Translation and folding of single proteins in real time

Significance

How proteins fold natively with efficient fidelity while being synthesized remains largely unexplored. Understanding protein synthesis on a single-molecule level is of particular interest to the life sciences and relevant for various diseases. Although protein synthesis and folding are well-studied subjects, cotranslational folding has been proven difficult to observe. Using optical tweezers, we measured the mechanics of synthesis and simultaneous folding in real time. We found that cotranslational folding occurs at predictable locations, exerting forces on the nascent polypeptide. Furthermore, we show that transient pauses and gradual slowing of translation occur in particular locations along the protein sequence, facilitating native secondary-structure formation. Thus, the rate of synthesis is inherently coupled to cotranslational folding, assuring reliable and fast native folding.

Abstract

Protein biosynthesis is inherently coupled to cotranslational protein folding. Folding of the nascent chain already occurs during synthesis and is mediated by spatial constraints imposed by the ribosomal exit tunnel as well as self-interactions. The polypeptide’s vectorial emergence from the ribosomal tunnel establishes the possible folding pathways leading to its native tertiary structure. How cotranslational protein folding and the rate of synthesis are linked to a protein’s amino acid sequence is still not well defined. Here, we follow synthesis by individual ribosomes using dual-trap optical tweezers and observe simultaneous folding of the nascent polypeptide chain in real time. We show that observed stalling during translation correlates with slowed peptide bond formation at successive proline sequence positions and electrostatic interactions between positively charged amino acids and the ribosomal tunnel. We also determine possible cotranslational folding sites initiated by hydrophobic collapse for an unstructured and two globular proteins while directly measuring initial cotranslational folding forces. Our study elucidates the intricate relationship among a protein’s amino acid sequence, its cotranslational nascent-chain elongation rate, and folding.

The overall rate of translation during synthesis is limited by decoding, peptidyl transfer, and translocation rates (1), and depends on several factors, such as the availability of tRNAs, amino acids, and translation factors, as well as the mRNA unwinding rate (2). In addition, interactions of the polypeptide with the ribosomal tunnel affect the speed of translation. If the rate of synthesis is greater than that of the exit tunnel emergence, bunching of the polypeptide occurs within the tunnel and the overall translation rate is reduced (3, 4). Folding of the nascent chain into its native structure is driven by free-energy minimization and occurs cotranslationally in a vectorial fashion, predominantly outside of the ribosomal tunnel (5, 6). Cotranslational folding within, as well as outside, the ribosomal exit tunnel exerts pulling forces on the nascent polypeptide that can prevent and even rescue translational stalling (7–9). During the early stages of peptide elongation, cotranslational folding compacts the nascent chain while it is still confined within the ribosome’s exit tunnel (10). A polypeptide’s amino acid sequence and interactions with itself and the solvent (solutes) determine its native folded structure. During synthesis, the spatial constrains within the ribosome’s peptide tunnel define the number of possible cotranslational folding pathways (10). However, the connection of cotranslational folding and rate of synthesis to a protein’s amino acid sequence is still not completely understood. The process of cotranslational protein synthesis and folding in the crowded environment of the cell is difficult to study due to the stochastic nature of mRNA translation and is usually investigated in vitro. These ensemble studies confirm that larger proteins collapse fast into a polyglobular conformation, fold through native-like intermediates in a distinct pathway, and emerge slowly as native structures (11, 12). With a few exceptions, stable tertiary structure formation requires the presence of hydrophobic amino acids (13, 14).

Single-molecule techniques have proven useful in the study of protein synthesis and subsequent folding (15–18), because these are asynchronous processes that are difficult to be observed using ensemble methods. Optical tweezers have been used to observe stepping of motor proteins (19–23), DNA–protein complexes (24), as well as unfolding and refolding of RNA molecules and proteins (25, 26). This powerful single-molecule method has provided information on (i) the translation machinery by reporting on the strength of interactions between the ribosome and mRNA (27), (ii) its translocation along a short hairpin-forming mRNA molecule (28), as well as (iii) the release of an arrested nascent chain (7). The ribosome has been shown to modulate the folding rate of nascent chains (5, 26).

Dual-trap optical tweezers provide the necessary stability and sensitivity to isolate, confine, and measure the activity of individual macromolecules in their native environment, largely decoupling the system under study from environmental noise (29). Here, we observed synthesis and folding of single nascent polypeptides in the form of interbead distance variations using dual-trap optical tweezers in real time, while the individual nascent chains were held at various constant forces. Thus, the tension exerted on the nascent chain during the initial steps of cotranslational folding of synthesizing hydrophobic sequence stretches could be measured directly. It could also be shown that successive prolines caused transient translation pauses, while positively charged residues slowed the rate of synthesis under tension. Using this approach, we could observe the minimal hydrophobicity that was required by a nascent-chain section to overcome a certain applied tension, initiating folding. It follows that this technique could be used to pinpoint the sequence location of a protein where cotranslational folding begins for a range of applied forces.

Results

Experimental Setup.

