1.1. SRP cycle

      The SRP cycle is initiated in the cytoplasm by the low affinity binding of the particle to all translating ribosomes (Figure 1, step 1). If the newly synthesised nascent chain bears a signal sequence, SRP remains tightly associated with the ribosome-nascent chain complex through interaction of SRP54 with the signal sequence as soon as it emerges from the ribosome (Walter and Blobel, 1981b; Walter and Blobel, 1981a; Walter et al., 1981). SRP interaction with the ribosome leads to an arrest or a retardation of the polypeptide chain translation (step 2), which is referred as "elongation arrest" (Walter and Blobel, 1981b; Ibrahimi, 1987; Lipp et al., 1987; Rapoport et al., 1987; Wolin and Walter, 1989). The SRP-mediated translation arrest increases the efficiency of protein translocation by enlarging the time window during which the nascent chain is in a translocation-competent form. The SRP-ribosome-nascent chain complex is then addressed to the endoplasmic reticulum membrane through interaction with an SRP-specific receptor (SR) on the membrane (step 3 and 4) (Meyer and Dobberstein, 1980a; Meyer and Dobberstein, 1980b; Gilmore et al., 1982a; Gilmore et al., 1982b). The interaction between SRP54 and the alpha subunit of SR stimulate both proteins to bind GTP (step 4).

      At the reticulum membrane, SRP bound to its receptor is released from the ribosome-nascent chain complex (Connolly and Gilmore, 1989) and translation resumes at its normal speed (step 5). The ribosome, which is now engaged to the translocon, translates its nascent chain across the reticulum membrane; directly through the translocon (step 6) (Pohlschroder et al., 1997; Matlack et al., 1998; Johnson and van Waes, 1999; Menetret et al., 2000; Beckmann et al., 2001). SRP is released from its receptor through GTP hydrolysis that dissociates the complex (step 6) (Connolly et al., 1991; Miller et al., 1993 ). The particle returns then to the cytoplasm where it initiates new targeting rounds (step 7). Hence, by a strict coupling of protein synthesis to translocation, SRP ensures that nascent chains, destined to be inserted into the membranes, or secreted cannot misfold in the cytosol.

1.2. Structure and functions

      The mammalian particle is composed of a 300 nucleotide-long RNA and of six polypeptides named according to their apparent molecular mass: SRP9; SRP14; SRP19; SRP54; SRP68; SRP72 (for recent review see Keenan et al., 2001; Koch et al., 2003).

      The particle can be divided into two functionally distinct domains by micrococcal nuclease digestion (Ullu et al., 1982; Gundelfinger et al., 1983) (Figure 2). The S domain is the functional unit required for signal sequence recognition. It comprises of the central part of the SRP RNA, the SRP19, SRP54, SRP68 and SRP72 proteins. The other domain, the Alu domain, contains the 3' and 5' end of SRP RNA as well as the SRP9 and SRP14 proteins. The sequence of the SRP RNA Alu domain constitutes the phylogenetic precursor of the Alu family of repetitive DNA sequences in Rodents and Primates (Ullu and Tschudi, 1984).

1.2.1. The different subunits of mammalian SRP

---

1.2.1.1. SRP RNA

      Canine SRP RNA was among the first one to be discovered and is the most extensively characterised. To date the RNA has been found in all SRPs except in chloroplasts whose particle is composed of only two proteins (Keegstra and Cline, 1999). The RNA size varies from 519 nucleotides in yeast Saccharomyces cerevisiae (Hann and Walter, 1991) to 77 nucleotides in Mycoplasma mycoides (Samuelsson and Guindy, 1990). The role of the RNA in the particle remains poorly characterised. It may either serve as a scaffold to ensure proper three dimensional orientation of SRP proteins, or it may play an active role in the particle as expected by the extensive structural rearrangements it undergoes during the SRP cycle (Zwieb and Ullu, 1986; Andreazzoli and Gerbi, 1991). Based on the bacterial SRP crystal structure, the SRP RNA was proposed to play a role in signal sequence binding (Batey et al., 2000).

      Based on phylogenetic analyses, the SRP RNA can be divided into four different structural domains (I-IV) using the 300-nucleotide long 7S RNA from human SRP as reference (Gundelfinger et al., 1984; Poritz et al., 1988a). This secondary structure comprises eight helices numbered from 1 to 8 (Larsen and Zwieb, 1991) (Figure 3).

      

Figure 3: Secondary structure comparison for SRP RNAs from Bacteria to Mammals.

      Domains are numbered in Roman numerals according to Poritz et al., 1988b, helices are numbered according to Larsen and Zwieb, 1991, and missing helices are depicted in red.

      Domain I comprises two short hairpins and a short stem, forming a three-way junction, as well as a single-stranded region and it comprises helices 1 to 4. The domain I comprises the SRP9/14 binding sites (Strub et al., 1991).Interestingly, this domain is absent in almost all bacterial SRP RNAs (Larsen and Zwieb, 1991). Domain II is a long helix, helix 5, linking domain I and domain III and IV. Domain III (comprising helix 6) and domain IV (comprising helix 8) can be folded as hairpin structures. Domain IV is universally conserved and contains the majority of deleterious mutations (Althoff et al., 1994).

      Each of the bacterial homologues of canine SRP RNA is missing at least one structural domain. In Escherichia coli the SRP RNA, named 4.5S RNA, is 114 nucleotide-long and folds into a single hairpin which shares significant sequence identity with domain IV of human SRP RNA (Hsu et al., 1984; Poritz et al., 1988b; Struck et al., 1988), but lacks the structure of domain I and III. rRNA phylogeny comparisons showed that bacterial SRP RNA derived from a molecule containing at least three of the four domains of mammalians SRP RNA, and that it was reduced in size via at least three independent events throughout bacterial evolution (Althoff et al., 1994).

      In contrast to bacterial SRP RNA which shows great size diversity, archaeal RNAs closely resemble mammalian SRP RNA and are quite homogenous in size with all known examples containing around 300 nucleotides. Archaeal SRP RNAs can be folded into secondary structures with helices identical to the human RNA, except for the presence of an additional helix formed by the pairing of the 3' and 5' ends: helix 1 (Kaine, 1990). This structure is also shared with Bacillus subtilis.

      In the three fungi Saccharomyces cerevisiae, Schizosaccharomyces pombe and Yarrowia lipolytica, only one part of domain I is conserved. A severe truncation of domain I occurred subsequently to the divergence of the Fungi from the other Eukaryotes lineage, in which most of the first and second hairpin loops were lost. The ubiquity of the domain IV structure together with the presence of conserved nucleotides implies that this domain constitutes the essential core of the particle.

      Mammalian SRP RNAs are 300 nucleotides long and comprise all four domains, lacking only the archaeal specific helix 1.

      Interestingly, despite the differences in size and sequences, archaeal RNA can functionally replace Escherichia coli 4.5S RNA (Brown, 1991). This is also true for human SRP RNA, though with less efficiency, nevertheless it was shown that human SRP RNA can extend the viability of cells in which 4.5S RNA synthesis is repressed (Ribes et al., 1990).

      In the trypanosomal particle, an additional t-RNA like RNA molecule is associated with the SRP RNA; its role remains unknown (Beja et al., 1993; Liu et al., 2003).

      Electron microscopy imaging has been used to localise the SRP RNA within the particle (Andrews et al., 1987). The mass of the RNA was found to be concentrated at the two ends of the particle, suggesting that the RNA spans the length of SRP, forming an extended stem structure which serves as a backbone for SRP assembly.

---

1.2.1.2. SRP proteins

      SRP68 and SRP72

      Little information is available to date concerning the two biggest protein subunits of SRP: SRP68 and SRP72. This lack of information is mostly due to the failure of all attempts made to produce recombinant proteins. These proteins are only found in Eukaryotes, no homologues were found in Archaea despite the conservation of their binding site in the RNA.

      Both proteins have an overall basic character. In SRP68, basic amino acids cluster in a region near the amino-terminus, whereas the regions outside are weakly acidic (Lütcke, 1995). In SRP72, positively charged amino acids are found clustered in the carboxyl-terminus (Lütcke et al., 1993).

      SRP68/72 can be detached from SRP as a stable heterodimer in solutions of high ionic strength and reassembled alone with 7S RNA at reduced ionic strength (Walter and Blobel, 1983a; Scoulia et al., 1987). Limited proteolysis experiments made on canine SRP revealed that a very basic fragment near the amino-terminus of SRP68 remained bound to the particle and was hence concluded to constitute an RNA binding domain (Scoulia et al., 1987; Lütcke et al., 1993). This was confirmed by the use of SRP68 truncated mutants that positioned the RNA binding domain in the amino-terminal region of the protein, coinciding exactly with the high positively charged region found in this part of SRP68.

      Canine and yeast SRP68 proteins have a stretch of conserved amino acids near the carboxyl-terminus (Brown et al., 1994), which for the canine SRP68 is important for the assembly of SRP72 (Lütcke et al., 1993). While SRP68 alone is able to interact with SRP RNA, it seems that SRP72 is only linked to the particle through its interaction with SRP68 (Scoulia et al., 1987). This interaction of SRP72 with SRP68 involves the carboxyl-terminal part of both proteins (Lütcke et al., 1993).

      The SRP68/72 heterodimer binds the SRP RNA in the S domain independently of SRP19 and of SRP54 (Siegel and Walter, 1988b). The function of the heterodimer in the particle remains elusive, but it might be involved in translocation since inactivation of SRP68/72 by alkylation's results in SRP that has lost the ability to promote translocation but which can still arrest elongation (Siegel and Walter, 1988c; Siegel and Walter, 1988d). In the central region and near the carboxyl-terminus of the canine SRP68, three motifs have been identified which resemble loosely conserved motifs of guanine nucleotide dissociation stimulators of Ras-related GTPases (Althoff et al., 1994). SRP68 and SRP72 are detected in the nucleolus of transfected rat fibroblasts (Politz et al., 2000) and are therefore assembled early in the particle. Interestingly a 6 kDa carboxyl-terminal fragment of SRP72 is cleaved off during programmed cell death, but this does not interfere with the protein transport activity (Utz et al., 1998).

      SRP54

      The SRP54 subunit is by far the most important of all subunits. This subunit represents the core of the signal recognition particle, as illustrated by its conservation in all SRPs investigated to date, even chloroplast SRPs, which lack the RNA, have a homologue of SRP54. The protein is subdivided in three domains: The carboxyl-terminus M domain and the amino-terminus consisting of the N domain followed by the G domain. The amino-terminal part of the protein comprises of a GTP binding region, the G domain, which allows the interaction of SRP54 with the SRP receptor a subunit (SRa) on the reticulum membrane. This interaction requires both proteins to be in their respective GTP bound state (Rapiejko and Gilmore, 1992). The carboxyl-terminus of the protein is enriched in methionine residues and is responsible for signal sequence recognition and for RNA binding (Kurzchalia et al., 1986; Römisch et al., 1990; Lütcke et al., 1992; Zopf et al., 1993). In bacteria, the protein is the only one present and together with the 4.5S RNA represents a minimal particle. In eukaryotes SRP54 binding to the RNA requires the previous binding of the SRP19 subunit (Walter and Blobel, 1983a).

      Bernstein et al., 1989 postulated that the methionine residues of the M domain reside on a single face of several a helices, which form together a hydrophobic signal sequence binding pocket. The presence of these methionines endows a great structural plasticity which may explain how the particle is able to bind signal sequences of variable lengths and sequences. This methionine "bristles" model was later supported by the crystal structure of the Thermus aquaticus M domain (Keenan et al., 1998), revealing a domain consisting of four a helices organised around a small hydrophobic core and a large hydrophobic groove formed by three of the a helices together with a flexible loop of 19 amino acid residues (Figure 4).

      

Figure 4: Crystal structure of Thermus aquaticus Ffh M domain from Keenan et al., 1998.

      The groove is lined almost entirely with large hydrophobic amino acids, and is clearly large enough to bind signal peptides that are in the a helical conformation. Bacterial and human SRP54 M domains contain both a hydrophobic pocket proposed to bind the signal peptide, but significant differences are observed. In Ffh, as aforementioned, the groove binding the signal peptide is wide and short and binds a loop (Keenan et al., 1998), whereas in human SRP54 M domain this groove is deep and elongated and binds a helices (Clemons et al., 1999).

      The M domain is also responsible for the binding of SRP54 to the SRP RNA. This is initiated by a conserved, arginine-rich helix-turn-helix motif located opposite the hydrophobic groove, and the binding of SRP54 to SRP RNA stabilises the structure of the M domain (Kurita et al., 1996). The bacterial protein Ffh uses the first helix, as well as the turn, to recognise a specific minor groove structural element in the SRP RNA. The interaction is realised through non-canonical base pairs and through one potassium ion (Batey et al., 2000). RNA binding is achieved in a similar manner in bacteria and human. In fact the comparison of both structures revealed that the majority of RNA-protein contacts are conserved from bacteria to human and this despite the differences in RNA secondary structure and in the protein sequences (Batey et al., 2000; Kuglstatter et al., 2002).

      Isolated M domains exhibit a lower affinity for signal sequences than the intact protein, which suggests that the NG domain of the SRP54 subunit also plays a role in signal sequence recognition (Zopf et al., 1993). Crystal structures are available for Thermus aquaticus (Freymann et al., 1997) and for Escherichia Coli (Montoya et al., 1997). These structures show that the NG domain possesses a subdomain not found in any other GTPase. It facilitates the formation of a network of interactions within the active site residues, which stabilise the nucleotide-free state of the protein. The N domain is a four-helix bundle which is closely associated with the G domain. The function of this N domain is unknown, but mutations in it result in a loss of signal sequence recognition (Newitt and Bernstein, 1997). A function of the N domain might be to link the binding of external factors to the SRP GTPase cycle (Lu et al., 2001; Millman et al., 2001). The G domain contains four conserved sequence motifs involved in nucleotide binding (Bourne et al., 1991) and it has a b/a fold with a five-stranded b sheet surrounded by a helices (Figure 5). A characteristic feature of the SRP subfamily of small GTPases is a unique insertion sequence within the G domain, called insertion box (Ibox), which has an aba structure (Freymann et al., 1997). This insertion is also present in the G domain of the SRP receptor and has been implicated in nucleotide exchange as well as in SRP-SR interactions (Freymann et al., 1997; Moser et al., 1997).

      

Figure 5: Thermus aquaticus Ffh NG domain (Freymann et al., 1997).

      SRP54 and SRa interact primarily through their respective NG domains. Neither SRP RNA, nor the SRP54 M domain are required for the stimulation of the GTPase activity in vitro (Macao et al., 1997; Peluso et al., 2000). One of the conserved GTPase motifs of the G domain, which interacts with the guanine base of the bound nucleotide, is adjacent to the NG domain interface, suggesting that one function of the N domain could be to sense or to regulate the nucleotide-state of the G domain. One function of the I-box domain could be to sequester residues away from the active site prior to formation of the SRP-SR complex. In this way, the GTP-dependent formation of the complex could be coupled to conformational changes leading to reciprocal stimulation of GTP hydrolysis in SRP54 and SRa (Moser et al., 1997; Jagath et al., 2000).

      Purified complexes of SRP54-SRa and of Ffh-FtsY are capable of multiple rounds of GTP hydrolysis in absence of any additional component (Miller et al., 1993; Miller et al., 1994). Hence no external GTP exchange factor (GEF) is required for the SRP-related GTPase, the stabilisation of the nucleotide-free state (a hallmark of GEF activity) seems to be achieved intrinsically in SRP GTPases, as suggested by the comparison of GDP-bound structure of Ffh NG domain (Freymann et al., 1997; Freymann et al., 1999) and of the Apo structure of Escherichia Coli FtsY (Montoya et al., 1997).

