Exploration of the Arrest Peptide Sequence Space Reveals Arrest-enhanced Variants*

Background: The stalling efficiency of translational arrest peptides (APs) is sensitive to mechanical pulling forces on the nascent chain. Results: We identify new APs with enhanced stalling efficiency. Conclusion: Mechanical pulling forces reduce stalling induced by diproline stretches in efp− cells. Significance: Our results provide new insights into AP-induced translational stalling and offer new in vivo force sensors. Translational arrest peptides (APs) are short stretches of polypeptides that induce translational stalling when synthesized on a ribosome. Mechanical pulling forces acting on the nascent chain can weaken or even abolish stalling. APs can therefore be used as in vivo force sensors, making it possible to measure the forces that act on a nascent chain during translation with single-residue resolution. It is also possible to score the relative strengths of APs by subjecting them to a given pulling force and ranking them according to stalling efficiency. Using the latter approach, we now report an extensive mutagenesis scan of a strong mutant variant of the Mannheimia succiniciproducens SecM AP and identify mutations that further increase the stalling efficiency. Combining three such mutations, we designed an AP that withstands the strongest pulling force we are able to generate at present. We further show that diproline stretches in a nascent protein act as very strong APs when translation is carried out in the absence of elongation factor P. Our findings highlight critical residues in APs, show that certain amino acid sequences induce very strong translational arrest and provide a toolbox of APs of varying strengths that can be used for in vivo force measurements.

During protein synthesis, the nascent polypeptide chain moves through the ϳ100-Å-long exit tunnel in the large ribosomal subunit (1). Strong interactions between the nascent chain and the tunnel might adversely affect protein synthesis and can even lead to complete blockage of the ribosome, a mechanism exploited by many antibiotics (2). It has therefore been postulated that the ribosome exit tunnel has a "Teflonlike" surface that minimizes interactions with the nascent chain to avoid adverse effects on translation (1,3).
Nevertheless, some nascent chain segments are able to interact with the ribosomal exit tunnel in ways that block or slow down translation (4,5). In bacteria, such translational arrest peptides (APs) 2 are often used to regulate the translation of downstream open reading frames in polycistronic mRNAs (6,7). APs interact with distinct ribosomal RNA and protein components within the ribosomal exit tunnel (8), inducing conformations at the ribosome active site that can block the peptidyl transfer reaction (9 -11).
In the case of the AP-containing SecM protein in Escherichia coli, the arrest of nascent chain elongation can be overcome by the activity of the motor protein SecA, which presumably breaks the AP-tunnel interactions by mechanically pulling on the nascent chain (12). On the basis of this notion, we hypothesized that APs might find general use as transplantable in vivo force sensors, as indeed turned out to be the case (13)(14)(15)(16).
The precise AP-tunnel interactions that lead to strong translational arrest are only partially known, and the arrest potency of related APs has not been systematically explored. We now report an extensive mutational screen of an unusually strong AP, a mutant variant of the SecM AP from Mannheimia succiniciproducens with the sequence HPPIRGSP, identifying substitutions that lead to substantial increases in resistance to pulling forces and stronger translational arrest. By combining the most potent mutations, we have been able to construct APs that induce efficient ribosome stalling in E. coli even under conditions of extremely strong pulling forces on the nascent chain.
In E. coli cells lacking elongation factor P (EF-P), diproline stretches in nascent polypeptide chains can also cause ribosome stalling (3). To test whether this kind of stalling mechanism is sensitive to pulling forces, we measured the strength of the translational arrest induced by diproline stretches with different flanking residues in an efp Ϫ strain, demonstrating their possible use as exceptionally short in vivo force sensors.

