Chain Dynamics of Nascent Polypeptides Emerging from the Ribosome

Experimental design

Nascent polypeptides of increasing length derived from the apomyoglobin (apoMb) sequence (Figure 1, panel a) were analyzed in an E. coli cell-free system. The polypeptide length was chosen so that variable fractions of the chain are expected to be buried in the ribosomal exit channel (Figure 1, panel b). The natively unfolded PIR₉₀ control protein was examined as a reference due to its low nonpolar content (Supporting Figure 3) and its intrinsic inability to adopt a compact structure in aqueous medium. Small angle X-ray scattering (SAXS) of pure full-length PIR shows that this sequence behaves as an unfolded self-avoiding Gaussian chain (17).

Initiation of protein biosynthesis is neither synchronized nor rigorously synchronizable across the ensemble of ribosomes in intact cells and cell-free systems. This limitation, which hampers both sensitivity and resolution, was overcome by generating ribosome-bound nascent chains of increasing length and monodisperse molecular mass. Creation of these homogeneous equilibrium populations of apoMb-RNCs is key to the success of our experiments (Figure 1, panel c). Given that conformational sampling is typically much faster than translation (16), these equilibrium populations may be regarded as working models for transiently elongating polypeptides. Cell-free expression was carried out in the presence of BODIPY-FL-tagged $\mathrm{Met}-{\mathrm{tRNA}}_{\mathrm{f}}^{\mathrm{Met}}$, leading to the biosynthesis of N-terminal BODIPY-FL-labeled nascent chains. This strategy provides the advantage of ensuring that all nascent chains are labeled with only one fluorophore at a unique sequence site (the amino group of N-terminal methionine). Due to its fluorescence lifetime in the low ns range (τ_F = 5.9 ns), BODIPY-FL is an ideal fluorophore here due to its small size and its ability to probe motions on a timescale close to the independent tumbling time of small globular proteins.

Biochemical characterization

As a complement to the spectroscopy, the amounts of chaperones associated with resuspended RNCs were characterized. Immunoblotting (Figure 2, panel a) shows that the average amount of TF per active ribosome increases progressively with chain elongation, with a sharp rise near completion of biosynthesis. The absolute amounts of TF associated with apoMb₁₅₃-RNCs after ultracentrifugation are comparable to the estimated amounts of translating ribosomes (data not shown). Given the known 1:1 stoichiometry for substrate:TF binding (18), the above results suggest that TF displays a higher affinity for nascent apoMb chains as they get longer and acquire a higher nonpolar content, in agreement with recent studies on other proteins (19, 20).

While the TF chaperone associates most tightly with RNCs bearing full-length apoMb, the biosynthesis of the natively unfolded control sequence PIR₉₀ requires reduced amounts of TF (Figure 2, panel a). In addition to its low hydrophobicity, PIR₉₀ has very few aromatic residues (Supporting Figure 3). Both nonpolar content and aromatic residues of the substrate are important for TF affinity in solution (21, 22). Therefore, the observed low TF content in PIR₉₀-RNCs is consistent with the prediction that this chaperone should not significantly interact with PIR₉₀.

Immunoblotting also shows that the bacterial Hsp70 chaperone DnaK associates with apoMb-translating ribosomes more than with the PIR₉₀ control sequence, in wild-type strains (Figure 2, panel b). However, in contrast with the behavior of TF, the amount of detectable ribosome-apoMb-associated DnaK is independent of substrate chain length within experimental error. The presence of residual DnaK even at short chain lengths is not entirely surprising given the known affinity of DnaK for E. coli ribosomes (23). Therefore some of the DnaK detected here may be due to association with the ribosome. DnaK concentrations increase for cell-free systems prepared from Δtig strains (lacking the gene for TF), in agreement with the overall complementary roles of DnaK and TF (8) and the likelihood that they compete for similar binding sites (8, 24). There is, however, no significant detectable difference in the amount of DnaK associated with RNCs carrying the short apoMb₁₆ and apoMb₃₅ nascent chains prepared from wild type and Δtig strains, due to the high experimental error. The two-tailed Student's t-test shows that the probability that the wild type and Δtig results are equal (p = 0.27 for apoMb₁₆ and p = 0.36 for apoMb₃₅) is far above the standard p<0.05 or 0.1 required to demonstrate that they are different.

The dynamic fluorescence depolarization approach

Initial steady-state anisotropy experiments on RNCs bearing apoMb polypeptides of increasing length (Supporting Figure 4) yielded no insights into the motions of nascent chains. Dynamic fluorescence depolarization (10, 11) is a more powerful approach that relies on the identification of fluorescence anisotropy decays in the frequency domain and enables the explicit discrimination of different types of motions (Figure 3, panel a). This methodology therefore provides specific information on the dynamics of nascent polypeptides in the context of the ribosomal complex.

