Ribosome-nascent chain interaction regulates N-terminal protein modification

Abstract

Numerous proteins initiate their folding, localization, and modifications early during translation, and emerging data show that the ribosome actively participates in diverse protein biogenesis pathways. Here we show that the ribosome imposes an additional layer of substrate selection during N-terminal methionine excision (NME), an essential protein modification in bacteria. Biochemical analyses show that cotranslational NME is exquisitely sensitive to a hydrophobic signal sequence or transmembrane domain near the N-terminus of the nascent polypeptide. The ability of the nascent chain to access the active site of NME enzymes dictates NME efficiency, which is inhibited by confinement of the nascent chain on the ribosome surface and exacerbated by signal recognition particle. In vivo measurements corroborate the inhibition of NME by an N-terminal hydrophobic sequence, suggesting the retention of formylmethionine on a substantial fraction of the secretory and membrane proteome. Our work demonstrates how molecular features of a protein regulate its cotranslational modification and highlights the active participation of the ribosome in protein biogenesis pathways via interactions of the ribosome surface with the nascent protein.

Graphical Abstract

Introduction

Proteins carry out the majority of biological functions in the cell. To acquire their functionality, newly synthesized proteins need to undergo adequate modifications, fold into the correct structures, and reach their proper cellular destinations. These processes, collectively termed protein biogenesis, often occur when the nascent polypeptide is being synthesized on the ribosome. The surface of the ribosome surrounding the nascent polypeptide tunnel exit serves as a hub to recruit various ribosome-associated protein biogenesis factors (RPBs) [1–3]. In bacteria, these include targeting factors such as Signal Recognition Particle (SRP) and SecA, cotranslational chaperones such as Trigger Factor (TF), and protein modification enzymes. The nascent protein encodes molecular signals to initiate dedicated protein biogenesis pathways. For instance, SRP recognizes a transmembrane domain (TMD) on membrane proteins, or a strong signal sequence containing a continuous stretch of hydrophobic amino acids [4] on a subset of secretory proteins. TF broadly engages hydrophobic sequences on cytosolic proteins as well as a subset of secretory proteins [5–8]. During translation elongation, emergence of new molecular signals on the nascent protein could allow distinct RPBs to be recruited or repositioned on the ribosome and potentially guide the sequence of cotranslational biogenesis events [1,2].

In addition to the recruitment of RPBs, the ribosomal surface also directly interacts with diverse nascent polypeptides. The surface of the 50S ribosome subunit is highly acidic due to the ribosomal RNA [9] and able to constrain nascent chains via electrostatic interactions [10,11]. In addition, multiple nonpolar regions are decorated on the ribosome surface at uL23/uL29, uL24, and bL17/bL32 [9,12]. Several nascent chains bearing hydrophobic signal sequences were reported to interact with uL23/uL29 adjacent to the tunnel exit [13–16]. Recently, it was reported that even an intrinsically disordered nascent protein crosslinks strongly to uL23 and the neighboring uL29 in a Mg²⁺-dependent manner [12]. These ribosome-nascent chain interactions have been shown to stabilize the nascent protein in an unfolded state until downstream sequences emerge to recruit cotranslational chaperones or binding partners[10,11,17,18]. This ‘holdase’ activity of the ribosome allows nascent proteins to avoid local kinetic traps and thus fold productively [18]. While the contribution of the ribosome in cotranslational protein folding has been actively studied, whether ribosome-nascent interactions participate in other biogenesis pathways remains unclear.

The covalent modification of nascent proteins constitutes an essential branch of cotranslational protein biogenesis. In bacteria, translation is initiated with formylmethionine (fMet), which is sequentially removed via the N-terminal methionine excision (NME) process catalyzed by two enzymes, peptide deformylase (PDF) and methionine aminopeptidase (MAP)[19,20]. Both enzymes are essential in bacteria and important targets for developing antibiotic drugs [19]. While PDF is promiscuous, MAP preferentially cleaves nascent polypeptides with small amino acids at the second residue after initiator methionine (iMet) [21,22]. Under physiological conditions, the NME reactions are further reshaped by the ribosome and other RPBs [7,23–25]. Using substrates bearing structureless N-terminal sequences from a membrane protein FtsQ and a cytosolic protein luciferase, our previous work demonstrated that the association of MAP with the ribosome leads to rate enhancements of 10²–10⁴ fold compared to reactions on peptide substrates, allowing diffusion-limited nascent chain processing by MAP. On the other hand, SRP and TF selectively restrict the time window for the reaction of suboptimal substrates with large side chains at the second residue, and thus ensure the local sequence specificity of cotranslational NME [23]. These observations underscore multiple mechanisms that can influence NME.

These previous studies were carried out on nascent polypeptides largely devoid of structural features near the N-termini, raising questions as to whether and how molecular features on the nascent chain in addition to the size of the second residue following iMet contribute to NME. Indeed, SRP was found to inhibit the deformylation of ribosome-nascent chain complexes (RNCs) bearing TMD or signal sequences on the nascent chain [24]. However, ~50% of the SRP-engaging proteins identified by selective ribosome profiling experiments underwent NME in vivo [26,27]. Moreover, FtsQ, a bona-fide SRP substrate, was efficiently processed by both PDF and MAP on the ribosome, suggesting that the SRP-induced inhibition of deformylation is conditional on other molecular determinants [23]. Whether and how molecular features on the nascent protein are sensed by the ribosome and other RPBs to determine the N-termini status of nascent proteins remain to be understood.

In this work, a combination of enzymatic kinetics, fluorescence measurements, and crosslinking analyses show that a contiguous stretch of hydrophobic residues, from either a TMD or a strong signal sequence, located near the N-terminus of the nascent chain inhibits both PDF and MAP reactions. This inhibition is due to confinement of the nascent chain on the ribosomal surface and is further exacerbated by SRP. Consistent with the biochemical measurements, this inhibitory effect remains dominant over the intrinsic sequence preference of the enzymes in vivo, leading to the retention of fMet on ~10% of the bacterial proteome. Our work reveals a previously unrecognized mechanism by which NME is regulated under physiological conditions, and provides a novel example of the active participation of the ribosome in protein biogenesis via its interaction with the nascent protein.

Results

An early TMD or signal sequence inhibits cotranslational NME.

Previous work based on peptide substrates showed that MAP strongly prefers substrates with a small residue following iMet [22]. To uncover additional features on the nascent protein that regulate NME efficiency and to test whether NME differs for proteins destined for different biogenesis pathways, we in vitro translated a series of model substrates in the E. coli S30 extract in the presence of ³⁵S-methionine [23](Fig. 1A). The model substrates consist of N-terminal sequences derived from proteins of interest fused to a cytosolic protein, thioredoxin (TrxA), and are free of methionine residues other than the iMet. To control for the specificity of MAP for small second-site residues, we replaced the second residue of all substrates with alanine (Figs. 1B and S1A). These substrates are subject to NME by endogenous PDF and MAP and to regulation by additional factors in the lysate during translation, and their NME efficiency is determined by the reduction of the iMet signal (Fig. 1A).

Figure 1: An early TMD or hydrophobic signal sequence inhibits cotranslational NME of nascent proteins.