EXPERIMENTAL PROCEDURES
DNA Manipulations-Site-directed mutagenesis was performed using partially overlapping oligonucleotides (17) using constructs generated previously (13,16). Truncations were introduced by using fixed forward oligonucleotides and variable reverse oligonucleotides in a PCR reaction. Exchanges were performed by using the same partially overlapping oligonucleotide approach but using both randomized positions in the forward oligonucleotides as well as specific sequences at the respective positions. Insertions were performed similarly by using the inserted sequence as the partially overlapping sequence. The resulting product was DpnI-treated and transformed into E. coli MC1061 cells. All products were transformed into MC1061 E. coli cells and confirmed by sequencing.
Pulse-labeling-E. coli MC1061, MG1655, and MG1655 ⌬efp cells, transformed with the respective constructs, were grown overnight at 37°C in M9 minimal medium. The minimal medium contained 19 amino acids at a concentration of 1 g ml Ϫ1 but no methionine, 100 g ml Ϫ1 thiamine, 0.1 mM 0.4% (w/v) fructose, and 100 g ml Ϫ1 ampicillin. The overnight cultures were back-diluted 1:10 and grown for 3 h to an A 600 of 0.3. Protein expression was induced by the addition of 0.2% (w/v) arabinose for 5 min prior to the addition of [ 35 S]methionine. After 2 min of pulse labeling, 0.5 sample volumes of ice-cold 50% (v/v) trichloroacetic acid was added to the sample and the samples were incubated for 30 min on ice. The samples were then centrifuged for 10 min at 20,800 ϫ g at 4°C, and the pellet was washed with ice-cold acetone and centrifuged again at 20,800 ϫ g at 4°C. The supernatant was removed, and the pellet was resolubilized in 2% SDS Tris buffer (10 mM Tris-Cl (pH 7.5) and 2% (w/v) SDS) by vortexing and heating to 95°C for 10 min. The sample was spun again to remove insoluble material, and Lep protein was immunoprecipitated using a LepB polyclonal antibody. Samples were resolved on SDS-PAGE and visualized in a Fuji FLA-3000 PhosphorImager. The full-length and arrested forms were quantified using Fujifilm Image Gauge and Easyquant software.

RESULTS
To screen for APs that can withstand strong pulling forces, we used a previously identified mutant version of the M. succiniciproducens SecM AP, called SecM(Ms-Sup1) (13,16,18), as the starting template. The AP was inserted close to the C terminus of the E. coli inner membrane protein leader peptidase (LepB) (Fig. 1a). LepB has two N-terminal transmembrane helices (TM1 and TM2) and a large C-terminal periplasmic domain that is cotranslationally translocated across the inner membrane by the SecYEG translocon. We used a modified LepB construct that contains a stretch of five aspartic acid residues (5D) in the periplasmic domain. The negatively charged 5D stretch is subjected to a strong pulling force generated by the electric membrane potential as it translocates across the inner membrane. The pulling force is detected as an increase in the fraction full-length protein (f FL ) in constructs when the length, L, of the linker that connects the 5D stretch to the P-site in the ribosome is 42-49 residues (Fig. 1b) (16).
We chose a linker length L of 46 residues that corresponds to a local minimum in the f FL profile (Fig. 1b, orange dot). The rationale behind this choice was that all minor changes in the location of the AP and linker region in the ribosomal tunnel and any attendant changes in the location of the 5D stretch relative to the membrane caused by mutations in the AP would result in an increase in f FL and, therefore, score as a weakening of the AP, whereas any increase in the arrest potency of the AP would lead to a decrease in f FL . As shown in Fig. 1c, after immunoprecipi-