      Interestingly, the binding of signal peptides to Escherichia Coli Ffh destabilises both M and NG domains, suggesting that both domains of SRP54 are in physical contact with the peptide (Zheng and Gierasch, 1997).

      Consistent with its critical role in SRP functions, the SRP54 subunit is found exclusively in the cytoplasm and might be assembled last into the particle (Grosshans et al., 2001).

      SRP19

      The SRP19 protein subunit is present in all organisms that contain helix 6 of SRP RNA (Zwieb and Larsen, 1997), which has been identified as the primary protein binding site. Helix 6 is closed by a tetranucleotide GNAR hairpin loop (where N represent any nucleotide and R either G or A); the adenosine in the third position is strictly conserved and is essential for hSRP19 binding (Zwieb, 1992). The protein is able to bind alone to SRP RNA (Lingelbach et al., 1988) and can be found in the nucleus (Politz et al., 2000). In addition to its essential role in mediating SRP54 binding, no other functions of SRP19 have been identified so far (Walter and Blobel, 1983a). The mediation of SRP54 binding is achieved by an SRP19-induced conformational change in the RNA (Gowda and Zwieb, 1997).

      No SRP19 has been found in bacterial SRP, but the protein is present in all Archaea characterised and sequenced so far (Eichler and Moll, 2001). In Archaea, like in Eukaryotes, SRP19 seems to be required for assembly of SRP, causing also a conformational change in the SRP RNA which favours the binding of SRP54 (Diener and Wilson, 2000).

      

Figure 6: Human SRP 19 X-ray structure resolved byWild et al., 2001.

      SRP19 has been structurally characterised from human (Wild et al., 2001; Kuglstatter et al., 2002) and from two archaeal species: Archaeoglobus fulgidus (Pakhomova et al., 2002) and Methanococcus jannaschii (Hainzl et al., 2002; Oubridge et al., 2002). These different structures revealed that the protein belongs to the a/b folding class of RNA-binding proteins. The core of the protein is structurally conserved and comprises three anti-parallel b strands packed against two a helices and four turns situated at two extended regions. (Figure 6)

      The crystal structure of human SRP19 bound to a distal fragment of helix 6, which includes the GNAR tetraloop has been resolved at 1.8 Å resolution (Wild et al., 2001). It shows that the protein-RNA recognition is achieved without direct base recognition. SRP19 binds the major groove of helix 6-tetraloop and the minor groove of helix 8 and clamps the two tetraloops together in agreement with biochemical studies (Zwieb, 1992; Zwieb, 1994). Binding of SRP19 to the RNA promotes the formation of an intricate network of hydrogen bonds that clamps the tetraloops of helices 6 and 8 together, resulting in both helices laying side by side (Figure 7). The juxtaposition of both helices upon binding of SRP19 stabilise the interaction of SRP54 M domain with the RNA, explaining why SRP19 binding is required for the interaction of SRP54 with the RNA.

      

Figure 7: Methanococcus jannaschii SRP19 bound to SRP RNA S domain (Oubridge et al., 2002).

      The crystal structure of a tertiary complex comprising the human SRP19, the M domain of human SRP54 and the S domain of 7SL RNA showed no direct contact between SRP19 and SRP54 (Kuglstatter et al., 2002).

      SRP9 and SRP14

      SRP9 and SRP14 are the sole protein components of the SRP Alu domain; in Mammals they exist as a heterodimer that forms in absence of SRP RNA (Strub and Walter, 1990).The heterodimerisation of the two proteins is required for binding to SRP RNA. Binding occurs to a region comprising both 3' and 5' ends of mammalian SRP RNA through a primary binding site consisting of a highly conserved single-stranded region with the consensus sequence UGUAA (Strub et al., 1991). The primary binding site forms a U-turn in SRP9/14-Alu RNA complex (Weichenrieder et al., 2000). This binding is stoichiometric and of subnanomolar affinity (Walter and Blobel, 1983a; Strub et al., 1991; Janiak et al., 1992; Bovia et al., 1994). The heterodimer has been shown to be implicated in the elongation arrest activity of the particle (Thomas et al., 1997; Mason et al., 2000). Truncation in the carboxyl-terminal part of SRP14 leads to a specific loss of the elongation arrest function both in vitro and in vivo. The proteins seem to be assembled early in the particle as suggested by their nuclear colocalisation with SRP68, SRP72, SRP19 and SRP RNA (Politz et al., 2000; Politz et al., 2002).

      

      

Figure 8 A et B: Human SRP9/14 heterodimer (Weichenrieder et al., 2000). SRP9 is depicted as a red labelled ribbon and SRP14 as a green labelled ribbon.

      No SRP9 and SRP14 homologues have been identified in the prokaryotic kingdom, although their binding sites are conserved in archaeal SRP RNA. In Bacillus subtilis, an additional protein component has been identified. The HBsu protein is a 10 kDa protein, originally identified as member of the histone-like protein family, and has been shown to bind the Alu domain of Bacillus subtilis SRP RNA (Nakamura et al., 1999b). In Rodents and Primates there is an excess of SRP9/14 over SRP RNA, and a fraction of the protein is bound to other cytoplasmic particles notably the Alu particles (Chang and Maraia, 1993; Chang et al., 1994; Bovia et al., 1995).

      SRP9 and SRP14 have very limited sequence homology but are indeed structural homologues. They both form a six-stranded anti-parallel b sheet stacked against four a helices (Birse et al., 1997; Weichenrieder et al., 2000; Weichenrieder et al., 2001) (Figure 8). The heterodimer interface is composed mainly of residues from b₁ strand and a₂ helix of each protein, as well as of the carboxyl-terminal part of SRP14 (Birse et al., 1997; Bui et al., 1997; Weichenrieder et al., 2000). The b sheet concave surface is highly positive due to the abundance of exposed basic residues (Birse et al., 1997). Two domains of SRP14 are not ordered in the crystals: the fifteen last amino acids of the carboxyl-terminus and an internal loop between b₁ and b₂, which was shown to be important for specific binding to SRP RNA (Bui et al., 1997) (Figure 8). Both proteins contribute to the formation of the heterodimer RNAbinding domain explaining why heterodimerisation is a prerequisite for RNA binding. Despite their structural homology the proteins differ in their requirements for dimerisation: an internal fragment of forty three amino acids in SRP9 being sufficient to allow dimerisation, whereas SRP14 tolerates only minor changes, such as removal of the internal loop region (Bui et al., 1997). A mutational analysis identified two regions critical for RNA binding, comprising a₁ and the adjacent turn in SRP9 and the first half of SRP14 internal loop (Bui et al., 1997).

      The positively charged concave b sheet of the heterodimer is the RNA primary binding site. Contacts with the RNA are achieved through basic residues interaction to the RNA backbone; no specific base-stacking interactions are found (Weichenrieder et al., 2000).

      No SRP9 homologue has been identified in yeast were the single stranded region constituting SRP9/14 primary binding site is conserved. In fact the SRP14 homologue which is present has been shown to bind to the RNA as a homodimer in Saccharomyces cerevisiae (Strub et al., 1999; Mason et al., 2000).

---

1.2.1.3. SRP receptor

      The mammalian SRP receptor is composed of two subunits SRa and SRb. Like SRP54, both subunits of SR are GTPases. This receptor is localised at the endoplasmic reticulum membrane (Gilmore et al., 1982a; Gilmore et al., 1982b). It functions as a docking protein that links the SRP-ribosome-nascent chain complex to the translocon, allowing the cotranslational targeting of membrane and/or secreted proteins. The SRa subunit is a 69 kDa protein and homologues have been found in Eubacteria, Archaea and Eukaryotes (see SRP database at: http://psyche.uthct.edu/dbs/SRPDB/SRPDB.html). It is peripherally associated to the endoplasmic reticulum membrane via interactions with the smaller SRb subunit (30 kDa), which is an integral membrane protein (Tajima et al., 1986). The GTPase domains of SRP54 and SRa are related and constitute the SRP-related GTPase subfamily, whereas SRb GTPase domain is distinct and belongs to the Arf GTPase subfamily (Miller et al., 1995). In Escherichia coli the homologue of the SRa subunit, FtsY, is the only one present. Like SRP54, FtsY is subdivided into two domains: an amino-terminal acidic domain (A) and a carboxyl-terminal GTPase domain (NG) (Montoya et al., 1997; Moser et al., 1997). Comparison of the crystal structure of Ffh and FtsY NG domains reveals a similar shape (Freymann et al., 1997; Montoya et al., 1997) (Figure 5).

      

Figure 9: Saccharomyces cerevisiae SRa/SRb complex from Schwartz and Blobel, 2003.

      The crystal structure of Saccharomyces cerevisiae SRb-GTP in complex with the SRa interacting domain has been resolved recently (Schwartz and Blobel, 2003). According to this structure, the SRb subunit consists of six b sheets surrounded by five a helices (Figure 9), sharing canonical features of the Arf GTPase subfamily as previously predicted (Miller et al., 1995). The structure revealed an intricate hydrogen bond network which ties both proteins together, allowing high affinity binding. As previously shown, the binding of SRa to SRb requires both hydrophobic and hydrophilic sequences located within the amino-terminal part of SRa (Young et al., 1995).

      In order to interact with each other, SRP54 and SRa have to be in their GTP-bound state (Rapiejko and Gilmore, 1992). In this state, both proteins act as GTPase-activating proteins for each other, this is a unique feature among GTPases (Powers and Walter, 1995). Interaction between the two subunits of SR is nucleotide dependent and requires SRb in its GTP-bound state (Legate et al., 2000). A chemical cross-link between SRb and a 21 kDa ribosomal protein has been identified (Fulga et al., 2001). This cross-link is only obtained with the GDP-bound state of SRb and suggests that the ribosome might play the role of SRb GTPase-activating protein (GAP) (Bacher et al., 1996; Bacher et al., 1999). SRa and SRP54 do not require any additional factors in order to dissociate GDP (Miller et al., 1993; Miller et al., 1994). Contrary to the case of SRa and SRP54, SRb is a rather typical GTPase and requires therefore a GAP and a GEF in order to fulfil its function. As aforementioned it is likely that the ribosome is responsible for SRb GAP activity, and the translocon could serve as GEF according to a model presented by Keenan et al., 2001.

1.3. Elongation arrest activity

1.3.2. Structure of the Alu domain

      It has been shown previously that SRP interacts with ribosomes after the transpeptidylation reaction and before the peptidyl-tRNA is translocated from the A-site to the P-site (Brown, 1989; Ogg and Walter, 1995). SRP54, which binds signal sequence is located near the polypeptide exit site, as seen by its cross-link to ribosomal proteins L23 and L29 (Pool et al., 2002; Gu et al., 2003).With regard to the shape and size of the particle observed by electron microscopy (50-60 Å wide and 220-240 Å long) (Andrews et al., 1985), the location of the S domain protein SRP54 near the exit site is consistent with a position of the Alu domain at the interface of ribosomal subunits, possibly near the A-site (Bui et al., 1997; Bui and Strub, 1999). This position of the Alu domain was confirmed by cross-linking experiments (Terzi et al., see manuscript below). At the ribosomal subunit interface the Alu domain might impair translation by interfering with tRNA and/or elongation factor functions. Hence, it was suggested that elongation arrest activity was achieved through a mimicry-based mechanism of the Alu domain with tRNA or elongation factors. However this is not supported by the Alu RNP crystal structure (Weichenrieder et al., 2000).

      

Figure 10: Alu RNP crystal structures (Weichenrieder et al., 2000).

      A: 5' domain of Alu SRP RNA complexed with SRP9/14. B: 5' domain and 3' domain SRP RNA complexed with SRP9/14. The protein primary binding site (U-turn) and the two interacting hairpin loops (L2 and L1.2) are highlighted. C: proposed model for the assembly of the Alu RNP Weichenrieder et al., 2000. Upon binding of the SRP9/14, the RNA 3' domain folds back onto the 5' domain, resulting in a closed Alu RNP conformation. SRP9 and SRP14 are shown as red and green labelled ribbon, respectively.

      The model proposed for the native Alu domain structure has no obvious structure similarities with tRNA or elongation factors. The Alu domain RNA comprises two domains: the 5' domain and the 3' domain (Figure 10). The 5' domain contains the most highly conserved region of the Alu RNA (Strub et al., 1991) and is the primary binding site of the SRP9/14 heterodimer. The crystal structure showed that the 5' domain is compactly folded into two helical stacks, between helices H1.2 and H2, connected by a central U-turn (Figure 11): a tertiary structure the authors named t-junction (Weichenrieder et al., 2000). This structure is also found in two others three-way junction RNAs, the hammerhead ribozyme (Pley et al., 1994; Scott et al., 1995) and the rRNA 23S binding-site for the ribosomal L11 protein (Conn et al., 1999; Wimberly et al., 1999).

      

Figure 11: Structure of the Alu RNA 5' domain.

      A: secondary structure of the Alu SRP RNA 5' domain. B: three-dimensional ribbon representation of A. Helices and loops are named according to Weichenrieder et al., 2000.

      The 3' domain consists of two helical segments connected by an internal asymmetric loop and contributes also to SRP9/14 binding, as previously observed by footprints (Strub et al., 1991). As suggested by their conserved potential for base pairing in Mammals, Archaea, Clostridium and Bacillus subtilis, the stem loops of the mammalian Alu RNA are connected by three tertiary interactions.

      The protein binds asymmetrically to the 5' domain, with the majority of contacts made to the concave b sheet surface of SRP14 (Figure 10A). These interactions are mediated primarily through positively charged side chains that directly contact the RNA phosphate backbone. G₂₄ and U₂₅ are key residues of the t-junction, these nucleotides are the most extensively contacted and the only base-specifically recognised nucleotides, highlighting the previous observation of G₂₄ being crucial for high efficient binding of SRP9/14 (Chang et al., 1997). In the model of the Alu domain native structure, SRP9/14 clamps the 5' and the 3' domain within its b sheet surface (Figure 10B), resulting in the two domains lying side by side. The 3' domain is packed against the 5' domain and makes extensive contacts with the concave b sheet surface of SRP9, consistent with chemical footprints (Strub et al., 1991). Most of these 3' domain contacts are made between the minor groove of the asymmetrical loop of the RNA and the b₂-b₃ loop of SRP9. According to the model, in complex with the SRP9/14, the RNA folds back on itself such that the central stem of SRP RNA, leading to the S domain of the particle, emerges from the SRP14 carboxyl-terminal side of the heterodimer (Figure 10C). A sequential assembly pathway of the SRP Alu domain has been proposed (Weichenrieder et al., 2001). After SRP9 and SRP14 heterodimerisation, the heterodimer associates with the free Alu RNA 5' domain. Binding of the protein to the U-turn rigidifies the structure and promotes the formation of tertiary Watson-Crick interactions between the loops. Subsequently, the 3' domain flips to contact a surface of the 5' domain complex composed of both SRP9 and of the 5' domain RNA. The specific behaviour of the 3' domain shed light on the previously observed requirement of flexibility between the 5' and the 3' domains, in order to achieve high affinity binding of the heterodimer to the RNA (Weichenrieder et al., 1997).

2.1. SRP14

      Despite their structural homology, SRP9 and SRP14 respond equally to the mutation of their sequences. SRP14 relies on a largely intact protein sequence for its functions and tolerates only minor changes in its sequence, whereas in SRP9 an internal fragment of forty three amino acids has been shown to be sufficient for dimerisation (Bui et al., 1997). In fact only two regions of the protein can be modified without impairing the SRP14 dimerisation function: the carboxyl-terminus and an internal loop of eighteen amino acid residues located between b₁ and b₂. Mutation of any other region of SRP14 leads to a dimerisation defect, even if the regions modified are not part of the dimer interface; amino and carboxyl-terminal a helices for example cannot be removed.