(A, B) Scheme of the cotranslational NME assay. Model substrates (depicted in (B)) were translated in E. coli cell extracts containing ³⁵S-Met and processed by endogenous PDF and MAP. POI-N: N-terminal sequence of protein of interest. (C) Scheme of the PhoA variants. The hydrophobic core of the signal sequence was replaced with the indicated sequences and colored in dark green. The number of the residues N- and C-terminal to the signal sequence in POI-N are indicated. (D) Representative SDS/PAGE autoradiography of the cotranslational NME of the PhoA variants. Red arrows indicate the substrates translated in the extracts. Black arrows indicate the loading control. Reactions in the presence of actinonin (ACT, 5 µM) provided controls for the iMet signal on substrates without NME. (E) Efficiency of cotranslational NME of the PhoA variants, determined from the data in (D) and their replicates by the reduction of the iMet signal in the reactions without ACT relative to that in the presence of ACT. (F) Scheme of the FtsQ variants. The NTE of FtsQ (first 27 residues) preceding the TMD (residues 28–49) was systematically truncated from the C-terminus, as indicated. (G) Representative SDS/PAGE autoradiography for cotranslational NME of the FtsQ variants. (H) Efficiency of cotranslational NME of the FtsQ variants, determined from the data in (G) and their replicates. All values are reported as mean ± SEM, with n ≥ 3.

Despite bearing the same second site residue, different NME levels were observed for the model substrates tested (Figs. S1B, C). Substrates bearing the N-terminus of cytosolic proteins, luciferase (Luc) and chloramphenicol acetyltransferase (CAT), were completely processed by PDF and MAP. Complete NME was also observed for integral membrane proteins FtsQ and RodZ, which can cotranslationally engage SRP and SecA [5,28,29,27]. NME was modestly inhibited on the secretory proteins OmpA and PhoA, which harbor weakly hydrophobic signal sequences [5,7,30], and strongly inhibited on DsbA, a secretory protein with a more hydrophobic signal sequence. EspP, an autotransporter bearing a 38-residue N-terminal element (NTE) preceding a hydrophobic signal sequence [31], was efficiently processed, whereas removal of its NTE (EspP dNTE) abolished NME. These results demonstrate additional layers of regulation of NME efficiency contributed by the N-terminal signals beyond the size of the second residue.

A common feature of the strongly inhibited substrates, DsbA and EspP dNTE, is a hydrophobic signal sequence that closely follows iMet. To test the effect of signal sequence hydrophobicity on NME, we replaced the hydrophobic core of the PhoA signal sequence with leucine and alanine and systematically varied the Leu/Ala ratio [32] (Fig. 1C). NME was strongly inhibited on the more hydrophobic variants, 3A7L and 2A8L (Figs. 1D and E). Despite having a hydrophobic TMD, FtsQ and RodZ underwent complete NME (Figs. S1B and C). We reasoned that the long NTE preceding their TMD might allow efficient processing by PDF and MAP before the TMD emerges from the ribosome tunnel exit. To test how the position of the TMD affects NME, we systematically truncated the NTE of FtsQ (Fig. 1F). Substrates with NTE longer than 12 residues were efficiently processed, but NME efficiency decreased drastically upon further truncation, indicating that a threshold distance of the hydrophobic signal sequence or TMD to the N-terminus is required for efficient NME (Figs. 1G and H).

To decipher whether the inhibition is exerted on the PDF or MAP reaction, we analyzed the N-terminus of the NME-inhibited substrates after their coupled translation and processing in the S30 extract (Figs. S1D–F) using thin layer chromatography (TLC). Substantial remaining fMet was detected, indicating partial inhibition of the deformylation reaction (Figs. S1E and F). However, deformylation was reduced by only 50% on 2A8L, EspP dNTE, and DsbA and cannot fully account for the inhibition of NME, suggesting that both the PDF and MAP reactions were compromised on these substrates (Fig. S1F).

The results above show that a hydrophobic signal sequence or TMD near the N-terminus of a nascent protein prevents its processing by PDF and MAP. This inhibition could arise from the intrinsic properties of these sequences, or from the action of additional RPBs in the cell lysate such as SRP. To distinguish between these possibilities, we prepared ribosome-nascent chain complexes (RNCs) with a defined nascent chain sequence and a chain length previously determined to be optimal for processing by PDF and MAP [23]. The RNCs were generated via in vitro translation in S30 extract, and purified via a stringent sucrose-gradient centrifugation to remove ribosome-associated factors [23]. We carried out pre-steady-state kinetic measurements of the PDF and MAP reactions on the purified RNCs under single-turnover conditions at sub-saturating enzyme concentrations (Fig. 2A) [23].

Figure 2: NME reactions are directly inhibited by sequence features on the nascent chain.

(A–C) Scheme of the RNC substrates used in the NME reaction measurements. An 8 amino acid stall sequence at the C-terminus, derived from Ms-sup1, allows the generation of translationally stalled nascent chains with defined lengths. The total length of the nascent chains is 67 residues for FtsQ variants (B) and 69 residues for PhoA variants (C). The release of ³⁵S-labeled (asterisk) iMet provides readout for the extent of the reactions. (D) and (E) Summary of the observed rate constants of the PDF (D) and MAP (E) reactions on purified RNCs with the FtsQ variants. The PDF reactions (D) were measured with 10 nM RNC and 100 nM PDF for RNC_FtsQ and RNC_{FtsQ N12} or 500 nM PDF for RNC_{FtsQ N8} and RNC_{FtsQ N5}. The MAP reactions (E) contained 10 nM RNC and 100 nM MAP. Apparent k_cat/K_m is the observed pseudo-first-order rate constant divided by enzyme concentration. (F) and (G) Summary of the observed rate constants of the PDF (F) and MAP (G) reactions on purified RNCs with the PhoA variants. The reaction rates of RNC_Luc (open bars) were included for comparison with a substrate without a signal sequence. The PDF reactions (F) were measured with 10 nM RNC and 100 nM PDF for RNC_Luc, RNC_PhoA and RNC_5A5L or 500 nM PDF for RNC_3A7L and RNC_2A8L. The MAP reactions (G) contained 10 nM RNC and 100 nM MAP. All values are reported as mean ± SEM, with n ≥ 2. ‘*’, ‘**’ and ‘***’ represent p ≤ 0.05, 0.01, and 0.001, respectively, based on unpaired t-tests. ‘ns’, p > 0.05.

Both the PDF and MAP reactions are sensitive to the location and hydrophobicity of the membrane targeting signals. For FtsQ variants in which the NTE preceding the TMD is systematically truncated (Fig. 2B), the PDF and MAP reactions were inhibited by 30- and 10-fold, respectively, when the NTE was shortened to 8 residues, indicating that a flexible NTE of at least 9–12 residues is required for efficient NME (Figs. 2D and E). For PhoA variants in which the hydrophobicity of the N-terminal signal sequence was varied (Fig. 2C), the PDF reaction was slowed ~10 fold by a signal sequence with moderate hydrophobicity (5A5L vs. Luc, Figs. 2F) and 100–200 fold by more hydrophobic signal sequences (3A7L and 2A8L, Fig. 2F). The MAP reaction was also susceptible to this inhibition, with 4 fold and 40–80 fold slower rates on substrates with moderate and high hydrophobicity, respectively (Fig. 2G).

In addition to hydrophobicity, a signal sequence or TMD could adopt helical structures that prevent the N-terminus from entering the enzyme active site. However, helical contents of the signal sequences, calculated from the prediction algorithm AGADIR, did not correlate with the observed NME reaction rates [33] (Figs. S2). For instance, the 5A5L signal sequence has a significantly higher helical propensity than PhoA, but the reaction rate of 5A5L is comparable to or even slightly faster than that of PhoA (Figs. 2F and G). Therefore, secondary structure propensity is not sufficient to explain the inhibitory effect of an N-terminal signal sequence or TMD.