Mutagenesis of the SecM(Ms-Sup1) Arrest
Peptide-Using the LepB[5D, L ϭ 46] construct, we generated an extensive mutant library in which every residue of the eight-residue Sec-M(Ms-Sup1) AP HPPIRGSP, as well as two flanking residues on either side of the arrest peptide (SS and GS; these residues were introduced during cloning and are YF and QR in the wild-type SecM(Ms) AP), were changed to every other amino acid (see "Experimental Procedures"). The 228 mutated constructs were transformed into E. coli MC1061 cells, and f FL values were determined in triplicate. All sequences and f FL values are listed in supplemental Tables S1 and S2, and the results are summarized in Fig. 2. The seven most C-terminal residues of the AP (residues Ϫ1 to Ϫ7) are all critical, and every substitution in this part of the AP, except for the Ser Ϫ2 3 (Asp, Pro) mutations, the Pro Ϫ6 3 (Lys, Asn, Arg) mutations, and the Pro Ϫ7 3 Tyr mutation resulted in very weak arrest (i.e. led to a strong increase in f FL ). Positions Ϫ8 to Ϫ10 showed a more graded response, with mutations to hydrophobic and especially tryptophan residues causing significant increases in the strength of the AP. Mutations in positions ϩ1 and ϩ2, immediately C-terminal to the AP, had no or little effect on f FL . .
Three mutations (Ser Ϫ2 3 Pro, Ser Ϫ9 3 Trp, and Ser Ϫ10 3 Trp) resulted in very strong arrest, with f FL reaching baseline levels (Ͻ0.2). We introduced the two Ser 3 Trp mutations (plus the His Ϫ8 3 Trp mutation) individually and jointly into a construct with a stretch of 10 aspartic acid residues, LepB[10D, L ϭ 42], that generates a very strong pulling force (16). As seen in Fig. 3, f FL Ϸ 1 in all three constructs with a single Ser 3 Trp exchange; i.e. these mutated APs can withstand the pulling force generated by 5 but not by 10 aspartic acid residues. Combining the three Trp mutations reduced f FL for the LepB[10D, L ϭ 42] construct to 0.2, indicating that their effects are additive.
To further characterize the Ser Ϫ9 3 Trp mutant, we obtained a full f FL profile for the LepB[10D, L ϭ 23-61] set of constructs (Fig. 4). Compared with the weaker SecM(Ms) and SecM(Ms-Sup1) APs, the SecM(Ms-Sup1;Ser Ϫ9 3 Trp) AP produces a much sharper main peak at L ϭ 41-42 residues, allowing us to determine the point of maximal pulling force for the 10D stretch with single-residue accuracy. On the other hand, the f FL values obtained with the Ser Ϫ9 3 Trp mutant are surprisingly high in the region L ϭ 23-36 residues, i.e. when the 10D stretch has not completely emerged from the ribosomal tunnel.  Table S2).

Enhanced Arrest Peptides
The Ser Ϫ2 3 Pro mutation appeared to dramatically increase the strength of the AP because only arrested protein could be observed for the LepB[5D, L ϭ 46] construct (f FL Ϸ 0) (Fig. 2). In this particular case, RNase-sensitive bands migrating slower than the full-length protein during SDS-PAGE were evident (Fig. 5a). These could either represent species with tRNA still bound to the nascent chain or, possibly, covalently attached tmRNA added by the SsrA rescue system (19). When introduced into the LepB[10D, L ϭ 42] construct, the Ser Ϫ2 3 Pro mutation yielded f FL Ϸ 0.3 (Fig. 5b), and no RNase-sensitive bands were observed. We further performed an alanine scan of positions Ϫ1 and Ϫ3 to Ϫ8 of the SecM(Ms-Sup1;Ser Ϫ2 3 Pro) AP in the LepB[10D, L ϭ 42] background. As expected, all of these mutations led to a strong increase in f FL (Fig. 5c), and no RNase-sensitive bands were apparent on the gels (data not shown). Therefore, SecM(Ms-Sup1;Ser Ϫ2 3 Pro) behaves like a strong AP under a force load, except that it has a more stably attached RNA than other APs.
As a final control, we changed the codon for the serine residues in positions Ϫ9 and ϩ2 to all six leucine codons in the LepB[5D, L ϭ 46] construct (Fig. 6). f FL varied only within the margin of error between the different mutants, indicating that the observed changes in arrest-peptide strength depend on the introduced amino acid rather than on the specific codon.
AP Function of Oligo-proline Stretches in an efp Ϫ Background-In the absence of elongation factor EF-P, bacterial ribosomes stall efficiently already at two consecutive proline residues (3). It is not known, however, whether stalling on oligo-proline stretches is sensitive to pulling forces on the nascent chain. To address this question, we used both a LepB construct that contains a stretch of 19 alanine residues at L ϭ 39 residues, i.e. at a linker length that places the [19A] stretch in the translocon, where it exerts only a weak pulling force on the nascent chain (13), and the LepB[10D, L ϭ 43] construct in which, as shown above, there is a strong pulling force (Fig. 7a). These constructs contained no internal oligo-proline stretches (the only naturally occurring PP doublet in LepB was replaced by AA).
As a control, we first tested the SecM(Ms) AP HAPIRGSP in the LepB[19A, L ϭ 39] construct in efp ϩ and efp Ϫ strains. In both cases, we observed a strong arrest (Fig. 7b). Upon replacement of the eight-residue-long SecM(Ms) AP by eight glycine and serine residues (GSGSGGSS), only full-length protein could be observed in either the presence or absence of EF-P, as expected. Replacement of the C-terminal residue in the GSGSGGSS stretch by proline (lane SG-1P) led to no arrested protein in either the presence or absence of EF-P. However, the introduction of two or three proline residues (lanes SG-2P and SG-3P) led to robust stalling of LepB[19A, L ϭ 39] in the absence but not in the presence of EF-P (Fig. 7b).
In contrast, the SG-2P sequence only partially stalled the LepB[10D, L ϭ 43] construct in the absence of EF-P (f FL ϭ 0.5), as seen in Fig. 7c. Stalling induced by oligo-proline stretches in the absence of EF-P can therefore be overcome at least partially, but only by very strong pulling forces.
Given the strong arrest potency and the short length of the PP motif, we decided to test sequence variants of this motif by mutating the two flanking residues to every other amino acid