We used three DNA constructs for translation experiments, where the encoded protein under study was preceded by a N-terminal presequence and succeeded by a C-terminal linker. These constructs were synthesized by ribosomes biotinylated in vivo (30) at the uL4 ribosomal protein. The presequence consisted of an amber stop codon, followed by a 35-aa linker spanning the length of the ribosomal tunnel and six histidines (6×His) (Fig. 1A). Adding a histidine-depleted cell-free transcription/translation system triggered synthesis of nascent chains up to the 6×His tag, as confirmed in ensemble control experiments using the GFP variant Emerald (GFPem) construct (SI Appendix, Fig. S1). Concomitantly, biotin was cotranslationally incorporated at the N-terminal amber stop codon using the suppressor tRNA technique (31, 32) (Fig. 1A, reaction 1). Thus, the ribosome–nascent-chain complex (RNC) featured two biotin tags, one at the end of the stalled nascent chain, just appearing outside of the ribosomal tunnel and another biotin molecule linked to the uL4 ribosomal protein, as shown previously (SI Appendix, Fig. S2A) (33). The RNC was tethered in situ between two optically trapped polystyrene beads (0.84 μm in diameter) via two identical streptavidin-DNA handles (34) (0.33 μm in length) in a microfluidic chamber (SI Appendix, Fig. S3). Both beads were held in orthogonally polarized optical traps of equal stiffness (35) (0.3 ± 0.03 pN/nm for all measurements) (Fig. 1B). Translation could be resumed by adding a cell-free translation reaction mix containing His. The nascent polypeptide remained bound to the ribosome after synthesis due to the C-terminal SecM arrest peptide (AP) following the C-terminal linker (Fig. 1A, reaction 2).

Fig. 1.

Experimental design and translation traces of proteins under different forces. (A) The three constructs used in this study were identically designed. Two structured proteins DHFR (187 aa, red) and GFPem (239 aa, green) and an intrinsically disordered protein hTau40 (441 aa, blue) were chosen. The SecM moiety ensured that the fully synthesized protein remained bound following translation. Reaction 1 incorporated biotin at the N-terminal amber stop codon halting synthesis at the His tag. (B) A constant-force F was applied on ribosome nascent-chain constructs, and reaction 2 was injected orthogonally to the measurement axis (SI Appendix, Fig. S4). The extension of the nascent polypeptide was measured during and after its synthesis. Not drawn to scale. (C) Comparison of typical translation traces for hTau40 (blue, 7 pN), DHFR (red, 10 pN), and GFPem (green, 10 pN), as well as a –His control (gray) (2 Hz). hTau40 showed the longest elongation free from stalling or folding (blue arrow) (n = 33 hTau40 traces). (D) Comparison of translation trajectories at different forces of 7, 10, and 20 pN for the DHFR construct showing typical extensions between 20 and 40 nm. Red arrows mark a sudden change in extension rate during synthesis at 10 and 20 pN. Red stars mark partial unfolding (20 pN) and folding (10 pN) events. Upon partial folding under 10-pN tension, the DHFR construct acquires the same extension reached after synthesis at 7 pN (gray arrow) (2 Hz) (n = 35 DHFR traces). (E) Reproducibility of translation trajectories of DHFR (7 pN). (F) Translation displacement traces (2 Hz) with the GFPem construct held at a constant force of 10 and 20 pN (n = 20 GFPem traces) resulting in typical extensions of 15–30 nm.

We have previously shown that individual GFPem constructs folded natively while remaining bound to surface-tethered ribosomes (33). In addition, control measurements demonstrated that translation could be stalled efficiently at the His-tag, resuming as soon as the missing histidine was introduced with the transcription/translation mix (SI Appendix, Fig. S1). Further control experiments were performed to determine the influence of fluid flow on our measurements and to test His-tag stalling on the single-molecule level. Streptavidin-DNA handles were tested with and without stalled RNCs at a range of different forces (2–48 pN) (SI Appendix, Fig. S4). As demonstrated in these control experiments, injections had no impact on translation measurements due to the orthogonal fluid flow configuration. The RNCs introduced extra positional fluctuations, observable with the small bead size used here. These signature fluctuations were not seen in the DNA-only case. Thus, we could confirm whether or not a double tether with the correct length (∼680 nm at 10 pN) included an RNC by performing a short constant force measurement before each translation measurement, comparing the measured positional noise to the two fits in SI Appendix, Fig. S4B.

Full Protein Synthesis by Single Ribosomes Under Constant Force.

We followed the synthesis of an intrinsically disordered protein, hTau40 (36), and two globular proteins (SI Appendix, Fig. S2 B and C), dihydrofolate reductase (DHFR) and the GFPem, under constant applied forces in the range of 7–20 pN (Fig. 1 C–F).

Measured displacement changes between the optically trapped beads during and after protein synthesis could be converted from elongation in nanometers to the number of translated amino acids for unstructured linear sections of a nascent chain by using the extensible worm-like chain model (eWLC) (37). The eWLC-derived nanometer-to-residue conversion is dependent on the applied forces (Table 1) and is valid for unstructured parts of the polypeptide chain.

Table 1.

