Co-translational membrane association of the Escherichia coli SRP receptor

Many integral membrane proteins (IMPs) are synthesized by membrane-bound ribosomes and are transferred directly from the ribosomes into the membrane by protein translocases or insertases (du Plessis et al., 2011; Facey and Kuhn, 2010; Johnson, 2009; Kudva et al., 2013; Neugebauer et al., 2012; Rapoport, 2007; Zimmermann et al., 2011). The targeting of ribosomes translating IMPs to the translocases or insertases is proposed to be mediated mainly by the evolutionarily conserved signal recognition particle (SRP) system composed of the SRP (Ffh protein and 4.5S RNA in Escherichia coli) and its receptor (FtsY in E. coli) (Walter and Johnson, 1994). All three components of the E. coli SRP system are essential for growth (Brown and Fournier, 1984; Gill and Salmond, 1987; Luirink and Dobberstein, 1994; Phillips and Silhavy, 1992) and for the proper expression and/or assembly of IMPs (Yosef et al., 2010).

As proposed, the E. coli SRP pathway begins when ribosomes translating IMPs interact with SRP in the cytosol and then the complex targets the membrane-associated SRP receptor, FtsY (Akopian et al., 2013; Luirink et al., 2012). Another view of the pathway suggests a different order of events (Bibi, 2011) in which the receptor also plays a central role. The functional core of FtsY, which is termed NG+1 (Eitan and Bibi, 2004), contains two structurally defined domains (Fig. 1A) – an ∼84-amino-acid-long N-domain, which forms a four-helix bundle, and a C-terminal ∼217-amino-acid-long catalytic G-domain (GTPase). Previous studies of FtsY revealed the receptor interaction with membrane-bound ribosomes (Herskovits et al., 2002) and with the SecYEG translocase (Angelini et al., 2005), and showed how FtsY is regulated on the membrane by acidic lipids (Bahari et al., 2007; Braig et al., 2009; Erez et al., 2010; Parlitz et al., 2007). Other studies showed that FtsY plays a central role in vivo in ribosome targeting to the membrane (Herskovits and Bibi, 2000) and in the productive expression of IMPs (Seluanov and Bibi, 1997; Yosef et al., 2010). Taken together, it is generally accepted that FtsY functions as a membrane-associated protein, as has been shown recently (Mircheva et al., 2009). However, unlike FtsY from some ancient Gram-positive bacteria, homologous FtsY proteins in other species are not IMPs (Bibi et al., 2001). Therefore, further attempts to elucidate the mechanism underlying the membrane targeting and association of FtsY might be crucial for a better understanding of the SRP pathway and the pathway of membrane protein biogenesis.

In the past, indirect studies have raised the possibility that FtsY is targeted to the membrane co-translationally (Bibi et al., 2001; Herskovits et al., 2001; Zelazny et al., 1997). Here, we investigated this hypothesis directly, by utilizing FtsY-derived translation intermediates in vivo. The results show that FtsY can target the membrane co-translationally. We also identify the receptor domains that are required for targeting and membrane association. These results lend support to the hypothesis that the SRP receptor delivers ribosomes to the membrane during its own translation (Bibi, 2011; Bibi, 2012) and offer new insights into the SRP pathway and its regulation.

