in Order to Facilitate Posttranslational Translocation in Bacteria
Damon Huber,1Nandhakishore Rajagopalan,1Steffen Preissler,1,2Mark A. Rocco,1,3Frieder Merz,1,4Gu¨nter Kramer,1 and Bernd Bukau1,*
1Center for Molecular Biology of the University of Heidelberg (ZMBH), ZMBH-DKFZ Alliance, Im Neuenheimer Feld 282, D-69120 Heidelberg, Germany
2Present address: Department of Biology, University of Konstanz, D-78457 Konstanz, Germany
3Present address: Department of Biomedical Engineering, Cornell University, Ithaca, NY 14853, USA
4Present address: Molecular Partners AG, Wagistrasse 14, CH-8952 Schlieren-Zurich, Switzerland
*Correspondence:bukau@zmbh.uni-heidelberg.de DOI10.1016/j.molcel.2010.12.028
SUMMARY
In Escherichia coli, translocation of exported proteins across the cytoplasmic membrane is dependent on the motor protein SecA and typically begins only after synthesis of the substrate has already been completed (i.e., posttranslationally).
Thus, it has generally been assumed that the translo-cation machinery also recognizes its protein substrates posttranslationally. Here we report a specific interaction between SecA and the ribo-some at a site near the polypeptide exit channel.
This interaction is mediated by conserved motifs in SecA and ribosomal protein L23, and partial disrup-tion of this interacdisrup-tion in vivo by introducing muta-tions into the genes encoding SecA or L23 affects the efficiency of translocation by the posttransla-tional pathway. Based on these findings, we propose that SecA could interact with its nascent substrates during translation in order to efficiently channel them into the ‘‘posttranslational’’ translocation pathway.
INTRODUCTION
In E. coli, the translocation of most soluble periplasmic and outer-membrane proteins across the cytoplasmic membrane is carried out by the Sec machinery and begins only after all or a significant portion of the substrate protein has already been synthesized (i.e., posttranslationally) (Driessen and Nouwen, 2008; Rapoport, 2007). The posttranslational translocation machinery consists minimally of two components: the SecYEG translocon and the ATPase SecA. SecYEG forms an hour-glass-shaped hydrophilic pore in the cytoplasmic membrane through which proteins must pass in an unfolded conformation
repeated rounds of ATP binding and hydrolysis (Brundage et al., 1990; Erlandson et al., 2008). Periplasmic and outer-membrane proteins are targeted to the Sec translocation machinery by an N-terminal signal sequence, which is proteolyt-ically removed from the protein during translocation (Rapoport, 2007). Alternatively, a small subset of proteins with highly hydro-phobic signal sequences target proteins for efficient, cotransla-tional translocation, which is dependent on the signal recognition particle (SRP) (Huber et al., 2005).
Although much is known about the mechanism of posttransla-tional translocation, substrate recognition by this pathway is much less well understood. One possibility is that SecA directly recognizes substrates by interacting with their N-terminal signal sequences (Gelis et al., 2007). Alternatively, it has been sug-gested that the dedicated Sec chaperone SecB recognizes posttranslational substrates by specifically interacting with low-affinity sites in the unfolded protein and passing it to SecA (Fekkes et al., 1998; Hartl et al., 1990).
Regardless of which component is responsible for recog-nizing substrate proteins, all current models for posttransla-tional translocation agree that recognition of substrate proteins occurs posttranslationally without involvement of the ribosome (Driessen and Nouwen, 2008; Rapoport, 2007). However, a number of recent genetic studies indirectly suggest that substrate recognition could occur cotranslationally. For example, the rate of signal sequence processing is faster in strains lacking the ribosome-associated chaperone trigger factor (TF) and translocation defects caused by mutations in the Sec machinery can be suppressed in strains expressing ribosome-binding-deficient TF (Lee and Bernstein, 2002; Ullers et al., 2007). Likewise, increasing overexpression of wild-type (WT) TF slows the rate of protein translocation (Lee and Bern-stein, 2002).
In the present study, we report that SecA binds to ribosomes specifically in a 1:1 stoichiometry. SecA binds to a site on the ribosome that includes ribosomal protein L23, which is situated near the opening of the polypeptide exit channel on the large subunit of the ribosome. In addition, we identified two conserved
the posttranslational pathway in vivo. Our findings suggest that SecA could cotranslationally interact with its substrates on the ribosome in order to channel them into the posttranslational translocation pathway.