2.1.1. SRP14 carboxyl-terminus

      We know from previous studies in our laboratory, that the removal of amino acid residues 91110 at the carboxylterminus of murine SRP14 abolished elongation arrest activity of reconstituted SRP (Thomas et al., 1997). In the meantime, the crystal structure of an Alu RNP complex revealed that amino acid residues 93, 94 and 95 are part of the heterodimer interface (Weichenrieder et al., 2000). It is therefore conceivable that the observed defect of elongation arrest activity might be indirectly due to conformational changes occurring in the complex. Alternatively, the conserved basic amino acids lying outside of the Alu RNA binding site (Figure 12) may directly contact the ribosome. In order to reinvestigate the role of this critical region of the SRP14 protein, we engineered a human SRP14 which lacked all amino acids not ordered in the structure. We decided to truncate all amino acids of the SRP14 carboxyl-terminus that are not ordered in the crystal structure. This was achieved by the introduction of a stop codon in the protein sequence after lysine 95 (h14K95, Figure 13). Based on the structure of the Alu RNP complexes, this h14K95 comprises all the elements required for the assembly of the SRP Alu domain.

      

Figure 12: Sequence alignment of SRP14 carboxyl-terminal part. Region 96 to 105 comprises at least six basic amino acid residues in all species.

      Recombinant proteins were produced in bacteria and the heterodimeric complexes h9/14 and h9/14K95 purified to homogeneity (Materials and Methods). As expected, the truncated protein h9/14K95 could efficiently bind to SRP RNA and was able to compete with the wild type protein for assembly into SRP.

      Particles were assembled with recombinant SRP19, SRP54, SRP9/14, canine SRP68/72 and canine SRP RNA using either h9/14 or h9/14K95 (RCwt, RC14K95, Materials and Methods). The reconstituted particles were added to wheat germ translations programmed with synthetic mRNAs of bovine preprolactin, a secreted protein, and sea urchin cyclin D, a cytoplasmic protein. The samples were analysed by SDS-PAGE and quantified by phosphorescence imaging (Figure 13).

      As expected for an active reconstituted particle, the RCwt specifically inhibited the accumulation of full-length preprolactin as compared to the accumulation of full-length cyclin (Figure 13, compare lane 1 to lanes 2 and 3). In contrast, RC14K95 did not specifically reduce the preprolactin signal as compared to the cyclin signal indicating a lack of elongation arrest activity (Figure 13, compare lane 1 to lanes 4 and 5).

      

Figure 13: Truncation of SRP14 lysine 95 abolishes the elongation arrest activity.

      Particles, reconstituted with canine and recombinant SRP proteins, including wild type hSRP9/14 (RCwt) or hSRP9/14K95 (RC14K95) were assayed for elongation arrest and translocation activities. In translocation assays, SRP-depleted canine microsomes were added to the translation reactions. In elongation arrest assays, we determined the relative inhibition of preprolactin synthesis as compared to cyclin. For the processing assays, the percentage of prolactin compared to the total amount of preprolactin and prolactin was determined. Processing [%] Stand. to -RC: Amount of processing due to endogenous SRP in the microsomes (lane 1) was set to 0. Lane 1: buffer; lanes 2 and 3: 100 and 50 nM RCwt; lanes 4 and 5: 100 and 50 nM RC14K95.

      Since elongation arrest activity is dependent on the signal recognition function of SRP, we investigated the signal recognition as well as the targeting functions of the reconstituted particles. To that end we assayed their capacities to translocate preprolactin into SRP-depleted canine microsomes. RCwt and RC14K95 can promote preprolactin translocation across the membranes. Particles that are defective in elongation arrest activity exhibit only half the translocation efficiency of wild type particles (Thomas et al., 1997). Consistent with its defect in elongation arrest, we observed that the translocation activity of RC14K95 was intact, but its efficiency was reduced to half the one of RCwt (Figure 13 compare lanes 2 and 3 with lanes 4 and 5, respectively). Hence, the signal recognition and targeting functions of RC14K95 were intact, and it lacked exclusively elongation arrest activity.

      To further establish that it is the presence of the truncated protein in the reconstituted particle that caused the defect in elongation arrest activity, we set up competition experiments. SRP components including wild type h9/14 were combined at equimolar ratios in the presence of increasing amounts of the truncated heterodimer. Similar competition experiments were previously done with wild type SRP9/14 protein without affecting the elongation arrest activity of the particle, confirming that the excess of protein by itself did not affect the function of the particle (Thomas et al., 1997). The h9/14K95 is expected to compete with the wild type protein for SRP RNA assembly, thereby abolishing elongation arrest activity of the particle. Indeed increased competitor concentrations diminished the elongation arrest activity of the reconstituted particle (Figure 14). It also reduced the translocation efficiency to almost the same level as observed for particles lacking h9/14 (Figure 14, compare lanes 2 and 7). Hence, the presence of h9/14K95 in the particle specifically abolished its elongation arrest activity.

      

Figure 14: h9/14K95 competition experiments.

      A: increasing amounts of h9/14K95 were introduced into the reconstitution reactions in presence of a constant amount of h9/14, equimolar to the other SRP proteins concentration. The relative inhibition of preprolactin synthesis in presence of the reconstituted particles is shown. Lane 1: buffer; lane 2: partially reconstituted particle lacking the SRP9/14; lane 3: reconstitution with wild type h9/14; lane 4 to 7: 1x, 3x, 9x and 21x molar excess of the truncated h9/14K95 in the reconstitution assay, respectively. The relative inhibition obtained for the wild type reconstitution, in absence of any truncated h9/14K95 (lane 3) was standardised to 100%. B: same competition assay as in A, except for the presence of SRP-depleted canine microsomes in the translation reaction. Processing activity is determined as the percentage of prolactin compared to the total amount of preprolactin and prolactin. The ratio obtained with the wild type reconstitution, in absence of the truncated heterodimer, is standardised to 100%.

      The effects on elongation arrest and on translocation efficiency were lower than expected for a competitor with equal affinity for SRP RNA. This might be due to a slightly reduced affinity of particles lacking elongation arrest activity for the ribosome. Nevertheless, our results, as well as previous results from Thomas et al., 1997, clearly showed that truncation in the carboxyl-terminus of SRP14 did not abolish the capacity of reconstituted SRP to bind to ribosomes and to promote translocation. We found later that h9/14K95 was contaminated by a small amount of wild type h9/14 (below 5%), resulting from ribosomal read through at the stop codon in bacteria (Materials and Methods).

      In presence of SRP9/14, two hydroxyl radical protection sites were observed on the SRP RNA central stem (G₄₈-G₅₁ and G₅₈-G_62, Strub et al., 1991). According to the native structure model of the Alu domain, derived from the crystal structure (Weichenrieder et al., 2000), G₄₈-G₅₁ protection results from protein-induced RNA tertiary contacts, whereas G₅₈-G₆₂ protection is caused by a combination of both induced RNA tertiary contacts and direct protein contacts. Since SRP14 carboxyl-terminus might be close to the RNA central stem, we decided to investigate the structure of a particle reconstituted with our truncated heterodimer using hydroxyl radical cleavage assays. For that purpose, we used a synthetic Alu RNA consisting of 151 nucleotides (Figure 15). This RNA was designed to contain all features of the SRP RNA Alu domain.

      

Figure 15: Secondary structure of Alu151 synthetic RNA. Helices are displayed as Roman numerals and numbered according to Larsen and Zwieb, 1991.

      Wild type SRP9/14, truncated SRP9/14K95 and bovine serum albumin as a control, were incubated with [³²P]-labelled Alu151 SRP RNA under conditions allowing the binding of the proteins to the RNA. Subsequently, the RNA was cleaved by hydroxyl radicals (see Materials and Methods) and the cleavage sites were investigated on polyacrylamide sequencing gel. The hydroxyl radicals cleaved the RNA at almost every nucleotide in the control reactions with the RNA alone (Figure 16, lane R) or with RNA in presence of bovine serum albumin (lanes 3 and 4). As previously observed, the wild type complex, composed of wild type SRP9/14 bound to the RNA, exhibits two sites on the 3' domain that are protected from cleavage by the radicals (Figure 16, lanes 1 and 2, Strub et al., 1991). In presence of the truncated SRP9/14K95 heterodimer, the same protection sites are observed as for the wild type complex (Figure 16, compare lanes 5 and 6 with 1 and 3). This indicates that the complexes, formed between h9/14K95 and SRP RNA, were also able to fold into a compact structure.

      The finding that h14K95 abrogated elongation arrest activity of the particle provided evidence that the small region outside the Alu RNA-binding domain in SRP14 plays a direct role in effecting elongation arrest. A critical functional role for this region is also supported by its high conservation (Figure 12). This region, together with the SRP9/14 binding site in SRP RNA (Strub et al., 1991), represents the most highly conserved features in the Alu domain. Its basic character is consistent with a role in binding ribosomal RNA. In fact, Terzi et al. (see manuscript below), showed that the carboxyl-terminal part of the SRP14 is critical for binding to the ribosomal RNA of the GTPase associated region. The loss of a critical contact with this ribosomal RNA through truncation of this region of the SRP14, could explain the strong effect observed on elongation arrest activity.

      

Figure 16: Alu RNP hydroxyl radical footprints.

      A: hydroxyl radical cleavage pattern of SRP Alu151 RNA in presence of SRP9/14 proteins. [³²P]-labelled Alu151 RNA was incubated with wild type SRP9/14, truncated SRP9/14K95, or with bovine serum albumin, under conditions allowing complex formation. Alu RNPs were subsequently submitted to hydroxyl radical cleavage. The cleavage products were separated on an 8M Urea-6% polyacrylamide gel and displayed by autoradiography. A sequencing reaction, using RNases T₁, U₂, B. cereus, as well as an alkaline hydrolysis RNA ladder, were run in parallel to map the extension (data not shown). Regions protected from cleavage in the presence of the heterodimer are labelled I and II. Lane R: cleavage pattern of the naked RNA; Lanes 1 and 2: cleavage pattern of the RNA in presence of a 1:1 or 1:3 molar ratio of wild type SRP9/14, respectively. Lanes 3 and 4: same pattern as for lanes 1 and 2 except that bovine serum albumin replaces the SRP9/14; Lanes 5 and 6: same pattern as lane 1 and 3, except that the truncated SRP9/14K95 replaces the wild type protein. B: secondary structure of the Alu151 RNA; regions protected from cleavage in the presence of either wild type or truncated heterodimer are outlined in grey.

2.1.2. SRP14 internal loop

      Through extensive mutational analysis, Bui et al., 1997 showed that although they are structural homologues, the SRP9 and SRP14 have different requirements for dimerisation. As mentioned before, many changes in SRP9 do not affect its function, whereas SRP14 tolerates only minor changes. The internal loop between b₁ and b₂ (Figure 8) is dispensable for the hSRP14 dimerisation function and shows little primary sequence conservation, except for four basic residues at position 31, 32, 42 and 43 (Figure 17). Furthermore, most of the loop is not ordered in the protein crystal structure, residues from position 36 to 53 are not seen (Birse et al., 1997). Mutations, or removal of all four of the conserved basic residues (K31, K32, K42 and K43 in human SRP14) leads to a loss of RNA-binding activity (Bui et al., 1997); however, replacing only two of these conserved residues, failed to diminish the RNA binding activity. Since the missing part of the internal loop contains two of the four conserved basic residues, it is thought to be involved in contacting the SRP RNA. Still, in the Alu RNP crystal structure, this internal loop is also not ordered, despite the presence of the SRP RNA (Weichenrieder et al., 2000). In the context of the translating ribosome we wondered whether this internal loop of SRP14 might contribute to the elongation arrest function of the particle. To test this hypothesis we engineered a human SRP14 protein where we deleted the residues of the loop that are not ordered in the structure. Only two of the conserved basic amino acid residues (K42 and K43) were removed in this truncated protein and according to Bui et al., 1997, it should not interfere with the binding to the SRP RNA. The truncated protein (hSRP14sloop) results in the truncation of eighteen amino acids, from arginine 36 to aspartate 53, within the internal loop (Figure 17).

      

Figure 17: Sequence alignment of SRP14 internal loop between b₁ and b₂.

      Deletion of amino acids R36 to D53 leading to the hSRP14sloop construct are marked by the grey box.

      Recombinant hSRP14sloop protein was produced in bacteria and the heterodimeric complexes h9/14sloop purified to homogeneity (Materials and Methods). Particles were assembled through combination of recombinant and canine proteins, together with in vitro transcribed synthetic SRP RNA (Materials and Methods). The reconstituted particles were added to wheat germ translations programmed with synthetic mRNAs of preprolactin, and sea urchin cyclin D, as for SRP9/14K95 (see Figure 13). The samples were analysed by SDS-PAGE and quantified by phosphorescence imaging (Figure 18).

      

Figure 18: Truncation of amino acid residues R36 to D53 in the internal loop of SRP14 do not affect the elongation arrest activity of reconstituted particles.

      Particles, reconstituted with canine and recombinant SRP proteins, including wild type hSRP9/14 (RCwt) or hSRP9/14sloop (RCsloop) were assayed for elongation arrest (A) and translocation activities (B). The particles were added to an in vitro translation reaction at 100 nM final concentration. In translocation assays, SRP-depleted canine microsomes were added to the translation reactions. In elongation arrest assays, we determined the relative inhibition of preprolactin synthesis as compared to cyclin synthesis. For the processing assays, the percentage of prolactin compared to the total amount of preprolactin and prolactin was determined. Processing Stand. to -RC: Amount of processing due to endogenous SRP in the microsomes (Buffer) was set to 0. RC-9/14: wild type particles reconstituted in absence of the hSRP9/14 heterodimer. RCwt: particles reconstituted with wild type proteins; RC-10Ct: particles reconstituted with a truncated SRP14, lacking ten amino acids in the carboxyl-terminus; RCsloop: particles reconstituted with the hSRP9/14sloop protein; SD: standard deviation.

      As expected, the RCwt specifically inhibited the accumulation of full-length preprolactin as compared to the accumulation of full-length cyclin, leading to a 58% inhibition of preprolactin synthesis (Figure 18A). Particles reconstituted with a truncated SRP14, lacking ten amino acids in its carboxyl-terminus, leads to a 61% inhibition of preprolactin synthesis, corresponding to wild type particle levels, as previously shown (Bovia et al., 1994). Adding RCsloop to the translation reaction, results in a 76% inhibition of preprolactin translation as compared to cyclin (Figure 18A, lane RCsloop). Hence, its elongation arrest activity is comparable to the two controls (RCwt and RC-10Ct).

      We also investigated the capacities of the particle reconstituted with the SRP14sloop protein to translocated preprolactin into SRP-depleted canine microsomes (Figure 18B). As expected from the results obtained for the elongation arrest activities, wild type and truncated particles could promote preprolactin translocation across the membranes to similar levels, whereas a particle lacking the SRP9/14 heterodimer had only 50% translocation efficiency as compared to the wild type (Figure 18B, compare RC-9/14 with RCwt).

      These results showed that the truncation of amino acids R36 to D53 in the SRP14 internal loop between b₁ and b₂ did not affect the functions of the Alu domain; elongation arrest, and translocation are as efficient in the context of RCsloop as they are for wild type reconstituted particles. As a result we can conclude that the SRP14 internal loop does not play a critical role in the SRP mediated translational arrest.