To test whether FtsY targets the membrane co-translationally (Bibi, 2011) and to investigate systematically which one or more domains of FtsY mediate the targeting, we used a functional A-domain-deleted NG+1 version of FtsY (Fig. 1A) (Eitan and Bibi, 2004) and produced in vivo stalled ribosome-nascent chain complexes of selected lengths (translation intermediates, TIs). The translation intermediates were constructed by utilizing the SecM translation-stalling sequence (TSS) (Nakatogawa and Ito, 2001). The SecM sequence was initially inserted into two different sites that allow emergence from the ribosome of either only a portion of the N-domain (TI-N_1–3) or all of the N-domain (TI-N_1–4). As a control, we utilized the unrelated cytosolic protein FruK (TI-FruK) with the SecM TSS inserted at the same distance from the N-terminus of this protein as in TI-N_1–4 (Fig. 1B). We then examined the growth of cells expressing the various translation intermediates. Fig. 1C (left panel) shows that, unlike the translation intermediate TI-FruK, expression of the FtsY-derived intermediates TI-N_1–4 and TI-N_1–3 conferred toxicity in wild-type cells. Notably, however, the observed effect was serious but not completely detrimental. A possible explanation is that the SecM-mediated arrest is not very tight in wild-type E. coli. Therefore, we examined the toxic effect of the translation intermediates in a strain whose transfer-messenger (tm)RNA system (SsrA and SmpB) was deleted; this strain is defective in the release of stalled ribosomes, including those that are SecM mediated (Collier et al., 2004). Fig. 1C (right panel) shows that, indeed, in ΔssrAsmpB cells, the toxicity of TI-N_1–4 and TI-N_1–3 was more pronounced and no growth was detected, whereas cells expressing TI-FruK grew well. These results raised the possibility that the FtsY translation intermediates are dominant negative and compete with the essential native receptor by sequestering factors involved in FtsY biogenesis and/or function. To address this issue, we overexpressed in trans a functional receptor (NG+1) or an unrelated protein (CspE), and observed that only NG+1 significantly relieves the toxicity of TI-N_1–4 (Fig. 1D, right panel), strongly suggesting that the toxic effect of TI-N_1–4 results from competition with the wild-type receptor.

Next, we investigated whether the translation intermediates are targeted to the membrane as ribosome-nascent chain complexes (RNCs). This was analyzed by cell fractionation and density gradient centrifugation (Fig. 2A). As shown, the translation intermediates are synthesized and migrate either with the cytosolic ribosomal fractions (fractions 10 and 11) or with membranes (pellet). Importantly, the results show that only TI-N_1–4 migrates almost exclusively with the membrane fraction. In contrast, TI-N_1–3 migrates mainly together with cytosolic ribosomes, as does the control TI-FruK. These results show that emergence from the ribosome of the entire N-domain (TI-N_1–4) is required for co-translational targeting and for association of the RNC with the membrane. This was confirmed by isolating membranes using flotation in a step-wise sucrose gradient, which showed the same results (Fig. 2B).

Typically, a portion of the SecM-stalled translation intermediates remains covalently attached to a tRNA molecule (Nakatogawa and Ito, 2001). To ascertain this in our preparations, an enriched sample with TI-N_1–4 RNCs was treated with RNaseA to cleave the peptidyl–tRNA ester bond. Fig. 2C shows that, indeed, RNaseA treatment released an immunoreactive peptide that migrated as expected, according to the theoretical mass of TI-N_1–4 (∼16 kDa). In order to establish the identity of the stalled peptide, we used affinity-based enrichment of stalled ribosomes and analyzed TI-N_1–4 by mass spectrometry. Briefly, cells expressing 6His-tagged TI-N_1–4 or an empty vector were treated as described in Materials and Methods, and samples were separated by Tris-tricin SDS-PAGE (Fig. 2D). The stained gel shows mainly ribosomal proteins, which nonspecifically bound to the NTA beads, as expected (Schaffitzel and Ban, 2007). Notably, however, above the ribosomal protein background, there were two specific bands, which appeared only in samples of TI-N_1–4. These bands were analyzed by mass spectrometry. The results identified band 1 as the full-length NG+1-SecM, which escaped the translation arrest. Band 2 was identified as the expected translation intermediate TI-N_1–4, containing only sequences of SecM and the N-domain of FtsY.