Fractional extension and length-to-residue conversion for extended polypeptides and dsDNA

Force, pN	x/L polypeptide	nm/aa polypeptide	x/L dsDNA	nm/bp dsDNA
7	0.51	0.2	0.95	0.323
10	0.61	0.24	0.96	0.326
15	0.72	0.29	0.97	0.329
20	0.80	0.32	0.99	0.336

Here, the eWLC-derived fractional extension with the corresponding nanometer-to-residue conversion are shown for a number of different forces. These conversion factors are valid for unstructured parts of a polypeptide chain with a residue length of 0.4 nm/aa, a persistence length of 0.66 nm, and a stretch modulus of 200 pN (SI Appendix). The dsDNA mechanics are indicated with a base pair length of 0.34 nm/bp, a persistence length of 50 nm, and a stretch modulus of 1,000 pN (SI Appendix).

During translation, hTau40’s overall N- to C-terminal length gain was greater than that of DHFR and GFPem (all traces, Fig. 1C). Following injection of the reaction mix containing His, the His-tag stalled hTau40 construct exhibited an extension increase of ∼98 nm in the measured interbead displacement at forces as low as 7 pN. The observed displacement changes corresponded to the expected length of the fully translated and unstructured hTau40 construct (491 residues 6×His-SecM), with an eWLC-derived nanometer-to-residue conversion factor of 0.2 nm/aa at 7-pN tension (Table 1).

On the contrary, the measured interbead displacement changes for both DHFR and GFPem were generally smaller than the expected length of the fully unfolded constructs. The DHFR construct showed displacement increases between 19 ± 5 nm at 7 pN (95% of traces) and 30–40 nm (82%) up to 76 ± 3 nm (18%) at 20-pN tension (Fig. 1D). DHFR’s cotranslational N- to C-terminal extension was larger than the 15- to 30-nm displacement increase of the GFPem construct at forces between 10 and 20 pN (all traces, Fig. 1F), although DHFR’s sequence is shorter than that of GFPem (SI Appendix, Fig. S5).

We conducted cotranslational measurements for the DHFR construct under different constant tensions between 7 and 20 pN and compared the observed bead displacement traces. Interestingly, one particular transition resulted in the same displacement change at different forces (SI Appendix, Fig. S6). At 7-pN tension, the nascent chain extended at a markedly lower rate of 0.3 ± 0.1 nm/s than at higher forces, tapering off at ∼20 nm from N- to C-terminal after ∼80 s (Fig. 1E). For comparison, a fully folded DHFR protein is expected to have a N- to C-terminal extension of ∼15 nm following translation, including the unstructured N-terminal and C-terminal linkers used here. By increasing the tension to 10–20 pN, the nascent chain extended at rates up to four times greater than at 7 pN and a sudden change in the measured extension rate took place at a later point, ∼45 s after injection near sequence position 74 (red arrows in Fig. 1D and SI Appendix, Fig. S5). This was an indication that cotranslational folding was taking place continuously during synthesis at 7 pN, because increasing the tension to 10–20 pN appeared to counteract the folding forces. This allowed us to observe translation decoupled from folding for more than just a few seconds at higher applied tensions. The sudden drop in extension rate between 20 and 25 nm during synthesis at 10- and 20-pN tension was followed by compaction at 10 pN and sudden partial unfolding at 20 pN (red stars in Fig. 1D). After compaction at 10 pN, DHFR reached a similar extension obtained after synthesis at 7 pN (gray arrow).

Posttranslational Unfolding and Refolding of Proteins.

To investigate whether the DHFR construct was fully translated and partially folded during synthesis under tension, we conducted a number of posttranslational unfolding experiments immediately following synthesis under an applied force of 20 pN (N = 37). By continuously varying the force between 0 and 40 pN with a loading rate of 6 pN/s, we observed unfolding from a partially folded state to DHFR’s expected fully unfolded length (Fig. 2 A and D), clearly indicating that DHFR partially folded while being fully translated. Relaxing the construct for ∼1 min at 0 pN and subsequently increasing the force again revealed that the constructs refolded. The unfolding DHFR allowed resolving three intermediate unfolding steps, resulting in mean unfolding forces of ∼26 pN (n = 37; Fig. 2E). The intermediate states I₁, I₂, and I₃ correspond to N- to C-terminal extensions of 66, 128, and 199 ± 12 aa, respectively. These displacement changes in the extension did not occur during our control experiments, where only DNA handles linked by a single streptavidin protein were present (SI Appendix, Fig. S7).

Fig. 2.

Unfolding/refolding of newly synthesized proteins. (A) Partial unfolding of DHFR (directly after translation under 10-pN tension to an extension of 136 ± 12 aa; green eWLC curve). Red data/arrows represent decreasing, and blue data/arrows denote increasing the tension on the partially folded intermediate. (0.5 pN/s, 1 kHz) As illustrated, the construct unfolded at 15 pN by 142 ± 12 aa to 272 ± 12 aa (black arrow, purple eWLC curve), corresponding to the fully unfolded length of DHFR (including Gly/Ser linker and SecM). Continued 0.5 pN/s force variation between 5 and 18 pN did not result in immediate refolding. (B) The tension on this unfolded construct was subsequently lowered to 0 pN. After ∼1 min at 0 pN, increasing the force (blue data/arrow, 10 pN/s) revealed partial refolding to an N- to C-terminal extension of 206 ± 12 aa (gray eWLC curve I₃). (C) hTau40 construct measurement initiated directly after translation at 10-pN tension to a total length of 527 ± 15 aa (eWLC curve, including linkers and SecM), several minutes after reaction 2 (Fig. 1A) injection. The force was continuously varied between 5 and 18 pN (0.5 pN/s, 1 kHz) as illustrated. No folding or unfolding was observed for hTau40. (D) DHFR construct measurements starting directly after translation under tension of 20 pN. The force was continuously varied between 0 and 40 pN. The newly synthesized DHFR constructs were allowed to refold at 0 pN and subsequently extended at an extension rate of 6 pN/s, showing native-like unfolding at forces between 20 and 30 pN, featuring three intermediate states I₁–I₃. These intermediate states correspond to N- to C-terminal extensions of 66, 128, and 199 aa, respectively. The fully unfolded state U has an extension of 272 aa. (E) Histogram of measured unfolding forces of newly synthesized, SecM-stalled, and refolded DHFR constructs, showing mean unfolding forces of 26 pN (n = 37; bin size, 5 pN). Extensions shown here correspond to measured end-to-end distance of the newly synthesized proteins (including DNA handles).