Collectively, the results in this section show that a hydrophobic signal sequence or TMD that emerges within 12 amino acids of the N-terminus inhibits the cotranslational modification of nascent proteins by both PDF and MAP. The observation of enzymatic inhibition in the purified system further indicates that the molecular properties of the nascent chain are sufficient to regulate the PDF and MAP reactions.

Nascent chain-ribosome interactions regulate NME.

To understand the mechanisms by which the nascent chain sequence regulates its cotranslational NME, we considered a model in which binding of the enzymes to the RNC is followed by docking of the nascent chain N-termini at the enzyme active site, which mediates hydrolysis of the peptide bond (Fig. 3A). Any effect of the N-termini at the hydrolysis step can be excluded, as residues 2–4 following iMet that directly interact with the enzyme active site are identical within each set of the variants tested (Fig. 3A, Step 3; Figs. 2B and C) [21,22]. We therefore examined whether the N-terminal hydrophobic signal sequence exerts regulation by altering the ability of NME enzymes to bind the RNC (Fig. 3A, Step 1) or the access of the nascent chain N-termini to the enzyme active site (Step 2).

Figure 3: Interaction between the ribosome and the nascent chain regulates NME efficiency.

(A) Proposed model of PDF or MAP reactions on the RNC. (B) Scheme of the FRET-based binding assay. The ribosomal protein bL17 was genetically modified with a ybbR tag at the C-terminus and labeled with BODIPY-FL (green star). The nascent chain length is 67 residues for Luc and the FtsQ variants, and 69 residues for the PhoA variants. TMR (red star) was labeled at the N-terminus of PDF(E133A) or the C-terminus of MAP(H79A). (C) and (D) Summary of the K_d values between the indicated RNCs and PDF(E133A) (C) or MAP(H79A) (D), determined from equilibrium titrations of TMR-labeled PDF(E133A) or MAP(H79A) to 10 nM BODIPY-FL-labeled RNCs. Values are reported as mean ± SEM, with n ≥ 2. (E, F) Observed rate constants of the PDF (E) and MAP (F) reactions as a function of the hydropathy score of the nascent chains, defined as the GRAVY score [41]of the most hydrophobic stretch of 11 residues that begins within the first 11 amino acids. Apparent k_cat/K_m values are from Figs. 2D–G. Lines are fits of the data to Eq. 6 in Methods, which describes the two-state model in (A). (G) Scheme of the RNCs used in the crosslinking experiments. The ribosomal protein uL23 carries a mutation S21C. A cysteine is placed near or in the TMD/signal sequence (dark green) of the model substrates, as detailed in Methods. The nascent chain length from the methionine to the PTC is 67 residues for FtsQN8 and DsbA, and 69 residues for the PhoA variants. (H) Representative western blots of the crosslinking experiment. The RNCs (500 nM) were incubated with or without 0.8 mM 1,4-bismaleimidobutane (BMB). Crosslinked products between the nascent chain and uL23 (red arrows) were detected with anti-Strep (for nascent chain) and anti-uL23 antibodies. Asterisks denote non-specific bands detected by the anti-uL23 antibody.

To test the first model, we established a fluorescence-based assay to measure the binding of the ribosome with PDF and MAP (Fig. 3B and Fig. S3). We incorporated a ybbR tag at the C-terminus of ribosomal protein bL17, which allows labeling of a donor dye (BODIPY-FL) via Sfp-mediated reactions [34]. An acceptor dye (tetramethylrhodamine, TMR) was labeled at the N-terminus of PDF or the C-terminus of MAP. Based on structural information, the dye pairs are ~40 Å apart in the ribosome•PDF or ribosome•MAP complex [35]. An affinity tag (3X Strep) followed by ubiquitin (Ub) was fused N-terminally to the nascent peptide of interest, allowing affinity purification of the RNCs followed by Usp2-catalyzed removal of the N-terminal Ub fusion to generate nascent chains with the native iMet [36]. To bypass potential complications from enzymatic processing of the nascent chain, we used catalytically inactive mutants, PDF(E133A) or MAP(H79A), for ribosome binding measurements [37–39].

The equilibrium dissociation constants (K_d) of PDF(E133A) and MAP(H79A) for the non-translating 70S ribosome are 7.3 ± 0.7 µM and 5.4 ± 0.8 µM, respectively, comparable to albeit slightly higher than the reported values of 2.7 µM and 2.4 µM using the wild-type enzymes [25,40] (Figs. 3C and D). The presence of the nascent chain does not affect the ribosome binding affinity of PDF (Fig. 3C). The FtsQ, PhoA and Luc nascent chains modestly stabilized the ribosome-MAP interaction by 2–3 fold (Fig. 3D). Nevertheless, substrates with 10–100 fold difference in reaction rates differ in K_d by < 3 fold (cf. FtsQ and FtsQN8; Luc and 2A8L; Fig. 3D). Thus, differences in enzyme-ribosome binding are insufficient to account for the strong inhibition of the PDF and MAP reactions (Figs. 2D–G).

These results strongly suggest that the inhibitory effect of the hydrophobic N-terminal sequences on NME arises primarily from differential access of the nascent chain N-termini to the enzymes on the ribosome. To test this model, we used an established fluorescence assay in which an environmentally-sensitive probe, coumarin, at the fifth residue of RNC_FtsQ undergoes an increase in fluorescence upon docking at the MAP active site ([23] and Fig. S4A). In contrast to RNC_FtsQ, the coumarin fluorescence signal of RNC_FtsQN8 remained unchanged, suggesting a failure of proper positioning of the FtsQN8 nascent chain at the MAP active site (Fig. S4B).

Based on these results, we hypothesized that hydrophobic interactions with the ribosome surface confine the nascent chain N-terminus at a position inaccessible to the enzyme active sites. The interaction biases the equilibrium toward the “inactive” state, whereas weaker interaction allows the nascent chain to sample the “active” state more frequently and complete the hydrolysis step (Fig. 3A). To test if this model can quantitatively explain the differential reaction kinetics of the model substrates, we derived an analytical equation according to the two-state model, which describes the hydrophobic interaction as an additional activation barrier for the enzymatic reactions by disfavoring the docking equilibrium (see Methods). Considering the threshold length of NTE required for efficient processing (≥ 12 residues, Figs. 2D and E), we calculated the grand average of hydropathy (GRAVY) score of the most hydrophobic stretch of amino acids that begins within 11 residues from the N-terminus [41,42]. The experimentally measured reaction rates fit well to the two-state model (Figs. 3E and F, Methods): the observed rate constants of both the PDF and MAP reactions correlate negatively with the calculated GRAVY scores for substrates with an N-terminal hydrophobic TMD/signal sequence, for which the docking equilibrium is unfavorable, but plateau at a maximum velocity for nascent chains below a threshold GRAVY score, for which the maximal occupancy of the docked state is achieved. The good agreement with data demonstrates hydrophobic contacts of the nascent chain, presumably with the ribosome surface, as a key determinant that drive the activation of NME reactions.