The translation intermediate experiments demonstrated that co-translational targeting of FtsY to the membrane in vivo is feasible, and showed that the targeting requires exposure of the N-domain from the ribosome. By utilizing fully translated FtsY mutants, we investigated which one or more domains of FtsY mediate its targeting and membrane association in vivo under non-arrested conditions. We have previously shown that the A-domain is not essential for targeting and function (Bahari et al., 2007; Eitan and Bibi, 2004). Therefore, we utilized the functional NG+1 for the following studies (Fig. 3A). Fig. 3B shows that deletion of the first helix of the N-domain was detrimental and NG+1ΔN₁ did not support the growth of FtsY-depleted cells. This is not surprising because even deletion of the most N-terminal amino acid of NG+1 (NG) is known to be deleterious (Eitan and Bibi, 2004) (Fig. 3B). In addition, we deleted the entire G-domain and, as expected, this mutant (N_1–4) was also unable to complement FtsY depletion (Fig. 3B). When the mutated proteins were expressed in wild-type E. coli (Fig. 3C), we observed that, like NG (Erez et al., 2010), NG+1ΔN₁ also conferred a serious growth defect. Importantly, in contrast to TI-N_1–4 (Fig. 1C), the expression of non-arrested N-domain (N_1–4) had no detectable effect on the growth of wild-type cells (Fig. 3C), demonstrating that competition with the wild-type receptor happens only during the co-translational targeting. We then examined whether the mutated proteins associate with the membrane. All the mutants were found in the membrane fraction (Fig. 3D), suggesting that neither N₁ nor the G domain is essential for the association of the receptor with the membrane. Fig. 3E shows that the N-domain also targets a fused foreign protein (GFP) to the membrane. Next, given that NG+1ΔN₁ targets the membrane, we asked whether N₁ is required for targeting during translation. By utilizing an additional translation intermediate, TI-N_2–4, (Fig. 3F, upper panel), we show that, in agreement with the deletion analysis described above, N₁ is not essential for co-translational membrane localization of the RNC. These results also suggest that helices 2–4 of the N-domain might form an independent structural domain (Fig. 3G) that mediates the receptor interaction with the membrane. As shown in Fig. 3H, the N_2–4 domain of FtsY does not contain hydrophobic segments such as those that exist in typical membrane proteins.

Next, to gain additional insight into the co-translational nature of FtsY targeting, we fused the SecM-TSS to the C-terminus of the otherwise functional NG+1 mutant (TI-NG+1). As a control, we utilized a mutated version of the SecM TSS (Pro166Ala) (termed mTI-NG+1), which had lost most of its translation arrest capacity (Murakami et al., 2004) (Fig. 4A). The expression of the translation intermediate was assessed by RNaseA treatment of cytosolic ribosomal fractions and western blot analysis, where the smeared tRNA-attached nascent peptide has been converted to a sharp band representing the translation intermediate (Fig. 4B). Next, we investigated whether TI-NG+1 is targeted to the membrane as an RNC, and the results show that only a small amount of mTI-NG+1 migrated with cytosolic ribosomes, and most RNCs migrated with the membrane fraction (Fig. 4C, upper panel). In contrast, a large portion of TI-NG+1 migrated with the cytosolic ribosomes (Fig. 4C, lower panel), suggesting defective targeting. The difference between mTI-NG+1 (mostly released from the ribosome) and TI-NG+1 (remains tethered) suggests that membrane association is disrupted when the fully translated G-domain remains tethered at the ribosomal exit tunnel.

We then asked whether the effect is specific for the G-domain or whether another polypeptide at that location and with the same size as the G-domain might also inhibit membrane association of the RNCs. To address this, we replaced the G-domain-encoding sequence with that of GFP, which has a similar size (G-domain has 217 and GFP has 238 amino acids), and examined the membrane targeting of the new construct TI-N_1–4–GFP. As shown in Fig. 4D, the GFP in TI-N_1–4–GFP folds properly and produces fluorescent RNCs, which target the membrane. This indicates that the ribosome-anchored G-domain is specifically responsible for disrupting the ability of the N-domain to target and/or sustain the association of TI-NG+1 RNCs with the membrane. Obviously, this situation cannot happen with the wild-type receptor in vivo. The G-domain is located at the C-terminus of the receptor and, as such, it is released from the ribosome immediately at the end of translation. Nevertheless, the observed phenomenon might teach us about the organization of FtsY domains in the docking complex. Therefore, we further explored the reason for the inhibition of targeting by the ribosome-tethered G-domain.