The frequency-domain dynamic fluorescence depolarization approach enables detection of complex anisotropy decays at very good resolution. The following reasons contribute to the high performance of this tool (25). In the frequency domain, differences between the polarized components of the decay are measured directly, at different frequencies of the sinusoidally amplitude-modulated excitation light source. On the other hand, in the more widely used time-domain methodology, which monitors polarized fluorescence emission decays in real time, the desired anisotropy decay profile is calculated from two separate intensity components, I_⊥ and I_∥. Furthermore, frequency domain data collection and fitting are based on two concurrently collected observables (i.e., the differential polarized phase angle and the ratio of polarized modulated amplitudes), each providing information on multiple anisotropy components and resulting in a better resolution than the observation of either quantity alone.

The BODIPY fluorophore, selectively linked to the N terminus amino group of the nascent chains, serves as a reporter for its own fast local dynamics (F), the local dynamics of the nascent chain (I) and the global dynamics of the whole RNC, dominated by the slow motions of the ribosome. As shown in the Figure, local and global motions are present concurrently within the same species (Figure 3, panel a).

Representative raw data collected by frequency domain fluorometry for short and long ribosome-bound nascent chains are shown (Figure 3). There is a dramatic difference between the dynamics of short and long nascent chains (Figure 3, panel b). The short chains experience a very fast motion, given that the sinusoidal maximum of the phase change falls at frequencies even higher than experimentally detectable. This fast motion has a similar apparent τ_C to free BODIPY (0.12 ± 0.04 ns, measured in the same medium), and must be rotationally restricted since the anisotropy does not completely decay over the experimental timescale.

The featureless modulation ratio profile further confirms the presence of local segmental dynamics and a very slow global tumbling (≥ 0.5 μs). Upon curve fitting (Figure 3, panel a, and Methods section), the slowest anisotropy phase, which corresponds to the residual anisotropy at time infinity, is assigned to the global motions of the ribosome complex.

Remarkably, the long chains show one additional local motion, which is detectable as a characteristic spectral feature (i.e., a curved bump) of the phase–change data at intermediate modulation frequencies. This feature reveals the presence of a local motion with a 3–7 ns apparent rotational correlation time (τ_c). The above dynamic component is clearly detectable (Figure 3, panels b, d) and statistically well beyond experimental error (Table 1). Spectral simulations further confirm the correlation between the fraction of depolarization decay assigned to the ns motion and its matching and readily identifiable spectroscopic signature (Figure 3, panels e, f).

Data analysis and interpretation

A detailed analysis of the fluorescence depolarization data for all the apoMb chain lengths (Figure 4 and Table 1) reveals that the number and type of observed motions are significantly dependent on the peptide elongation stage. Only two distinct classes of dynamics are identified for short nascent chains while three types are detected for longer chains.

We assign the slow phase to the chain motions associated with the overall tumbling of the higher order complex, dominated by the size of the ribosome. Accordingly, the τ_c for this motion was set to 1 μs. While this motion cannot be resolved directly, as its timescale is much longer than the fluorophore lifetime, its presence is inferred from the fact that a non-decaying anisotropy component needs to be introduced for the fitting algorithm to converge. The fraction of anisotropy decay for this motion is treated as a floating parameter in the fits.

The phase corresponding to motions occurring on an intermediate timescale (τ_c = 3–7 ns) provides evidence for the presence of an independent conformation of the ribosome-bound nascent protein. This phase is detected only for the longer nascent chains (apoMb_89-153), and its timescale corresponds to the approximate apparent rotational correlation time of a small structured domain. The apparent rotational correlation time for the intermediate timescale motion changes significantly upon release of apoMb₁₅₃ from the ribosome in the crowded medium, from 4.5 ± 0.7 to 41±14 ns. As expected, the slow motion due to global RNC tumbling is no longer present, for this species. Therefore it is clear that apoMb undergoes some structural and/or shape changes upon release from the ribosome. The fast timescale motion is detected for all samples, including ribosome-released apoMb₁₅₃.

The fact that the 3–7 ns intermediate timescale motion preserves a constant fraction and a similar τ_C for the 89- to 153-residue chains suggests that the general nature of this motion does not significantly vary as the protein emerges from the ribosome. While high resolution structural/dynamic details cannot be drawn by fluorescence depolarization alone, the above observations support the fact that this motion is a persistent feature of all RNCs bearing the first ≈60 N-terminal residues of apoMb.

