Regulation by a chaperone improves substrate selectivity during cotranslational protein targeting

Significance Correct protein biogenesis is crucial for all cells. Numerous factors including molecular chaperones, modification enzymes, and protein-targeting machineries bind near the ribosome exit site and can access the nascent protein. How nascent proteins are accurately selected into the correct biogenesis pathway in such a crowded environment is an emerging question central to accurate protein biogenesis. Using chemical biology and biochemical and biophysical tools, we show that the major cotranslational chaperone, trigger factor, and cotranslational targeting machinery, signal recognition particle, regulate each other at multiple stages, including initial binding, ribosome delivery to the membrane, and enforcement of a timer for cotranslational protein targeting. Together, these mechanisms enhance accurate substrate selection into both cotranslational and posttranslational protein targeting pathways. The ribosome exit site is a crowded environment where numerous factors contact nascent polypeptides to influence their folding, localization, and quality control. Timely and accurate selection of nascent polypeptides into the correct pathway is essential for proper protein biogenesis. To understand how this is accomplished, we probe the mechanism by which nascent polypeptides are accurately sorted between the major cotranslational chaperone trigger factor (TF) and the essential cotranslational targeting machinery, signal recognition particle (SRP). We show that TF regulates SRP function at three distinct stages, including binding of the translating ribosome, membrane targeting via recruitment of the SRP receptor, and rejection of ribosome-bound nascent polypeptides beyond a critical length. Together, these mechanisms enhance the specificity of substrate selection into both pathways. Our results reveal a multilayered mechanism of molecular interplay at the ribosome exit site, and provide a conceptual framework to understand how proteins are selected among distinct biogenesis machineries in this crowded environment.

The ribosome exit site is a crowded environment where numerous factors contact nascent polypeptides to influence their folding, localization, and quality control. Timely and accurate selection of nascent polypeptides into the correct pathway is essential for proper protein biogenesis. To understand how this is accomplished, we probe the mechanism by which nascent polypeptides are accurately sorted between the major cotranslational chaperone trigger factor (TF) and the essential cotranslational targeting machinery, signal recognition particle (SRP). We show that TF regulates SRP function at three distinct stages, including binding of the translating ribosome, membrane targeting via recruitment of the SRP receptor, and rejection of ribosome-bound nascent polypeptides beyond a critical length. Together, these mechanisms enhance the specificity of substrate selection into both pathways. Our results reveal a multilayered mechanism of molecular interplay at the ribosome exit site, and provide a conceptual framework to understand how proteins are selected among distinct biogenesis machineries in this crowded environment.
SRP is a universally conserved ribonucleoprotein complex responsible for the cotranslational targeting of proteins to the eukaryotic endoplasmic reticulum (ER), or the bacterial plasma membrane (31). SRP recognizes ribosome-nascent chain complexes (termed RNC or cargo) carrying strong signal sequences and delivers them to the SecYEG or YidC translocation machinery on the target membrane. SRP binds RNC via two interactions: a helical N domain in the SRP54 protein (called Ffh in bacteria) binds the ribosomal protein L23, and a methioninerich M domain binds hydrophobic signal sequences on nascent proteins as they emerge from the translating ribosome (Fig. 1A). Both SRP and SRP receptor (called FtsY in bacteria) also contain a conserved NG domain, comprised of a GTPase (guanosine 5′-triphosphate hydrolase) G domain and the N domain, whose direct interaction mediates the delivery of cargo to the target membrane.
Biophysical analyses (32)(33)(34) showed that membrane targeting is a two-step process in which SRP and FtsY first associate via their N domains to form a transient early intermediate (31,32,35). GTP (guanosine 5′-triphosphate)-driven rearrangements then bring the G domains of both proteins into close contact, giving a stable closed complex (36,37). This rearrangement also exposes a membrane-binding helix of FtsY and thus is coupled to the membrane targeting of cargo (38). Importantly, SRP•FtsY assembly contributes extensively to the fidelity of SRP (39). The initial recognition of RNC by SRP is insufficient to reject suboptimal cargos bearing weak signal sequences (40,41). Instead, a correct cargo strongly stabilizes the otherwise labile early intermediate and thus accelerates formation of the SRP•FtsY closed complex over 10 3 -fold, whereas suboptimal cargos provide much less stimulation (34,40,42). This enables rapid delivery of the correct cargos to the target membrane and provides kinetic discrimination against suboptimal cargos (Fig. S1).