One possible reason is that the GTPase activity of TI-NG+1 plays a role. To examine this, we constructed TI-NG+1(A335W), which is a GTPase-activity-compromised mutant (Shan et al., 2004). As a control, we constructed mTI-NG+1(A335W) harboring a mutated TSS, as described for mTI-NG+1. We then examined whether these translation intermediates are targeted to the membrane. The results show that mTI-NG+1(A335W) and TI-NG+1(A335W) behave identically to mTI-NG+1 and TI-NG+1, respectively (compare Fig. 4E with Fig. 4C). Given that both TI-NG+1 and TI-NG+1(A335W) migrated mainly with cytosolic ribosomes, we conclude that the GTPase activity is not responsible for TI-NG+1 mis-targeting.

Another possible explanation for why the G-domain interferes only when tethered to the ribosome is that its dual interaction with the ribosome exit tunnel and with the N-domain might restrict the orientation of the N-domain in such a manner that its association with the membrane is disrupted. Therefore, we destabilized the N–G interaction by deleting from the G-domain each of the two main peptides that mediate this interaction, L448-Q463 or/and K484-D497 (Fig. 4F). In contrast to TI-NG+1, these mutants were detected predominantly in the membrane fraction (Fig. 4G). We further mutated a key residue in the interacting region, D238R, which is located at the N-domain (Fig. 4F), and we observed that also this mutation partially restored membrane association of the translation intermediate (Fig. 4G, lower panel). These results suggest that FtsY targeting might be regulated at the translational level to allow timely interaction of the N-domain with its membrane target during translation, prior to translation termination (see Discussion).

It is well established that FtsY performs its activity on the membrane. However, how FtsY targets the membrane has thus far remained obscure. More specifically, it is unknown how the receptor targets the membrane – post- or co-translationally – and whether it has a primary docking site. E. coli FtsY is not an IMP (see also Fig. 3H) and, yet, several studies have raised the possibility that it might be targeted to the membrane in a co-translational manner (summarized in Bibi, 2011). The targeting of FtsY in vivo was shown to be SRP independent (Herskovits et al., 2002), as also shown in vitro for the mammalian SRP receptor (Young and Andrews, 1996). Certain bacterial SRP receptors are, in fact, membrane proteins, which most likely target the membrane co-translationally (Bibi et al., 2001; Shen et al., 2012). Similarly, fusing an active NG-domain version of the E. coli receptor to a membrane protein (LacY) does not abolish its function in vivo (Zelazny et al., 1997). Also in that case, the receptor is targeted to the membrane during translation following LacY (Macfarlane and Müller, 1995; Seluanov and Bibi, 1997). In accordance with this hypothesis, our current in vivo studies with FtsY translation intermediates demonstrate that a co-translational mode of membrane targeting for FtsY is feasible. We also show that emergence from the ribosome of the N-domain of FtsY is necessary for targeting and association of TI-N_1–4 RNCs with the membrane. Moreover, we show that, within the N-domain, helices 2–4 (N_2–4) harbor the targeting information and form an independent membrane-interaction domain. Finally, our results imply that the co-translational targeting of FtsY might be regulated (Fig. 4). By analyzing ribosome-profiling data and the observed ribosome occupancy along the FtsY mRNA (Li et al., 2012), we noted that there are three relatively strong pause sites (supplementary material Fig. S1). The first pause site is located at the beginning of the G-domain coding sequence, where the entire N-domain has already emerged from the ribosome. The other two pause sites are located towards the end of the G-domain mRNA, just before the two N–G interface peptides are translated. Notably, in all three pause sites, we identified Shine-Dalgarno-like sequences that were implicated in translational pause (Li et al., 2012). Pausing at these internal Shine-Dalgarno-like sites could regulate the targeting process by temporally coordinating the co-translational folding of the N- and G-domains, and thus allowing proper interaction and assembly of the N-domain on its membrane target. Interestingly, previous in vitro studies identified a stem-loop structure that might function as a pause site during translation of the α-subunit of the mammalian SRP receptor, SR (Young and Andrews, 1996).