The fully unfolded DHFR construct, tethered by two 1-kbp dsDNA handles followed predictable eWLC behavior. This was demonstrated by overlaying the experimental force-extension data with a dsDNA eWLC function in series with a polypeptide eWLC function (Fig. 2, dashed purple curve and SI Appendix). Similar unfolding experiments were carried out after synthesis of the hTau40 construct (n = 17). Here, we did not observe further unfolding, nor any refolding of the construct upon constant-rate force variations between 5 and 18 pN, because the protein was most likely already unfolded (Fig. 2C). A similar overlay of a combined eWLC function (Fig. 2C, dashed black curve, and SI Appendix) with an unstructured polypeptide contour length of 527 residues (hTau40 construct with linkers and SecM) demonstrated that it also matched the force-extension profile of the dsDNA-tethered, SecM-stalled, and unstructured hTau40 construct.

Translation Rate and Stalling.

The longest observed translation pauses for unfolded segments of nascent polypeptides, lasting on the order of several seconds, coincided with sequence stretches featuring successive Pro residues downstream of several positively charged amino acids (Figs. 3 and 4 and SI Appendix, Figs. S8 and S9). In these instances, during incorporation of successive Pro, a number of positively charged Arg/Lys residues were confined inside the ribosomal tunnel, as illustrated in Fig. 3 C and D. DHFR’s fastest stretches of synthesis of 9–15 aa/s at tensions of 7- to 20-pN tension were observed within the first 10 s of translation (Figs. 3A and 4 D–F). The fastest observed translation rates of up to 16–22 aa/s were measured during synthesis of hTau40 constructs, at forces ranging from 7 to 20 pN (Figs. 3B and 4 A–C). In both constructs, they occurred along segments with low numbers of positively charged amino acids and Pro residues.

Fig. 3.

Translation trajectories of individual proteins. (A) DHFR translation trajectory at 20 pN (light gray, 1 kHz; red/gray overlay, 2 Hz; 0.32 nm/aa) showing stalling in regions with successive Pro–Pro (green lines) and slowing in regions with positively charged Arg, Lys (red lines). The red shaded zones depict the number of Arg, Lys confined within the ribosomal tunnel for an extended nascent chain (red curve, SI Appendix, Fig. S9). His (blue lines) did not significantly contribute to slowing here. Gray overlay depicts cotranslational compaction following translation of strongly hydrophobic segment indicated by the yellow horizontal bars (1 of n = 35 DHFR traces). (B) Translation trace of hTau40 (7 pN; light gray, 1 kHz; blue/gray overlay, 2 Hz; 0.2 nm/aa) demonstrating stalling at Pro–Pro and subsequent slowing after multiple positively charged amino acids were incorporated (1 of n = 33 hTau40 traces). Color coding as in A. The Inset magnifies the trajectory segment where Pro–Pro stalling occurred during translation. (C) Unfolded sequence stretches rich in positively charged residues within the ribosomal tunnel featured lower translation rates, resulting in slower observed increasing bead displacements (arrow) than segments with low Arg/Lys (His) ribosomal tunnel occupation. (D) Prolonged stalling occurred during Pro–Pro incorporation concurring with a high positively charged residue ribosomal tunnel occupancy in DHFR and hTau40. Hydrophobic collapse followed by cotranslational compaction could cease or momentarily reverse observable increases in bead-to-bead displacements (arrow) despite polypeptide growth.

Fig. 4.

The effect of force on translation rates. (A–I) The translation rate traces were determined by averaging the 1-kHz acquired bead displacement data to 0.5 Hz, and then taking the first derivative smoothed with a 2-s adjacent averaging window. The maximum translation rate (full horizontal line), average rate for translation before initial compaction (dashed line), and average rate over 40 s with compaction (dotted line) are shown. Pro–Pro stalling during translation free from compaction is highlighted with a green background. Pie charts depict the fraction of nascent chains remaining unfolded (colored) and folded (gray) during synthesis under tension. Increased tension on the nascent chain during synthesis reduces the average rate of translation before initial compaction for hTau40 and DHFR (n = 61). Because GFPem already folds at the N-terminal under the tensions used here, only its maximum rate and overall average rate convoluted with folding/unfolding could be determined.