TF is a major cotranslational chaperone in bacteria, with an estimated cellular concentration of 50-80 μM (6). With a dissociation constant (K d ) of ∼1 μM for ribosomes (43), TF is bound to virtually every ribosome in the cell. Like SRP, TF contacts the ribosome via the L23 and L29 proteins near the Significance Correct protein biogenesis is crucial for all cells. Numerous factors including molecular chaperones, modification enzymes, and protein-targeting machineries bind near the ribosome exit site and can access the nascent protein. How nascent proteins are accurately selected into the correct biogenesis pathway in such a crowded environment is an emerging question central to accurate protein biogenesis. Using chemical biology and biochemical and biophysical tools, we show that the major cotranslational chaperone, trigger factor, and cotranslational targeting machinery, signal recognition particle, regulate each other at multiple stages, including initial binding, ribosome delivery to the membrane, and enforcement of a timer for cotranslational protein targeting. Together, these mechanisms enhance accurate substrate selection into both cotranslational and posttranslational protein targeting pathways. ribosome exit site (3,5,44). Also analogous to SRP, TF preferentially interacts with hydrophobic sequences on the nascent polypeptide (1,2,4,45,46), mediated by a large concave surface rich in hydrophobic residues (1,(3)(4)(5)(6). Despite these similarities with SRP, TF directs substrate proteins to distinct biogenesis pathways: It exhibits synthetic lethality with DnaK/J and facilitates the productive folding of cytosolic proteins (1,4,7,9,11). It also interacts with a subset of secretory and outer membrane proteins and interfaces with the posttranslational SecA/B pathway (8,10,(12)(13)(14).
SRP and TF are two distinct biogenesis pathways that a nascent protein must commit to. This raises intriguing questions: How do these two factors, which have overlapping substrate preferences, compete and/or collaborate at the ribosome exit site? How are nascent proteins sorted between them and committed to the correct pathway in a timely and accurate manner? Extensive past work to address these questions has led to different (and sometimes contradictory) models, including (i) TF and SRP compete for binding to the RNC (10,15,16,18); (ii) TF and SRP can bind to the same RNC simultaneously (17,(19)(20)(21); (iii) FtsY rejects TF from SRP-bound ribosomes (17); and (iv) TF preferentially occupies longer nascent chains (13,(45)(46)(47) and, by inference, SRP preferentially binds short nascent chains. A unifying model that reconciles all these observations and explains how nascent chains on the ribosome are selected by TF or SRP is still lacking. Most importantly, most of the previous studies have focused on the initial binding of SRP or TF to the nascent polypeptide, which may not be the step at which nascent proteins are committed to their respective biogenesis pathways.
In this work, we used high-resolution biochemical and biophysical analyses to investigate the interplay between TF and SRP at the ribosome exit site in molecular detail. We show that TF regulates SRP function by three distinct mechanisms, which together enhance the ability of the SRP pathway to reject suboptimal substrates. Our results establish a comprehensive and cohesive model that explains previous observations, delineates the complex interplay between protein biogenesis factors at the ribosome exit site, and provides a conceptual foundation to understand how timely and accurate selection of substrates is achieved in this crowded environment.

Results
Anticooperative Binding of SRP and TF to RNC. We first asked how TF affects cargo recognition by SRP as the nascent polypeptide emerges from the ribosome exit tunnel. To this end, we used amber suppression technology to incorporate a fluorescent nonnatural amino acid, 7-hydroxycoumaryl ethylglycine (Cm), into the nascent polypeptide downstream of the signal sequence ( Fig. 1 A and B, asterisks) (48). When paired with SRP labeled with BODIPY FL at residue 421 in the Ffh M domain, efficient Förster resonance energy transfer (FRET) was observed (48), providing a highly specific and sensitive assay to report on the interaction of SRP with nascent polypeptides on RNC.
To test whether TF helps enhance the specificity of SRP, we used a range of substrates with varying dependence on SRP (Fig.  1B). FtsQ, a bona fide SRP substrate, uses an integral transmembrane domain (TMD) as the signal sequence (18,49). An engineered substrate, 3A7L, contains a modestly hydrophobic signal sequence more representative of SRP-dependent secretory proteins. (40). As suboptimal substrates, we used EspP, a SecA/B substrate that is rejected by SRP due to its N-terminal extension (40,50,51), and phoA, a primarily SecA-dependent substrate whose signal sequence is less hydrophobic (52,53). For all of the experiments in Figs. 1-4, we purified homogeneous stalled RNCs with 80-85 amino acids between the start of signal sequence and the peptidyl transferase center of ribosome. This mimics the stage at which the signal sequence emerges from the ribosome exit tunnel and is optimal for recognition by SRP (47,54).
