRP14 and SRP9 with the aim of identifying determinants for RNA binding and heterodimerization (N.Bui et al., 1997). These results do not positively identify the basic surface described above as being involved in specific Alu RNA binding, nor do they exclude it since neither the basic residues cited above, nor SRP14‐Phe27, have yet been investigated in a conclusive fashion. Furthermore, it may be that multiple mutations are required to get a significant deterioration of RNA binding ability. On the other hand, it is shown that deletions of the first part (residues 33–43) of the long SRP14 loop as well as point mutations of SRP9‐Glu15 on 9α1 and SRP9‐Asp21‐Pro22, all of which are on one edge of the molecule, are detrimental to specific Alu RNA binding without affecting heterodimerization (N.Bui et al., submitted). These results, together with the ability of Alu RNA to rescue certain dimerization‐deficient mutants, point to the fact that the RNA binds across both subunits of the heterodimer. Consideration of the size of the RNA relative to that of the SRP9/14 heterodimer, suggests that several regions, including those identified by mutation, and the β‐sheet surface most probably interact with the SRP Alu RNA, but it should be borne in mind that SRP9/14 may functionally interact, not only with Alu RNA, but also with other RNAs, for example rRNA.
The growing family of RNA binding modules
The SRP9/14 heterodimer is the latest member of a growing family of small α/β RNA binding proteins examples of which are: the ribonucleoprotein (RNP) domain (Nagai et al., 1990; Oubridge et al., 1994); the double‐stranded RNA binding domain (dsRBD) (Farrandon et al., 1994; Bycroft et al., 1995; Kharrat et al., 1995); the K homology (KH) domain (Musco et al., 1996); the coat protein of bacteriophage MS2 (Valegård et al., 1990); the translational initiation factor IF3 (Biou et al., 1995); the S1 RNA binding domain (Bycroft et al., 1997); and many ribosomal proteins (Nagai, 1996). The RNP and KH domains, as well as several ribosomal proteins (Liljas and Garber, 1995), belong to the so‐called split α‐β‐α motif differing from the dsRBD, MS2 and SRP9/14 where the sheet is a β‐meander. In aminoacyl‐tRNA synthetases, a number of different tRNA anti‐codon binding modules have also been characterized (Cusack, 1995; Moras and Poterszman, 1996). Interestingly, RNA and DNA binding modules appear to be in general structurally distinct and therefore to have evolved independently.
The α‐β‐β‐β‐α topology of SRP9 and SRP14 proteins is similar to the α‐β‐β‐β‐α motif of the dsRBDs but differs in α‐helical connectivity and stacking of β‐strands with α‐helices (Kharrat et al., 1995). A single dsRBD does not bind double‐stranded RNA (dsRNA) in a sequence‐specific manner (Farrandon et al., 1994; Bycroft et al., 1995), and it has been proposed that multiple modules may be necessary for specific RNA recognition and binding (St Johnson et al., 1992; Farrandon et al., 1994) as for instance found for dsRBDs in stau protein (Kim‐Ha et al., 1995) or KH domains in Bic‐C protein (Mahone et al., 1995). Furthermore, it has been suggested that dsRBDs interact with dsRNA solely via the loop regions (Farrandon et al., 1994; Bycroft et al., 1995; Kharrat et al., 1995), since there are few positively charged residues on the β‐sheet surface accessible to interact with dsRNA. The occurrence of a highly positively charged β‐sheet surface in SRP9/14 thus suggests that RNA binding properties of SRP9/14 differ from dsRBDs, although the results of N.Bui et al. (1997) show that one edge of the molecule is also involved in specific interactions with Alu RNA.