The nature of the amino acids preceding and following a successive proline sequence position also appeared to play a role. As shown in the Inset of Fig. 3B and SI Appendix, Fig. S8B, the first Pro–Pro incorporation into hTau40’s sequence at ¹⁵⁸APPG¹⁶¹ took longer than that of the following Pro–Pro incorporations in the linearly extended (unfolded) regime (white background in Fig. 4): ¹⁷⁵TPPA¹⁷⁸, ¹⁸¹TPPS¹⁸⁴, ¹⁸⁷EPPK¹⁹⁰, and ²¹⁷TPPT²²⁰ (SI Appendix, Fig. S5).

Cotranslational Nascent-Chain Compaction.

As seen in Fig. 3A and SI Appendix, Figs. S8A and S9, the first instance of cotranslational compaction for the DHFR construct, being synthesized at a tension of 10–20 pN, resulted in a sudden change in the rate of bead displacement at around sequence position 74, near a cluster of hydrophobic residues composed of Leu, Ile, and Val. After partial cotranslational folding had occurred, we did not observe further fast increases of bead–bead displacements even though translation likely continued under the applied tension. The observed equilibrium position after synthesis of position 74 was probably the result of synchronous compaction during elongation due to folding (Fig. 3D), as demonstrated by posttranslational unfolding to the full length of the construct (Fig. 2 A and D).

In the N-terminal half of hTau40’s sequence, amino acids with lower hydrophobicity, such as Ala and Gly, form the bulk of all hydrophobic residues. The overall hydrophobicity of the C-terminal half of the hTau40 sequence starting at Val²²⁶ is greater, because Ala and Gly become rarer. During translation of the hTau40 construct at a tension of 7–10 pN, we observed cotranslational compaction starting from sequence positions 228–307 (84% of traces at 7 pN, 71% at 10 pN) (Figs. 3B and 4 B and C, and SI Appendix, Figs. S8B and S10A).

In previous ensemble folding experiments, it had been shown that hydrophobic collapse is the first step in folding of proteins, occurring before the formation of significant secondary structure (38, 39). A gaugeable definition of hydrophobic collapse probability as a function of sequence position was required to quantitatively describe the initiation of cotranslational compaction that was observed as a sudden change in the measured bead displacement during synthesis of highly hydrophobic segments (yellow horizontal bars, Fig. 3 and SI Appendix, Fig. S8) followed by slowed extension and/or shortening due to cotranslational nascent compaction. These compacted states could be unfolded posttranslationally as shown in Fig. 2.

The presence of a minimum number of closely spaced hydrophobic residues, from six to seven, is sufficient to induce cotranslational long-distance folding initiated by hydrophobic collapse in the absence of externally applied forces (14). Thus, the sequence positions where compaction is likely to take place depend on the number of hydrophobic residues along the sequence and their corresponding hydrophobicity, as well as the relative distances between them. The likelihood of cotranslational folding for a given sequence stretch can be expressed as the sequence-dependent hydrophobic collapse index (HCI) (SI Appendix) (14). The magnitude of the applied force on the growing nascent chain also influences cotranslational nascent chain compaction due to folding. The greater the applied force, the more closely spaced hydrophobic residues would be required to overcome it. A force-dependent threshold value, $HC{I}_s\left(f\right)$, predicting a sequence position, s, where cotranslational hydrophobic collapse is likely to occur, was estimated as shown in Fig. 5A. Segments with a HCI value greater than $HC{I}_s\left(f\right)$ were identified as potential cotranslational folding sites, fitting with the positions where cotranslational compaction was observed.

Fig. 5.

(A) Prediction of hydrophobic collapse for DHFR (red), GFPem (green), and hTau40 (blue), considering only the hydrophobic amino acids Leu, Ile, Val, and Phe. Black horizontal lines highlight four estimated force-dependent values of $HC{I}_s\left(f\right)={10}^{-2}f+0.43$ ($\pm \hspace{0.17em}0.05$) at sequence position$s$, where hydrophobic collapse is likely to occur at 0, 7, 10, and 20 pN (SI Appendix). (B–F) Translation rate model fits (black curves, Eq. 1) for DHFR and hTau40 with (dashed) and without (full) the HCI term $C_i\left(f\right)$, considering positive residue–rRNA electrostatic interactions and successive Pro–Pro stalling. U indicates the unstructured state (colored curve; synthesis decoupled from folding), and PF indicates the partially folded (dark gray) state of the growing polypeptide. Light gray curve depicts unfolding of the SecM-stalled polypeptide. Dark-yellow and light-yellow bars represent unfolded segments with HCI values of 0.55–0.65 and 0.45–0.55, respectively. The dashed green line denotes the predicted length of the fully translated and unstructured construct under tension, as determined with eWLC model (SI Appendix). (B) Fitting Eq. 1 to a 20-pN DHFR trace predicted a translation time of ∼78 s. Compaction occurred cotranslationally after synthesis of a strongly hydrophobic segment (HCI peak 3 in A). (C) A DHFR translation extension curve at 20 pN without cotranslational compaction of the nascent chain. Translation terminates at the SecM AP after 81 s. (D) Overall predicted translation time of ∼63 s at 10 pN, featuring cotranslational compaction at HCI peak 3 (A and B). (E) hTau40 trace at 7 pN, fit resulted in an estimated overall translation time of ∼64 s. Cotranslational nascent chain compaction was observed after synthesis of strongly hydrophobic segment 228–307 (HCI peaks in A). (F) A synthesis time of ∼65 s for hTau40 at a tension of 7 pN resulted from fitting the model to a translation trace showing cotranslational compaction after synthesis of the sequence containing HCI peak at position 228.

