The ribosome stabilizes partially folded intermediates of a nascent multi-domain protein

In vivo production of site-selective 19F-labelled RNCs

To explore co-translational folding at high sensitivity by 19F NMR spectroscopy, we used the non-canonical amino acid 4-trifluoromethyl-l-phenyl alanine (tfmF), exploiting the three-fold degeneracy of the 19F nucleus within its rotationally mobile CF3 group32. Using an evolved orthogonal amber suppressor transfer RNA (tRNA)/aminoacyl-tRNA synthetase pair33,34, a single tfmF residue was biosynthetically incorporated into the FLN5 sequence by adapting our previously described protocol for in-frame amber suppression (Fig. 1a and Methods)22,23. In addition, an arrest-enhanced variant of the SecM motif was developed (Extended Data Fig. 1) to stall translation at a specified position and thereby produce homogenous samples of 19F-labelled RNCs that remained stable for the duration of NMR data acquisition, as confirmed by western blot analysis and 19F NMR measurements of translational diffusion (Extended Data Fig. 2). The one-dimensional (1D) 19F NMR spectrum of FLN5 RNC showed a single resonance, which, following selective proteolysis to release the FLN5 domain, was retained in the NMR spectrum of the cleaved nascent chain component (Fig. 1b and Extended Data Fig. 1). By contrast, the purified, parent ribosome did not produce a detectable 19F NMR signal (Fig. 1b), confirming the background-free and high selectivity of 19F incorporation by amber suppression.

Fig. 1: Site-specifically 19F-labelled RNCs report on the folding of FLN5 on and off the ribosome.figure 1

a, Schematic of production of 19F-labelled RNCs (Methods). CmR, cloramphenicol resistance gene; araBAD, l-arabinose operon; ampR, ampicillin resistance gene; T7lac, T7 promoter inducible by isopropyl ß-d-1-thiogalactopyranoside (IPTG). b, The 19F NMR spectra of a RNC with a cleavable FLN5 domain, before and after addition of tobacco etch virus (TEV) protease and purification of component parts (Extended Data Fig. 1). c, The 19F NMR spectra of isolated FLN5 and FLN5 + 110 RNC, and isolated FLN5 Y719E and FLN5 + 21 RNC. Observed and fitted spectra are shown in grey and red/blue respectively (298 K, 500 MHz). δF, 9F chemical shift. RNC spectra magnified by a factor of ×2. d, The 2D 1H,15N NMR (selective optimized flip angle short transient (SOFAST) heteronuclear multiple quantum coherence (HMQC)) spectra of 15N-labelled and 15N/19F-labelled isolated FLN5 and FLN5 Y719E (298 K and 283 K, respectively; 800 MHz). δN, 15N chemical shift; δH, 1H chemical shift. e, Crystal structure of FLN5 (Protein Data Bank (PDB) no. 1QFH) coloured by residue-specific 1H,15N amide backbone chemical shift perturbations (CSP) observed following 19F incorporation at position 655 (Extended Data Fig. 3). The N and C termini are shown.

Source data

Detecting folding on the ribosome using 19F NMR

To test the ability of 19F NMR to distinguish different conformations of FLN5, we examined the conservative substitution of a solvent-exposed tyrosine residue to tfmF at position 655 on β-strand A, where the nascent chain in its disordered conformation does not significantly interact with the ribosome and thus remains sufficiently dynamic for NMR observation17. We initially produced isolated FLN5, labelled uniformly with 15N and site-selectively with 19F at position 655, and assessed the impact of fluorination. Minimal changes in thermodynamic stability (difference in Gibb's free energy (∆∆G) ≈ +0.4 kcal mol−1; Extended Data Fig. 3) and 1H,15N-correlated chemical shift perturbations (∆δHN < 0.15 ppm; Fig. 1d,e and Extended Data Fig. 3) were observed. The absence of the Y655 resonance in the fluorinated protein 1H,15N spectrum (Fig. 1d) confirmed the high tfmF incorporation efficiency (>95%).