RESULTS
SecA Binds to Ribosomes in a Specific Fashion
In order to test if SecA can bind to ribosomes, we incubated puri-fied SecA with ribosomes at equimolar concentrations and sepa-rated ribosome-bound SecA from unbound SecA by pelleting ribosomes through a 30% sucrose cushion by ultracentrifuga-tion. The majority of the SecA present in the binding reaction cosedimented with the ribosomes, indicating that SecA can bind to ribosomes (Figure 1A). Increasing the concentration of potassium chloride in the binding reaction to 500 mM completely disrupted ribosome binding (Figure 1A), suggesting that the binding interface is dominated by hydrophilic interactions. In addition, increasing the concentration of SecA in the binding reaction resulted in a saturation of binding at an approximately 1:1 stoichiometry of SecA to ribosomes as determined by Coo-massie staining and quantitative western blotting (Figure S1A, available online), which suggests that there is a single SecA-binding site on the ribosome. Serial dilutions of SecA-binding reactions containing equimolar concentrations of SecA and ribosomes suggested a KDfor the SecA-ribosome complex in the submicro-molar range (Figure S1B).
We analyzed ribosome binding in greater detail by fluores-cence anisotropy with SecA labeled with the fluorophore Ru (bpy)2(dcbpy) (den Blaauwen et al., 1997). When 700 nM Ru (bpy)2(dcbpy)-labeled SecA was incubated with ribosomes, we observed an increase in anisotropy that equilibrated with increasing concentrations of ribosomes. We calculated a KDof around 900 nM by fitting the anisotropy data to the quadratic equation and assuming a 1:1 binding stoichiometry (Figure 1B).
Scatchard analysis revealed a similar KD and confirmed that the binding stoichiometry was 1:1. Because the cellular concen-trations of ribosomes and SecA are20mM and8mM, respec-tively (Akita et al., 1991; Lill et al., 1988), our results suggest that SecA binds to ribosomes in a physiologically relevant fashion.
Identification of the Ribosome-Binding Domain of SecA SecA normally forms dimers in solution that are in dynamic equi-librium with a KDof around 1mM (Akita et al., 1991; Doyle et al., 2000; Or et al., 2002; Woodbury et al., 2002). The 1:1 stoichiom-etry for ribosome binding therefore suggested that SecA might interact with the ribosome along its dimer interface. Several different dimer interfaces have been suggested by different structural models. However, when we compared the interfaces in three structural models of the SecA dimer (PDB files 2FSF, 2IBM, and 2IPC) (Papanikolau et al., 2007; Vassylyev et al., 2006; Zimmer et al., 2006), we noticed that many of the residues participating in hydrophilic bonds between the two SecA subunits were located in or near nucleotide binding domain-2 (NBD2;residues 419–615). Moreover, in two of the structural
We therefore purified a fragment of SecA consisting of amino acid residues 418–668 (SecA418–668) in order to test if it could bind to ribosomes. This protein fragment includes all of NBD2 and thea-helical linker domain and is nearly identical to the frag-ment of SecA used byGelis et al. (2007)for structural studies of Figure 1. SecA Binds to Ribosomes in a 1:1 Stoichiometry in a Salt-Sensitive Fashion with a KDof900 nM
(A) SecA cosediments with ribosomes in a salt-sensitive fashion. Two micro-molar purified SecA was incubated with 2mM purified ribosomes in the pres-ence of 50 mM or 500 mM KCl, as indicated, and the ribosome-bound SecA was separated from unbound SecA by pelleting ribosomes through a 30%
sucrose cushion. The supernatant (S) and pellet (P) fractions were separated on a 10% SDS-PAGE gel and analyzed by Coomassie staining. The running positions of SecA and the ribosomal proteins are indicated.
(B) Plot of the fluorescence anisotropy of 700mM Ru(bpy)2(dcbpy)-labeled SecA versus the concentration of ribosomes and the fitted binding curve.
The inset displays the Scatchard plot analysis of the anisotropy data. Both methods yielded a KDof900 nM.
N-terminal affinity tag that was cleaved from the protein during purification. Full-length SecA418–668 and the largest truncation product bound ribosomes in cosedimentation experiments (Fig-ure 2B). However, the smaller truncation products could not bind ribosomes. Because the N-termini of these truncation products were identical, we could determine the C-termini of the frag-ments by mass spectrometry. The smallest truncation product that retained ribosome-binding activity terminated with gluta-mine-644 (Figure 2C), and the largest fragment that could not bind to ribosomes terminated with alanine-626. Because both of these residues are located in the N-terminal region of the a-helical linker domain, these results suggested that this region is important for ribosome binding.