Translation Rate Model with Estimation of Force-Dependent Cotranslational Folding.

To describe the optical tweezers-observable translation rate ${\dot{x}}_i\left(t\right),$ as a function of sequence position $i$, we introduced the following simple model:

$${\dot{x}}_i\left(t,f\right)=C_i\left(f\right)k{\dot{x}}_0\left(t\right)\exp \left[l{\rho }_i+m{N}_i^{R,K}+n{N}_i^H\right],$$ [1]

with a scaled maximum translation rate $k{\dot{x}}_0\left(t\right)$ and a nascent chain compaction factor $C_i\left(f\right)$ (SI Appendix). The term ${\rho }_i$ represents Pro–Pro positions along the sequence, whereas the terms $N_i^{R,K}$ and $N_i^H$ represent the approximate number of positively charged residues Arg/Lys and His, respectively, which can be found inside the ribosomal tunnel at position $i$, assuming a tunnel length of 30 extended residues (9, 40). Coefficients $l$, $m$, and $n$ determine the degree of stalling for ${\rho }_i$, $N_i^{R,K}$, and $N_i^H$, respectively.

Fitting this simplified translation model to DHFR translation traces suggested that hydrophobic collapse and continued compaction could occur cotranslationally near sequence positions 9, 51, and 74 (Fig. 5 B–D, yellow horizontal bars), where the respective HCI values peak at 0.52, 0.53, and 0.61, respectively (Fig. 5A). Below 10 pN, cotranslational folding could occur early on near position 9, resulting in GFPem-like translation traces (SI Appendix, Fig. S10B). At 10- and 20-pN tension, the construct extended in an unstructured fashion during synthesis of the first 70–80 residues, collapsing near the HCI peak of 0.61 at sequence position 74 (Fig. 5 B and D). After that point, no further bead displacement changes decoupled from folding were observed, even though translation likely continued. In some cases, DHFR did not exhibit cotranslational collapse at 20-pN tension, extending to its full contour length (18% of traces, Fig. 5C). The cotranslational collapse at applied forces of 10–20 pN compacted the nascent chain at the strongly hydrophobic segment ⁶⁸LKDRINIVLSREL⁸⁰ (Fig. 5 B and D, lowest dark yellow horizontal bar), which coincides with the sequence position of DHFR’s fourth native β-sheet ⁷²INIVL⁷⁶. At 20-pN tension, the construct shown in Fig. 5B partially unfolded posttranslationally after 93 s by 23 ± 9 residues (3.2 Å/residue, SI Appendix). Because folding in GFPem already took place near sequence position 15 (HCI = 0.72, Fig. 5A), even at 20 pN, there were no segments decoupled from folding long enough for adequate fitting of the rate model.

Fitting the same model to hTau40 translation traces indicated that nascent-chain compaction at 7 pN initiated between sequence positions 228–307 near the three HCI peaks of 0.48 (Fig. 5 E and F, light-yellow horizontal bars), where transient β-sheet formation had been previously observed between positions 224 and 315 in NMR bulk studies (36) (SI Appendix, Fig. S5). Following translation, the construct unfolded to its full unstructured length under the applied tension, as demonstrated in Fig. 5E and SI Appendix, Fig. S10A. The lack of a strong hydrophobic core forming during translation coupled with a fast initial translation rate could prevent the occurrence of stable native contacts. This would explain the limited formation of secondary structure and the lack of a well-defined tertiary structure for hTau40 (36).

For the measured bead displacement traces of DHFR and hTau40, we numerically determined the negative stalling coefficients $l$, $m$, and $n$ of the model for the observed translation rate as a function of the sequence position $i$. In all fits, these coefficients were found to be negative, where $\left|l\right|>\left|m\right|>\left|n\right|$ (Table 2). Thus, individual Pro–Pro positions had the greatest immediate retarding effect on the observed translation rate, followed by individual positive residues within the ribosomal tunnel, with His having only a minor contribution. Of course, these coefficients also depend on other situation-specific factors (SI Appendix).

Table 2.

Fitting parameters for the translation rate model

Fitting parameters		DHFR	hTau40
Pro–Pro dependent	l	−2.6 ± 0.5	−2.4 ± 0.5
Arg/Lys dependent	m	−0.21 ± 0.05	−0.17 ± 0.04
His dependent	n	−0.05 ± 0.02	−0.06 ± 0.02

Numerically determined negative coefficients l (stalling due to Pro–Pro peptide bond formation), m [stalling due to the number of Arg/Lys residues confined within the ribosomal tunnel (RT)], and n (stalling due to the number of His residues confined within the RT) for the DHFR and hTau40 translation trajectories (Fig. 4), assuming a RT length of 30 residues. A scaled maximum burst rate of $k{\dot{x}}_0\left(t\right)=5$ nm⋅s⁻¹ was chosen for all fits (SI Appendix).