2.2. SRP9

      The crystal structure of the Alu RNP revealed a striking feature. The RNA 3' and 5' domains are align side by side underneath the curved b sheet structure of the protein (Weichenrieder et al., 2000). In complex with the SRP9/14, the RNA of the 3' domain folds back onto the 5' domain, resulting in a closed conformation where the S domain bends up to 180° and emerges from the SRP14 carboxyl-terminus side of the heterodimer (Figure 10C). The packing of the 3' domain onto the 5' domain is consistent with the observed chemical footprints (Strub et al., 1991). It also explains the requirement for flexibility between the 5' and the 3' ends of the SRP RNA in order to achieve high affinity binding of the SRP9/14 (Weichenrieder et al., 1997). The SRP9 protein plays a key role in the folding of the 3' domain, since the folded structure is stabilised through contacts established between SRP9 and the RNA. The b₂-b₃ loop of SRP9 is inserted into the minor groove of helix H3.2 (Weichenrieder et al., 2000). In the closed conformation of the Alu RNP, the side chain of SRP9 lysine 41 is positioned between the 5' domain and the 3' domain RNA backbones. Although the crystal resolution is not sufficient to precisely map the interactions, the e-amino nitrogen of lysine 41 is close to U₂₃ of the 5' domain, and G₅₈ of the 3' domain (Figure 19).

      Weichenrieder et al., 2001 showed that mutations of lysine 41 and cysteine 39 into alanine, in SRP9, results in a complex which is specifically deficient in the Alu RNA 3' domain binding. Competition experiments with RNA constructs where the 3' and 5' ends are flexibly or rigidly bound, as well as electromobility shift assays, revealed that the complex formed with the SRP9K41A/C39A is in a more open conformation than the wild type complex. This behaviour is most likely due to the mutation of lysine 41 into alanine, since cysteine 39 is located more than 5.5 Å away from the RNA in the crystal structure and its substitution with serine does not have any effect (Weichenrieder et al., 2001).

      We decided to investigate the function of these open or closed conformations of the Alu domain in the elongation arrest activity of the particle. To that end we produced mutated SRP9 proteins in which we substituted lysine 41 into alanine, to disrupt the interaction with the RNA, or arginine, to restore the interaction (Figure 20B and C).

      A: wild type Alu RNP complex. B: mutated SRP9K41A complex. C: mutated SRP9K41R complex. The side chains of residues at position 41 of SRP9 and nucleotides U₂₃ and G₅₈ of the 3' and 5' domain are depicted as wireframe. The structure of the complexes with the mutated SRP9 (B and C) were realised with the Swiss PDB Viewer program.

      Recombinant SRP9 proteins were produced in bacteria and the heterodimeric complexes h9K41A/14 and h9K41R/14 purified to homogeneity (Material and Methods). Particles were assembled with a combination of recombinant and canine proteins, together with canine SRP RNA using either h9/14, h9K41A/14 or h9K41R/14 (RCwt, RCK41A, RCK41R, see Materials and Methods for details). The reconstituted particles were added to wheat germ translations programmed with synthetic mRNAs of bovine preprolactin, a secreted protein, and sea urchin cyclin D, a cytoplasmic protein. The samples were analysed by SDS-PAGE and quantified by phosphorescence imaging (Figure 21).

      Particles, reconstituted with canine and recombinant SRP proteins, including wild type hSRP9/14 (RCwt), hSRP9K41A/14 (RCK41A) or hSRP9K41R/14 (RCK41R) were assayed for elongation arrest and translocation activities. In translocation assays, SRP-depleted canine microsomes were added to the translation reactions. In elongation arrest assays (A), the relative inhibition of preprolactin synthesis as compared to cyclin was determined. For the processing assays (B), the percentage of prolactin compared to the total amount of preprolactin and prolactin was determined. Processing Stand. to -RC: Amount of processing due to endogenous SRP in the microsomes (Buffer) was set to 0. Reconstituted particles, as labelled on top of each lane, were added to the translation reaction at a final concentration of 100 nM.

      The wild type reconstitution, RCwt, when introduced into the translation reaction leads to a specific 96% inhibition of full length preprolactin synthesis, as compared to the cytoplasmic cyclin D (Figure 21A). Despite the fact that it is expected to weaken the binding of the 3' domain (Weichenrieder et al., 2001), the SRP9K41A protein had no significant effect as its elongation arrest activity was only slightly reduced from 96 to 88% as compared to wild type (Figure 21A, lane RCK41A). The loss of the 3' domain interaction only slightly reduces the elongation arrest activity of the particle (Figure 21A, compare the 88% inhibition obtained with RC41A to the 96% inhibition obtained with RCwt). Restoration of the interaction with the 3' domain through the K41R substitution, allowing the formation of the closed conformation of the Alu RNP, leads to a 96% inhibition of preprolactin translation, as for wild type reconstitution (Figure 21A, compare RCK41R with RCwt).

      To further establish the phenotype of this SRP9K41A mutated proteins, we investigated the processing activity of reconstituted particles (Figure 21B). Wild type and mutated particle were able to translocate preprolactin into SRP-depleted canine microsomes. We observed that the translocation efficiencies were similar for wild type and mutated particles (Figure 21B, compare the 59% obtained with RCwt to the 63% and 69% achieved with RCK41A and RCK41R respectively).

2.3. SRP RNA

      Except for the sole chloroplast SRP, all SRPs contain a RNA moiety. The role of the RNA in the particle remains unclear but it seems to be more than just a scaffold holding the proteins together. Indications that mammalian SRP RNA could play an active role in particle functions derived from the observation that the presence of SRP RNA enhances SRP54 and SRa mediated GTP hydrolysis (Miller et al., 1993). Furthermore, bacterial SRP RNA has been shown to facilitate the assembly and disassembly of the Ffh-FtsY complex (Peluso et al., 2000; Peluso et al., 2001).

      The crystal structure of bacterial SRP to a very high resolution (1.8 Å) indicated that the signal peptide binding site may lie alongside the RNA backbone (Batey et al., 2000), suggesting that the signal sequence recognition surface is composed of both SRP54 (or Ffh) and of SRP RNA. Such a binding site of both RNA and protein will neatly accommodate the binding to signal sequences which vary in sequence but are characterised by a hydrophobic region of six to fifteen amino acids flanked by two or five positively charged residues. The hydrophobic residues may then be recognised through interaction with the protein hydrophobic pocket, whereas the positively charged amino acids may bind to the SRP RNA.

      In addition to its role in SRP function, bacterial 4.5S RNA is also implicated in the translation process. Hence, a threshold level of 4.5S RNA is required to maintain a normal rate of translation (Bourgaize and Fournier, 1987; Jensen et al., 1994). The role of the 4.5S RNA in translation is due to its binding to the bacterial homolog of elongation factor 2 (Nakamura et al., 1999a; Suzuma et al., 1999; Nakamura et al., 2001). Cross-linking studies identified two distinct interactions of the 4.5S RNA with bacterial ribosome, one is obtained with the 23S rRNA and is dependent on the presence of Ffh and of the nascent chain, whereas the other one, achieved with the 16S rRNA, is independent of Ffh (Rinke-Appel et al., 2002 Brunelli et al., 2002).

      The RNA of the Alu domain folds into two hairpins that are joined by a stem (Figure 3). These two hairpins have a potential for anti-parallel base pairing which is phylogenetically conserved in Bacillus, Clostridium, Archaea, and in Metazoans (Larsen and Zwieb, 1991; Zwieb et al., 1996; Gorodkin et al., 2001). The tertiary interaction between the two distant loops of the SRP Alu RNA was confirmed in the Alu RNP crystal structure (Weichenrieder et al., 2000). Three Watson-Crick base pairing were observed to occur between G13, G14, C15 and C37, C34 and G33, respectively (Figure 22).With regard to the structure, the SRP9/14 binds to the opposite site of the base paired loops (Figure 22B). These loops are solvent exposed in the crystal and could therefore play a role in the events leading to the arrest in the nascent chain elongation. Consistent with a role of SRP RNA in elongation arrest, Andreazzoli and Gerbi, 1991 showed that 7SL RNA undergoes conformational changes upon binding of SRP proteins, and upon binding of the particle to polysomes. Thomas et al., 1997 showed that in addition to loss of elongation arrest function, a particle reconstituted with human SRP14 lacking twenty amino acids of its carboxyl-terminal part, exhibits an increase sensitivity of nucleotides from the loop to hydroxyradical cleavage.

      In order to address the role of these tertiary base pairs in the SRP-mediated translation arrest, we produced synthetic SRP RNAs with individual mutations in each loop that would disrupt base pairing, or with mutations in both loops that would restore base pairing with a different primary sequence. The same approach has been successfully used to defined interactions between RNAs in pre-mRNA splicing (Kandels-Lewis and Seraphin, 1993), or to probe tertiary structure interactions in Escherichia coli 16S rRNA decoding domain (Vila-Sanjurjo and Dahlberg, 2001), as well as to demonstrate Watson-Crick base pairing interactions between tRNA and 23S rRNA in the peptidyl transferase centre (Noller et al., 1995).

      A: secondary structure of SRP RNA Alu domain. L2 and L1.2 loops that are base paired in the RNA protein complex are named according to Weichenrieder et al., 2000. Helices H2, H1.2 as well as the conserved primary sequence which links them (U-turn) are shown. Nucleotides protected from cleavage by hydroxyl radicals in the RNA protein complex are shown in bold, nucleotides that get base paired between the loops are displayed in red. Stretches of ten nucleotides are marked in blue. B: crystal structure of the SRP Alu domain (Weichenrieder et al., 2000). SRP9 and SRP14 are displayed as red and green ribbon respectively. Nucleotides from loops L2 and L1.2 that are involved in tertiary interactions between the loops are displayed as wireframe. C: enlargement of the tertiary interactions between loops L2 and L1.2. Nucleotides from the loops, oriented facing the opposite loop are displayed as wireframes. Hydrogen bond between the bases involved in base pairing are displayed as green dashed lines.

2.3.1. SRP9/14 binding to mutated SRP RNAs

---

2.3.1.1. Mutations introduced in the RNA sequence based on biochemical data

      Elongation arrest activity of the particle requires the presence of the SRP9/14 heterodimer (Siegel and Walter, 1986; Strub et al., 1991). Therefore, the first step in the investigation of the mutated SRP RNAs consisted in studying their ability to bind SRP9/14. Before the resolution of the Alu RNP crystal structure, the tertiary model was based on phylogenetic conservation of four nucleotides in each of the loops in higher Eukaryotes (Zwieb et al., 1996). Comparative sequence analysis of SRP RNAs from different species leads to a model of anti-parallel base pairing between the loops of the SRP RNA Alu domain. These interactions were supposed to take place between nucleotides U₁₂, G₁₃, G₁₄ and C₁₅ of loop L2 and nucleotides G₃₃, C₃₄, U₃₅ and A₃₆ of loop L1.2 respectively. Biochemical data corroborated this sequence comparison analysis. Hence, nucleotides G₁₃, G₁₄, G₁₆ and U₃₅ were shown to be sensitive to kethoxal or CMCT cleavage in naked RNA, but were protected from the same cleavage in SRP (Andreazzoli and Gerbi, 1991). These nucleotides are not part of the SRP9/14 protection sites (Strub et al., 1991) and their protection could be reasonably explain by formation of interaction with other nucleotides upon SRP9/14 binding.

      The investigation of the RNA tertiary interaction was started in the laboratory by Y. Thomas. Through a SELEX study of the loop sequences, he showed that nucleotides of the loops can influence the binding of the SRP9/14 to the RNA, although they do not belong to the protein binding site (unpublished results).

      Based on this information we engineered mutations in the RNA loop sequences, in order to disrupt two of the four putative base pairs between the loops (Figure 23). The binding efficiencies of the mutated RNAs for SRP9/14 were assayed through the use of biotinylated RNA and in vitro synthesised human SRP14 complemented with recombinant human SRP9 (see Materials and Methods). SRP9/14 bound to the different RNAs was analysed by SDSPAGE and quantified by phosphorescence imaging. The binding efficiencies obtained were normalised to the value of wild type synthetic RNA which was set to 100%. Consistent with the location of the loops on the opposite side of the SRP9/14 binding site, we found out that all the mutated RNAs tested were able to bind to the protein (Figure 23). A majority of the RNAs had similar binding efficiencies as wild type, indicating that these mutations had no effect on the protein binding. Only three RNAs exhibited decreased SRP9/14 binding efficiencies, reaching at maximum, a value that is half that of the wild type (Figure 23, RNAs AC, ACGU and AG having respective binding efficiency of 67, 48 and 45% as compared to the wild type). Effects on binding efficiencies were only observed with L2 and not with L1.2 mutations.

      After the crystal structure became available (Weichenrieder et al., 2000), not four but three Watson-Crick base pairs were observed between the loops. Base pairing occurs between nucleotides G₁₃, G₁₄, C₁₅ and C₃₇, C₃₄ and G₃₃, respectively (Figure 22) and differs slightly from the one predicted (Figure 23). With regard to the structure, most of our mutations only affect one of the base pairs observed in the structure, instead of the two we supposed to disrupt. Each of the three mutations that affect the binding of SRP9/14 (Figure 23, lanes AC, ACGU and AG) included a modification of the U₁₂, which is a key component for the proper structure of the L2 loop, where it forms a U-turn which stabilises the loop structure. The destabilisation of the L2 structure resulting from the mutation of U₁₂ may explain why these three constructs reduced the protein binding efficiency.

      

Figure 23: Mutations introduced in the Alu RNA loops sequence.

      The mutations depicted here are based on the anti-parallel model of tertiary interactions between the loops proposed by Zwieb et al., 1996. The mutated RNAs are named accordingly to the modification of their primary sequence. The loops are named according to the crystal structure of the Alu RNP (Weichenrieder et al., 2000). Putative base pairing between nucleotides from both loops are displayed as vertical lines, the modified nucleotides are depicted in bold and the supposedly restored interactions are displayed by dashed lines. Binding efficiencies of the mutated RNAs for the SRP9/14 are shown. In vitro synthesised [³⁵S]-labelled SRP14 combined with recombinant SRP9 were bound to different biotinylated mutated SRP RNAs as indicated. The RNA-bound proteins, separated from free proteins with magnetic streptavidin beads, were submitted to SDS-PAGE and quantified by phosphorescence imaging. The percentage of protein binding obtained for wild type SRP RNA was standardised to 100%.

---

2.3.1.2. Mutations based on the crystal structure

      Once the Alu RNP structure became known, we constructed new mutated RNAs taking into account the three Watson-Crick interactions observed in the crystal. We introduced mutations in one loop that would disrupt base pairing, or in both loops that would restore base pairing with a different primary sequence. Specifically, we mutated two and three positions in each loop individually or in both loop together (Figure 24A). Guanidine was replaced by cytidine and vice versa. Hence, the complementary mutations restored three G-C base pairs.

      

Figure 24: Binding efficiency of hSRP9/14 for mutated SRP RNAs.

      A: summary of the mutations introduced into the SRP RNA to disrupt and/or restore the base pairing between loop L2 and L1.2 (see Materials and Methods for nomenclature). The wild type sequence of the interacting bases is shown as a guideline on the right side of the table. B: in vitro synthesised [³⁵S]-labelled human SRP14 combined with recombinant SRP9 were bound to biotinylated mutated RNAs as indicated on top of each lane. The RNA-bound proteins were submitted to SDS-PAGE followed by autoradiography. The input represents 1/3 of the total [³⁵S]-labelled SRP14 used in the experiments. The lower panel represents the phosphoimager quantification of the SRP9/14 binding efficiency. The percentage of protein binding, normalised to wild type RNA, is indicated for each lane. Neg.: rat 4.5S RNA previously shown not to bind the SRP9/14 heterodimer (Bovia et al., 1997).