In order to determine if thea-helical linker domain alone was sufficient to direct ribosome binding, we tested whether a frag-ment of SecA consisting of residues 616–668 could target the small ubiquitin-like modifier (SUMO) of Saccharomyces cerevisiae, which has no detectable ribosome-binding activity, to the ribosome. Similar to SecA418–668, a fusion protein consist-ing of SecA616–668 fused to the C terminus of a Strep-tagged SUMO protein (Strep-SUMO-SecA616–668) purified as a series of C-terminally truncated fragments. Although the truncation products could not bind to the ribosome, full-length Strep-SUMO-SecA616–668 cosedimented with ribosomes, albeit weakly, through a 30% sucrose cushion (Figure 2D), suggesting that the linker domain is both necessary and sufficient to target SecA to the ribosome.
Lysines 625 and 633 of SecA Are Required for Ribosome Binding
Two residues in the a-helical linker domain, lysine-625 and lysine-633, raised our interest as being potentially involved in ribosome binding based on the following characteristics: (1) both residues are located in the region of SecA418–668 that appears to be important for ribosome binding; (2) the side chains of these residues are completely solvent exposed and located on the same face of anahelix; (3) despite the general sequence vari-ability in this region of SecA (Papanikolau et al., 2007), all of the orthologs of SecA (but not SecA2) from widely diverged bacterial species in the COG database (http://www.ncbi.nlm.nih.gov/
COG/) contain a positively charged residue at either position 625 or 633 or both, which could interact with the negatively charged ribosome; (4) lysine-625 and lysine-633 participate in hydrophilic bonds between subunits in one or more of the struc-tural models of the SecA dimer (Table S1); (5) to our knowledge, these residues have not previously been identified as being important for any known activity of SecA.
When we repeated the ribosome sedimentation experiments by using SecA variants containing alanine substitutions at one (SecA[K625A] or SecA[K633A]) or both (SecA[K625A/K633A]) of these positions, we found that the individual K625A and K633A substitutions greatly interfered with ribosome binding and the double K625A/K633A substitution even further reduced the amount of SecA that copelleted with ribosomes (Figure 2E).
These substitutions do not appear to significantly affect the SecA
analytical size-exclusion chromatography indicated that SecA (K625A/K633A) dimerized to a similar extent as the wild-type protein (Figure S2B).
SecA Crosslinks to Ribosomal Protein L23
We sought to identify the SecA-binding site on the ribosome by using nonspecific crosslinking. Incubation of SecA with ribo-somes in the presence of EDC, which nonspecifically catalyzes the formation of a peptide bond between carboxyl groups and primary amino groups, resulted in the appearance of a single high-molecular weight crosslinking adduct that was visible by Coomassie staining (Figure 3A) and could be recognized by anti-SecA antibodies. The crosslinking adduct was not present when either SecA alone or ribosomes alone were incubated with EDC, indicating that the band was an adduct between SecA and a component of the ribosome.
In order to facilitate purification of the crosslinking adduct, we fused a short (15-amino acid) biotin-attachment peptide (Beckett et al., 1999) to the C terminus of SecA (SecA-biotin). SecA-biotin competed with wild-type SecA for ribosome binding in vitro (Fig-ure S3) and could complement aDsecAmutation in vivo, indi-cating that it behaves like wild-type SecA. Using SecA-biotin, we purified the SecA-containing crosslinking adduct by using streptavidin-coupled magnetic beads (Figure 3B). When we analyzed the purified crosslinking adduct by western blotting against each of nearly all of the ribosomal proteins, we found that it cross-reacted only with the L23 and SecA anti-bodies (Figure 3C), and we could confirm the presence of L23 in the crosslinking adduct by LC-MS/MS (Table S3). These data suggested that SecA binds to the ribosome near L23, which is located at the polypeptide exit channel on the large subunit of the ribosome.