There has been extensive debate on whether TF and SRP compete with one another for binding RNCs (Fig. S2A), or whether they bind the same RNC ( Fig. S2B) (10, 15-17, 19, 20). Well-established theoretical treatments (55) and kinetic simulations (Fig. S2B) show that these alternative models can be rigorously distinguished by quantitatively analyzing the effect of TF on SRP-RNC binding using our conformation-sensitive FRET assay. If binding of TF and SRP to RNCs is mutually exclusive, then TF will deplete the free RNCs available to bind SRP, necessitating higher SRP concentrations to reach saturation (Fig. S2A). Further, when SRP is allowed to bind RNC at saturating concentrations, TF cannot alter the conformation and, hence, the FRET value of the RNC•SRP complex (Fig. S2A). In contrast, if TF alters the FRET value of the RNC•SRP complex, then this rules out a strictly competitive model and could only be explained by a model in which TF and SRP bind to the same RNC and alter each other's interaction with the nascent polypeptide (Fig. S2B). TF could either strengthen (cooperative) or weaken (anticooperative) SRP binding to the RNC; the latter model is simulated in Fig. S2B.
A recent study measured how SRP (or TF) altered RNC binding to TF (or SRP) using FRET between ribosomal protein L23 and TF (or SRP). Nevertheless, a cobinding model was assumed, rather than tested, in this analysis (20). Indeed, the low signal-to-noise ratio of this assay (often 10-20%) led to ambiguous interpretations: Although some data appeared consistent with a competitive model (Fig. S2A), others appeared consistent with a cobinding model (Fig. S2B) (20). The robust FRET between SRP (or TF) and signal sequence allowed us to more unambiguously test and distinguish between these models. Equilibrium titrations of RNC-SRP binding showed that, with all four substrates, increasing amounts of TF induced two significant changes: (i) Increasingly higher SRP concentrations are required to reach saturation, and (ii) there are successive reductions in FRET efficiency when the RNC•SRP complex is formed at saturating concentrations ( These results provide strong evidence for anticooperative binding of TF and SRP to the RNC, but are incompatible with models in which their binding to RNC is mutually exclusive (Fig.  S2A). In addition, the effect of TF on the observed RNC-SRP binding affinity (app K d SRP ) is saturable (Fig. S2E). This is also consistent with predictions from the model in which TF and SRP bind to the same RNC, but is incompatible with models in which their binding is mutually exclusive (Fig. S2F).
If TF weakens the affinity of SRP for RNC, then reciprocally, SRP would weaken the affinity of TF for RNC. To test this prediction, we developed a FRET assay to measure RNC-TF interaction. We labeled TF with BODIPY FL at an engineered cysteine (C377) which, when paired with Cm-labeled RNC, induced efficient FRET ( Fig. 2 A and B). Equilibrium titrations using this FRET assay showed that TF binds tightly to all of the RNCs tested, with K d TF values ranging from 2.5 nM to 8.4 nM ( is saturable (Fig. S3D). This provides independent evidence for anticooperative binding between SRP and TF at the RNC (Fig. S2B).
Collectively, these results provide more conclusive evidence that TF and SRP can bind to the same RNC, on which they alter the binding affinities and conformation of one another (17,19,20). Importantly, these modulations did not lead to significant consequences on the occupancy of SRP or TF on the RNCs. As the cellular TF concentration (>50 μM) is over 10 3 -fold higher than the weakest RNC-TF affinity observed here, the effect from SRP has negligible consequences on the occupancy of TF on the RNCs. Although SRP is less abundant (∼400 nM), all of the RNC-SRP dissociation constants are still below the cellular SRP concentration in the presence of TF (Fig. 1E). A numerical analysis based on this information shows that TF would modestly influence SRP occupancy on RNC phoA but not for the other RNCs tested (see more discussion in TF Enhances the Specificity of SRP-Dependent Protein Targeting). Further, binding of SRP to RNC EspP is comparable to RNC FtsQ or RNC 3A7L in the presence of TF (Fig. 1E). Therefore, at cellular concentrations, significant amounts of RNC•SRP•TF ternary complexes accumulate for both the SRP and SecA/B pathway substrates, and substrate commitment to SRP or TF does not primarily occur at the RNC-binding stage.