The apoMb₅₇-RNC has a peculiar experimental behavior. It displays 3-component dynamics in only 75% of the experiments, while the remaining 25% fit best to 2-state dynamics, lacking the ns timescale motion (Figure 4). The poor data reproducibility at this chain length may reflect the fact that the independent ns motions of nascent apoMb₅₇ (50–100 % buried in the ribosomal tunnel, Figure 1, panel b) could arise from an unstable structure whose population is strongly affected by small sample-dependent changes in experimental conditions.

In summary, three distinct dynamic regimes are identified for nascent apoMb-RNCs, fast (sub–ns), intermediate timescale (3–7 ns) and slow (>1 μs). These motions are assigned to global tumbling of the ribosomal complex, nascent chain dynamics, and local fluorophore dynamics, respectively. The short nascent chains (16, 35 residues) display only the fast and slow motion while the longer chains (89–153 residues) also display the ns motion.

The behavior of the nascent chains in the absence of molecular chaperones was analyzed by creating chaperone-free RNCs in a TF-deficient cell-free system (Δtig strain, −G conditions). We then treated the resuspended ribosomes bearing either apoMb₁₅₃-or PIR₉₀-RNCs with GrpE and excess ATP (+G conditions, Figure 5, panel a). This procedure ensures release of any bound DnaK from the nascent chains, as confirmed by immunoblotting. The presence of additional chaperones (DnaJ or SecB) in the RNC has been excluded by immunoblotting (Supporting Figure 2). As shown by the frequency-dependent phase changes (Figure 5, panel b, modulation frequency ≈70 MHz), release of DnaK from full length apoMb does not lead to a decrease in the fraction of the ns motion. In fact, an increase in fraction (from 4–95%) is observed. This result is important because it shows that the 3–7 ns motion is not due to the binding of ribosome-bound nascent chains to DnaK but is a unique feature of the nascent chain. In other words, this independent motion is sampled by the nascent chains regardless of the presence of chaperones.

Not surprisingly, no changes in nascent chain dynamics are detected for PIR₉₀ under +G conditions.

Models for RNC dynamics

Based on the above data, some important insights into the nature of the independent nascent chain motions can be gained. We will focus the discussion on the nascent chains of sufficient length to emerge from the ribosomal tunnel, as they exhibit the most interesting behavior.

Representative cartoons (Figure 6) highlight the known presence of a dynamic exchange between chaperones and nascent chains and provide a framework for interpretation of the data on the RNCs generated in wild type cell-free systems. Three important facts need to be considered. Namely: (a) the existence of a reversible interaction between nascent chains and trigger factor or DnaK, (b) the fact that TF is effectively “frozen on the ribosome” (i.e., ribosome-bound) on the timescale of TF binding to nascent chains (20, 26), and (c) the μs–s on/off rates for nascent chain interaction with TF (20, 26) and μs–min on/off rates for interaction with DnaK (27). The above observations support the idea that nascent chain emerging from the ribosome are transiently bound to TF and/or DnaK on a slow timescale (μs–s) relative to the dynamic fluorescence depolarization timescale (ns). Therefore, while chaperone-bound nascent chain species are likely to exist, their motions remain spectroscopically unresolved here and merge into the fraction of non-decaying anisotropy of the entire RNC (Figure 4). We conclude that this study does not provide information about the dynamics of chaperone-bound species 1 and 3 (Figure 6). Conversely, BODIPY depolarization is sensitive to the independent chain motions of the chaperone-free species of type 2 or 4.

The possibility that ∼ns chemical exchange processes between species of types 1 (or 3) and 2 (or 4) are responsible for the observed 3–7 ns timescale motions is ruled out on the grounds that this type of dynamics is also detected in the absence of any cotranslationally active chaperones (Figure 5). In addition, as discussed above, the literature reports much longer lifetimes (μs–s) for TF- and DnaK-nascent chain complexes.

In summary, we interpret the apoMb_89-153 3–7 ns motions as the signature of an independently tumbling structural domain of nascent chains emerging from the ribosome. Shorter apoMb chains buried in the ribosomal tunnel and natively unfolded PIR₉₀ lack this feature.

Nascent chain structure and dynamics

Which specific structure is compatible with such independent ns dynamics? It is clear that this question cannot be answered based on fluorescence depolarization data alone. On the other hand, it is possible to broadly define the class of structures consistent with the spectroscopic results.

The putative size and shape compatible with the observed ns dynamics of apoMb₁₅₃-RNCs (Figure 7) prompt some broad speculations. The expected rotational correlation times corresponding to rigid extended chains, semi-compact ellipsoids and spheres are shown for different chain lengths. The effective viscosity (required to generate the plots and measured by us in the resuspended ribosome medium, see also Supporting Information) sensed by the nascent chain depends on both crowding of the medium and local structural confinement (28). Two representative limiting viscosities (η) are considered here: η of free BODIPY (panel a) and η of ribosome-released apoMb₁₅₃ (panel b), measured in RNC solutions.