The 19F NMR spectrum of FLN5 showed a single resonance as expected (Fig. 1c and Extended Data Fig. 3). Similarly, the 19F spectrum of natively folded FLN5 + 110 RNC, in which FLN5 is tethered to the ribosome by 110 linking residues15, contained a single peak with an identical chemical shift (Fig. 1c and Extended Data Fig. 2). A shorter linker of 21 residues (FLN5 + 21 RNC) shifts the 19F NMR peak by +0.8 ppm (Fig. 1c and Extended Data Fig. 2), a similar chemical shift to that of the isolated, unfolded variant of FLN5, having the Y719E point mutation (Fig. 1c,d; ref. 15). The chemical shift of tfmF655 is therefore a simple, direct reporter of the folding of FLN5, both on and off the ribosome.

Identification of co-translational intermediates populated during biosynthesis

The co-translational folding of FLN5 has previously been examined by specifically measuring its unfolded and folded state NMR resonances using 15N labelling and selective 13C-methyl labelling, respectively15. We explored whether 19F NMR could be used to directly observe the folding transition, and so produced eight additional 19F-labelled FLN5 RNCs, varying the number of linking residues deriving from the subsequent FLN6 domain (Fig. 2a,b and Extended Data Fig. 2; ref. 15), with each reporting as a representative biosynthetic snapshot at equilibrium.

Fig. 2: Co-translational folding of FLN5 monitored by 19F NMR spectroscopy.figure 2

a, Design of FLN5 RNCs in which FLN5 is tethered to the PTC via a linker sequence comprising a variable number of FLN6 residues and an arrest-enhanced SecM stalling motif. b, Anti-hexahistidine western blot of purified FLN5 RNCs, with and without ribonuclease A (RNase A) treatment. Representative data shown from two independent repeats. c, The 19F NMR spectra of FLN5 RNCs with increasing distance from the PTC. Observed spectra shown in grey were fitted and peaks assigned to U, I1, I2 or N states (coloured), with the sum of the fits shown in black. NMR data were multiplied with an exponential window function (10 Hz line broadening factor) before Fourier transformation. d, The 19F NMR spectrum of FLN5 + 34 RNC, processed with a line broadening factor of 5 Hz. Residual spectrum after fitting is shown below. e, Folding of FLN5 on the ribosome, measured using 19F NMR line-shape fits. f, Line-widths measured by line-shape fits of spectra as shown in c. All error bars indicate errors calculated by bootstrapping of residuals from NMR line-shape fittings.

Source data

The nascent chain remains unfolded with linker lengths of 21 and 28 residues (Fig. 2c). However, within the 19F spectra of longer RNCs (FLN5 + 31 to FLN5 + 67), we observed multiple peaks that altered in their apparent line-widths and signal intensities, indicative of a folding transition (Fig. 2c,d and Extended Data Fig. 2). Analysis of the spectra, in both the frequency and time domains, showed that FLN5 populates four distinct states during co-translational folding (Fig. 2c,d and Extended Data Fig. 2). The peak integrals are directly related to the concentrations of each state (and thus the total integral to the sample concentration; Extended Data Fig. 2) and so were used to quantify their relative populations (Fig. 2e).

The sharpest peak at −61.8 ppm, corresponding to the unfolded state (denoted U), is found in the spectra of RNCs with linker lengths of 21 to 42 residues (Fig. 2c). However, its population begins to significantly reduce beyond 28 linking residues from the PTC (Fig. 2e). Concurrently, a slower progressive increase in natively folded FLN5 (denoted N, at −62.6 ppm) is found from FLN5 + 31 to FLN5 + 110 RNCs (Fig. 2c,e). These data are consistent with previous observations of U and N by two-dimensional (2D) 1H,15N-correlated and 1H,13C-correlated NMR spectroscopy, respectively15.