TF Slows the Recruitment of SRP Receptor to Suboptimal Cargos. We therefore examined the next step of the SRP pathway, cargo delivery to the target membrane mediated by SRP•FtsY assembly. Kinetic regulation of this process plays a key role in the ability of SRP to reject suboptimal substrates (34,40). We asked whether TF increases substrate discrimination by SRP during this process.
Stable SRP•FtsY assembly is a two-step process in which a transient early intermediate is initially formed, followed by a GTP-dependent rearrangement to form a stable closed complex (32). The stability of the early intermediate directly correlates with the rate at which the closed complex is formed, and represents a major commitment step for correct cargos (33,40,41). We therefore tested how TF affects formation of the early targeting   (32,33,56). Formation of the SRP•FtsY complex was monitored using FRET between N-(7dimethylamino-4-methylcoumarin-3yl)-maleimide (DACM)-labeled at SRP C153 and BODIPY-labeled at FtsY C345 (Fig. 3A) (32,33). Equilibrium titrations using this assay showed that, with a bona fide SRP substrate such as RNC FtsQ , a highly stabilized SRP•FtsY early complex is formed, and TF has negligible effects on its stability or FRET efficiency (Fig. 3B). As the signal sequence becomes weaker, the RNC•SRP•FtsY early complex becomes less stable, as reported previously (40). Importantly, TF further weakens the early complex, and this effect is more substantial as the signal sequence becomes weaker (Fig. 3 C-E; summarized in Fig. 3F). Further, TF lowers the FRET efficiency when the early targeting complex is formed at saturating FtsY concentrations; this effect also becomes more pronounced as the signal sequence becomes weaker (Fig. 3 B-E). This indicates that TF alters the conformation of the SRP•FtsY early complexes formed with weaker SRP substrates, such that the G domains of SRP and FtsY (where the FRET probes are located) are positioned farther apart.
If TF induces the formation of a weaker and distorted early targeting complex for RNCs bearing weaker signal sequences, then assembly of the stable closed SRP•FtsY complex with these substrates will be significantly slowed by TF. To test this hypothesis, we compared the kinetics of SRP•FtsY closed complex assembly between preformed RNC•SRP and RNC•SRP•TF complexes. Complex assembly was measured using the same FRET assay but in the presence of GMPPNP (5′-guanylyl imidodiphosphate), which allows the early intermediate to proceed to the closed complex (Fig. 4A). Independent measurements using a fluorescent probe that specifically detects the closed complex have yielded rate constants comparable to the FRET assay (40). With strong SRP substrates such as FtsQ, complex formation is rapid and unaffected by TF (Fig. 4 B and F). With 3A7L, TF slows SRP•FtsY closed complex assembly threefold (Fig. 4 C and F). With SRP-independent substrates such as EspP and phoA, TF slows closed complex assembly more substantially, 10-to 12-fold (Fig. 4 D and F). Thus, although SRP can strongly discriminate between authentic and suboptimal cargos via the kinetics of FtsY recruitment, TF enhances this discrimination by an additional order of magnitude.
TF More Effectively Inhibits SRP as the Nascent Chain Elongates. It has been a long-standing observation that, as the nascent polypeptide elongates beyond a critical length of 130-140 amino acids, SRP loses its ability to target even the RNCs bearing SRPdependent secretory proteins such as preprolactin (47,57). This imposes a limited time window for SRP and FtsY to complete targeting, especially in bacteria where translation elongation occurs rapidly and SRP does not delay translation. However, the molecular basis underlying this phenomenon has been controversial. Some studies suggest that RNC-SRP binding affinity is reduced as the nascent chain becomes longer, whereas others suggest that SRP loses the ability to target longer nascent chains to the membrane (47,57,58).
To address this question, we prepared RNCs bearing longer nascent chains, with 130-135 amino acids from the N terminus of the signal sequence to the peptidyl transferase center. Using the FRET assays described above, we tested whether and how a longer nascent chain length affects cargo recognition by SRP (K d SRP ) and assembly of the closed RNC•SRP•FtsY complex (k on ; Fig. 5A). Intriguingly, a longer nascent chain length exerted modest effects on RNC-SRP binding affinities (Table 1 and Fig.  S4). Importantly, the SRP-RNC 130-135 binding affinities observed here (Table 1) and elsewhere (58) are still in the low nanomolar range and sufficient for SRP to occupy the RNC at its cellular concentration. Likewise, a longer nascent chain slows SRP•FtsY closed complex assembly by less than threefold (Table 2). These differences, although statistically significant and reproducible, are insufficient by themselves to lead to rejection of long-chain RNCs from the SRP pathway.