How the targeting proceeds and what is the primary docking site for FtsY RNCs on the membrane are unknown. There are several known FtsY binding partners, including lipids, SecY, YidC and maybe other membrane-associated proteins (Angelini et al., 2005; Cristóbal et al., 1999; Erez et al., 2010; Welte et al., 2012). Regarding lipids, we show here that the previously identified lipid-responsive and interactive domains of FtsY (Bahari et al., 2007; Braig et al., 2009; Parlitz et al., 2007) are not required for RNC targeting (for example in TI-N_2–4). It was suggested that the translocon serves as a docking site (Angelini et al., 2005); however, because the A-domain is required for stable interaction of FtsY with the translocon (Kuhn et al., 2011) and our translation intermediates do not contain this domain, it is less likely. Attempts to identify additional putative membrane attachment site(s) are currently in progress.

Intriguingly, co-translational targeting of the receptor to the membrane means that both the receptor and its translating ribosome reach the membrane simultaneously. There are several indications that link the targeting of the SRP receptor to the targeting of its translating ribosomes (Bibi, 2011), although there is currently no direct empirical evidence indicating that these very ribosomes remain membrane bound at the end of FtsY translation.

E. coli TOP10 was used for propagation and preparation of various plasmid constructs. E. coli BW25113 [Δ(araD-araB)567, ΔlacZ4787::rrnB-3, λ⁻, rph-1, Δ(rhaD-rhaB)568, hsdR514] served in growth experiments and in vivo studies. E. coli BW25113ΔssrAsmpB was used for overexpression of FtsY translation intermediates and growth inhibition experiments. E. coli IY28 (BW25113-araCFUP-ftsY::kan) (Erez et al., 2010) was used for depletion of FtsY. Typically, cells were cultured in LB medium supplemented with the appropriate antibiotics (kanamycin was used at 30 µg/ml and ampicillin at 200 µg/ml). All constructs were expressed from the ampicillin-resistance plasmid pT7-5 under regulation of the tac promoter. Cultures were grown overnight at 37°C in LB, diluted to an A₆₀₀ of 0.01–0.05, and induced at an A₆₀₀ of 0.5 with 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside). Cultures were then grown for 1 h or 5.5 h for FtsY translation intermediates or studies with FtsY mutants, respectively.

All the mutations in FtsY were generated using standard PCR techniques with pT7-5-tacP-NG+1 as a template, and the plasmids were verified by sequencing. NG+1ΔN₁ was constructed with the forward primer 5′-AAAAAAATCGACGATGATCTGTTTGAG-3′ and reverse primer 5′- CATGGAATACTGTTTCCTGTGTG-3′. For constructing mutants NG+1(A335W) and NG+1(A335W)ΔN₁, we used the forward and reverse primers 5′-TGGGCGGTTGAACAGCTTCAGG-3′ and 5′- TGCACGGAAAGTATCACCCGCCGCC-3′, respectively. NG+1(D449N) was constructed with forward primer 5′-GCATCACGCTAACGAAACTGAACGGCACGGCGAAAGG-3′ and reverse primer 5′-CCTTTCGCCGTGCCGTTCAGTTTCGTTAGCGTGATGC-3′ (Asa Eitan and E.B., unpublished). N₁–₄ was constructed with the forward primer 5′-TAACAACGATTCGCTTTGAACATGTC-3′ and reverse primer 5′- CGCCAGAATCTCGCCCATCTCTTC-3′. For constructing the plasmid pT7-5-tacP-FruK, we amplified the fruK gene by PCR from E. coli MG1655, utilizing the forward primer 5′-CCGCCGGAATTCATGCATCATCATCATCATCATAGCAGACGTGTTGCTACTATCACCC-3′ and reverse primer 5′-GCCGCGGATCCTCAGTTAAAAGGTTGTAATCGAC-3′. The PCR product was digested with EcoRI and BamHI and used to replace NG+1 in pT7-5-tacP-NG+1, which was digested with the same enzymes. For constructing the plasmid pT7-5-tacP-(N_1–4-GFP), the N-domain of FtsY was amplified from pT7-5-tacP-NG+1 by PCR utilizing the forward primer 5′-GGGTACCGGTCGCCACCATGTTCGCGCGCCTGAAACGCAG-3′ and the reverse primer 5′-CAGCTCCTCGCCCTTGCTCACGCCTTCAACATTCAGCGGC-3′. The PCR product was then fused to GFP. For the construction of translation intermediates, the SecM TSS-encoding gene was PCR amplified from E. coli MG1655 and inserted at various sites in NG+1 or FruK as detailed in supplementary material Table S1. The translation intermediate constructs were designed in such a manner that the entire ‘test’ protein is fully exposed from the ribosome. Given that the SecM stalling motif is only 17 amino acids long, we included an additional 33 amino acids of SecM that precede the stalling sequence. The final length of 50 amino acids ensures that the entire test protein is exposed out of the ribosome during translation.