The 19F NMR observations also reveal large populations of two putative intermediate states that have previously not been observed, to the best of our knowledge15. These states are detected as broad peaks, which persisted for the duration of the NMR experiments (Extended Data Fig. 3). The intermediates have chemical shifts similar to those of U and N, indicating the absence and presence of native-like tertiary contacts local to the 19F labelling site within these states, denoted I1 and I2, respectively (Fig. 2c). They are initially populated at 31 residues from the PTC (Fig. 2c), at which there is complete emergence of FLN5 from the exit tunnel15. I1 is maximally populated with 31–34 linking residues, while I2 is increasingly populated up to ~47 residues from the PTC before progressively reducing with linker length (Fig. 2e).

NMR peak line-widths can provide information on dynamic processes, reporting on processes such as chemical exchange and rotational tumbling35. To assess the effect of chemical exchange between the nascent chain states on the observed NMR line-widths, we acquired 19F on-resonance rotating-frame relaxation rate (R1ρ) measurements36 of FLN5 + 34 RNC (Extended Data Fig. 5); these data show that the I1 and I2 resonances are not the result of broadening of the U or N peaks. Line-widths are also affected by tumbling; in addition to structural conformations, line-widths of nascent chain resonances are therefore particularly sensitive to even transient, weak binding to the large ribosomal particle5,17. The line-widths of U remain generally sharp across all RNC lengths, indicating that the nascent chain remains mobile, at least locally to the 19F labelling site (Fig. 2f; ref. 15). By contrast, the N resonances are broad at short RNC lengths but narrow away from the ribosome (Fig. 2f) and can be attributed to faster tumbling of the globular FLN5 domain as it is extruded25. The line-widths of I1 and I2 are significantly broader than those of U and N (Fig. 2f), but progressively narrow with both nascent chain length (Fig. 2f) and with increasing ionic strength (Extended Data Fig. 4), indicating that they bind, partly through electrostatic interactions, to the ribosome surface, resulting in more limited mobility.

Moreover, the broad line-widths (that is, fast effective transverse relaxation rates R2) account for the absence of intermediate state resonances in previous NMR measurements using alternative labelling schemes; these require 2D experiments, which increases the dead time during which the signal relaxes and decays. Overall, the 19F NMR data identify two stable, structurally distinct intermediate states, which are populated outside the exit tunnel and are closely associated to the ribosome surface.

Slow interconversion between nascent chain conformations

We acquired 19F chemical exchange saturation transfer (CEST) measurements36 to investigate the kinetic interconversion between the four nascent chain states. By irradiating frequencies at particular offsets from an NMR resonance with a weak applied radiofrequency (B1) field, the resulting perturbation (that is, signal reduction) is transferred to the interconverting state via chemical exchange37. CEST measurements of FLN5 + 34 RNC (Extended Data Fig. 5) indicate that chemical exchange between all states occurs slowly (rate constant (kex) < 1.3 s−1, time constant (τex) > 0.8 s). By contrast, an isolated variant of FLN5 exchanges at a faster rate of 3.6 ± 0.4 s−1 between its unfolded and native-like intermediate structure that lacks G-strand contacts but is otherwise folded30 (Extended Data Fig. 5), suggesting that the effective folding rate is reduced on the ribosome and that additional processes may potentially be competing with folding. The observed slow exchange between RNC states, corroborated by the R1ρ measurements discussed above (Extended Data Fig. 5), also verify the presence of two distinct intermediate state peaks (rather than a single, highly broadened peak), since irradiating I1 did not result in a significant perturbation of I2, and vice versa (Extended Data Fig. 5).

Partially structured intermediates on the ribosome

Off the ribosome, truncation of the six carboxy-terminal (C-terminal) residues of isolated FLN5 (FLN5∆6) produces a population of a stable intermediate (Extended Data Fig. 3; ref. 30), previously characterized as having a native-like core with a detached terminal G-strand, and with the conserved cis-proline P742 in a trans conformation (Extended Data Fig. 3; ref. 30). Previous structural modelling has indicated that this conformation is sterically accessible on the ribosome with a linker length of at least 18 amino acids30, and so we sought to examine whether I1 and I2 adopted this structure.