We asked whether TF could contribute to this rejection. We first tested how TF affects the binding of SRP to RNCs with longer nascent chains. Analogous to observations with shorter nascent chains, we observed anticooperative binding between TF and SRP to RNCs with a nascent chain length of 130-135 amino acids (Fig. 5 B-D). Compared with RNCs with shorter nascent chains, TF exerts an approximate twofold larger weakening effect on the binding of SRP to RNCs with longer nascent chains (Fig. 5  B-D). As a result, the affinity of SRP for RNC FtsQ is weakened to 39 nM, that for RNC 3A7L is weakened to 53 nM, and that for RNC phoA is weakened to 200 nM (Fig. 5 B-D and Table 1).
We next tested whether TF also prevents the targeting of longchain RNCs to the membrane through SRP•FtsY assembly. With RNC FtsQ-135 , which contains a strongly hydrophobic TMD, SRP•FtsY complex assembly remains rapid at the longer chain length and is not significantly affected by TF (Fig. 5E). With RNC 3A7L , which contains a weaker signal sequence more representative of SRP-dependent secretory proteins, TF exerts a much larger inhibitory effect at the longer chain length, slowing the assembly of the targeting complex by 18-fold ( Fig. 5F and Table  2). Analogous results were observed with RNC phoA-130 (Fig. 5G). Thus, TF exerts a larger inhibitory effect on the membrane targeting of RNCs as the nascent chain elongates, and this effect occurs for both SRP-dependent and SRP-independent secretory protein substrates.
To evaluate the contribution of TF versus the intrinsic changes in SRP activity to the rejection of long-chain RNCs, we performed a numerical simulation using the thermodynamic and kinetic parameters obtained in Figs. 1, 4, and 5. First, we computed the fractional occupancy of SRP on RNCs (Fig. S5A). Second, an apparent rate constant was determined using the closed complex assembly rate constants and the estimated range of cellular FtsY concentration (Fig. S5 B and D) (59). Using these parameters, we calculated the amount of time required for SRP and FtsY to successfully deliver either short-(80mer) or long-chain (130mer) RNC 3A7L and RNC phoA to the membrane in the presence or absence of TF (Fig. S5 C and E). With RNC 3A7L-80 , targeting can be completed within 2 s. When the 3A7L nascent chain is lengthened to 130 residues, the amount of time needed to target RNC 3A7L-130 is lengthened, but only to ≥4 s in the absence of TF (Fig. S5 C and E, light pink). However, in the presence of TF, this value increases to 17-80 s, well beyond the time for the nascent protein to complete its synthesis on the ribosome (Fig. S5 C and E, dark pink and dashed line). Analogous observations were made with phoA: In the absence of TF, the intrinsic reductions in SRP activity at the longer nascent chain length would still allow RNC phoA-130 to be targeted within a reasonable time window (Fig.  S5 C and E, light pink). However, with TF present, SRP and FtsY require ≥53 s to target RNC phoA-130 (Fig. S5 C and E, dark pink). Thus, with the help of TF, long-chain RNCs exit the cotranslational targeting pathway effectively.
Collectively, the results in this section show that at longer nascent chain lengths, TF more strongly inhibits SRP functions including binding of the RNC and recruitment of SRP receptor. These allosteric inhibitions from TF could contribute significantly to the rejection of RNC from the SRP pathway once the nascent chain reaches a critical length. This effect is observed with both SRP-dependent and SRP-independent secretory proteins but, surprisingly, is not pronounced with integral membrane protein substrates such as FtsQ.
TF Enhances the Specificity of SRP-Dependent Protein Targeting. Our results strongly suggest that TF will enhance the discrimination of SRP against borderline substrates, such as EspP and phoA. To test this hypothesis, we determined the effect of TF on the targeting efficiency of proteins with different signal sequences. We used a well-established heterologous assay in which various signal sequences are fused to the mature region of an established SRP substrate, preprolactin (pPL). The pPL variants are translated in a wheat germ extract (60,61). The ability of Escherichia coli SRP and FtsY to mediate the cotranslational targeting of pPL variants to microsomal membranes is tested using cleavage of signal sequence to report on successful targeting and translocation (Fig. 6). Multiple factors render this assay an optimal choice for testing insights from biophysical studies of SRP in a complete targeting reaction: (i) Bacterial SRP and FtsY mediate pPL targeting as efficiently as their mammalian homologs despite the heterologous nature of this assay (61), highlighting the remarkable conservation of the SRP pathway; (ii) wheat germ extract is devoid of endogenous SRP and potential SRP regulators that may serve redundant functions as TF, which allows the effect of TF to be most readily unmasked; (iii) the microsomal membrane has much higher cotranslational translocation activity than bacterial membrane vesicles (62)(63)(64), so that the output of the assay is not limited by the downstream translocation steps; and (iv) rapid cleavage of signal sequences by signal peptidase allows both the premature and mature proteins to be quantified, providing the most accurate quantification of targeting efficiency; analogous readouts are not available with bacterial membrane vesicles. Finally, the Sec translocon has a more relaxed specificity than SRP and translocates both cotranslational and posttranslationally targeted substrates (65); thus, the overall assay reports on the more stringent substrate selection by the SRP pathway.