The ribosomal protein uL23, together with the neighboring uL29, has been shown to constitute an interaction site for multiple nascent protein sequences and contains a hydrophobic cleft on which an emerging TMD can dock [13–16]. The 50 Å distance between the uL23 cleft and PDF/MAP on the ribosome is also consistent with the threshold length of the NTE (~12–13 residues in an extended conformation) required for efficient enzymatic processing of the nascent chain N-termini (Figs. S4C and D). These observations suggest that uL23 could constitute one of the sites on the ribosome that confine the movement of the nascent polypeptide. To test this possibility, we carried out site-specific crosslinking to test the docking of the TMD/signal sequence on uL23, taking advantage of the fact that the vicinity of the ribosome exit site is free of native cysteines [16] (Fig. 3G). We engineered a single cysteine on uL23 (S21C) and a second cysteine at or near the TMD/signal sequence of the nascent chain. We observed specific and prominent crosslinks between the nascent chain and uL23 on RNCs bearing a TMD/signal sequence (Fig. S4E, F and Fig. 3H), indicating that nascent chains can dock on uL23 and sample a conformation that disfavors their enzymatic processing by PDF and MAP. Nevertheless, the efficiency of crosslink with L23 does not correlate with the NME reaction kinetics, strongly suggesting that additional sites surrounding the ribosome exit tunnel contribute to the sequestration of the nascent chain.

RPBs modulate the accessibility of nascent chains to regulate NME.

At the ribosome tunnel exit, multiple RPBs coordinate for access to the nascent protein and can potentially regulate NME [1]. This includes SRP, which recognizes N-terminal TMDs and strongly hydrophobic signal sequences to mediate membrane targeting, and TF, which engages cytosolic proteins and a subset of the secretory proteome [1,43]. To test this possibility, we measured the NME reactions on the model RNC_FtsQ and RNC_PhoA variants in the presence of SRP and TF (Figs. 4A–D).

Figure 4: RPBs regulate the NME reactions primarily by reorienting the nascent chain.

(A-D) The effect of RPBs (400 nM SRP, 5 µM TF, or both) on the PDF (A, B) and MAP (C, D) reactions with the FtsQ (A, C) and PhoA nascent chain variants (B, D). Reactions were carried out as described in Figure 2. Daggers indicate that (k_cat/K_m)_app is < 200 M-1s-1. Statistical tests were carried out on log(k_cat/K_m)_app values using unpaired t-tests, where ‘*’, ‘**’ and ‘***’ represent p < 0.05, 0.01, and 0.001, respectively. (E) and (F) The effect of RPBs (400 nM SRP or 5 µM TF) on the binding affinity of PDF(E133A) (E) and MAP(H79A) (F) for the indicated RNCs. The K_d measurements were carried out as described in Figs. 3B–D. All values are reported as mean ± SEM with n ≥ 2.

Although RNC_FtsQ is a bona fide SRP substrate, the presence of SRP affected the PDF and MAP reactions by less than two fold (Figs. 4A and C, blue bars), consistent with previous results and the efficient cotranslational NME of FtsQ in the cell extract [23] (Fig. 1G). With substrates in which the NTE is ≤12 residues, SRP induced 50 to >100-fold inhibition of the PDF and MAP reactions, respectively (Figs. 4A and C, blue bars), in agreement with the observations by Ranjan et al. [24]. SRP did not significantly affect the NME of substrates with weak signal sequences (RNC_5A5L and RNC_PhoA). However, reactions on the more strongly hydrophobic RNC_3A7L and RNC_2A8L substrates were slowed ~10- and > 50-fold by SRP, respectively (Figs. 4B and D, blue bars). The SRP-induced inhibition was not due to weakened binding of the enzymes to the RNC, as direct binding assays detected a <3-fold effect of SRP on the RNC binding affinity of PDF or MAP (Figs. 4E and F), consistent with previous observations [44]. These results show that SRP regulates the NME reactions on substrates with an N-terminal TMD/signal sequence, likely by confining the membrane targeting sequences to its signal sequence-binding M-domain at the ribosome tunnel exit and thus excluding the access of the N-termini to PDF and MAP [45,46].

Unexpectedly, TF accelerated the PDF reactions on substrates with a hydrophobic N-terminal sequence/TMD by 3–5 fold (Fig. 4A–B, FtsQ N8, 3A7L, and 2A8L, green bars). The RNC binding affinity of PDF and MAP was unaffected or slightly weakened (< 2-fold) by TF, arguing against facilitated recruitment of the enzymes as the source of the TF-induced rate enhancement (Fig. 4E). Importantly, TF induced cysteine-specific crosslinks between the signal sequence and an engineered cysteine on uL22 (S30C) (Fig. S5), suggesting that TF brings the nascent chain close to uL22 near the PDF and MAP binding sites [47,48]. The effects of SRP and TF on the PDF reaction counteract each other for a substrate with modest hydrophobicity (Figs. 4B, 3A7L), whereas the inhibitory effect of SRP dominates for substrates with strongly hydrophobic N-terminal targeting signals (Figs. 4A, B, FtsQ N8, FtsQ N5, and 2A8L). Thus, the interplay between SRP and TF on the RNCs enables intricate regulation of NME on a broad spectrum of substrates.

To further dissect the roles of the ribosome surface and RPBs in regulating the NME efficiency of the model substrates, we leveraged a previously established computational model [23], which uses parametrized PDF and MAP reaction rates on RNC based on experimental data to numerically calculate the extent of NME of model substrates during ongoing protein synthesis in cell lysate. This analysis showed that, in the absence of RPBs, the 2A8L and 3A7L nascent chains underwent ~10% and 20% cotranslational NME, respectively, when 300 residues are synthesized (Fig. S6A). NME was abolished in the presence of SRP, while the additional presence of TF allowed ~10% NME on 3A7L (Figs. S6B and C). These results indicate that during continuous protein synthesis, the N-terminal hydrophobic sequences on nascent proteins is sufficient to suppress their cotranslational NME, but the presence of SRP and TF leads to more complete inhibition.

Proteins with hydrophobic N-terminal targeting signals are resistant to NME in vivo

To quantitatively measure NME efficiency in vivo, we designed a reporter-based assay. We constructed an arabinose-inducible bicistronic plasmid, in which the model substrate and the control protein TrxA are encoded after two separate Shine-Dalgarno sequences (Fig. 5A). The proteins were expressed in E. coli cells, pulse-labeled with 35S-Met/Cys, and immunoprecipitated via the C-terminal FLAG tag. All the internal methionine and cysteine residues on the model substrates were removed such that the radioactive signal reports on the retention of iMet (Figs. 5B, D and F, “iMet” lanes). Wild-type substrates containing native methionines and cysteines were measured simultaneously (“6M” and “8M4C” lanes) to control for different expression levels of the model substrates. We first measured the in vivo NME of full-length FtsQ with varying residues at the second position (Figs. 5B and C). Consistent with the sequence specificity of MAP [22], the retention of iMet on these substrates strongly correlated with the size of the second residue, suggesting that the assay faithfully reports on the NME efficiency of model substrates in vivo.

Figure 5: Proteins with an early TMD or hydrophobic signal sequence are resistant to NME in vivo.