      As for the previous constructs, the effect of the mutations on SRP9/14 binding was examined using biotinylated RNA and in vitro synthesised human SRP14 complemented with recombinant human SRP9. SRP9/14 bound to the different RNAs was analysed by SDSPAGE and quantified with the phosphoimager. Bovia et al., 1994 showed that in the biotinylated binding assays, wild type SRP RNA binds only 30% of the protein (Figure 24B, compare Input and WT). All of the mutated SRP RNAs were able to bind h9/14 (Figure 23). Mutations in loop L1.2 and two compensatory mutations in both loops had no effect on the binding efficiency as compared to wild type RNA, whereas three compensatory mutations in both loops increased the binding efficiency (Figure 24B, 2L1.2, 3L1.2, 2 Comp and 3 Comp). Both SRP RNAs with mutations in loop L2 showed a two-fold decrease in the binding efficiency to SRP9/14 (Figure 24B, 2L2 and 3L2). Surprisingly, one of our compensatory mutations supposed to restore the tertiary interaction had a decreased binding efficiency for SRP9/14 (Figure 24B, 1 Comp). It bound only 15% of the heterodimer as compared to the wild type. This result was quite puzzling but might be explained by the different conformation adopted by this mutated RNA. It migrates faster than all the other mutated RNAs in native acrylamide gel electrophoresis, indicating that it folds into a more compact structure. The conformational change might have interfered with the protein binding. In general, the changes in the primary sequences we introduced into the RNA loops did not abolish protein binding, but they had a modulating effect on the binding efficiencies. To confirm this observation, we made a new series of mutated RNAs where we changed G₁₄C₁₅ and G₃₃C₃₄ into uridines and adenines in various combinations (Figure 25).

      

Figure 25: hSRP9/14 binding efficiency for additional mutated SRP RNAs.

      The RNAs are named accordingly to their changed nucleotides. The percentage of binding, measured with the use of biotinylated RNAs, is shown.

      With these additional mutations, we observed similar results as with previous mutations. All RNAs with complementary mutations and the L1.2/A₃₃A₃₄ RNA bound SRP9/14 like wild type. The L1.2/U₃₃U₃₄ RNA was reduced to 77% and the two RNAs with mutations in loop L2 were reduced by 41 and 44%. In addition, tertiary base pairing appears to be required for efficient binding, since the decreased binding efficiencies of 2L2 and 3L2 RNAs were rescued in the 2 Comp and 3 Comp RNAs.

      In summary, the formation of tertiary interactions between the RNA loops appears to be required for high-efficient binding of SRP9/14 to full length SRP RNA. This is illustrated by the fact that all RNAs with complementary mutations, except for the very specific 1 Comp RNA, bound the heterodimer like the wild type SRP RNA.

      In addition, mutations in loop L2 always reduced the efficiency by about 50%, whereas mutations in loop L1.2 had a smaller or no effect on protein binding.

---

2.3.1.3. Electromobility shift assay

      In order to quantify SRP9/14 binding, we decided to use competition experiments with electromobility shift assay. This method allows us to directly compare the dissociation constant of the mutated RNAs to the one of the wild type RNA. The complex was separated from free RNA by electromobility shift assay, but due to the relatively small size of the heterodimer as compared to the 300-nucleotide long SRP RNA, the protein-bound RNA will not be easily distinguished from the naked RNA in a native gel electrophoresis. Therefore, we introduced the mutations we wanted to study into a smaller SRP RNA of eighty six nucleotides (Sa86 RNA), described previously as the minimal Alu RNA folding domain (Weichenrieder et al., 1997). This RNA is sufficiently small to allow the separation between free and proteinbound RNA in a native acrylamide gel electrophoresis, as described in Weichenrieder et al., 2001. Specifically, [³²P]-labelled wild type Sa86 RNA, under conditions allowing all the RNA to be in the protein-bound form, was mixed with increasing concentrations of unlabelled mutated Sa86 RNAs as cold competitor. Free and protein-bound complexes were then separated in a native 8% acrylamide gel electrophoresis and the amount of RNA/protein complex evaluated by a phosphoimager system (see Materials and Methods for details). The bound fraction was measured for each concentration of competitor and was further integrated into a linear regression plot where the slope represents the ratio between the competitor and the wild type RNA Kd for SRP9/14 (Figure 26B, see Materials and Methods for calculation details). Consistent with what would be expected for a competitor with equal binding efficiency for SRP9/14, the value we obtained for wild type Sa86 as competitor was close to 1 (Figure 26B), confirming the validity of the method. The results obtained by the quantification of mutated RNAs dissociation constant were in agreement with the results obtained in biotinylated RNA binding assays. The 2 Comp RNA showed a ratio value of 0.65, indicating that this RNA had a higher binding affinity for SRP9/14 as the wild type SA86 RNA, which is in agreement with the biotinylated RNA binding assays (Figure 24B). We also obtained a lower binding affinity of 2L2 and 1 Comp mutated RNA, with ratios of 3 and 9.2 respectively. 3 Comp RNA showed a ratio value of 1.25, indicating that it bound to the SRP9/14 with the same affinity as the wild type RNA. Mutation 2L1.2, which leads to the disruption of two base pairings through mutations in the L1.2 loop sequence, had a 2.2 fold decreased value of Kd as compared to wild type (Figure 26B).

      Additional mutations that we introduced in the Sa86 sequence, but not in the full length RNA, highlighted the critical role of the nucleotide at position fifteen for SRP9/14 binding. Mutating only C₁₅ into G lead to a decrease binding of SRP9/14, as seen with a ratio of 1.49 (Figure 26C). A similar decrease was obtained with the G₁₅G₃₄G₃₇ mutation with a ratio of 1.97. Mutations of the sole G₃₃ into a C lead to a two-fold decrease in the Kd for SRP9/14 binding as compared to the wild type, whereas the C₁₃C₁₄C₃₃ behaved like wild type (Figure 27C).

      

Figure 26: Electromobility shift assay of small mutated Alu RNA complexed with human SRP9/14.

      [³²P]-labelled Sa86 RNA, in conditions allowing 100% of SRP9/14 binding, was mixed with Sa86-derived cold mutated competitor RNAs. Protein-bound RNA was separated from the free RNA on an 8% acrylamide native gel electrophoresis. The amount of RNA-protein complex was evaluated by a phosphoimager system and integrated to allow the calculation of the ratio between the Kd of competitor and wild type RNA for SRP9/14 binding. A: competition assay realised with increasing amount of cold 3 Comp RNA as competitor. The final concentrations of competitor used in the assay are marked on top of each lane. Free and protein-bound RNA are marked on the left. WT alone: wild type labelled RNA in absence of competitor and protein; WT + BSA: control without competitor where SRP9/14 was replaced with the same amount of bovine serum albumin; WT + h9/14: control in absence of competitor RNA. B: measure of dissociation constants for RNAs complexes comprising Sa86 mutated RNAs that shares mutations produced in the full length RNA. The value of the ratio between wild type and competitor dissociation complex, measured as described in Materials and Methods, are shown. C: measure of dissociation constant ratio obtained with additional mutated RNAs produced only in the small SA86 RNA.

      The remarkable effect on SRP9/14 binding we obtained with the biotinylated RNAs binding assay, despite the excess of RNA, might result from increased kinetic instability of the mutated RNAs where the base pairing are disrupted. Therefore the differences observed might have derived from a technical artefact resulting in the loss of binding during the extensive wash steps. No drastic changes in the mutated RNAs dissociation constant were observed in our electromobility shift assays. Alternatively, the effects of the mutations might be stronger in complete SRP RNA than in Alu RNA for reasons which are not yet understood. However, since the proteins have been shown to bind the RNA with subnanomolar affinity (Janiak et al., 1992; Bovia et al., 1994), the minimal defect observed on SRP9/14 binding to the mutated SRP RNAs should not interfere with the assembly of the particle.

2.3.2. Functional analysis of particles containing mutated RNAs

      We reconstituted SRP with canine and recombinant SRP proteins together with synthetic SRP RNAs. The RNAs were produced in large scale transcriptions and purified away from DNA and nucleotides under non-denaturing conditions (see Materials and Methods). The reconstitution conditions were optimised by testing different protein and RNA concentrations (results not shown). Translocation and elongation arrest activities of the reconstituted particles were assayed in a wheat germ translation system. For elongation arrest activity, the differential effects of the reconstituted particle on synthesis of the secretory protein, preprolactin, and of the cytoplasmic protein, cyclin D, were compared. For translocation activity, we followed the processing of preprolactin to prolactin in the presence of salt-washed canine microsomes (Materials and Methods). The translation reactions were analysed by SDS-PAGE and quantified with the phosphoimager.

      Dog purified SRP and particle reconstituted with purified canine SRP respectively inhibited preprolactin translation to 94% and 91% (Figure 27A, lane 2 and 3). However, with particles reconstituted with in vitro transcribed wild type SRP RNA the inhibition of preprolactin was only 67% (Figure 27, compare lane 4 with lanes 2 and 3). Purified SRP and canine RNA reconstituted particles had similar translocation efficiencies values (85% and 81% respectively, Figure 27A lower panel, lane 2 and 3), whereas particles reconstituted with synthetic SRP RNA have only 55% processing activity (Figure 27 A, lower panel lane 4). Reannealing the synthetic SRP RNA under different conditions always lead to a complete loss of translocation activity of the reconstituted particles, suggesting an interference with the S domain assembly. The presence of several RNA conformations might account for the reduced function of particles reconstituted with synthetic RNA.

      In the following experiments, we always normalised elongation arrest activities and translocation efficiencies to particles with synthetic wild type RNA.

      As compared to the synthetic wild type RNA, disrupting base pairing through mutations in loop L2 resulted in a significant but not complete loss of elongation arrest activity, independently of whether two or three base pairs were affected (Figure 27B, lanes 2L2 and 3L2). Restoring tertiary base pairing between the two loops with mutations in loop L1.2 completely rescued the negative effect of the 3L2 mutation and significantly improved the 2L2 phenotype (Figure 27B, lanes 3 Comp and 2 Comp). In contrast, two nucleotides mutation in loop L1.2 had no significant effect on the elongation arrest activity of the particle (lane 2L1.2). Three mutations in loop L1.2 slightly decreased elongation arrest activity of the particle as compared to the wild type (lane 3L1.2).

      To calculate the elongation arrest activity of the SRPs, the ratio of preprolactin to cyclin synthesis in the presence of buffer was used as 0% inhibition (see Materials and Methods for details). Compared to this buffer, the elongation arrest activity of SRP reconstituted without SRP9/14, used as negative control, was 20% as previously observed (Thomas et al., 1997).

      

Figure 27: Preprolactin elongation arrest and processing efficiency of particles reconstituted with mutated RNAs.

      A: autoradiogram of some reconstituted particles described in B. Lane 1: negative control, reconstitution in absence of SRP9/14; Lane 2: purified canine SRP; Lane 3: reconstitution with recombinant proteins and purified canine SRP RNA; Lane 4: particle reconstituted with synthetic wild type SRP RNA; Lane 5: particles reconstituted with 2L2 RNA; Lane 6: particles reconstituted with 2L1.2 RNA; Lane 7: particles reconstituted with 2 Comp. RNA. Upper panel: elongation arrest assay; lower panel: translocation assay.

      B: mutated particles elongation arrest assay. Reconstituted SRP, with mutated in vitro synthesised RNA, were added at 100 nM final to wheat germ translation reactions, programmed with sea urchin cyclin and bovine preprolactin mRNAs. The resulting [³⁵S]labelled proteins were analysed by SDS-PAGE and quantified by phosphoimager. Elongation arrest efficiency of the different reconstitutions, as marked for each lane, was evaluated as described in Materials and Methods. The specific inhibition of preprolactin synthesis observed in presence of wild type reconstituted particle was standardised to 100%. WT: particle reconstituted with wild type proteins and wild type synthetic RNA. All measures were done in duplicate. C: mutated particles processing efficiency. Reconstituted SRP are tested as in A, except for addition of SRP-depleted microsomes in the translation reactions. The efficiency of processing, as marked for each lane, was evaluated by exposure to a phosphoimager screen (see Materials and Methods for details). Preprolactin processing efficiency obtained with wild type reconstituted particles was standardised to 100%. WT: particle reconstituted with wild type proteins and wild type synthetic RNA.

      All SRPs containing mutated synthetic RNAs were active in signal recognition and targeting, since they are all able to translocated preprolactin into the microsomes. Particles lacking elongation arrest have their translocation capacities reduced by about 50% as seen previously (Figure 13A, Thomas et al., 1997).Indeed, the processing activities we observed for our mutated reconstitutions confirmed the results of their elongation arrest activities. SRPs comprising 2L2 and 3L2 RNAs had reduced translocation efficiencies as compared to particles reconstituted with wild type synthetic RNA (Figure 27C, lanes WT, 2L2 and 3L2). All the other SRPs had activities comparable to wild type particles. The processing activity of the microsomes in the absence of exogenous SRP was 20%, resulting from contaminating particles in the microsomes preparation.

      Hence, the particles reconstituted with in vitro transcribed mutated SRP RNA were all active in signal recognition and in targeting, whereas 2L2 and 3L2 SRPs were specifically deficient in their elongation arrest activities.

---

2.3.2.1. The elongation arrest activity observed with the mutated particles is not due to a reduced SRP9/14 binding

      Although unexpected, based on the RNA binding experiments, mutated RNAs which exhibited decreased binding efficiency for SRP9/14 were also the ones with defects in preprolactin translocation and elongation arrest. Thus we wondered whether it was conceivable that inefficient assembly of the h9/14 protein caused the phenotype. To test this hypothesis, we reconstituted 3L2 particles in the presence of increasing amounts of h9/14. Up to eight-fold excess of SRP9/14 over the other SRP components were used in the reconstitution experiments and this without improving elongation and translocation activities of these particles (Figure 28). Elongation arrest activity and preprolactin processing efficiency stagnated around an average value of 55 and 62% respectively, which is in agreement with the 46 and 55% elongation arrest and processing activities observed previously for the 3L2 RNA (Figure 27B and C). Thus, the reduced SRP9/14 binding is not responsible for the observed reduction in elongation arrest and translocation activities of 3L2 mutated SRP RNA. This interpretation is supported by the facts that the dissociation constant of the 3L2-SRP9/14 complex, like the wild type complex, is in the nanomolar range and that the protein concentration used in the assay was 200800 nM (Figure 24).

      As previously mentioned, reannealed RNA cannot be used to reconstitute functional particles, and several conformations were observed for these RNAs in native acrylamide gel electrophoresis (data not shown). We then wanted to see, whether the observed decrease in elongation arrest could result in the presence of RNAs trapped in a conformation which did not allow proper binding of the SRP proteins. In order to address this question, we decided to reconstitute particles in presence of increasing amounts of synthetic RNAs. Up to eight-fold excess of RNA over the proteins were used in the reconstitution experiments. This was done for wild type RNA as well as for 2L2 and 3L2 RNAs (Figure 29).

      

Figure 28: Effect of increasing hSRP9/14 concentration on 3L2 reconstituted particles.

      A: elongation arrest assay. 3L2 particles reconstituted with increasing amounts of human SRP9/14 were added to wheat germ translation reactions programmed with sea urchin cyclin and bovine preprolactin mRNAs. The elongation arrest efficiency is marked for each lane, as measured by phosphoimager. WT: reconstitution with wild type synthetic SRP RNA; 1x: standard reconstitution of 3L2 particles (see Materials and Methods); 2x, 4x, 8x: reconstitution of 3L2 particles in presence of a 2-fold, 4-fold, or 8-fold excess of recombinant hSRP9/14 as compared to 1x, leading to final SRP9/14 concentrations in the reconstitution of 1, 2 and 4 mM respectively. B: processing of 3L2 particles reconstituted as in A. The processing efficiency is marked for each lane, as measured by phosphoimager. WT: reconstitution with wild type synthetic SRP RNA; 1x: standard reconstitution of 3L2 particles; 2x, 4x, 8x: reconstitution of 3L2 particles in presence of a 2-fold, 4-fold, or 8-fold excess of recombinant hSRP9/14 as compared to 1x. All measures derived from duplicate experiments.

      

Figure 29: Effect of increasing RNA concentration in reconstituted particles function.