We first tested whether the putative co-translational intermediates possessed a stable structure by incubating 19F-labelled FLN5 + 37 RNC in 2 M urea (Fig. 3a). We observed a shift in the folding equilibrium towards U, while populations of I1 and I2 showed no discernible change. This indicates that the intermediates possess some stable structure that is largely resistant to mildly denaturing conditions. To assess this further, we introduced the destabilizing Y719E point mutation into 19F-labelled FLN5 + 47 RNC (Fig. 3b), which resulted in the collapse of its three 19F resonances into a single sharp peak (Extended Data Fig. 2), and in which its line-width and chemical shift are consistent with an unfolded state. Residue Y719 is natively solvent inaccessible, so the ability of a mutation to completely unfold both I1 and I2 indicates that they adopt partially folded structures. Additionally, we 19F-labelled FLN5 + 47 RNCs at positions natively buried in the hydrophobic core (Y715 and Y727; Extended Data Fig. 6). We found 19F NMR resonances attributable to a native-like structure, whose thermodynamic stabilities are higher than those found in RNCs labelled at position 655 (relative to isolated FLN5; Extended Data Fig. 6), suggesting the core is at least partially formed in the intermediates.

Fig. 3: The ribosome-bound intermediate states are partially folded.figure 3

a, The 19F NMR spectra of FLN5 + 37 RNC in the absence and presence of 2 M urea. Fractional populations shown below. b, The 19F NMR spectra of FLN5 + 47 and FLN5 + 47 Y719E RNC. Below, the line-width of FLN5 + 47 Y719E is compared against the line-widths of U determined for other RNCs (mean ± s.d.; Fig. 2f). c, The 19F NMR spectra of FLN5 + 47 and FLN5 + 47 P742A RNC. Analysis shown in Extended Data Fig. 4. d, The 19F NMR spectra of tfmF655-labelled FLN5∆6 + 47GS and FLN5 + 42GS RNCs (283 K, 500 MHz). Schematic depicts RNC construct design. Analysis shown in Extended Data Fig. 4. Unless stated otherwise, error bars indicate errors propagated from bootstrapping of residuals from NMR line-shape fittings.

Source data

Within the isolated FLN5 intermediate, the native-like folded core comprises the A- to F-strands, and accordingly the 19F chemical shift of residue 655 (residing on the A-strand) is native-like (Extended Data Fig. 3). Therefore, based on their chemical shifts (Fig. 2c), it is likely that the A-strand on I2 is also folded onto the hydrophobic core, whereas in the I1 state, native side chain contacts between the A-strand and its neighbouring residues are absent and thus the A-strand is unlikely to be completely associated.

Next, we examined isomerization of the conserved proline within the intermediates. Using populations determined from their 19F NMR integrals, we measured the free energy changes upon mutation of P742 to alanine, which destabilizes the cis conformation (Extended Data Fig. 4; ref. 30). The point mutation completely destabilizes I1 (∆∆GI1-U > 1.7 kcal mol−1), as indicated by the absence of its 19F resonance in the RNC spectra (Fig. 3c and Extended Data Fig. 4), showing that I1 possesses the native cis-P742. However, I2 and N are only mildly, but equally, destabilized (∆∆GI2–U = 0.8 ± 0.2, ∆∆GN–U = 0.9 ± 0.2 kcal mol−1 for FLN5 + 34; Fig. 3c and Extended Data Fig. 4), indicating they likely have the same P742 conformation. Although this destablization is less than that for isolated FLN5 (∆∆GN–U ≈ +4 kcal mol−1 (ref. 30)), previously observed 1H,13C-methyl resonance chemical shifts of RNCs show that N adopts the cis-proline conformation30; thus additional effects on the ribosome likely mitigate the destabilizing mutation within I2 and N. Overall, in contrast to the isolated intermediate (Extended Data Fig. 3; ref. 30), both I1 and I2 likely possess the cis conformer of P742, potentially rationalizing the observed slow exchange (Extended Data Fig. 5) between U and the intermediates to enable proline isomerization to occur.