The MS2 protein makes contact with an RNA hairpin via residues in the 10‐stranded anti‐parallel β‐sheet formed by a dimer of the MS2 coat protein (Valegård et al., 1994). This dimer bears some resemblance to the SRP9/14 heterodimer but includes two α‐helices, one from each subunit, interdigitating in an anti‐parallel fashion. Indeed, it has been shown that individual MS2 subunits do not fully fold, the final conformation depending on their mutual association (Peabody and Lim, 1996). In the case of SRP9/14 it is not known to what extent the individual proteins fold before heterodimerization, although the determination of the crystal structure of the murine SRP9 protein alone is in progress and may give some indications (Doublié et al., 1996). The MS2 coat protein binds the RNA hairpin primarily by hydrogen bonding with residues of the β‐sheet surface facilitated by a tyrosine which stacks with a cytosine base in the RNA. In the case of SRP9/14 it seems likely, considering the large number of exposed basic residues, that contacts to the phosphate backbone may play a more important role in RNA interactions.
The RNA binding properties of SRP9 and SRP14 are reminiscent of the DNA binding properties of a group of transcriptional activators which require the formation of homo‐ or heterodimers before binding with high affinity to specific DNA sequences (Strub and Walter, 1990; Nelson, 1995). Indeed, the saddle‐like form of SRP9/14 is reminiscent of known structures such as the DNA TATA binding protein (TBP) (Nikolov et al., 1992) and the bifunctional PCD/DCoH protein (Endrizzi et al., 1995; Ficner et al., 1995). Furthermore, it has been shown that certain dimerization‐deficient mutants of SRP14 can be rescued by Alu RNA binding (N.Bui et al., submitted).
In conclusion, the crystal structure of the SRP Alu RNA binding SRP9/14 heterodimer reveals a novel RNA binding motif, designated Alu binding motif (Alu bm), a new member in the growing family of small α/β RNA binding proteins. It also provides the first step towards a structure‐based understanding of how the SRP Alu domain functions in elongation arrest after signal peptide binding to the S domain of SRP. In order to fulfill its function in elongation arrest, the SRP Alu domain presumably has to interfere with the ribosome or other factors involved in elongation. It is currently unknown whether these interactions are made by the protein and/or RNA moieties of the SRP Alu domain. The next steps towards further understanding this stage of translational regulation are the determination of the atomic structure of the complete SRP Alu domain; the SRP9/14 heterodimer complexed with Alu RNA, and to identify with which components of the translation machinery the SRP Alu domain interacts.
Materials and methods
Expression and purification
The SRPΦ14‐9 protein was crystallized (Birse et al., 1996) by the hanging drop method in 2.0 M NaH2/K2HPO4, pH 7.7, 2% MPD, 1.0 mM NaN3 at 4°C with a final protein concentration of 5–8 mg/ml. Crystals formed over 2–3 weeks and were typically 150×150×300 μm3 in space group P4322 with cell dimensions a = b = 69.02 Å, c = 90.44 Å. There is one SRPΦ14‐9 polypeptide per asymmetric unit. The crystals diffract to beyond 2.5 Å resolution flash‐frozen at 100 K using 30% sucrose as cryoprotectant. Derivatives were prepared by soaking crystals in pre‐equilibrated cryoprotectant‐containing drops with 2.0 mM thimerosal (C9H9HgO2SNa) or 1.0 mM K2Pt(CN)4 for ∼24 h.
Data collection and processing
A native data set was collected to 2.5 Å resolution at ID2 (High Brilliance beamline‐ESRF). Two mercury (thimerosal) derivative data sets were collected; mercury I, to 3.1 Å resolution on a Siemens rotating anode generator (EMBL‐Grenoble) and mercury II, to 2.9 Å resolution at ID2‐ESRF. The selenomethionine data set to 2.9 Å resolution was collected on ID2‐ESRF. The native, mercury and selenomethionine data sets used 30 cm Mar Research image plate detectors (MAR Research, Hamburg, Germany). A platinum derivative data set was collected to 2.5 Å resolution on BM14 (MAD beamline‐ESRF) using an image intensifier/CCD detector. Data were processed using DENZO (Otwinowksi, 1993) and MOSFLM (Leslie, 1992) and scaled using the CCP4 suite of programs (Collaborative Computing Project No. 4, 1994). According to the Matthews coefficient (Matthews, 1968), applying protein density estimations, solvent content in the tetragonal bipyramid crystals is ∼34%. A single platinum site was located in difference Patterson maps. Three mercury sites were found by cross‐Fourier methods using single isomorphous replacement (SIR) protein phases from the platinum derivative. MIR phases from platinum and mercury were used to locate six ordered selenomethionine sites (out of 10 methionines in the SRPΦ14‐9 molecule). Heavy‐atom positions were refined and MIR phases to 2.8 Å resolution were calculated using MLPHARE (Collaborative Computing Project No. 4, 1994) to produce the initial MIR electron density map. The map was further improved by solvent flattening, histogram matching and Sayres' equation using the program DM (Collaborative Computing Project No. 4, 1994).