      A: elongation arrest assay. particles, reconstituted with 0.5 to 8-fold excess of mutated RNAs as compared to the other SRP components, were added to wheat germ translation reactions programmed with sea urchin cyclin and bovine preprolactin mRNAs. The elongation arrest efficiency is marked for each lane, as measured by phosphoimager. B: processing activities of mutated particles reconstituted as in A. The processing efficiency is marked for each lane, as measured by phosphoimager. White bars: particles reconstituted with 0.5 to 2-fold excess of wild type synthetic RNA; black bars: particles reconstituted with 0.5 to 2-fold excess of 2L2 RNA; grey bars: particles reconstituted with 1 to 8-fold excess of 3L2 RNA. 2L2 and 3L2 measures derived from duplicate experiments.

      Increasing the amount of 3L2 RNA up to eight-fold in the reconstitution did not improve the elongation arrest activity of the particle, which reached values oscillating around an average 40% specific inhibition of preprolactin synthesis (Figure 29A, grey bars). This is consistent with the 46% inhibition observed for this mutated particle under standard conditions (Figure 27B, lane 3L2). Preprolactin processing efficiency gave similar results, with a value fluctuating around 60% (Figure 29B, grey bars), as compared to the 55% observed previously (Figure 27C, lane 3L2). The elongation arrest activity seemed to be improved with a two-fold excess of RNA, reaching a value of 51% (Figure 29A, black bars). However, the preprolactin processing pattern showed that the excess of RNA did not rescue the phenotype of the 2L2 mutated particles, since the average value is close to 50% (Figure 29B, black bars) in agreement with the value obtained under standard condition (Figure 27C, lane 2L2).

      Hence, increasing the protein or the RNA concentrations in our reconstitution assays did not rescue the phenotype observed for 2L2 and 3L3 mutated particles. However, to further establish that the mutated RNAs affect the function of the SRP Alu domain directly and not through a reduced binding of SRP9/14, we decided to purify reconstituted particles. Since this purification requires relative large amounts of canine proteins, we limited our analysis to wild type, 2L2 and 2 Comp RNAs.

      

Figure 30: Purification of particles reconstituted with mutated RNAs.

      Particles reconstituted with canine proteins and in vitro transcribed RNA, were purified on a DE53 column as described in Materials and Methods. Aliquots, taken at each step, are submitted to 5-20% gradient SDS-PAGE and revealed by silver nitrate staining. A: DE53 purification pattern of particles reconstituted with canine proteins and wild type synthetic RNA. S: input of proteins used in the reconstitution prior to the column purification, corresponding to 5 pmole of each protein; FT: DE53 flow through; W: wash fraction; 1 and 2: first and second elution of the column at 600 mM potassium acetate, corresponding to the elution of complete particles; 3 and 4: first and second elution of the column at 1000 mM potassium acetate, corresponding to partially reconstituted particles. The position of each canine SRP protein is marked on the right side of the gel. B: comparison of the different reconstituted particles after purification on DE53 as labelled on the top of each lane. 2 pmole of each purified particles, as evaluated from comparison of the elution with the input fraction, are submitted to 5-20% gradient SDS-PAGE and dyed with silver nitrate. St: standard consisting of 2 pmole of recombinant SRP proteins. C: western blot analysis of the purified particles. A fraction corresponding to 0.5 pmole of the purified particle, as evaluated from the gel in B, were submitted to 5-20% SDS-PAGE, followed by western blot with antibodies against human SRP14. The concentrations of the respective purified particles were evaluated as compared to a standard consisting of 0.25, 0.5, 0.75 and 1 pmole of hSRP14.

      Particles, reconstituted with canine proteins and synthetic SRP RNA, are loaded onto a DE53 column to remove non-bound proteins and partially reconstituted particles (Materials and Methods). Fully assembled SRP is eluted from the column at a 600 mM potassium acetate concentration (Chang et al., 1997). Particles eluted from the column were analysed by silver staining and quantified by immunoblotting with anti-SRP14 (Figure 30).

      The purified particles were added at different concentrations into translation and translocation assays. Below 25 nM, none of the reconstituted particle exhibited a significant elongation arrest activity (Figure 31A). At higher concentrations, 2L2 particles had again a reduced elongation arrest activity as compared to wild type or to 2 Comp SRPs (Figure 31A, compare dotted line with filled and dashed lines). Interestingly, the 2 Comp SRP appeared to work better than wild type SRP (Figure 31A, compare dashed lines with filled lines), confirming the notion that base pairing, and not the primary sequence, is essential for activity. As previously observed, measurement of low level elongation arrest activities lead to large variations of the standard deviation for 2L2 particles (Figure 31A, dotted lines). In agreement with our previous experiments, the elongation arrest activity of the 2L2 particle is not completely abolished but reaches at maximum half the level of the 2 Comp activity.

      

Figure 31: Functional analysis of DE53 purified mutated particles.

      A: inhibition of preprolactin synthesis in presence of increasing amount of purified reconstitutions. Particles reconstituted with canine proteins and mutated synthetic SRP RNA were purified on DE53 prior to their addition into wheat germ translation reactions at final concentrations ranging from 20 to 80 nM. The produced [³⁵S]-labelled protein were analysed by SDS-PAGE and quantified by phosphoimager. For each reconstitution, the elongation arrest activities are displayed as function of the concentration of mutated particles used in the assays. Filled lane: particle reconstituted with wild type synthetic RNA; dotted lane: particle reconstituted with 2L2 RNA; dashed lane: particle reconstituted with 2 Comp RNA. B: preprolactin processing activities in presence of increasing amount of purified reconstitutions. Particles are tested for their processing activities by addition of SRP-depleted microsomes to translation reactions described in A. Filled lane: particle reconstituted with wild type synthetic RNA; dotted lane: particle reconstituted with 2L2 RNA; dashed lane: particle reconstituted with 2 Comp RNA.

      Elongation arrest and processing values displayed in the figure results from four different experiments.

      Bovine preprolactin processing, which is more accurate to monitor than elongation arrest activity at low SRP concentrations, yielded results in agreement with the elongation arrest assays. The processing activity of the 2L2 particle increased more slowly than the ones of wild type and 2 Comp SRPs (Figure 31B, compare dotted line with filled and dashed lines). Notably, 2 Comp and wild type particles reached maximal translocation activities at concentrations of 40 nM, whereas the one of 2L2 SRP had not yet reached its plateau at the maximal concentration of 80 nM used in the assay. The curve of the processing activity of 2L2 SRP, predicted that it could reach the same processing efficiency as the wild type SRP, confirming that the elongation arrest activity of SRP is not required for translocation (Siegel and Walter, 1985). However, a higher concentration of active particle is necessary to ensure efficient targeting in time to allow translocation.

3.1. Alu SRP proteins

      The mutational analysis we undertook in this study allowed us to confirm the critical role played by the SRP14 in the elongation arrest function. In agreement with previous studies (Thomas et al., 1997; Mason et al., 2000), we showed that deletion in the carboxyl-terminus of SRP14 leads to a complete loss of elongation arrest activity. Furthermore, we narrowed down the amino acids involved in the process to five residues located between lysine 95 and lysine 100. Deletion of ten residues in the carboxyl-terminus of SRP14 had no effect on the elongation arrest (Bovia et al., 1994), whereas deletion of twenty (Thomas et al., 1997) and of fifteen residues (Figure 13) results in a complete loss the elongation arrest function of the reconstituted particles. This loss of function is correlated with binding to ribosomal rRNA., Terzi et al., (see manuscript below) demonstrated that whereas wild type SRP14 is able to bind to a complex composed of the ribosomal L12 protein and of GAR rRNA, both representing the GTPase-associated centre of the ribosome, SRP14K95 did not bind anymore.

      The short carboxyl-terminal motif of SRP14 that we showed to be essential for elongation arrest activity is the most highly conserved feature in the SRP14 proteins, suggesting an essential functional role. Indeed we showed that four of these five conserved residues that were critical for the function are basic amino acids, which strongly suggests that they play a role in RNA binding. These residues are unlikely to bind SRP RNA, since they do not contact the SRP RNA in the Alu RNP crystal structure (Weichenrieder et al., 2000), and since all of the protein footprints observed on SRP Alu RNA (Strub et al., 1991) can be explained by the contacts seen in the crystal structure. Consistent with this, the hydroxyl radical footprints on the SRP RNA central stem that we obtained with SRP9/14K95 were exactly the same as the one obtained with the wild type heterodimer.

      The truncation of SRP14 did not abolish the binding of the reconstituted particles to ribosomes since translocation is still observed (Figure 13). Therefore it is conceivable that the loss of elongation arrest activity is correlated with the loss of rGAR RNA binding. Hence the arrest in nascent chain elongation might be achieved through an interaction of the SRP14 carboxyl-terminus with the ribosomal GTPase-associated centre, taking place between GAR rRNA and amino acids K96, R97, E98, R99 and K100 of human SRP14. Consistent with this putative interaction, the competition experiments we made with the truncated SRP14K95 resulted in elongation arrest activity and translocation efficiency that were lower than expected for a competitor with equal affinity for ribosome. Hence, the deletion of SRP14 carboxyl-terminus, abolishing one interaction of SRP with the ribosome, might result in a higher dissociation rate of particles reconstituted with the truncated protein as compared to the wild type particles. This might explain why up to nine-fold excess of SRP14K95 was required to observe decreased elongation arrest and translocation in our competition experiments (Figure 14). Whereas the mutated protein binds the SRP RNA as efficiently as the wild type protein, the particles containing wild type protein might be bound for a much longer time to ribosome as compared to the truncated particles. Thus higher concentrations of truncated particles are required to displace the wild type particle from the ribosome.

      The role of the SRP14 internal loop, between b₁ and b₂, remains unknown, but it does not seem to be involved in the elongation arrest activity. Unlike the carboxyl-terminal region of SRP14, the internal loop shows only little primary sequence conservation, which argues against a critical role in SRP functions. In fact, only four basic amino acid residues (K31, K32, K42 and K43 in human SRP14) are conserved and these residues have been shown to be involved in binding to SRP RNA (Bui et al., 1997). This is confirmed by the determination of SRP14sloop-Alu RNA dissociation constant by Biacore analysis. Whereas the wild type protein binds to the wild type RNA with a Kd of 0.55 nM ± 0.19, the SRP9/14sloop has a Kd value of 7.33 nM (J. Hasler, personal communication). Yet the internal loop of SRP14 is not entirely seen in the Alu RNP crystal structure despite the presence of the RNA (Weichenrieder et al., 2000). The missing residues (36 to 53) include two of the conserved basic amino acid residues (K42 and K43). The two others conserved basic residues (K31 and K32) are seen in the structure and they contact the SRP RNA. Thus, the internal loop of SRP14 might interact with the Alu RNA in the region of the bend, and therefore participate in the closed conformation of the Alu RNP. Hence the internal loop is not ordered in the crystal because the bending of the RNA 3' domain onto the 5' domain is not directly observed in the structure. In their crystals, Weichenrieder et al., 2000 observed two RNA molecules bound to one SRP9/14, one engaging its 5' domain, whereas the other one engaged its 3' domain. However, their model for the native Alu domain structure, where the 3' domain folds back onto the 5' domain is consistent with biochemical data. It explains the requirement for SRP9/14 dimerisation (Strub and Walter, 1990), the protection of SRP9 cysteines 39 and 48 from alkylation in the RNA-protein complex (Siegel and Walter, 1988c), as well as all the hydroxyl radicals footprints (Strub et al., 1991). Furthermore their model is in agreement with the size of the particle, measured by electron microscopy (Andrews et al., 1985).

      Thus, the internal loop of SRP14 might be involved in the assembly of the particle, leading to the closed conformation of the Alu RNP, but we showed that it is dispensable for elongation arrest. This might be because the interaction of the basic residues of SRP14 internal loop with the Alu domain SRP RNA is not the only one involved in the RNA bend. Thus, the truncation of the internal loop in SRP14sloop is not sufficient to abolish the formation of the Alu RNP closed conformation, explaining why this truncation did not significantly affect the elongation arrest activity of the particle.

      Furthermore, we saw that the mutation of SRP9 lysine 41, which contacts the 3' and the 5' domain of the Alu SRP RNA, did not influence the elongation arrest activity of the particle. This might also reflect that the weakened interactions produced by the SRP9 mutations, are not sufficient to completely abolish the closed conformation of the particle Alu domain.

      In Weichenrieder et al., 2001 investigations of the role of the RNA 3' domain in SRP9/14 binding were achieved by the use of circularly permuted Alu RNAs. In these RNAs, the original 3' and 5' ends of the SRP RNA were connected with a linker of one or four uridines, respectively. A linker of one uridine was too rigid to allow the formation of the closed conformation, whereas the four uridines linker was sufficiently flexible to allow it. With these RNAs, the authors showed that the rigidly bound Alu RNA had a 50-100 lower affinity for SRP9/14 binding as compared to the wild type RNA. The abolition of the closed conformation by the one uridine linker was confirmed by RNase V1 cleavage.

      However, we investigated SRP9K41 mutations in particles reconstituted with canine purified SRP RNA. Therefore, the RNA 3' and 5' ends are the native ones, meaning that flexibility is allowed between 3' and 5' domains. The presence of the internal loop of SRP14 and the flexibility of the RNA might be sufficient for the 3' domain to fold back onto the 5' domain, even in absence of lysine 41 interactions with the RNA central stem. Throughout the particle assembly process, conformational changes that allow the formation of the closed conformation might also be introduced in the RNA. Therefore the weakened interactions resulting from the SRP9K41A mutations might be overcome by these conformational changes, which bring the 3' and the 5' domain in closed proximity, allowing the formation of RNA-protein and/or RNA-RNA interactions. Indeed, in addition to SRP14sloop and SRP9K41 interaction with the RNA, the crystal structure revealed previously observed footprints (Strub et al., 1991) as protein induced RNA-RNA interactions (G₅-G₁₀; G₄₈-G₅₁) or as both induced RNA tertiary contacts and direct protein contacts (C₁₇-C₂₉; G₅₈-G₆₂). These interactions might be sufficiently strong to hold the Alu domain into a closed conformation in the full length particle, despite the disruption of SRP9 lysine 41 or SRP14sloop interactions.

      Furthermore, the folding of the 3' domain onto the 5' domain leads to a 25 Å shortening of the particle's length. Upon interaction with ribosomes, such a difference in the particle's length would most likely interfere with SRP functions.

      The involvement of the Alu domain switch, from an open to a closed complex, might be of functional importance for the elongation arrest activity of the particle. Since the mutation of SRP9K41 is not sufficient to completely abolish the closed conformation, we propose to achieve this by the use of a SRP RNA construct in which native 3' and 5' ends will be rigidly bound in order to abolish the folding of the 3' domain. Previous observations have shown that the flexibility between the two domains is required for high efficiency SRP9/14 binding (Weichenrieder et al., 1997). The reduced binding of the SRP9/14 for a RNA where native 3' and 5' ends are bound together, can be overcome through the purification of the reconstituted particles. We showed that the reduced efficiency of synthetic RNA in the particle reconstitution can be surmounted by purification of the fully reconstituted particles on DE53 chromatography. Wild type SRP9/14 has been shown to bind the RNA with subnanomolar affinity (Janiak et al., 1992; Bovia et al., 1994) with a Kd of 0.55 nM measured by Biacore as aforementioned. Thus a 100-fold reduced affinity for SRP9/14 of a rigidly bound RNA would result in a Kd value around 50 nM, which is relatively close to the 100 nM final concentration of SRP used in our elongation arrest assays. However, this can be overcome through the purification of the reconstituted particle and through the addition of an excess of SRP9/14 in the translation reaction, assuring that the effects that one might observe would not result from the reduced affinity of the rigidly bound RNA for SRP9/14.

3.2. Alu SRP RNA

      In this study we provided the first evidence that a specific structure of the SRP RNA Alu domain is required for elongation arrest activity. The presence of three base pairs between the two distant loops is important for high efficient binding of SRP9/14, as well as for proper elongation arrest function.