The terminal G-strand (I743 to I748) directly succeeds P742 and, as described above, is detached (after truncation) from the folded core of the isolated intermediate30. We thus investigated its role in co-translational folding by replacing the six C-terminal FLN5 residues with a stretch of poly(glycine–serine) residues in a RNC. We found that N was completely destabilized by the series of mutations (∆∆GN–U > 2.3 kcal mol−1; Fig. 3d and Extended Data Fig. 4). However, I1 and I2 both persisted, being less destabilized (∆∆GI1–U ≈ +1.5 ± 0.2 kcal mol−1; ∆∆GI2–U ≈ +1.9 ± 0.2 kcal mol−1; Fig. 3d), indicating that the G-strand contributes significantly less to their overall folding stabilities. We also observe narrower I1 and I2 resonances by modifying the FLN5 C terminus, suggesting that interactions between the ribosome and this nascent chain segment are reduced (Extended Data Fig. 4). We note that the G-strand resides within a ribosome-binding segment previously identified in U by 1H,15N-correlated NMR measurements17.

The combined NMR data (Fig. 3) therefore show that I1 and I2 possess a folded core, in which the G-strand is likely to be at least partly detached and interacting with the ribosome, while I1 is further characterized by incomplete association of the A-strand, which has been found to also be labile in folding intermediates off the ribosome30.

Corroborating structural evidence of intermediate states

We next performed coarse-grained (CG) MD simulations using structure-based models as an orthogonal means of examining the co-translational folding of FLN5, applying parallel biased metadynamics38 to enhance sampling transitions between nascent chain conformations using ten collective variables (Methods). The MD simulation temperature was calibrated to match populations of isolated FLN5 and its C-terminal truncations with those determined experimentally (Extended Data Fig. 7). The introduction of previously calibrated electrostatic interactions between FLN5 and the ribosome17 enabled us to accurately predict FLN5 + 31, from six RNCs (across FLN5 + 21 to FLN5 + 47), as the length at which folding begins (Extended Data Fig. 7). From the simulations, we generated and analysed the folding free energy landscapes, defined by native contacts between neighbouring β-strands, to determine the folding pathway. Consistent across the RNCs is the initial formation of native contacts within the A- to F-strands (Extended Data Fig. 7), which results in an ensemble of marginally stable intermediates (Fig. 4b), collectively characterized by a native-like core with a detached, transiently associating G-strand (Fig. 4a). Despite capturing only a single, lowly populated intermediate state (Fig. 2e and Extended Data Fig. 7), the simple CG models propose structures (Fig. 4b) that are qualitatively consistent with the 19F NMR data of I2 (Fig. 3). The reduced contacts observed between the A-strand and its neighbouring loop region (between strands F and G) within the same structures (Extended Data Fig. 7) may account for I1 within the structural ensemble.

Fig. 4: Structural ensemble of the FLN5 co-translational intermediate state determined by MD simulations.figure 4

a, Structural ensemble of the FLN5 + 34 intermediate from CG models. The 10 most populated intermediate conformations are superimposed with the native FLN5 crystal structure (orange; PDB no. 1QFH) and coloured from N terminus (red) to C terminus (blue). Ribosome and linker are not shown for clarity. b, Examples of the FLN5 + 34 intermediate structures, with the FLN5 crystal structure aligned. Colours as in a. Arrow indicates axis from C to N terminus. c, The bottom left plot shows the contact probability between the FLN5 + 34 in its unfolded, intermediate and native states and the ribosome from CG models. Contact probabilities of the intermediate and native states are coloured on the FLN5 structures (above) with regions of highest probability highlighted. The right depicts the contact probability between the FLN5 + 34 intermediate and the ribosome, mapped onto the ribosome surface.