Ms-Sup1
S -10 →W S -9 →W H -8 →W 3W f FL S -10 S -9 H -8 PPIRGSP   Table S3, whereas the Z residue (with X ϭ Ser) had only a minor effect on f FL (Fig. 8b). A similar analysis was published recently by Wilson and co-workers (20), who used a ␤-galactosidase assay as a readout (i.e. the assay was carried out in the absence of a pulling force). The results for the X residue were similar in the two assays (Fig. 8c), except for X ϭ Ala and X ϭ Pro, which both gave a stronger arrest in the ␤-galactosidase assay than in the pulling force assay. In contrast to our results, however, Wilson and co-workers (20) found a large variation in f FL when the Z residue was varied (Fig. 8d). It therefore appears that XPPZ-induced stalling has a different dependence on the Z residue in the absence or presence of a strong pulling force.
To test for the contribution of residues farther upstream of the PP motif, we mutated the four underlined residues in the preceding SGSGSGGS stretch to Trp, Arg, Pro, and Asp. Indeed, these mutations gave rise to a wide range of f FL values (Fig. 9), in most cases conforming to the pattern seen for the SecM(Ms-Sup1) AP in Fig. 2.
Finally, we generated a full f FL profile for the LepB[10D, L ϭ 23-52] series of constructs using PP-induced arrest in an efp Ϫ background. As seen in Fig. 7d, the APPK stretch withstands even stronger pulling forces than the SecM(Ms-Sup1;Ser Ϫ9 3 Trp) AP, and the main peak is sharper and has a clear maximum of f FL ϭ 0.7 at linker length L ϭ 42 residues. Notably, the shoul-

Enhanced Arrest Peptides
der at L Ϸ 23-36 residues seen for the SecM(Ms-Sup1;Ser Ϫ9 3 Trp) AP is not present in this case, suggesting that it represents a specific effect elicited on the SecM(Ms-Sup1;Ser Ϫ9 3 Trp) AP by the highly negatively charged 10D stretch.