      We showed that the disruption of two or three of the base pairs between loops L2 and L1.2 resulted in a decreased elongation arrest activity of the reconstituted particles. The effects observed are different depending on which loop the mutations are introduced into. Mutations in L2 decreased the binding of the SRP9/14, whereas mutations introduced in L1.2 had no effect on the protein binding.

      Based on Andreazzoli and Gerbi, 1991 experiments, the RNA tertiary structure is stabilised upon binding of the protein. Nucleotides of L2 and L1.2 loops were sensitive to chemical modifications in naked RNA, whereas they were protected from the same chemical modifications in the particle. Thus, binding of the SRP9/14 to the Alu domain stabilises the base pairing interactions between the loops. This results in an energetically favoured conformation, thereby explaining why disruption of the loops tertiary interaction can have a slight effect on SRP9/14 binding.

      The two loops have a different structure which might explain the differential effect of their mutations. L2 is constrained by a U-turn at position U₁₂ and by a shared G₁₁-G₁₆ base pair, whereas loop L1.2 lack any internal stabilisation. The flexibility of it structure might allow L1.2 nucleotides to form alternative base pairs with loop L2 (e.g. G₁₃ with U₃₅ in 2L1.2), whereas loop L2 is too constrained to allow alternative base pairing after mutations.

      The changes introduced in loop L2 results in the presence of four pyrimidines between both loops. These two pyrimidine rings would be too far away to form hydrogen bonds, which is not the case in the 2L1.2 RNA where four purines are facing each other, leading to a reduced distance which might allowed alternative tertiary interactions.

      In the crystal structure, positions 15 and 33 of the RNA are located in a region where L2 and L1.2 are the closest, suggesting that their mutations strongly affect the structure. Thus, mutation G₁₅ might interfere with the G₁₁-G₁₆, which together with the U-turn at U₁₂ participates in the stabilisation of the L2 loop. The G₁₅ and G₁₅G₃₄G₃₇ might have decreased SRP9/14 binding efficiencies because they destabilise the whole folding of the Alu RNA through an effect on the L2 loop. Consistent with this, the mutation of G₃₃ into a C might form a base pair with G₁₆, interfering then with the shared G₁₁-G₁₆ pair, resulting in a decreased SRP9/14 binding. C₁₃C₁₄C₃₃ on the other hand binds the protein to wild type levels, possibly because the interference of C₃₃ with the shared G₁₁-G₁₆ pair is counterbalanced by the presence of two additional Cs in the L2 loop. These Cs at position thirteen and fourteen could contribute to the stabilisation of the L2 structure through an internal G₁₁-C_x base pair. From that point of view, the 1 Comp mutation should result in a more compactly folded RNA because it results in the presence of five Gs among the six nucleotides of loop L2. These five guanines might allow a different stabilisation of the L2 loop since several nucleotides are available for a shared pair with G₁₁ as well as for G-C base pairing with loop L1.2. The resulting RNAs structure might be too compactly folded to allow proper binding of the heterodimer.

      We have seen that the effects on SRP9/14 binding were different in the full length SRP RNA and in a small eighty-six nucleotide-long Alu RNA. The higher effect on the protein binding observed with the full length RNA might derive from a technical artefact. SRP9/14 primary binding site is located in the RNA 5' domain (Strub et al., 1991; Weichenrieder et al., 2000), and the protein can bind to the RNA in absence of tertiary interactions between the loops. However, since the tertiary base pairs most likely stabilise the complex, it is feasible that their disruption increases the dissociation rate of the complex. Hence, the bound protein may be lost during the washing procedure in the biotinylated assay. In agreement, the same mutations had only a minimal effect in the electromobility shift assay. The absence of extensive wash steps as well as the polyacrylamide gel caging effect explains the reduced effect of the mutation on the protein-RNA dissociation constant. The differences might also be due to interference of the S domain RNA with the folding of the Alu domain. Binding assays were done in absence of proteins from the S domain and the S domain RNA may interfere with the folding of the Alu domain RNA. Specifically, helices 6 and 8 are free in absence of SRP19, whereas they get clamped together upon binding of SRP19 (Kuglstatter et al., 2002). The free movement of these helices may interfere with the folding of the Alu domain, especially once it has been weakened through the disruption of L2 and L1.2 base pairing. This is also consistent with previous data showing that the proteins bind the RNA in a cooperative manner (Walter and Blobel, 1983a).

      Nevertheless, except for the specific effect observed on the 1 Comp RNA, in which the mutations unexpectedly modified the RNA structure, the mutated RNAs have only at maximum half the SRP9/14 binding efficiency as compare to wild type RNA. Yet, the SRP9/14 heterodimer has been shown to bind the Alu RNA with subnanomolar affinity (Janiak et al., 1992; Bovia et al., 1994), and in the reconstitution assay the particles are assembled from individual component at a final 0.5 mM concentration. Hence, the reduced protein binding efficiency observed with our mutated RNAs should not impair the reconstitution of particles with these mutated RNAs. This is further supported by the result that 14sloop protein confers full elongation arrest activity to reconstituted particles despite the fact that the dissociation constant of 14sloop-Alu complex is 13-fold higher that wild type complex.

      Disruption of the base pairing between the loops reduces the elongation arrest activity of the reconstituted particles. However the contributions of loop L2 and L1.2 are not the same. Disruption of the tertiary interactions through mutations in L1.2 had only a minor effect on elongation arrest activity. The effects could be rescued by compensatory mutations, introduced in the opposite loop, restoring the three Watson-Crick base pairing but with a different primary sequence.

      The reduced elongation arrest activity did not result from the reduced binding of the SRP9/14, since neither increasing the amount of protein, nor increasing the amount of RNA in reconstitution are able to rescue the phenotype. Furthermore purified particles have the same reduced elongation arrest activity. We were only able to rescue the phenotype by restoring the tertiary interactions in the RNA loops, without conserving the primary sequence.

      The effect on elongation arrest function could result from conformational changes in the SRP9/14 protein due to the mutated loops. However, we think that this is unlikely to be the case since the formation of the RNA binding surface of SRP9/14 requires a largely intact protein complex (Weichenrieder et al., 2000) and only very small changes in the C-terminal regions of SRP14 and SRP9 are tolerated without interfering with RNA binding (Bui et al., 1997). It is therefore unlikely that the structure of the SRP9/14 protein in the complex was changed significantly by the mutations in the RNA. The tertiary structure is most likely induced by protein binding (Andreazzoli and Gerbi, 1991; Weichenrieder et al., 2001) and it therefore contributes to the stability of the complex. The small changes in the dissociation constants and in the RNA binding efficiency are therefore plausibly explained by a slightly diminished stability of the complex. It is also unlikely that the mutations in the two loops interfered with the formation of the closed complex. Alu RNA variants with mutations that interfere with the formation of the closed conformation have their dissociation constants reduced by 50 to 100fold (Weichenrieder et al., 2001), whereas in our case the dissociation constant was only reduced by 9-fold at maximum (Figure 26).

      In addition, the particles comprising the mutated RNAs have intact signal recognition and targeting functions, arguing against a significant influence of the mutations on the overall structure of SRP. Although it cannot be excluded entirely, it is therefore unlikely that conformational changes in SRP9/14 or in the overall structure of SRP account for the observed effects of the mutations on elongation arrest activity.

      We showed that the reduced efficiency of in vitro transcribed SRP RNA for reconstitution, as compared to purified canine SRP RNA, can be overcome through the purification of the particles on DE53 chromatography. Furthermore the reconstitution efficiency might also be increased by the addition of SRP protein in the transcription reaction. SRP9/14 primary binding site is located in the RNA 5' domain (Strub et al., 1991; Weichenrieder et al., 2000; Weichenrieder et al., 2001), and the proteins, together with SRP19, SRP68 and SRP72 are found in the nucleus (Politz et al., 2000; Politz et al., 2002). Therefore these proteins are expected to be assembled early in the particle. Conformational changes resulting from the sequential binding of the proteins that are involved in the early stages of the SRP assembly might favour the folding of our synthetic SRP RNA into a native form, closely related to the one achieved in vivo, thus increasing the efficiency of the reconstitution.

      The critical role of the tertiary interactions between loops L1.2 and L2 in SRP function is corroborated by the phylogeny conservation for potential base pairing between the loops. SRP RNA Alu domain of Metazoa, Plantae and Archaea can be folded into the same secondary structure as human 7SL RNA Alu domain. Zwieb et al., 1996 showed that the capacity to form anti-parallel base pairing between the loops of the RNA Alu domain is conserved. Taking into consideration the additional information that is now available since the Alu RNP crystal structure was solved, we revised the previously published alignment (Figure 32). This new alignment confirmed that the potential for base pairing between the loops is conserved in Metazoa, Plantae and Archaea. Despite this conservation, three groups can be distinguished. First, all metazoan SRP RNAs closely resemble the human 7SL RNA. Specifically, they contain all the features of the human RNA Alu domain: the U-turn residue and the shared G-G base pair that stabilises loop L2 (Figure 32A, star and green labelling, respectively), the base pairing potential (Figure 32A, red labelling) and the two helical stacks (Figure 32B).

      A: loop L1.2 and loop L2 sequence alignments of a representative sample of Metazoan, Plantae and Archaea species. Loop sequences are labelled with a black bar and the adjacent helix sequences are outlined in bold italic, dotted line: borders of loop sequences not unambiguously identified. The U-turn residue is labelled by a black star B: secondary structure models. Green: shared G-G base pair, red: base paired nucleotides in the loops, blue: conserved nucleotide in Archaea.

      Only minor sizes differences are observed in the second loop of Caenorhabditis elegans and in the two loops of Drosophila melanogaster.

      Second, loop L2 is highly conserved in plants, except the U-turn residue of Humulus japonicus, Humulus lupulus and Benincasa hispida which is replaced by a cytidine. However, a C-turn, sharing common feature with U-turn, has been shown in beet western yellow virus RNA pseudoknot structure (Su et al., 1999). Therefore the change of the U into a C in these three plants might results in the formation of a C-turn in the structure. Most specifically, loop L1.2 is always smaller than in Metazoa and helix 1.2 is one base pair shorter (Figure 32B). However, sequence conservation and complementary nucleotide changes support the formation of two base pairs between plants RNA loops. These two base pairs are adjacent to the G of the shared pair in loop L2 and the three bases are therefore expected to form a continuous stack like in human Alu RNA.

      Third, in Archaea, the U-turn residue is strictly conserved but the shared G-G pair is absent. However, the shared pair could be replaced by G-U or A-U base pair, which are often seen. The major features of this last group are the extended base pairing potential between the loops and the increased length of helix 1.2. The sequence complementarities allow four base pairing interactions between the loops. The remaining nucleotides of loop L1.2 have no apparent complementarities and may form a large loop of up to fifteen nucleotides. However, thermodynamic studies showed that loops containing four or five nucleotides are the most stable (Groebe and Uhlenbeck, 1988) and we therefore favoured a shorter loop (Figure 32B). Comparison of the Archaea Alu RNA sequences in this context exhibit an interesting conservation of a G residue, in helix 1.2. Its presence in all archaeal RNA sequences might highlight an important structural function. The structure we proposed might be a plausible hypothesis. Nucleotides of loop L1.2 that are not involved in base pairing with loop L2 may stack into a helical structure, thereby extending helix 1.2. In that case, the conserved G, by being bulged out of the helix might allow the proper bending of this longer helix 1.2, thus enabling base pairing between the loops despite the longer helix.

      Bacillus and Clostridium SRP RNAs have also an Alu-like domain which resembles the archaeal Alu domain. Protozoa and Fungi SRP RNAs have truncated Alu domain, lacking one or both stem-loop structures. Interestingly, Trypanosoma SRP acquired an additional RNA molecule that can be folded into a tRNA-like structure. The acquisition of this second RNA molecule might be an evolution compensation to the lack of specific tertiary structure in Trypanosoma SRP RNA Alu domain (Beja et al., 1993; Liu et al., 2003).

      The involvement of the Alu RNA base pairing structure in the elongation arrest function might not be surprising. In the crystal structure, these loops are located on the opposite side of SRP9/14, and both the proteins and the RNA loops are solvent exposed. The comparison of the Alu RNP structure (Weichenrieder et al., 2000) with Saccharomyces cerevisiae tRNA structure (Ladner et al., 1975) reveals that both complexes have a relatively similar length (Figure 33). This implies that the SRP Alu domain may fit into subunit interface of ribosomes like tRNA. Furthermore the CCA end and the anticodon loop of the tRNA are located on opposite side of the structure and exposed to solvent, allowing the interaction of the CCA end with the peptidyl transferase centre when charged with an aminoacyl residue, and of the anticodon with the mRNA. Since, the protein and the base paired RNA loops of the SRP Alu domain are also located on opposite side of the molecule; they may both interact with components of the translational machinery.

      Left panel: Alu RNP crystal structure fromWeichenrieder et al., 2000. The protein SRP9 and SRP14 are shown as red and green ribbon respectively. Right panel: Saccharomyces cerevisiae Phe-tRNA crystal structure (Ladner et al., 1975).

      An active role of the RNA in SRP function is consistent with the high conservation of the RNA among particles of all kingdoms. With the notable exception of chloroplast SRP, all particles comprise at minimum a RNA and a fifty four homologue protein.

3.3. Model for elongation arrest activity

      New information about the interaction of SRP with ribosomes has emerged during the last three years. Cross-linking studies in Mammals and Bacteria positioned the SRP54 subunit near the ribosome polypeptide exit site (Pool et al., 2002; Gu et al., 2003). In our laboratory we have shown that the Alu domain of mammalian SRP is located at the ribosome subunit interface, and that the SPR14 carboxyl-terminal region interacts with the ribosomal GTPase associated centre. Moreover, nucleotides 97-110 of 7SL RNA have been shown to be more protected from chemical modification in polysome-bound SRP than in soluble SRP (Andreazzoli and Gerbi, 1991), supporting an interaction of this region of the SRP RNA central stem with ribosomes. Compiling all this information provided us with some clues of how to position the Alu domain within the bacterial ribosomal structure described by Yusupov et al., 2001. I therefore introduced manually the structure of the Alu domain into the bacterial ribosome structure (Figure 34).

      The Alu RNP crystal structure (Weichenrieder et al., 2000) was placed in Thermus thermophilus ribosome crystal structure (Yusupov et al., 2001) according to the latest biochemical data available (see text for detail). 1) Alu SRP proteins 2) Alu SRP RNA 3) ribosomal protein S12 4) ribosomal protein L12 5) A site tRNA 6) 23S rRNA 7) 16S rRNA. The ribosomal protein L12 is part of the ribosome GTPase associated centre. This figure was made with Rasmol.

      When placed in the ribosome, the Alu domain fits the ribosomal subunit interface (Figure 34), which was expected from the size comparison with Phe-tRNA (Figure 33). Interestingly, placing the protein moiety of the Alu domain toward the large ribosomal subunit (Figure 34-1), directed the RNA loops toward the small ribosomal subunit (Figure 34-2). According to the SRP RNA central stem interaction with the ribosome (Pool et al., 2002; Gu et al., 2003), our modelling places the RNA loops in the vicinity of the ribosomal protein S12 (Figure 34-3). Our model leads to the orientation of the protein moiety of the Alu domain towards the large ribosomal subunit, whereas the RNA moiety faces the small subunit, a striking similarity with tRNA CCA end and anticodon localisation within the ribosome. Hence, SRP might interfere with the nascent chain elongation by blocking the elongation factors entry into the ribosome, or by impeding with the movement of the ribosomal subunits with respect to each other.