In agreement with previous observations, 2A8L-pPL and 3A7L-pPL are efficiently targeted and translocated in this assay (Fig. 6 A, B, and E), whereas borderline substrates such as EspP and phoA retained ∼20-25% targeting (Fig. 6 C-E). Importantly, the presence of TF reduced the targeting of EspP-pPL and phoA-pPL to ∼5-10% but has negligible effects on the targeting and translocation of 2A8L-pPL and 3A7L-pPL (Fig. 6 A-D; summarized in Fig. 6E). This provides direct evidence that the complex interplay between SRP and TF contributes to enhancing the specificity of cotranslational protein targeting.
A mathematical model has been established that calculates the fraction of substrates that can be targeted by the SRP pathway based on the kinetic and thermodynamic parameters of three stages of the targeting cycle: cargo binding, SRP•FtsY assembly, and kinetic proofreading using GTP hydrolysis (Fig. S1) (40). Using this model and the rate and equilibrium constants determined here, we numerically simulated how TF affects the fraction of substrates retained at each step of the SRP pathway. The calculation shows that TF exerts a very modest effect on the fraction of RNC bound by SRP: Although SRP occupies >95% of all four RNCs in the absence of TF (Fig. 7A, light gray bars), TF brings the occupancy of SRP on RNC FtsQ , RNC 3A7L , and RNC EspP to ∼90%, and that of RNC phoA to 75% (Fig. 7B, light gray bars). In the membrane targeting step, >90% of RNC FtsQ and RNC 3A7L are retained in the SRP pathway ( Fig. 7A) with or without TF. Without TF, ∼96% and 59% of RNC phoA and RNC EspP remain in the SRP pathway despite being SRP-independent substrates (Fig. 7A, dark gray bars) whereas, with TF present, the amount of RNC phoA and RNC EspP that remains in the SRP pathway is lowered to 42% and 9.2%, respectively, after the targeting step (Fig. 7B, dark gray bars). Additional substrates were rejected through competition between GTP hydrolysis and cargo unloading (dark bars) (40). The overall targeting efficiencies predicted by the numerical analysis agreed well with the results of the targeting and translocation assays in both the presence and absence of TF (Fig. 7 C and D), suggesting that this model reasonably captures the molecular determinants that govern the selectivity of the SRP pathway.

Discussion
Numerous factors, including molecular chaperones, modification enzymes, quality control complexes, and protein targeting  machineries, bind to the ribosome near the exit site and direct nascent polypeptides to distinct biogenesis pathways. Accurate selection of the nascent protein into the correct biogenesis pathway is essential to maintain the homeostasis of the proteome. In recent years, there has been an increasing number of structures of individual protein biogenesis factors bound to the ribosome exit site (3,18,26,56,(66)(67)(68)(69)(70)(71)(72)(73)(74), as well as efforts to globally catalog the nascent polypeptides they occupy (9, 13). However, the key question remains: Given the crowded environment at the ribosome exit site and a limited time window of action, how does a nascent polypeptide engage the correct set of factors and hence commit to its correct biogenesis pathway? Using SRP and TF as a model system, this work reveals a multitude of molecular mechanisms by which TF and SRP regulate each other's function. This interplay allows SRP to better discriminate against borderline cargos and thus ensures the correct partitioning of proteins to their respective biogenesis pathways. TF regulates SRP at three distinct stages (Fig. 7E). As the nascent polypeptide begins to emerge from the ribosome exit site (≤85 amino acids), TF weakens the SRP-RNC interaction by up to an order of magnitude (step 1). A mathematical analysis shows that this modulation translates into a modest change in the occupancy of SRP on the translating ribosomes, allowing TF to help remove a small fraction (∼20%) of SRPs from RNCs bearing some of the Sec-dependent substrates, such as phoA, at an early stage (Fig. 7E, step 1). In the second stage, TF selectively reduces the rates for stable SRP•FtsY complex assembly for RNCs bearing SecA/B-dependent, but not SRP-dependent, substrates (Fig. 7E, step 2), and thus slows the delivery of these RNCs to the target membrane. The kinetics of SRP•FtsY assembly is a crucial checkpoint in cotranslational protein targeting (34,40). Here, we find that TF further enhances the kinetic discrimination of SRP against suboptimal substrates during this step (Fig. 4F), significantly reducing the amount of RNC phoA and RNC EspP that remains in the SRP pathway after membrane targeting (Fig. S5B, dark gray bars).