A Genetic Screen for SecA-Binding-Deficient L23 Mutants
In order to identify which amino acid residues in L23 might be involved in SecA binding, we devised a genetic screen for mutations in the L23 gene (rplW) that cause a dominant protein translocation defect. We expressed different L23 variants from an IPTG-inducible promoter on a pTrc99b plasmid in a strain containing themalE-lacZreporter gene, which encodes a fusion protein between maltose binding protein (MBP) and b-galactosi-dase (MBP-LacZ) and has been used previously to screen forsec mutants (Gannon and Kumamoto, 1993; Kumamoto and Beck-with, 1983; Oliver and BeckBeck-with, 1981). Because the MBP-LacZ fusion protein is translocated across the cytoplasmic membrane by virtue of the N-terminal signal sequence of MBP, and because b-galactosidase is not functional in the periplasm, otherwise wild-type cells are normally phenotypically Lac-. However, the expression of mutant L23 proteins that cause a defect in protein translocation should result in the accumulation of b-galactosi-dase in the cytoplasm and a Lac+phenotype.
As expected, expression of wild-type L23 from a plasmid did not affect protein translocation in N48 as indicated by the low b-galactosidase activity in these cells. However, expression of
Molecular Cell
SecA Binds to Ribosomes
Figure 2. Ribosome Binding Is Mediated by Two Conserved Lysine Residues in thea-Helical Linker Domain
(A) Ribbon representation of theE. coliSecA structure from PDB file 2VDA (Gelis et al., 2007). Nucleotide binding domain-1 (NBD1; residues 9–228 and 375–418) a-helical
(L23[SDW]), E42A/I43A/K44A (L23[EIK]), F51A/E52A/E54A/E56A (L23[FEVEVE]), F51A/E52A/E54A/E56A/E89A (L23[FEVEVE/
E89A]), E42A/F51A/E52A/E54A/E56A (L23[E42A/FEVEVE]), and
E42A/I43A/K44A/F51A/E52A/E54A/E56 (L23[EIK/FEVEVE]). In addition, partial (K66-S78) or complete (V63-K81) deletion of the loop domain of L23 (Bornemann et al., 2008), which extends into the interior of the ribosome and forms a portion of the wall of the polypeptide exit channel, caused increasedb-galactosidase activities in N48.
These mutations cluster around three conserved regions in L23 (Figure 4B): (1) a previously identified patch of mostly acidic surface residues of unknown function beginning with phenylala-nine-51 and consisting of the glutamates in the sequence
51FEVEVE (Kramer et al., 2004). We considered glutamate-89 to be a member of this cluster because it appears to continue the ridge of conserved acidic residues formed by this sequence (Schuwirth et al., 2005); (2) a cluster of three conserved partially surface-exposed residues on the same face of L23 as51FEVEVE with the sequence42EIK; and (3) the loop domain of L23 (resi-dues 63–81). In contrast, expression of L23 variants that were defective for TF binding (that is, substitutions in the sequence
16VSEKAS, which is located on the opposite face of L23 from the above-described motifs) had no effect onb-galactosidase activity (Figure 4A), suggesting that the translocation defects were not the result of a defect in TF binding.
Ribosomes Containing Mutant L23 Proteins Are Defective for SecA Binding
Because a number of other factors involved in protein transloca-tion (such as the SRP, SecYEG, and YidC) also bind to the ribo-some at or near L23 (Gu et al., 2003; Kohler et al., 2009; Mitra et al., 2005; Schaffitzel et al., 2006), we wished to confirm that the residues we identified in our genetic screen were important for SecA binding. To this end, we purified L23(FEVEVE)-, L23 (EIK)-, L23(E42A/FEVEVE)-, and L23(FEVEVE/E89A)-containing ribosomes from strains whose sole copy ofrplWwas expressed from an IPTG-inducible promoter on a plasmid. Cells expressing the mutant L23 proteins grew slightly faster than those expressing wild-type L23 at 37C with the exception of those expressing L23(E42A/FEVEVE), which grew slightly slower. We could not obtain DrplW transductants of strains expressing L23(EIK/FE-VEVE), suggesting that this variant was not functional enough to support growth. The affinity of SecA for ribosomes containing the L23 variants was decreased, demonstrated by a 2–3-fold increase in the KD as determined by fluorescence anisotropy
(B) Two micromolar SecA418–668, which was purified as a series of C-terminally truncated protein fragments, was incubated in the presence or absence of 1mM ribosomes. Ribosome-bound SecA418–668was separated from unbound SecA418–668by pelleting ribosomes through a 30% sucrose cushion. The amount of SecA418–668in the supernatant (S) and pellet (P) fractions was analyzed by SDS-PAGE and western blotting against SecA. The masses (as determined by LC-MS) and the C-terminal five amino acids of each of the largest three protein fragments are displayed. The starred band is a ribosomal protein that cross-reacts with the anti-SecA antiserum.