(A) Scheme of the model substrates used in the in vivo NME assay. The protein of interest (POI) and the control protein TrxA was tagged with a C-terminal 3X FLAG tag and expressed from a bicistronic plasmid under the control of an arabinose promoter (P_ara) and their respective Shine-Dalgarno sequences (SD1 and SD2). (B) In vivo NME of model FtsQ-X2 variants (red arrows), in which the second residue (X2) was mutated to the indicated amino acids. The substrates were induced in the E. coil strain CAG12184 and pulse-labeled with ³⁵S-Met/Cys for 10 min, immunoprecipitated, and analyzed by autoradiography. The signal of the iMet constructs, in which all six internal methionines were replaced with leucine, indicates the retention of iMet. The signal of the 6M constructs, which retain the six native methionines, serves as the control for protein expression level. TrxA (black arrows) serves as the loading control. (C) Quantification of the data in (B) and their replicates. The iMet signal was divided by the 6M signal to account for variations in expression level and normalized to that of FtsQ-K2, which completely retains the iMet. Values are reported as mean ± SEM, with n = 3 biological replicates. (D) In vivo NME of FtsQ truncation variants, measured as in (B). (E) Quantification of the data in (D) and their replicates. Values are normalized to FtsQ-K2 and reported as mean ± SEM, with n = 3 biological replicates. (F) In vivo NME of PhoA-A21L variants (red arrows), measured as described in (B). The 8M4C constructs contain the 8 native methionines and 4 native cysteines in PhoA and serves as controls for variations in expression levels. The second residue is alanine in all PhoA variants. (G) Quantification of the data in (F) and their replicates. Values are normalized to that of 2A8L and reported as mean ± SEM, with n = 3 biological replicates. (H) Distribution of the hydropathy of the N-terminal sequences in the E. coli proteome, calculated as in Figs. 3G and 3H. Proteins bearing N-terminal sequences with GRAVY scores > 2.6 (comparable to or more hydrophobic than FtsQ N8) are colored in dark gray. (I) Enriched functional annotation terms (Gene Ontology and Uniprot Keywords) associated with the 426 proteins highlighted in (F).

To understand the effect of a hydrophobic N-terminal targeting signal on NME in vivo, we tested the FtsQ variants in which the NTE is systematically truncated. All FtsQ variants were successfully integrated into the membrane and, except for FtsQ-N5, adopted the correct Type-II topology, excluding artifacts such as protein misfolding (Figs. S7, A-C). Consistent with the analysis in the reconstituted system and in cell extract, we observed significant accumulation of iMet on substrates with an early TMD, comparable to the strongly disfavored NME substrate with lysine at the second residue (Figs. 5D and E). Significant iMet retention was also observed with FtsQ-N8 and FtsQ-N5 in cells lacking methionyl-tRNA formyltransferase (∆fmt), in which deformylation is bypassed for NME, providing direct evidence for inhibition of the MAP reaction in vivo (Figs. S7D and E). Similarly, we carried out in vivo NME assays on the PhoA signal sequence variants. The mutation A21L prevents cleavage of the signal sequence by leader peptidase, allowing us to monitor the N-terminal modification of the substrate [49]. We observed increased retention of iMet on 3A7L and 2A8L compared to 5A5L and PhoA (Figs. 5F and G), in good agreement with our in vitro experiments. Collectively, these results show that a hydrophobic membrane targeting sequence near the N-terminus of a protein is a strong inhibitory signal for NME in vivo.

Based on these in vivo NME measurements and the two-state model derived from in vitro kinetic analysis, proteins with N-terminal signal sequences comparable to or more hydrophobic than FtsQ-N8 and 3A7L are anticipated to retain fMet in vivo. We therefore calculated the GRAVY score of the most hydrophobic segment close to the N-terminus across the E. coli proteome (Fig. 5H), as described for the model substrates in Figs. 3E and F. This analysis predicted that 427 proteins, predominantly composed of inner membrane, periplasmic and outer membrane proteins, are subject to this regulation (Fig. 5I). Consistent with our prediction, several studies reported the retention of the formyl group on multiple membrane proteins with an early TMD [26,50–52]. These results strongly suggest that, despite the promiscuity of PDF, a non-trivial fraction of the proteome retains fMet, which may provide an embedded regulatory signal on the nascent protein.

Discussion

NME is a ubiquitous protein modification essential for cell viability and occurs in an obligatorily cotranslational mechanism. While the structure and enzymology of NME enzymes have been extensively studied based on peptide substrates and proteomics data, whether and how this modification is regulated at the ribosome tunnel exit in coordination with other protein biogenesis pathways remain unclear. Our work here demonstrates that hydrophobic membrane targeting signals on the nascent protein impose an overriding layer of regulation on NME in vitro and in vivo. Kinetic analyses combined with crosslinking experiments suggest the role of hydrophobic interactions between the nascent chain and the ribosomal surface in driving the activation of the enzymatic reactions. The NME regulation is further enforced by the targeting factor SRP, exemplifying the active role of the ribosome in modulating nascent protein biogenesis pathways.

Emerging data show that the ribosome is not only a protein synthesis machine, but also actively participates in diverse protein biogenesis pathways [1,2]. In addition to providing a platform for the recruitment of multiple RPBs, the ribosome interacts with diverse nascent polypeptides both within the exit tunnel and on the ribosome surface [5,11–15,18,53]. These interactions can regulate the rate of translation elongation as well as cotranslational protein folding [10,11,18,54,55]. Here, our results suggest that hydrophobic interaction of the nascent protein with the ribosome surface could also regulate their N-terminal modification. While the ribosome surface is mostly acidic, multiple proteins including uL23, uL29, uL24, and bL17/bL32 [9,12] present contiguous hydrophobic surfaces that could provide docking sites for signal sequence or TMDs ([12] and this work), which sequester the nascent chain N-terminus from access by the NME enzymes. Moreover, the ribosome can interact with nascent chains via electrostatic and metal ion-mediated interactions [10–12]. Recent studies showed that interaction of the ribosome surface with the nascent chain of phosphorylated insulin receptor (PIR) and ɑ-synuclein clusters to a specific region near the tunnel exit, where uL23 and uL29 are located [11,56]. Whether these interactions confine the conformation of other classes of nascent chains and whether such confinement influences their NME remains to be explored.

Together with previous studies, our work here explains how PDF and MAP coordinate with other RPBs at the crowded ribosome exit site. First, interaction with the ribosome dominates the binding affinity of PDF and MAP. The binding of either enzyme to the RNC is not significantly stabilized by the nascent polypeptide (this work), nor affected more than 3-fold by SRP or TF (Figs. 4E and F). Reciprocally, TF or SRP binding to the ribosome is weakened <2-fold by physiological amount of PDF and unaffected by MAP [25,44]. These results are consistent with structural studies showing that the NME enzymes can flexibly adopt alternative binding modes on the ribosome in response to SRP and TF binding [25,35,44,57]. Coupled with their high ribosome association and dissociation rates, both enzymes can rapidly scan a variety of translating ribosomes for substrates [23,25]. Secondly, efficient NME initiates for nascent chains as short as 45 amino acids [23,25], and the irreversible chemical steps of the PDF and MAP reactions are rapid for an optimal N-terminus (k_cat > 200 s⁻¹ for PDF and >30 s⁻¹ for MAP [23,58]). Thus, the action of the NME enzymes on optimal substrates are kinetically privileged compared to the other RPBs. Notably, although physiological concentrations of SRP reduced the rates of the PDF and MAP reactions on RNC_FtsQ-N12 by ~20 fold, FtsQ-N12 underwent efficient NME in the extract and in vivo, likely reflecting this kinetic privilege. Finally, NME of suboptimal substrates is further restricted to a limited time window during translation. Nascent chains with large side chains at the second residue are inhibited by SRP and TF at nascent chain lengths below 67 residues and above 82 residues [23]. Proteins with an early TMD or signal sequence are targeted to the plasma membrane as soon as their membrane targeting motifs emerge from the ribosome, and are thus quickly removed from access by the NME enzymes (this work).