Discussion

The single-molecule experimental setup depicted in Fig. 1 allowed us to gently usher the nascent polypeptide out of the ribosomal tunnel during translation. Depending on the amino acid sequence being synthesized, the observed bead-to-bead displacement changes were a function of either nascent polypeptide growth, partial folding and unfolding of the newly translated polypeptide chains, or a combination of both synthesis and folding. Increasing the tension on the nascent chain decoupled synthesis from folding for initial segments of hTau40 and DHFR. The applied forces were low enough to preserve subtle dynamics in our measurements that would be hidden from bulk experiments.

We investigated three cases in our study where DHFR was best suited to the constant cotranslational force range of 7–20 pN used here. The other two constructs represented two extremes. GFPem remained compacted and never unfolded, whereas hTau40 remained mostly unfolded at these forces. DHFR, on the other hand, exhibited dynamic behavior, remaining compacted at 7 pN after synthesis of the first HCI peak (SI Appendix, Fig. S10B), while unfolding cotranslationally at 20 pN for instance.

Proteins with a well-defined hydrophobic core, such as DHFR and GFPem, showed a stronger tendency to fold cotranslationally than the intrinsically disordered hTau40 construct, a protein lacking in hydrophobic residues. The natively very compact (β-barrel) GFPem construct requires considerably high forces (100–600 pN) to unfold (41). During our measurements, it assumed a shorter, energetically more favorable conformation under tension than DHFR, despite being a longer construct. Under tension between 7 and 20 pN, the proteins never fully folded into their lowest native free-energy state, which would correspond to a total increase in the interbead displacement of ∼15 nm including the peptide linker (SI Appendix, Fig. S5). Instead, they folded partially, evident from the longer-than-native N- to C-terminal extension that they adopted during translation. Given enough time under tension, DHFR and hTau40 could be fully unfolded to their expected unstructured lengths (Figs. 2 and 5 C and E). Both SecM-stalled unstructured constructs followed eWLC chain behavior, evident from the overlays in Fig. 2. Posttranslational partial refolding was only possible for DHFR, although it took up to a minute without tension. This is in agreement with previous observations (26), showing that the fully translated nascent polypeptide is unable to refold quickly once unfolded, probably due to its close proximity to the negatively charged rRNA backbone of the ribosome. Electrostatic interactions may slow long-range refolding times considerably (26).

Peptide bond formation between two consecutive Pro residues takes place at a much lower rate than between Pro and other amino acids (42). In addition, it had been shown previously that XPP motifs preceded by one of the following residues P, D, and A led to strong stalling. The same was also observed for PPX motifs followed by P, W, D, N, and G (43). Therefore, hTau40’s ¹⁵⁸APPG¹⁶¹ position should feature the strongest stalling, in full agreement with our observations (Fig. 3B and SI Appendix, Fig. S8B). Although the incorporation of individual successive Pro residues had a greater immediate retarding effect on the rate of synthesis than individual positively charged residues, apparent from $\left|l\right|>\left|m\right|>\left|n\right|$ (Table 2), each positively charged residue contributed to slowing throughout its traversal of the ribosomal tunnel. Several successive positively charged residues along a sequence offered a synergistic, additive effect to the slowing of the overall translation rate over a relatively long sequence stretch. Codon use and moderately stable mRNA structures appeared to play a minor role in stalling (SI Appendix, Fig. S11); in accordance with Charneski and Hurst (3). The average rate of synthesis of hTau40 and DHFR before initial compaction decreased with greater applied tension on the nascent chain (Fig. 4). At an applied force between 7 and 10 pN, the average rate of synthesis was 7 ± 1 aa/s for both constructs. When the tension was increased to 20 pN, however, the average rate before initial compaction dropped to 5 ± 1 aa/s. Although not inhibiting peptide bond formation, the increased force exerted on the nascent chain applied here seems to slow synthesis. Similar forces would also occur during folding within the tunnel (8). This feedback mechanism could serve as a folding aid, providing extra time for native contacts to form during synthesis.

DHFR folded cotranslationally at the same extremely hydrophobic segment (Figs. 3A and 5 B and D, and SI Appendix, Figs. S8A and S9) at 10 and 20 pN, indicating that the observed folding was initiated by hydrophobic collapse. Because similar cotranslational folding at a very hydrophobic segment was also observed for GFPem (Figs. 1F and 4 G–I) and hTau40 (Figs. 3B and 5 E and F, and SI Appendix, Fig. S8B), it was evident that initial folding here was also driven by hydrophobic collapse. These observations agree with a multitude of previous ensemble studies citing hydrophobic collapse as the first step in folding (11, 12, 38, 39). Subsequently, the nascent chain can self-interact and seems to prefer to stay compact and close to the ribosomal exit tunnel rather than elongated, resulting in a “stalled” bead-to-bead distance. A pictogram providing a graphical interpretation of these different contributing factors is shown in Fig. 3 C and D.

We determined the HCI values for the GFPem sequence and identified position 15 as a probable initiation site for cotranslational folding (Fig. 5A). Increasing the tension to 20 pN was not enough to counteract the early occurring compaction of the nascent chain. Higher constant forces could not be used to avoid rupturing of the tethers (33). Nevertheless, we estimated that the HCI peak of 0.72 was high enough that cotranslational folding would take place even against a counteracting tension of ∼30 pN. Therefore, the interbead displacement trace of GFPem synthesis was never decoupled from folding.