For growth inhibition experiments, E. coli BW25113ΔssrAsmpB was transformed with the indicated plasmid and plated on LB plates containing ampicillin (200 µg/ml) in the absence (as a control) or presence of an inhibitory concentration of IPTG (0.3 mM) needed for expression of the test constructs. For suppression of TI-N_1–4 toxicity, cells harboring either pCV3-araP-NG+1, pCV3-araP-CspE or empty pCV3 were co-transformed with plasmid encoding TI-N_1–4 and plated on LB plates containing ampicillin (200 µg/ml) and kanamycin (30 µg/ml), in the presence of a lethal concentration of the TI-N_1–4 inducer IPTG (0.5 mM) and 0.2% arabinose. FtsY complementation experiments were conducted by plating transformed IY28 cells, harboring the indicated constructs on LB agar plates with ampicillin (200 µg/ml) and kanamycin (30 µg/ml), with or without 0.2% arabinose (for induction of the chromosomal ftsY gene) and with IPTG, the inducer of the complementing construct. Plates were scanned after 17 h at 37°C.

Harvested cultures were washed in buffer containing 25 mM HEPES pH 7.5, 10 mM magnesium acetate, 50 mM potassium acetate, 5% sucrose, 250 mM NH₄Cl and 0.1 mM EDTA, and resuspended in the same buffer with 10 µg/ml DNaseI and 1 mM PMSF. Cell suspensions were then sonicated, incubated on ice for 20 min and subjected to low-speed centrifugation for the removal of cell debris. Membranes were collected by ultracentrifugation (1.5 h, 75,000 rpm, 4°C) using TLA100.3 rotor (Beckman) and then resuspended in the same buffer. The membranes were then purified by flotation in a three-layer solution (400 µl of 61% sucrose buffer, 680 µl of 53% sucrose buffer and 270 µl of buffer with no sucrose) by ultracentrifugation (17 h, 54,000 rpm, 4°C) using a TLS55 rotor (Beckman). Membrane-associated proteins were resolved by SDS-PAGE.

Harvested cultures were washed in buffer MNH [20 mM HEPES pH 7.5, 20 mM MgCl₂, 20 mM KCl, 5% sucrose, 60 mM NH₄Cl, 0.1 mM dithiothreitol (DTT)] and resuspended in the same buffer. Extracts were prepared by three cycles of brief sonication (5 s) at 1-min intervals on ice, followed by a low-speed centrifugation (16,000 g for 10 min) to remove cell debris. Ribosomes and membranes were collected by ultracentrifugation (90 min at 75,000 rpm at 4°C; TLA-100.2 rotor) in a tabletop Optima TLX ultracentrifuge (Beckman). Pellets were resuspended in ice-cold 5% sucrose solution in buffer MNH. Samples were loaded on top of a preformed sucrose density gradient (1.35 ml containing 0.27-ml layers of 25%, 20%, 15%, 10% and 7.5% sucrose; w/v). Following ultracentrifugation (52 min, 54,000 rpm at 4°C; TLS-55 rotor) fractions were collected from the top. The A₂₆₀ was measured for each fraction using a NanoDrop spectrophotometer. Every gradient experiment was repeated three times, and densitometry plots were generated for each western image. The density values of all bands were combined (100%) and the percentage density of each band was calculated.