Our results indicate that other RPBs could regulate NME by tuning the accessibility of the nascent chain N-termini to the enzyme. SRP, with an M-domain that engages hydrophobic signal sequences/TMDs at the ribosome tunnel exit, restricts the sampling of the nascent chain, such that only membrane proteins with a long extension preceding the TMD efficiently undergo NME. TF, which forms a cradle to protect patches of hydrophobic amino acids on an elongating nascent protein, brings the nascent chain near the PDF/MAP docking site on the ribosome and could modestly stimulate NME. The two RPBs counteract one another, but SRP dominates on substrates with strongly hydrophobic membrane targeting signals, reflecting the specificity of this targeting factor [43]. Moreover, regulation of NME by RPBs is substrate- and nascent chain length-dependent. While TF enhances the PDF reaction on substrates with N-terminal targeting signals, it inhibits NME on cytosolic proteins and unstructured nascent chains beyond the optimal length, suggesting its versatile role in modulating nascent protein modification [7,23,25]. In addition to SRP and TF, SecA and DnaK also interact with hydrophobic nascent polypeptides [59,60]. However, stable binding of SecA to RNCs requires the synthesis of >100 residues [59], including the periplasmic domain of RodZ, a membrane protein that depends on SecA for its cotranslational targeting [29]. DnaK does not bind to the ribosome and is generally considered to act downstream of TF during protein synthesis [61]. Considering their late engagement during protein maturation, SecA and DnaK are unlikely to significantly affect the NME reactions and their regulation by SRP and TF under physiological conditions.

Our results here, together with previous work, provide a more complete model for how the NME enzymes mediate efficient and selective cotranslational processing of a large fraction of the bacterial proteome (Fig. 6). Although both PDF and MAP are sub-stoichiometric to translating ribosomes in cells [62], their rapid ribosome binding and dissociation rates enable these enzymes to scan translating ribosomes for the emerging nascent polypeptide (Step 1) ([23,25,44] and this work). Ribosome binding greatly increases the effective enzyme concentration with respect to the nascent protein, enabling rapid processing of optimal substrates. On the other hand, the interaction of nascent proteins with the ribosome surface regulates the flexibility of the nascent chain and its ability to search for and access the enzyme active site. Cytosolic proteins and secretory proteins with weakly hydrophobic signal sequences can readily dock at the enzyme active site, where the nascent protein is deformylated by PDF (Step 2) followed by excision of iMet if the second residue is small (Step 3). In contrast, an early TMD or hydrophobic signal sequence confines the nascent chain on the ribosome surface at a site away from the enzymes, such as on uL23, and inhibits their access to and processing by both enzymes (Step 4). Recruitment of SRP, which engages the signal sequence/TMD in the M-domain at the ribosome tunnel exit, further precludes the N-termini from reaching the enzyme active sites (Step 5). Together, these effects lead to the retention of fMet on a significant fraction of the membrane and secretory proteome (Step 6).

Figure 6: Model for the regulation of NME by interaction of the nascent chain with the ribosome surface and with SRP.

PDF and MAP rapidly scan translating ribosomes for emerging nascent proteins (Step 1). Proteins without strongly hydrophobic sequences freely access the enzyme active site, allowing efficient processing by PDF (Step 2) and MAP (Step 3) given a small penultimate residue. On the other hand, an N-terminal TMD or hydrophobic signal sequence interacts with the ribosome surface and is precluded from reaching the NME enzymes (Step 4), which is exacerbated by SRP (Step 5). The NME inhibition leads to fMet retention on a significant fraction of the membrane and secretory proteome (Step 6), and may potentially enable their quality control in case of failed translocation (Step 7).

The function of fMet retention remains an outstanding question. The significant enrichment of membrane and secretory proteins in the NME-inhibited proteome suggests potential roles of fMet in the biogenesis and/or quality control of these proteins. Diverse roles in the folding, assembly, and quality control of proteins have been described for N-terminal acetylation, a modification that is chemically similar to the formyl group [63–66]. Previous studies also suggested that fMet can serve as a degradation signal for bacterial and yeast proteins [67,68]. Although not yet identified, the protease responsible for the fMet/N-degron pathway in bacteria resides in the cytoplasm or on the inner membrane and was proposed to be FtsH, a membrane-embedded AAA+ protease facing toward the cytoplasm [67]. It is plausible that fMet enables the degradation of membrane/secretory proteins that fail in targeting and/or translocation (Fig. 6, Step 7). While membrane proteins with long and flexible terminus can be recognized and degraded readily [69,70], the retention of fMet on proteins with an early signal sequence or TMD may provide an alternative recognition signal for quality control pathways and facilitate the clearance of mislocalized proteins.

Materials and Methods

Cotranslational NME assay

The POI-N sequences and TrxA were cloned into the pK7-derived plasmids using Gibson cloning and Fastcloning techniques to generate the pK7-POI-TrxA plasmids [23,71–73]. All internal methionine residues and the second residue of each POI-N were replaced with alanine. DNA fragments encoding the T7 promoter, T7 terminator and the model substrates were PCR-amplified from pK7-POI-TrxA, transcribed, and translated in vitro in E. coli S30 extracts containing ³⁵S-methionine as described [23], and analyzed by SDS/PAGE and autoradiography. Where indicated, the control reactions contained 5 μM actinonin (ACT). Cotranslational NME efficiency is quantified as $\left(1-\frac{I_{-ACT}}{I_{+ACT}}\right)\times 100\%$, where I indicates the ³⁵S signal from the substrate.

Cotranslational deformylation assay

A Strep tag was introduced to the C-terminus of thioredoxin in the pK7-POI-TrxA constructs for the indicated model substrates to generate the pK7-POI-TrxA-Strep plasmids. The substrates were translated as described for the cotranslational NME assay and immunoprecipitated using Strep-tactin Sepharose resin. Purified substrates were digested with proteinase K (1 mg/mL, 37 °C, overnight), and radioactive Met and fMet were analyzed by thin-layer chromatography (Millipore Silica gel 60, n-butanol: acetic acid: water = 5: 3: 2) and autoradiography [24]. Cotranslational deformylation efficiency is quantified as $\left(1-\frac{I_{-ACT}}{I_{+ACT}}\right)\times 100\%$, where I indicates the ³⁵S signal from the fMet species.

Protein expression and purification

Wild-type PDF, MAP, MAP-H79A, SRP and TF were expressed and purified as described [23,74,75].

To construct the expression plasmid for PDF-E133A, an 11-residue ybbR tag was fused to the protein coding sequence, and subcloned into pET28a vector containing a cleavable N-terminal His₆-SUMO tag [34]. To express PDF-E133A, BL21 star (DE3) cells were induced at OD = 0.6 with 0.5 mM IPTG at 30 °C for 3 h, and lysed by sonication in buffer A (20 mM Hepes-KOH, 300 mM NaCl, 0.2 mM CoCl₂, 10% glycerol, 1 mM TCEP, pH 7.5) containing protease inhibitor cocktail (GoldBio). Clarified lysates were purified with Ni-NTA resin, and the His₆-SUMO tag was removed by overnight dialysis in the presence of His₆-Ulp1 protease against buffer B (20 mM Hepes-KOH, 20 mM NaCl, 10% glycerol, pH 7.5). Tag-less proteins were further purified over a Ni-NTA column, supplemented with 100 mM NaCl, 0.2 mM CoCl₂, 1 mM TCEP, and 50% glycerol, and stored in –30 °C.