      An interference of SRP with the small ribosomal subunit function is in agreement to the particle's time of action. By the use of different antibiotics that block the elongation at different steps, Ogg and Walter, 1995 showed that SRP binds to ribosome when it has just completed the transpeptidylation reaction, but before it has undergone translocation of the peptidyl tRNA from the A site to the P site. In Moazed and Noller, 1989 hybrid state model for translation (Figure 35), binding and movement of tRNA between the A and P sites of the small ribosomal subunit is uncoupled from its movement between A, P and E sites of the large subunit. Thus tRNA movement in the hybrid state model occurs in two steps: 1) movement with respect to the large subunit; 2) movement of both tRNA along with their bound mRNA with respect to the small subunit. Furthermore, movement into the hybrid states is spontaneous and occurs independently of elongation factor two (EF2) and GTP. The EF2 however catalyses the movement of the anticodon ends of both tRNAs relative to the 30S subunit, moving peptidyl-tRNA into the P/P state and deacetylated tRNA in E state (Wilson and Noller, 1998). Hence a block of the elongation cycle, taking place between the transpeptidylation reaction and the translocation of the peptidyl tRNA from the A site to the P site is likely to be achieved by interference with the small ribosomal subunit function.

      The tRNA binding sites on the 50S and 30S subunits are represented schematically by the upper and lower rectangles, respectively. The 50S subunit is subdivided into A (aminoacyl), P (peptidyl), and E (exit) sites; the 30S subunit is subdivided into A and P sites. The tRNAs are represented by vertical bars and the nascent chain by a wavy line; aa represents the aminoacyl moiety and OH the deacylated 3' end of tRNA; the circles represents elongation factors 1 and 2, EF-1 and EF-2 respectively. The binding states, indicated at the bottom of each panel, indicate the state of each ribosomal subunit.

      The rescue of elongation arrest function of SRP RNAs mutated in loop L2 and, to a lower extent in L1.2, without conservation of the loops primary sequence, suggests that it involves an interaction with the RNA structure, and not specific nucleotide recognition. Due to the position of the Alu RNA in our modelling, we favour an interaction with a protein component of the translational machinery rather than an interaction with a ribosomal RNA. The interaction might be achieved through recognition of the Alu RNA loop L2 phosphate backbone. We think that L2 is the target of the interaction because only mutations in L2 had a drastic effect on the elongation arrest function; hence, compensatory mutations might rescue the function by restoring the base stacking of loop L2.

      Bacterial protein S12 and its mammalian homologue, S23, are good candidates for a putative interaction leading to the arrest of the nascent chain elongation. The S12 has been shown to interact with the aminoacyl tRNA in the A-site (Stark et al., 2002). In fact the protein modulates the translation fidelity by stabilising tRNA binding during initial selection (Rodnina et al., 2002 and references therein). Discrimination between cognate and non-cognate tRNA in the A site is coupled with EF1 GTP hydrolysis (Hopfield, 1974; Ninio, 1975). Furthermore the cognate rate of EF1 GTP hydrolysis is four orders of magnitude higher than the non-cognate rate (Rodnina et al., 1996). Hence, binding of the cognate tRNA causes an increase rate of EF1-dependent GTP hydrolysis, leading to the preferential release of EF1-GDP from the ribosome. Less information is available about EF2 interactions, but it can be cross-linked with mammalian S23 protein (Uchiumi et al., 1986). Domain IV of EF2 has structural similarities with the anticodon domain of the tRNA-EF1 ternary complex (Nissen et al., 1995; Nyborg et al., 1996). Indeed the orientation of EF2 in the ribosome is very similar to the one of EF1 (Stark et al., 1997), and EF2 domain IV is directed toward the decoding centre with its tip positioned closed to several rRNA elements of the small subunits tRNA sites (Wilson and Noller, 1998 and references therein). Furthermore the domain IV of EF2 is important for coupling GTP hydrolysis to movement of tRNA since an EF2 where domain IV is truncated results in a 50-fold reduction in the rate of translocation, without affecting the binding of EF2 to ribosome (Rodnina et al., 1997). EF2 dependent GTP hydrolysis precedes and greatly accelerates the rearrangement that leads to tRNA translocation. Thus, interfering with this step is likely to affect the rate of the nascent chain elongation. Furthermore, EF2 hydrolysis takes place immediately after binding to ribosome (Rodnina et al., 1997).

      We therefore propose the following model for SRP-mediated arrest of nascent chain elongation. The Alu domain of SRP binds ribosomes, allowing S domain binding near the nascent chain exit site. Through the interaction of SRP14 carboxyl-terminal with rGAR RNA, and the interaction of the base paired RNA loops with S23, EF2 is prevented from binding to the ribosome, blocking tRNA translocation in the small subunit A site to the P site, therefore arresting translation.

      Thus, according to this model, the simultaneous mutations of SRP14 carboxyl-terminus and of the RNA L2 loop should impair the ability of the SRP Alu domain to interact with the ribosome.

      The continuous increase of preprolactin translocation in 2L2 SRP, despite the reduced elongation arrest activity can be correlated to primary binding of SRP14 to rGAR RNA, allowing S domain interaction with the ribosome. Therefore the translocation efficiency increased in correlation with the concentration of 2L2 SRP. The key role of SRP14 carboxyl-terminal primary binding to ribosome is supported by the complete loss of elongation arrest function upon SRP14 carboxyl-terminal truncations.

      The interaction of the Alu RNA loops with S23 protein is supported by the conservation of the function without conservation of primary sequence. Interference of this interaction with translation is supported by a reduced rate of EF2 GTP hydrolysis, similar to that observed in non-cognate tRNA-EF1 (Rodnina et al., 1996). Elongation arrest is just transient in a homologous system (Lipp et al., 1987; Wolin and Walter, 1989), which is in agreement with our model since an EF2 where domain IV is truncated results in a reduction of tRNA translocation rate (Rodnina et al., 1997). Therefore, interfering with EF2 domain IV-mediated coupling of GTP-hydrolysis with tRNA movement in the small subunit should result in a reduced speed of the nascent chain translation.

      The model is an interesting working hypothesis in that it supposes an interaction of the SRP Alu domain with two partners which are critical for the GTP hydrolysis of elongation factors: the GTPase centre of the large subunit, and the decoding centre of the small subunit.

      Furthermore this model, where interactions with both ribosomal subunits lead to the arrest of nascent chain elongation, is supported by the role of bacterial 4.5S SRP RNA in translation. In Escherichia coli, 4.5S RNA has two different roles, one related to SRP and requiring Ffh binding and another one independent of Ffh and not related to SRP, required to maintain a normal rate of translation via its binding to EF2. Hence mutations in the large ribosomal subunit GTPase centre and in the 16S rRNA, altering the decoding fidelity, suppress 4.5S RNA requirement for translation (Brown, 1989; Prescott and Dahlberg, 1990; O'Connor et al., 1995; Brunelli et al., 2002).

      Indeed, our model implies an SRP evolutionary process where the acquisition of the Alu domain results in the acquisition of an additional role of the particle that affects the nascent chain elongation. Interestingly, whereas bacterial 4.5S RNA was required for translation independently of SRP, the mammalian particle arrests the nascent chain elongation. It is even more striking to notice that the translational role of 4.5S RNA is achieved through binding with EF2, whereas in our model the elongation arrest is achieved by impairing EF2 binding to the ribosome.

4.1. Mutated RNAs genesis

      Mutations at specific sites in the RNA sequences were introduced by polymerase chain reaction (PCR) using the QuickChange^ä method (Stratagene). DNA oligonucleotides, containing the desired mutation, flanked by 15 nucleotides, were used to produce the mutated SRP RNAs (Interactiva, Germany).

      The PCR reaction was done in 20 mM Tris-HCl pH 8.8, 10 mM KCl, 10 mM (NH₄)₂SO₄, 2 mM MgSO₄, 0.1% (v/v) Triton^® X-100, 0.1 mg/ml bovine serum albumin (BSA) containing 4 ng of template DNA, 100 ng of sense and of antisense oligonucleotides, 0.25 mM dNTPs and 2.5 U of Pfu DNA polymerase (Promega). The following clones were produced, either in a small 86-nucleotide Alu RNA previously described: pSA86H Weichenrieder et al., 1997, or in the full length SRP RNA :

      pSaG₁₅, pSaC₃₃, and pSa2L1.2 were produced by mutation of pSA86H as template. pSa 1 Comp was produced by mutation of pSaG₁₅ as template. pSa 2 Comp was produced by mutation of pSa2L1.2 as template. pSaG₁₅G₃₄G₃₇ was produced by mutation of pSa2L1.2 as template. pSa 3 Comp was produced by two step mutations: first on pSA86H, then on the clone obtained in the first step. pSa2L2 was derived from p2L2 by ligation: the fragment of 2L2 containing the desired mutation is exchanged with the corresponding fragment in the pSA86H. pSaC₁₃C₁₄C₃₃ was produced by mutation of pSa2L2.

      p2L2, p2L1.2, p3L2, p3L1.2 clones were obtained by the introduction of mutations p7Sswt sequence (Strub et al., 1991). p2 Comp was generated using p2L2 as template. p3 Comp was generated using p3L2 as template. p1 Comp was produced from pSa1 Comp and p2Sswt by ligation: the fragment of pSa1 Comp containing the desired mutation was exchanged with the corresponding fragment of p7Sswt.

      pAC, pGU, pGG, pAG, pCU, pU₁₄U₁₅, pA₃₃A₃₄, pA₁₄A₁₅, and pU₃₃U₃₄ were produced from p7Sswt as template. pACGU, pCCGG, pAGCU, pU₁₄U₁₅A₃₃A₃₄, and pA₁₄A₁₅U₃₃U₃₄ were derived from the following template respectively: pAC, p2L2, pAG, pU₁₄U₁₅ and pA₁₄A₁₅.

      All clones sequences were verified by automatic sequencing using Sp6 primer (5'accttatgtatcatacacat-3').

4.2. Nomenclature used for the mutated RNAs

      To facilitate the understanding of the mutations introduced in the SRP RNA sequence we used a specific nomenclature. The names are given such that the first part refers to the number of base pairing disrupted (or restored), whereas the last part referred to the loops in which the mutations where introduced (e.g. 2L2 means that 2 base pairing where disrupted in loop L2). In case of compensatory mutations the first part referring to the number of base pairing involved in the mutation is followed by the comp suffix.

4.3. Large scale transcription

      Plasmids obtained in Escherichia coli strain XL1-blue, purified on Qiagen Midi cartridge, were linearized with Xba (Gibcobrl) before in vitro transcription with T7 RNA polymerase. Plasmid pMR7wt (kind gift from Dr Mallet Arnaud et al., 1997) was used to produce T7 RNA polymerase. Reactions were done in 100 µl of 40 mM TrisHCl, pH 8.1, 20 mM MgCl₂, 1 mM spermidine, 5 mM dithiothreitol (DTT), 4 mM of each rNTPs, 1 mg/ml BSA, 100 ng/ml DNA template and 22 U/ml T7 RNA polymerase. Nucleotides were obtained from Sigma and stored as 50 mM solutions at -80°C.

      After two hours at 37°C, the transcription reaction was submitted to RQ1 RNase free DNase treatment (Promega). The mix was then phenol extracted and purified through a 1 ml Sephadex G50 fine column (Pharmacia). After ethanol precipitation, the RNAs are kept in sterile water. Their concentrations were measured by OD₂₆₀ nm, and verified on a denaturing acrylamide gel.

4.4. Protein purification

4.5. SRP reconstitutions

      All SRP were tested in at least two independent experiments.

4.6. RNA binding assay

      hSRP14 was synthesised in 10 ml wheat germ translation reactions in presence of 1 pmole of recombinant hSRP9 as described (Bovia et al., 1997). The heterodimer h9/14 was bound to 2 pmole of biotinylated SRP RNAs in 50 mM HEPES-KOH pH 7.5, 350 mM potassium acetate pH 7.5, 3.5 mM magnesium acetate and 0.01% (v/v) Nikkol. The bound protein was purified away from unbound protein by the use of Streptavidin magnetic beads (Dynal) as described in Bovia et al., 1994. Bound protein was analysed by 15% SDS-PAGE, visualised by autoradiography and quantified by phosphoimager (BioRad).

4.7. Hydroxyl radical cleavage reaction

      Alu151 RNA was labelled at its 3' end with [³²P]pCp (g[ ³²P] 370 kBq/ml) and RNA ligase as described (England et al., 1980). Binding of labelled Alu151 RNA to recombinant hSRP9/14K95 or hSRP9/14 was performed in 4 ml reactions in final salt concentrations of 20 mM HEPES-KOH pH 7.5, 300 mM potassium acetate pH 7.5, 3.5 mM magnesium acetate, 1 mM DTT and 0.01% (v/v) Nikkol. In parallel sample, which serves as negative control, SRP9/14 was replaced by equal amount of bovine serum albumin. Protein-to-RNA ratios were 1:1 or 1:3. After binding, the samples were diluted to 20 ml in final salt concentrations of 20 mM HEPES-KOH pH 7.5, 300 mM potassium acetate pH 7.5, 3 mM magnesium acetate, 4 mM DTT and 0.01% (v/v) Nikkol. To perform the cleavage of the RNA, the following four solutions were combined in a drop at the side of the tube in the given order: 1) 25 mM ferrous ammonium sulfate [Fe(NH₄)₂(SO₄)₂], 2) 50 mM EDTA, 3) 125 mM sodium ascorbate, and 4) 2.5% (v/v) hydrogen peroxide. The drop was spun into the samples, which were then incubated on ice for 3 minutes. The reaction was stopped by the addition of 25 ml of 0.1 M thiourea and 10 ml of 3 M sodium acetate, 1 mg of Escherichia coli tRNA, and 40 ml of TE, followed by phenol extraction and precipitation in ethanol. The samples were dissolved in 4 ml of sample buffer and the reaction products were analysed by autoradiography following separation on a 6% sequencing gel. To trace the sequence of the RNA, sequencing reaction were created with either 0.1 U RNase T1 (Pharmacia) for guanosine, 1 U RNase U2 (Pharmacia) for adenosine, and RNase B. cereus (Pharmacia) for cytosine and uracile. A complete ladder was also obtained by partial alkaline hydrolysis in 0.1 M NaHCO₃, pH 9.5 for 3 minutes at 95°C. All sequencing samples contained 1 mg of unlabeled Escherichia coli tRNA.

4.8. Electromobility shift assay

      The electromobility shift assay was derived from Weichenrieder et al., 2001. Sa86 RNA, labelled by transcription with a [³²P] UTP (24.10¹⁵ Bq/mmol) was used as tracer. To set up the conditions of the competition experiments, 200 nM cold Sa86 RNA (containing 20 000 Cpm of labelled SA86 RNA as tracer) was titrate against SRP9/14 protein with concentration ranging from 25 nM to 600 nM. Conditions allowing 100% of protein-RNA binding were achieved with final concentration of 200 nM RNA and 190 nM SRP9/14.

      Individual reconstitution assays were done in 20 mM Hepes pH 7.5, 300 mM potassium acetate, 10 mM magnesium acetate, 5 mM DTT, 0.015% (v/v) Nikkol. For reconstitution, 5 µl of RNA pSA86H (containing 20 000 Cpm of labelled RNA) were recombined with 5 µl of competitor RNA and 10°µl of human SRP9/14 heterodimer. The binding reaction was allowed to proceed for 10 minutes on ice, then for 10 minutes at 37°C. Labelled and unlabelled RNAs were annealed in assay buffer without DTT and Nikkol, immediately before use at 4x of final concentration. Competitor RNA final concentrations ranged from 50 to 800 nM. As negative control, samples consisting of Sa86 in absence of protein, and Sa86 in presence of bovine serum albumin were loaded on the gel.

4.9. Computer programs

      The figures of proteins or RNA structure were realised by the generation of povscript file with SwissPDB viewer or Ribbons. The povscript files were then rendered with Povray.

      Pdb files that contain only alpha carbon cannot be opened with Swisspdb or Ribbons, figures derived from such files were realised with Rasmol.