In the third regulatory event, TF helps reject substrates from SRP when the nascent chain is elongated beyond a critical length (Fig. 7E, step 3). Rejection of long-chain RNCs is critical for maintaining the selectivity of SRP, as it sets a "timer" that irreversibly removes substrates that fail to complete targeting within a given time window. Indeed, slower translation elongation, which gives SRP a longer time window, allows substrates with weaker signal sequences to enter the SRP pathway (75). The results of this and previous work (47,57,58) show that multiple mechanisms together impose this timer. First, RNC-SRP binding affinity could weaken modestly with long-chain RNCs. Second, recruitment of FtsY to RNC•SRP complexes slows twofold to threefold when the nascent chain becomes longer. Importantly, TF exerts a larger inhibitory effect on both of these steps at the longer nascent chain length, and thus aids in the rejection of long-chain RNCs from the SRP pathway (Fig. S5). Finally, we found that even this timer is set differentially: Targeting of SRP substrates containing integral TMDs was less sensitive to the nascent chain length, whereas secretory proteins with strong and weak signal sequences are strongly discriminated.
As would be predicted from these biophysical properties, TF enhances the selectivity of SRP and helps prevent suboptimal substrates, such as phoA and EspP, from leaking into the SRP pathway ( Fig. 7 A-D). This could help prioritize limited amounts of SRP in a cell for strictly cotranslational substrates. The results of the targeting and translocation reactions, both in the presence and absence of TF, agree well with predictions from a mathematical analysis based on the kinetic and thermodynamic parameters for different stages of the targeting cycle ( Fig. 6 and Fig. S1) (40). This analysis highlights the importance of kinetic control at the SRP•FtsY assembly step as a major checkpoint that discriminates against suboptimal substrates. It also shows that TF allows suboptimal substrates to be rejected at an earlier stage (membrane targeting) and reduces the fraction of substrates that reaches the membrane and has to be rejected by proofreading (Fig. 7 A-D).
Reciprocally, SRP helps TF engage the correct substrates. At early stages, TF binds with comparable affinities to RNCs bearing nascent chains belonging to both the cotranslational and posttranslational targeting pathways, and even exhibits twofold  to fourfold higher affinity for RNCs bearing FtsQ and 3A7L than the phoA and EspP substrates ( Fig. 2E and ref. 20). This is not surprising given the preference of TF for hydrophobic sequences (4,45). It is via the regulation by SRP and FtsY, rather than the intrinsic affinity of TF for RNCs, that substrates such as FtsQ and 3A7L are removed from TF, allowing it to select and commit to Sec-dependent substrates. Collectively, the results here demonstrate a multilayered mechanism for the selection of nascent proteins into distinct biogenesis pathways (Fig. 7E). Given the high cellular concentration of TF, it is bound to >98% of translating ribosomes. There, TF alters the affinity and conformation of SRP binding to nascent polypeptides that emerge from the ribosome (step 1). This effect is modest, and removes a small fraction of SRP from some ribosomes bearing SRP-independent substrates. The majority of RNCs accumulate as ternary complexes with SRP and TF. For RNCs bearing SRP-dependent substrates, delivery to the membrane via FtsY recruitment is rapid and unaffected by TF; these substrates are hence selected into the cotranslational targeting and translocation pathway (Fig. 7E). In contrast, for RNCs bearing SRP-independent substrates, membrane targeting is slow and further slowed by TF (step 2). While these RNC•SRP•TF complexes bide for membrane targeting, ongoing protein synthesis quickly elongates the nascent polypeptide beyond a critical length, at which TF more effectively inhibits SRP functions (step 3). These proteins are hence rejected from SRP and sorted to either cytoplasmic folding or the posttranslational SecA/B pathway (Fig. 7E).