Fluorescence labeling

MAP-H79A containing a C-terminal GLPATGG tag was labeled with TMR (Invitrogen) via sortase-mediated reactions as described [23,76].

PDF-E133A was labeled with TMR at the N-terminus using Sfp-mediated reaction [34]. The labeling reaction was carried out in Sfp buffer (50 mM Hepes-KOH, 10 mM MgCl₂, pH 7.5) containing 50 µM PDF-E133A, 10 µM His₆-tagged Sfp and 50 µM CoA-TMR conjugates, and incubated at room temperature for 1 h. The reaction mixture was passed through TALON resin to remove Sfp. Excess dye conjugates were removed using a PD-10 column (GE). Labeling efficiency was ~ 80%.

RNC purification

RNCs for enzymatic reactions.

Radiolabeled RNCs were generated and purified as described previously [23]. Briefly, the DNA templates were in vitro translated in E. coli S30 extracts containing ³⁵S-methionine and 5 µM actinonin. The reaction mix was purified through 10–50% sucrose gradient fractionation centrifugation in a SW32 rotor (Beckmann Coulter, 23,000 rpm, 15 h, 4 °C) and the 70S fractions were collected.

RNCs for Nascent chain-MAP docking.

Coumarin-labeled RNCs were generated via the amber suppression technique by in vitro translating the DNA templates in S30 extracts of an E. coli strain KC6 expressing tRNA_CUA in the presence of 220 µM L-(7-hydroxycoumarin-4-yl)ethylglycine (Cm), 12 µM purified Cm tRNA synthetase and 5 µM actinonin, as described [77]. 70S ribosomes were isolated via sucrose gradient fractionation as described above.

RNCs for the fluorescence binding assay.

The coding sequence for human ubiquitin, UBA52, was subcloned from pHUE [36] and inserted between a 3X Strep tag and the nascent chain sequence on pK7-derived plasmids via Gibson cloning. An 8-residue Ms-sup1 stalling sequence was placed at the C-terminus of the nascent chain sequence [78]. DNA templates were in vitro translated in S30 extracts from the E. coli strain containing a ybbR tag at the C-terminus of bL17 (KC6 ΔrplQ::kan pL17_ybbR), and the reactions were loaded on a Strep-tactin column. Purified RNCs were labeled with BODIPY-FL (Invitrogen) in reactions containing 1 µM RNC, 4 µM Sfp and 4 µM BODIPY-FL-CoA at room temperature for 2 h. The reactions also contained 1 µM deubiquitinylating enzyme (Usp2-cc) [36] to remove the 3X Strep-ubiquitin moiety and expose the methionine on the nascent chain. Labeled RNCs were further purified through a sucrose cushion centrifugation.

RNCs for the crosslinking assay.

The plasmids for RNCs used in the crosslinking assay were constructed by placing the nascent chain sequences between an N-terminal 3X Strep tag and a C-terminal SecM stall sequence in pK7-derived plasmids. Cysteine mutations were placed at the end of the signal sequences for the PhoA variants (T16C), in the middle of the TMD for FtsQ N8 (T14C), and in the middle of the signal sequence of DsbA (G15C) via site-directed mutagenesis. The DNA templates were in vitro translated in S30 extracts from the E. coli strains carrying a cysteine mutation at uL22 (KC6 ΔrplV::kan pL22_S30C) or uL23 (KC6 ΔrplW::kan pL23_S21C), and the RNCs were purified via Strep-tactin resin.

To generate E. coli strains KC6 ΔrplQ::kan pL17_ybbR, KC6 ΔrplV::kan pL22_S30C and KC6 ΔrplW::kan pL23_S21C, the strain KC6 [79] was transformed with the plasmid pEK20 encoding C-terminal ybbR-tagged bL17, uL22-S30C, or uL23-S21C. The respective genomic ribosomal protein genes were subsequently knocked out via lambda-red recombination [16,80].

Enzymatic assays on purified RNC.

Deformylation assay.

The measurements were carried out manually or using an RQF-3 Quench-flow instrument (KinTek) as described previously [23]. The reaction was initiated by mixing 20 nM ³⁵S-labeled RNC with an equal volume of PDF at sub-saturating concentrations in assay buffer (50 mM Hepes-KOH, 150 mM KOAc, 10 mM Mg(OAc)₂, 0.1 mM CoCl₂ and 1 mM TCEP, pH 7.5), and quenched with an excess amount of ACT at specific time points. Excess MAP was added subsequently to release deformylated methionine, which remains soluble after treatment with 15% trichloroacetic acid (TCA). The mixture was centrifuged, and the TCA-soluble ³⁵S-Met signal in the supernatant was quantified by scintillation counting.

Methionine cleavage assay.

Measurements were carried out as described previously [23]. Briefly, 20 nM of ³⁵S-labeled RNC was first deformylated by incubation with 100 nM PDF. The reaction was initiated by adding equal volume of MAP at sub-saturating concentrations and quenched with 15% TCA at specific time points. The released ³⁵S-Met was quantified as described above and previously [23].

The amount of released methionine in the deformylation and methionine cleavage assays was normalized to the total amount of nascent chains and fit to Eq. 1,

$$\frac{\left[Met\right]}{{\left[RNC\right]}_0}=C\left(1-e^{-k_{obs}t}\right)$$ Eq. 1

where k_obs is the observed rate constant, and C is the maximum fraction of RNC that can be processed. The k_obs was then divided by the enzyme concentration used during the reaction to obtain an apparent k_cat/K_m value.

Fluorescence measurements

All fluorescence measurements were performed on Fluorolog-3 (HORIBA).

Equilibrium titrations to measure ribosome-PDF and ribosome-MAP binding.

The measurement was carried out with 10 nM BODIPY-labeled RNC in assay buffer with 1 mg/mL BSA to prevent nonspecific binding. Increasing amount of TMR-labeled PDF-E133A or MAP-H79A was added into the reaction, and the fluorescence emission at 515 nm was recorded at the excitation wavelength of 485 nm. The FRET efficiency was calculated according to Eq. 2,

$$FRET=1-\frac{F_{DA}}{F_D}$$ Eq. 2

where F_DA and F_D are the fluorescence signals with and without the acceptor, respectively, after corrections for dilution and environmental sensitivity using parallel titrations with unlabeled PDF-E133A or MAP-H79A.

The FRET efficiency was fit to the Michaelis-Menten equation (Eq. 3),

$$FRET=A\times \frac{\left[E\right]}{K_d+\left[E\right]}$$ Eq. 3

Where A denotes the FRET efficiency at saturated condition, [E] is the concentration of PDF-E133A or MAP-H79A, and K_d is the dissociation constant between the RNC and PDF or MAP.

Nascent chain-MAP docking.

The experiment was carried out using Cm-labeled RNC_FtsQ and RNC_{FtsQ N8} as described previously [23]. Briefly, 10 nM of RNC and indicated amount of MAP-H79A were mixed, and the emission spectrum was recorded at the excitation wavelength of 360 nm.