At 7-pN tension, DHFR sequence positions 9 and 51 were recognized as potential cotranslational folding sites with HCI values of 0.52 and 0.53, respectively. These positions correspond to the first and third native β-sheets of DHFR. By increasing the tension to 10–20 pN, the site of first cotranslational nascent chain compaction was moved to position 74 with a HCI value peak at 0.61 (fourth native β-sheet of DHFR, Fig. 5 B–D and SI Appendix, Fig. S10B). The transient Pro–Pro pause near position 27, observed during translation at 10- to 20-pN tension, could provide extra time for N-terminal secondary-structure formation after compaction of the hydrophobic segment with HCI peak at position 9. Likewise, the Pro–Pro stalling predicted at position 84 would also allow for more time to properly fold the N-terminal half after rapid translation and collapse of the segment containing HCI peaks 2 and 3 at positions 51 and 74, respectively. The positive residues in DHFR’s N-terminal domain gradually slow translation (Fig. 3A), particularly after synthesis of the third HCI peak, enabling equilibrium-like sampling of the ribosomal tunnel-restricted conformational space (10, 44).

Similarly, HCI values determined for hTau40 gave three possible hydrophobic collapse sites. Interestingly, all three sites coincide with the positions of three transient β-sheets at the C-terminal half of hTau40. Following the initial rapid translation of the N-terminal domain, Pro–Pro stalling could facilitate the formation of long-range contacts between the N terminus and middle domain (SI Appendix, Fig. S10A). Our results corroborate that pauses during synthesis coordinate translation kinetics with the cotranslational folding of single domains (45).

Although the HCI values of a certain sequence stretch serve as an indicator for the likelihood of cotranslational folding along the sequence of a protein, it is not a definitive predictor. This is evident from the DHFR and hTau40 trajectories in Fig. 5 that show continued N- to C-terminal extension of the nascent polypeptide chain following synthesis of strongly hydrophobic sequence stretches. Thus, the HCI peaks determined for DHFR and hTau40 are just reaching the threshold hydrophobicity required for cotranslational folding to occur under tension between 7 and 20 pN. Because GFPem never unfolds under tension, even during posttranslational pulling experiments, its HCI peak is well above the threshold limit required for cotranslational folding to occur under tension at the applied range of constant forces.

Initial cotranslational hydrophobic collapse into a compact state likely occurs already inside the ribosomal tunnel, close to the exit vestibule (9, 10). The folding forces during initial compaction observed in our study were on the order of ∼7 pN for hTau40 during collapse of its second-half hydrophobic patch and 10–20 pN for DHFR’s two HCI peaks between sequence positions 50 and 100. GFPem’s folding forces exceeded 20 pN. Although the forces applied to the GFPem nascent chain during translation were limited to 20 pN, the folding forces pulling on the nascent chain within the ribosomal tunnel could have been greater during the initial collapse of the N-terminal sequence stretch with HCI value of 0.72 (Fig. 5A). A minimal HCI value of 0.43 ± 0.05 required for cotranslational folding without an applied tension was extrapolated from the apparent linear relationship between HCI and applied force (Fig. 5A). This would serve as a baseline for determining cotranslational folding sites of other proteins in vivo.

In summary, single-molecule analyses by dual-trap optical tweezers revealed cotranslational features of translation rate and protein folding, namely (i) the correlation of the number of positive amino acids and Pro–Pro locations with the translational rate, resulting in translation trajectories consisting of rapidly translating sequence stretches with intermittent pauses; (ii) evidence that stalling during synthesis provides extra time for the cotranslational formation of long-range contacts in N-terminal domains; (iii) the relationship between the density of strictly hydrophobic amino acids along the sequence and sites of possible cotranslational folding; and (iv) the magnitude of the forces exerted on the nascent chain during initial hydrophobic collapse. The forces exerted on the nascent chain during cotranslational folding also prevent stalling, while slowing the rate of synthesis at tensions above 10 pN. Not only does the genetic code contain the information for native folding, it dictates the relative speed of translation, assuring efficient cotranslational folding with high fidelity. Cotranslational folding itself applies a tension on the nascent chain within the ribosomal tunnel that in turn prevents unscheduled stalling events. This coupling of rate and folding would ensure optimal synchronized translation in a polysome complex.

Methods

His-tag stalled RNCs were coupled to beads in a multistage reaction. First, ribosomes were biotinylated (46, 47) and coupled to streptavidin-DNA (1,000 bp)–covered beads (34). Synthesis of the biotinylated 35-aa linker was then carried out with the desired construct sequence using a customized version of the PURE cell-free transcription/translation system (48) (PURExpress Δ amino acids, histidine, tRNA, ribosomes; NEB E3315Z; New England Biolabs). The experiments were conducted with a custom dual-trap optical tweezers instrument using backfocal interferometry with differential detection (29). Trap stiffness was kept at ∼0.3 pN/nm, and data were acquired at a rate of 62.5 kHz. The translation reaction occurred at ∼26 °C in TICO buffer [20 mM Hepes-KOH, pH 7.6, 6 mM (Ac)₂Mg, 30 mM AcNH₄, 4 mM β-mercaptoethanol]. A detailed description of the methods used is given in SI Appendix.