12 mM EDTA was added to the cytosolic ribosome fractions, which were then incubated on ice for 20 min. Next, 50 µg/ml RNaseA (Sigma) was added for 1 h at 25°C. The samples were analyzed by SDS-PAGE and western blotting with anti-FtsY antibodies.

Harvested cultures of cells expressing TI-N_1–4 were washed in buffer MNH and resuspended in the same buffer. The cell suspensions were sonicated, incubated on ice for 20 min and subjected to low-speed centrifugation (20 min, 70,000 rpm at 4°C; SS34 rotor) to remove cell debris. Ribosomes were collected by overnight ultracentrifugation (16,000 rpm at 4°C; Ti-70 rotor). The pellets were solubilized in ice-cold Tris-HCl buffer pH 7.5, and subjected to low-speed centrifugation to remove non-soluble material. Supernatants were mixed with an equal volume of a solution containing 6 M LiCl and 10 M urea. Following incubation at −20°C for 24 h to precipitate the rRNAs, samples were centrifuged (20 min, 16,000 g at 4°C) and supernatants were dialyzed against urea buffer to dilute the lithium (6 M urea, 20 mM HEPES pH 7.5, 500 mM NaCl). Next, samples were incubated with pre-equilibrated nickel nitrilotriacetic acid beads (Ni-NTA; QIAGEN) in urea buffer supplemented with 5 mM imidazole for 1 h at 4°C with agitation. The beads were washed twice with wash buffer (6 M urea, 20 mM HEPES pH 7.5, 100 mM NaCl, 20 mM imidazole), and then four times with 0.1 ml of 6 M urea, 20 mM HEPES pH 7.5, 100 mM NaCl, 20 Mm imidazole). Proteins were eluted from the beads with a buffer containing 6 M urea, 20 mM HEPES pH 7.5 and 300 mM NaCl supplemented with 200 mM imidazole. The eluted material was concentrated 25-fold using Vivaspin 10 kDa cutoff and separated by 14% Tris-Tricine SDS-PAGE. Specific protein bands were excised from the Coomassie-stained gel. The protein bands were subsequently reduced, alkylated and in-gel digested with bovine trypsin (sequencing grade, Roche Diagnostics, Germany). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was performed and raw data files were searched with MASCOT (Matrix Science) against a NCBInr database.

SDS-PAGE and western blotting were performed as described previously (Herskovits et al., 2002) using antibodies against FtsY or GFP. India HisProbe-horseradish peroxidase was purchased from Pierce. Anti-GFP antibodies were obtained from BAbCO (Richmond, CA). Goat anti-rabbit-IgG and goat anti-mouse-IgG antibodies conjugated to horseradish peroxidase were used as secondary antibodies (Jackson Immunoresearch).

TI-N_1–4–GFP-containing sucrose gradient samples were separated by SDS-PAGE and the gels were washed thoroughly with water and scanned using the Typhoon 9400 scanner (Molecular Dynamics), using a fluorescence mode (excitation wavelength, 488 nm; emission wavelength, 520 nm).

We are grateful to Elena Bochkareva for experimental advice and Michal Danino for technical assistance. We thank Nir Fluman for help in analyzing ribosome-profiling data, and Asa Eitan and Liat Bahari for preliminary experiments with NG+1 mutants. We thank Nir Fluman and Daniel Ben Halevy for critical comments on the manuscript, and all laboratory members for fruitful discussions.