Crosslinking assays

500 nM RNCs carrying Cys mutations were reduced in crosslinking buffer (50 mM Hepes-KOH, 150 mM KOAc, 10 mM Mg(OAc)₂, pH 7.0) with 5 mM TCEP at room temperature for 30 min, followed by incubation with 0.8 mM 1,4-bis(maleimido)butane (BMB) or 1,4-bis(maleimido)hexane (BMH) for 1 h. The reaction was quenched with 50 mM DTT, and analyzed by Western blot using indicated antibodies. Anti-Strep and anti-HA were purchased from Genscript and anti-uL23 was customized from Genscript as previously described [16].

Two-state model

Considering a two-state model (depicted in Fig. 3A), in which the active and inactive states are rapidly exchanging before proceeding to the irreversible chemistry step,

$$inactive\stackrel{K_{dock}}{\rightleftharpoons }active\stackrel{k_c}{\to }product$$ Eq. 4

The free energy change (∆G) between the inactive and active states is driven by the hydrophobic interaction between the nascent chain N-terminus and uL23, and thus related to the hydropathy (Φ) of the N-terminus,

$$\Delta G=\Delta G_0+a\cdot \Phi$$ Eq. 5

where ∆G₀ describes the intrinsic free energy term independent of N-terminus hydropathy, and a is a constant of proportionality. The observed rate constant (k_obs) is proportional to the fraction of the active state and can be described by Eq. 6,

$$k_{obs}=k_{max}\times \frac{1}{1+e^{-\frac{\Delta G}{kT}}}=\frac{k_{max}}{1+e^{-\frac{\Delta G_0}{kT}-\frac{a\cdot \Phi }{kT}}}$$ Eq. 6

where k_max denotes the maximal rate constant when the active fraction approaches unity, k is the Boltzmann constant, and T is the temperature in the kelvin scale.

To define Φ, we used the Kyte and Doolittle scale [41] to calculate the grand average of hydropathy (GRAVY) and searched with an 11-residue rolling window from the N-terminus of the nascent chain, given that the hydrophobic cleft on uL23 can accommodate a helical peptide of 11–13 residues [16]. Based on the distance between the cleft and the enzyme active sites (Figs. S4C and D), the GRAVY score of the most hydrophobic segment that begins at or before the 11^th residue from iMet was plotted against the observed rate constant of the PDF or MAP reaction, and fit to Eq. 6 (Figs 3E and F).

Simulation for cotranslational NME

Numerical simulations of cotranslational NME under single turnover conditions were carried out as described previously [23]. This model considers sequential PDF and MAP reactions during protein translation in S30 lysates, in which the concentration of ribosome is in excess of actively translated nascent proteins. The irreversible pseudo-first order rate constants for PDF and MAP reactions are defined as (k_cat/K_m)_app × [enzyme]_free, where [PDF]_free = 130 nM and [MAP]_free = 63 nM. (k_cat/K_m)_app values for RNCs below 45 aa are set to zero, and those longer than 69 aa are set to the measured rate constants in Figs. 3 and 4. Values of (k_cat/K_m)_app between 45 and 69 aa are linearly interpolated from experimental data. The translation elongation rate is set to 2 aa/s.

In vivo NME assay

DNA fragments encoding full-length FtsQ and PhoA were amplified from E. coli genomic DNA, appended with a C-terminal 3X FLAG tag, and integrated into pBAD/His C (Invitrogen) under control of the arabinose promoter [81] via Gibson cloning. A second open reading frame (ORF) encoding thioredoxin with a C-terminal 3X FLAG tag was cloned downstream. Both the FtsQ/PhoA and thioredoxin ORFs are preceded by a Shine-Dalgarno sequence in the 5’UTR. The iMet constructs for the FtsQ variants contain the mutations M137L, M159L, M177L, M180L, M193L, M217V. The iMet constructs for PhoA variants contain the mutations M26L, M75L, M157A, M286L, M324L, M420L, M422L, M464L, C190S, C200S, C308S, C358S.

Plasmids were transformed into the E. coli strain CAG12184 (λ-rph-1 tolC210::Tn10 (tet)) or KPS73 (λ-rph-1 tolC210::Tn10 (tet) ∆fmt ::Km) [67,82]. Cells were grown in Luria-Bertani medium containing 0.2% glucose and 100 µg/mL ampicillin at 37 °C until OD = 0.5–0.7, washed and resuspended to OD = 0.05 in pre-induction medium (M9 medium containing 0.5% glycerol, 0.2% glucose, 100 µg/mL ampicillin and 40 μg/mL of 20 amino acids except methionine), and grown to OD = 0.5. Cells were collected, washed 3 times with prewarmed induction medium (M9 medium containing 0.5% glycerol and 40 μg/mL of 20 amino acids except methionine and cysteine), and resuspended with induction medium corresponding to 1/20 volume of the cell culture. Cells were induced with 0.02% arabinose at 37 °C for 30 min, labeled with 200 µCi/mL ³⁵S-Met/Cys EasyTag mix (PerkinElmer) for 10 min, and flash-frozen in liquid nitrogen. The frozen cells were thawed, spun down, and lysed in 25 µL lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 2 mM DTT, 5 mM EDTA, 4% SDS, pH 8.0) containing 1X protease inhibitor at 95 °C for 10 min. The lysates were diluted with 500 µL IP buffer (50 mM Tris-HCl, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, pH 8.0), incubated at room temperature for 10 min, and centrifuged at 12,000 × g for 5 min. Clarified lysates were further diluted with 500 µL TBS (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) and incubated with 10 µL Anti-FLAG M2 magnetic beads (Sigma-Aldrich) at 4 °C for 1.5 h with rotation. The beads were washed twice with IP buffer and once with TBS, and the FLAG-tagged proteins were eluted by incubating the beads with 25 µL 2X SDS sample buffer at 95 °C for 10 min. The radioactive signals were subsequently analyzed by SDS/PAGE and autoradiography. To calculate iMet retention, the substrate/TrxA signal ratio from the iMet construct was divided by that of the corresponding 6M or 8M4C construct, and normalized to the value of FtsQ K2 or 2A8L in each experiment.

Cell fractionation

CAG12184 cells expressing the FtsQ variants were harvested and fractionated as described previously [29]. In brief, cells were subject to osmotic shock and lysozyme treatment to release the periplasm, and the spheroplasts were further lysed by freeze-thaw cycles in liquid nitrogen. The lysates were centrifuged in a TLA120.1 rotor (Beckmann Coulter) at 63,000 rpm for 1 h to separate the membrane and cytoplasm. The cellular localization of the FtsQ variants was determined by Western blot of the different fractions using anti-FLAG antibody (Genscript).

Proteinase K protection assay

Spheroplasts prepared as described above were resuspended in Buffer PK (100 mM Tris-HCl, 20% sucrose, 20 mM MgSO₄, pH 8.0). 0.5 mg/mL of proteinase K was added into the reaction and incubated on ice for 1h with or without 1% of Triton X-100 as control. The reaction was stopped by adding 5 mM PMSF, and analyzed by Western blot using anti-FLAG antibody.

Highlights.

• N-terminal methionine excision (NME) is inhibited on proteins with an early membrane-targeting signal.

• Ribosome-nascent chain interaction dictates substrate access to the NME enzymes.

• SRP exacerbates the NME inhibition on membrane and secretory proteins.

• Molecular interplay at the ribosome exit allosterically regulates nascent protein modification.