Our model reconciles many previous observations. The model that SRP and TF accumulate as ternary complexes with RNCs is consistent with cosedimentation, crosslinking, and FRET analyses (17,19,20). The ability of TF to alter the interaction of SRP with signal sequences is also consistent with less efficient crosslinks of SRP to nascent chains in the presence of TF (15). These data also show that, in the presence of other RNC-interacting factors, the initial interaction of TF/SRP with RNCs is more dynamic at short nascent chain lengths and that commitment of nascent proteins to TF occurs at longer nascent chain lengths. This explains the observation in ribosome-profiling experiments, where TF was found to stably associate with RNCs bearing SRP-independent substrates at nascent chain lengths >100 residues (13).
Our results emphasize the crowded environment at the ribosome exit site. The effect of TF in weakening RNC-SRP binding does not follow a hyperbolic concentration dependence expected for binding of a single TF molecule to each RNC, but rather displays a complex pattern (Fig. 5 B-D) indicative of the binding of multiple TFs to the RNC (Fig. 7E) as proposed previously (45). Further, the observation that TF alters the FRET value of SRP•FtsY early complexes strongly suggests that all three factors are present on the RNC at this stage. Our findings also suggest a remarkable degree of structural plasticity for cotranslational biogenesis machineries bound at the ribosome exit site, wherein each factor could alter its conformation to accommodate another factor bound in the vicinity. In support of this model, multiple conformations of SRP bound to RNCs have been observed recently in fluorescence and single-molecule studies (34,58). Regulation of SRP conformation by TF presumably leads to the observed weaker binding to RNCs and the slower recruitment of FtsY in the membrane-targeting step. How the ribosome exit site accommodates multiple factors, how SRP and the SRP•FtsY complex transition between alternative conformations in response to the signal sequence, and how this conformational flexibility is exploited by TF to enhance selectivity remain intriguing questions for future investigations.
Accurate selection of nascent proteins into distinct fates is at the heart of proper protein biogenesis. In additional to TF, additional factors that interact with the RNC could exert analogous regulations on SRP, TF, or other protein biogenesis machineries. The results here provide an example for how this regulation can help enhance the fidelity of substrate selection, and a conceptual framework for considering how nascent proteins are accurately sorted among other biogenesis machineries at the ribosome exit site, or in other crowded cellular environments.

Methods
Materials. The E. coli Ffh, FtsY, 4.5S RNA, and TF were expressed and purified using established protocols (76,77). Single cysteine mutations were introduced via Quikchange mutagenesis (Stratagene). Mutant proteins were purified using the same procedures as wild-type protein. Ffh (C153) was labeled with DACM and FtsY (C345), Ffh (C421), and TF (C377) were labeled with BODIPY-FL-N-(2-aminoethyl)-maleimide (Invitrogen) as described previously (32). Labeled proteins were purified from unconjugated dye using gel filtration chromatography on Sephadex G-25 resin (Sigma) as described (32). Labeling efficiencies were usually >90%. Homogenous RNCs were generated by in vitro translation and stalled using the sequence SEKGYR-IDYAHFTPQAKFSTPVWISQAQGIRAGPQRLS from the E. coli SecM protein (78). Translation stalling occurs at the underlined G, and the C-terminal PQRLS residues are not incorporated into the nascent chain. Short-chain and long-chain RNCs contain 80-85 and 130-135 amino acids, respectively, between the N terminus of signal sequence and the peptidyl transferase center of ribosome. RNCs were affinity-purified via the strep 3 tag on the nascent polypeptide, as described (41,48,78). The quality of RNC was documented in ref. 78. RNCs purified through this procedure are identical in their ability to bind SRP and accelerate SRP•FtsY assembly as RNCs further purified by sucrose gradient centrifugation (33,40). Amber suppression technology to incorporate Cm into the nascent polypeptide is described in detail in ref. 48. Cotranslational Protein Targeting and Translocation Assay. The protein targeting efficiency of SRP, with or without TF, was determined by a cotranslational translocation assay using [ 35 S]methionine-labeled pPL as a substrate, as described previously (60). The signal sequences of 3A7L, EspP, and phoA were fused to the mature region of prolactin. Reactions were carried out using 345 nM SRP, 0 μM or 16 μM TF, varying concentrations of FtsY (specified in figure legends), and 1.5 equivalent of trypsin-digested EDTA and salt-washed ER microsomal membranes. Reactions were analyzed by SDS/ PAGE followed by autoradiography. Mathematical modeling of targeting efficiency is described in SI Methods.