Source data

Contacts made by the nascent chain with the ribosome surface in the MD simulations (Fig. 4c and Extended Data Fig. 7) correlate well with previous NMR measurements: trajectories for U show strong (up to 80% contact probability), predominantly electrostatic interactions at its C-terminal binding site (residues N728–C747) and weak contacts elsewhere17, while contacts between N and the ribosome occur at the domain’s C-terminal hemisphere and are largely steric with only small electrostatic contributions (Fig. 4c and Extended Data Fig. 7; ref. 25). We find that a significant proportion (~50%) of the intermediate ensemble contacts the ribosome through charge interactions (Extended Data Fig. 7). The interactions identified (Fig. 4c) are localized at the C terminus, as observed for U although less strong, and are consistent with experimental data (Fig. 3d and Extended Data Fig. 4). Contacts are also found at the more positively charged, amino-terminal (N-terminal) hemisphere of FLN5, centred at residues K646 and K680, which preferentially orients the partially folded domain towards the RNA-rich side of the ribosome vestibule (Fig. 4b), predominantly contacting rRNA helices H24, H47 and H50 (Fig. 4c).

We subsequently re-examined cryo-EM data obtained for FLN5 + 45 and FLN5 + 47 RNCs31, previously fitted with all-atom density-guided MD simulations with exclusively native structures defined within structure-based models. Having discovered that these RNCs predominantly populate partially folded intermediates in this work (Fig. 2), we used the previously obtained electron densities as restraints to fit structures with inter-residue contacts characterizing I2 (Fig. 4a) instead (Extended Data Fig. 8). These new models showed cross-correlations that were quantitatively similar to those obtained for natively folded structures (Extended Data Fig. 8). Additionally, the intermediate conformations also showed binding to the ribosome surface at the N-terminal loop regions and the G-strand of FLN5 (Extended Data Fig. 8), as identified in the CG models (Fig. 4c). We conclude that the cryo-EM data corroborate the proposed intermediate state structures and their interactions with the ribosome.

Mechanism of intermediate state stabilization on the ribosome

We next sought to experimentally examine the effect of the identified binding site on co-translational folding. We thus replaced residues that are predicted to strongly bind to the ribosome, K646 and K680 (Fig. 4c), found natively in the loop regions, with glutamic acid residues to reverse their charge. The 19F NMR spectrum of the FLN5 + 34 K646/K680E RNC shows that folding remains four-state (Fig. 5). However, the N is stabilized on the ribosome by 0.6 ± 0.3 kcal mol−1 relative to U, despite the mutations destabilizing the FLN5 domain off the ribosome by ~0.4 kcal mol−1 (Extended Data Fig. 3). Moreover, both I1 and I2 are each destabilized relative to N by 0.2–0.3 kcal mol−1. This shift in co-translational folding, together with a small reduction in the line-widths of I1 and I2 (Fig. 5), is therefore consistent with disruption of ribosome interactions that contribute to the stabilities of the intermediates. The folding equilibrium is also shifted towards N in a longer nascent chain possessing the same mutations (Extended Data Fig. 4), although to a lesser extent, indicating that the interactions mediated by K646 and K680 are strongest closest to the ribosome surface. However, the persistence of broad NMR resonances attributable to the intermediate states suggest that I1 and I2 possess additional stabilizing binding sites or other modes of interactions that were not defined within the CG models.

Fig. 5: Electrostatic interactions with the ribosome surface stabilize partially folded nascent chains.figure 5

The 19F NMR spectra of FLN5 + 34 and FLN5 + 34 K646E/K680 RNCs. Line-widths and populations of each RNC state determined by analysis of the spectra are shown on the right. Error bars indicate errors (propagated) from bootstrapping of residuals from NMR line-shape fittings.

Source data

Electrostatic interactions between the nascent chain and the ribosome can also be mediated via magnesium ions

留言 (0)

沒有登入
gif