DISCUSSION
SecM translational APs are sensitive to pulling forces acting on the nascent polypeptide chain (13)(14)(15)(16). But just how much pulling force can they be made to withstand? Which residues in an AP are the most critical to the arrest potency and, hence, interact most strongly with the ribosomal tunnel? Here we show that the interactions between the SecM(Ms-Sup1) AP and the ribosomal exit tunnel can be tuned to considerably increase the strength of stalling. We also demonstrate that diproline stretches, which induce stable translational stalling in EF-P deletion strains, effectively function as very short force-sensitive arrest peptides.
For the SecM(Ms-Sup1) AP, we substituted every amino acid for every other amino acid in the sequence SSHPPIRGSPGS (where the underlined segment is the AP as defined by Ala scanning (18)) and analyzed the arrest potency under a rather strong pulling force. In positions Ϫ7 to Ϫ1, only two mutations were found to increase the strength of the AP, namely Ser Ϫ2 3 Pro and Pro Ϫ6 3 Arg (Fig. 2). Mutations of residues Ser Ϫ10 , Ser Ϫ9 , and His Ϫ8 to hydrophobic residues, and to Trp in par-ticular, also markedly increased the arrest potency up to a point where the mutated AP could completely withstand the force generated by the electric potential across the inner membrane acting on a stretch of 5 but not of 10 consecutive aspartic acid residues (Figs. 3 and 4). These positions are located in the vicinity of the constricted region of the ribosomal tunnel (8), and the effect of tryptophan in these positions may be similar to the tryptophan-binding site seen in the stalled TnaC AP, where bound L-tryptophan interacts with residue Ile Ϫ10 in the AP (9). By simultaneously mutating the three positions Ϫ10 to Ϫ8 to tryptophan, we were able to generate an AP (WWWPPIRGSP) that is able to stall translation even in the presence of the pulling force elicited by 10 aspartic acid residues (Fig. 3). This AP can, presumably, be made even stronger by introducing the Ser Ϫ2 3 Pro and Pro Ϫ6 3 Arg mutations, but, because the pulling force from a [10D] stretch is the strongest we have been able to generate so far (16), we cannot determine whether this is the case.
Two previous studies have reported AP motifs obtained from randomized sequence libraries using two related selection schemes (21,22). The most potent motifs found were GI(R/ H)XPP, FXXYXIWPPP, R(S/A)PP, and HGPP. All four motifs are broadly consistent with the data reported in Fig. 2 and include the strong Ser Ϫ2 3 Pro mutation. Arg Ϫ4 , which is absolutely critical in the SecM(Ms-Sup1) AP, is present in two of the four motifs, as is Ile Ϫ5 . Aromatic residues in position Ϫ7 (Tyr) and Ϫ10 (Phe) also have counterparts in the SecM(Ms-Sup1) AP mutational screen. A third study (18) identified revertants of a weakened variant of the E. coli SecM AP. Again, the results agree with ours (in particular, the strong stalling potency of a revertant AP with cysteine in position Ϫ10 was evident in this study, cf. Fig. 2). Finally, we note that the 10 C-terminal residues in the E. coli SecM AP (SQAQGIRAGP) differ from the Sec-M(Ms-Sup1) AP by sequence changes that, according to the data in Fig. 2, should make it less potent, as indeed is the case (13,18). The recent discovery that oligo-proline stretches stall translation in the absence of EF-P (3,20) prompted us to investigate whether this kind of stalling also could be overcome by pulling on the nascent chain. Similarly to the "classic" APs discussed above, we found that translational stalling at a diproline stretch was relieved only by a very strong pulling force [10D] (Fig. 7). Furthermore, we corroborate earlier evidence that the potency of the PP stalling motif is sensitive to the identity of the immediate N-terminal flanking residue (23, 24) (Fig. 8), and can therefore be modulated to some extent.
In summary, we find that it is possible to design translational APs that are considerably more resistant to mechanical force than naturally occurring SecM APs and that might even be expected to fully stall translation under almost any conceivable pulling force that can be generated in vivo. We further find that translational stalling induced by the most potent diproline stretches in the absence of EF-P can be overcome, but only by very strong pulling forces. The collections of SecM and diproline APs reported in Figs. 2 and 8 considerably expand the toolbox of transplantable force sensors with different "spring constants" that can be used to measure mechanical forces generated on a nascent chain by various cotranslational processes, such as protein folding and membrane insertion/ translocation.