(C) ClustalW alignment of thea-helical linker domain (residues 622–668) fromE. coli,Bacillus subtilis, andThermus thermophilus. The proteolytic cleavage sites, which result in the two largest truncation products of SecA418–668, are noted with arrows. Conserved residues lysine-625 and lysine-633 are boxed.
(D) Two micromolar Strep-SUMO-SecA616–668was incubated in the presence or absence of 1mM wild-type ribosomes or ribosomes containing L23(E42A). Ribo-some-bound SecA was separated from the unbound protein by pelleting ribosomes through a 30% sucrose cushion. The amount of Strep-SUMO-SecA616–668in the supernatant (S) and pellet (P) fractions was determined by SDS-PAGE and western blotting with alkaline phosphatase-coupled Strep-Tactin, which recognizes the Strep tag. The band representing full-length Strep-SUMO-SecA616–668is noted with an arrow.
(E) One micromolar wild-type SecA, SecA(K625A), SecA(K633A), or SecA(K625A/K633A), respectively, was incubated with 1mM ribosomes and the ribosome-Figure 3. SecA Crosslinks to Ribosomal Protein L23
(A) Binding reactions containing 1mM SecA or 1mM ribosomes were incubated in the absence or presence of the nonspecific crosslinker EDC (2 mM), as indi-cated. Samples were analyzed by SDS-PAGE and Coomassie staining. The running positions of the ribosomal proteins and full-length SecA are noted.
The running position of the high-molecular weight crosslinking adduct is marked with a (*).
(B and C) SecA-biotin was crosslinked to ribosomes by using 2 mM EDC and purified by binding to streptavidin-coupled magnetic beads and washing under denaturing conditions. The purified product was resolved on a 10%
SDS-PAGE gel. Gels were analyzed either by Coomassie staining (B) or by western blotting (C) with rabbit antiserum against SecA (red) and sheep anti-sera against the noted ribosomal proteins (green).
Molecular Cell
SecA Binds to Ribosomes
experiments (Figure 4C). In line with these results, 20%–60% less SecA cosedimented with the mutant ribosomes through a sucrose cushion when incubated at an equimolar concentration of 1mM (Figure 4D). In addition, the Strep-SUMO-SecA616–668construct, which we used to identify the binding domain in SecA, could not bind to ribosomes containing L23(E42A) (Figure 2D). Ribosomes containing L23 variants that caused a SecA-binding defect were not defective for SRP binding in ribosome sedimentation
experi-Figure 4. Mutations in the L23 Gene that Disrupt SecA Binding to the Ribosome Cause a Dominant Translocation Defect In Vivo
(A)b-galactosidase activities of N48 cells (malE-lacZ) expressing the indicated L23 variants from a plasmid. Error bars represent the range of measured activities.
(B) Structural model of the 50S ribosomal subunit (PDB file 2AW4;Schuwirth et al., 2005). A view is shown of the surface of L23 (light blue) distal (left) and proximal (right) to the polypeptide exit channel. The locations of conserved surface resi-dues that affect SecA binding (yellow), TF binding (red), and the loop domain (dark blue) are indi-cated. Other ribosomal proteins are depicted as ribbon representations (green). The backbone trace of the ribosomal RNA is colored gray.
(C) Plot of the fluorescence anisotropy of 700mM Ru(bpy)2(dcbpy)-labeled SecA versus the concen-tration of the respective mutant ribosomes and the fitted binding curves. The calculated KDs for the ribosomal variants are shown (below left).
(D) One micromolar SecA was incubated with 1mM ribosomes containing the indicated mutant L23 protein. Ribosome-bound SecA was separated from the unbound SecA by pelleting ribosomes through a 30% sucrose cushion. The amount of SecA in the pellet fractions was determined by quantitative western blotting against SecA and the percentage of SecA that copelleted with the respective mutant ribosomes compared to the amount of SecA that copelleted with wild-type ribosomes is given below. As a loading control, the amount of ribosomal protein L1 in the pellet fractions was determined by using a sheep anti-L1 antibody.
The Interaction between SecA and the Ribosome Appears to Facilitate Posttranslational Translocation In Vivo
In order to test whether ribosome binding by SecA was important for protein trans-location in vivo, we examined the
In order to test whether ribosome binding by SecA was important for protein trans-location in vivo, we examined the