Model building and refinement
Using DM solvent‐flattened maps and density skeletons created using the program MAPMAN (Kleywegt and Jones, 1994, 1996), the SRP9 and SRP14 polypeptide chains could be traced. Strand and helical density were clearly interpretable. A polyalanine chain was built into the density for SRP9 and SRP14, followed by the addition of 161 side chains. The characteristic density from the single tryptophan, SRP9‐Trp7, provided a starting point for model building. The mercury sites provided sequence markers to locate three cysteines in the model (SRP9‐Cys39, SRP9‐Cys48 and SRP14‐Cys56). The fourth cysteine (SRP9‐Cys39) is covalently bound to a β‐mercaptoethanol molecule (see ‘Putative Alu RNA binding surface’). The single platinum site was found coordinated between SRP9‐Arg34 and SRP9‐Lys41. Difference Fourier maps located selenomethionine density to position six methionines, SRP9‐Met23, SRP9‐Met70, SRP9‐Met73, SRP14‐Met1, SRP14‐Met81 and SRP14‐Met91 (Figure 5A). The model was refined to 2.5 Å resolution using simulated annealing, positional refinement, B‐factor refinement and manual rebuilding using the programs XPLOR (Brünger, 1992) and O (Jones et al., 1991). The model includes 77 residues of SRP9 (4–81) and 84 residues of SRP14 (1–34) and (47–97). The SRP14 loop electron density connecting 14β1–14β2 (34–47) is weak suggesting that the loop is flexible. Five C‐terminal and three N‐terminal residues of SRP9 and 13 C‐terminal residues of SRP14 have poorly defined density. In addition, the model does not include the SRPΦ14‐9 N‐terminal Φ extension (18 residues) nor the entire linker region of which only eight of 17 residues are ordered. In total, disordered regions making up 10% of the SRP9/14 heterodimer (22% of the SRPΦ14‐9 fusion protein) could not be modelled. As defined by PROCHECK (Laskowski et al., 1993), there are no residues in disallowed main‐chain torsion angle regions and only two residues in the generously allowed regions. The model includes one phosphate ion, a β‐mercaptoethanol molecule and 39 water molecules. An electron density map, illustrating a hydrophobic core region of the final refined model, is shown in Figure 5B. The final Rcryst for all data is 24.8% and Rfree is 29.9% for the SRPΦ14‐9 model.
Coordinates and structure factors for SRPΦ14‐9 referred to in this paper will be deposited in the Brookhaven Protein Database within 1 year of publication. Sequence accession numbers from GenBank are as follows: 464807, 1362938, 1363236, 201063, 586035, 1362463. Arabidopsis thaliana (Y10116), Oryza sativa (Y10118) and Zea mays (Y10117) sequences (N.Bui, N.Wolff and K.Strub, unpublished results).
We would like to thank Andy Thompson (EMBL) and Sean McSweeney (EMBL) for help on the MAD Beamline (BM14), Bjarne Rasmussen (EMBL) at the High Brilliance Beamline (ID2), Michael Wulff (ESRF) at the White Beamline (ID9) at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. We also wish to thank Sylvie Doublié for helpful discussions. K.S. and S.C. are both members of SRPNET which is supported by the European Union Training and Mobility of Researchers programme and the Swiss Government. K.S. is a fellow of the START programme of the Swiss National Science Foundation. Correspondence and requests for material should be addressed to Stephen Cusack ().
- Copyright © 1997 European Molecular Biology Organization