Both panels a and b exclude the presence of fully extended chains, given that the typical polypeptide persistence length is only 4-6 residues. Less elongated ellipsoids and spheres are feasible for a wide range of molecular sizes. Panel a predicts longer chain lengths (90–130 residues) than could possibly emerge form the ribosome in apoMb₁₅₃. Panel b, conversely, estimates the possible presence of 10- to 15- residue-long spheres and semi-compact ellipsoids.

Large nascent chains emerging from the ribosome may sense local crowding/confinement to a similar extent as ribosome-released apoMb₁₅₃ while smaller structures may sense a viscosity closer to that of the free fluorophore. While it is impossible to differentiate between the two scenarios based on the currently available information, it follows from the above discussion that the observed tumbling unit should be at least partially compact and possibly correspond to a fairly small stretch of residues. In summary, we speculate that the independent ns dynamics of full length apoMb₁₅₃ may correspond to a semi-compact ellipsoidal shape of chain length > 8 residues.

Conclusions

We have presented direct spectroscopic evidence for the presence of independent nascent protein conformations during intermediate and late stages of biosynthesis. These species are absent in the case of a natively unfolded control protein incapable of assuming a well defined structure.

The direct biophysical evidence presented here moves a step forward to understand conformational sampling during the earliest stages of a protein's life. The presence of ribosome-bound nascent chain structure has been detected only indirectly before, (9, 29-32) and high resolution information has only been gained for the outer domains of multidomain proteins (33). Here, for the first time, we (a) present direct evidence for the conformational dynamics of nascent polypeptides encoding a single-domain protein both inside and outside the ribosomal exit tunnel, and (b) demonstrate the distinctive dependence of nascent chain structural dynamics on the stage of chain elongation and amino acid sequence. While more investigations are needed to unveil further details at higher resolution, this work takes an initial step at exploring the structure/dynamics of nascent proteins.

In contrast with purified N-terminal fragments, which become structured only extremely close to chain completion or heavily aggregate (34), we show here that ribosome-bound apoMb explores independent conformations even when a large C-terminal fraction is missing.

Finally, this investigation introduces dynamic fluorescence depolarization as a tool to study cotranslational protein folding. The power of this technique stems from its simplicity. Only one site-specific label is needed, in contrast with other approaches (e.g., FRET), which typically require multiple probes. In addition, dynamic fluorescence depolarization enables the resolution of multiple concurrent motions in a single experiment.

Bacterial strains and plasmids

The E. coli K12 strain A19 (35) was supplied by G. Kramer. The MC4100Δtig::cat E. coli strain was provided by P. Genevaux and F.U. Hartl (36). Construction of the E. coliA19Δtig strain was carried out by transferring the Δtig::cat allele from MC4100Δtig::cat into the A19 strain by P1 transduction using the P1vir phage. Transductants were selected at 37°C on Luria-Bertani (LB) agar containing chloramphenicol (15 μg/ml) and 10 mM sodium citrate. Colonies were further purified by restreaking on the same plates. The absence of TF was confirmed by Western blot.

The plasmid containing the sperm whale Mb gene was described previously (15). A pET-28b (Novagen) plasmid carrying the gene for Grb14 PIR domain was supplied by I. Broutin (17). The PIR stop codon (TGA) was mutated to GAT with the QuikChange^® Site-Directed Mutagenesis Kit (Stratagene). The pTf16 plasmid containing the wild-type tig gene was purchased from Takara Mirus Bio.

Generation of BODIPY-FL Met-tRNA_f^Met

BODIPY^® FL (Invitrogen) and tRNA_f^Met (Sigma-Aldrich) were used without further purification. Aminoacylation of tRNA_f^Met was performed as described (37) with overexpressed and purified E. coli Methionyl tRNA synthetase (38). Met-tRNA_f^Met was conjugated to the succinimidyl ester of BODIPY-FL (10.0 mM) according to known protocols (39, 40). The reaction was quenched with tris-HCl (pH 7.5, 0.3 mM), brought to pH 5 with acetic acid (AcOH), and diluted 6× with aqueous NaOAc (0.3 mM, pH 5.4) followed by phenol extraction and ethanol precipitation twice. The pellet was resuspended in 0.25 volumes of H₂O. Fluorescent labeling was confirmed by NaOAc-urea-PAGE (7 M urea, 9.7% acrylamide, 0.3% bis-acrylamide) adapted from (41), followed by fluorescence imaging (UC 4×4 scanner from Genomic Solutions, λ_ex 488 nm, emission channel bandpass filter centered at 512 nm) and staining with StainsAll (Sigma-Aldrich).

Cell-free protein synthesis

E. coli S30 extracts were prepared from either wild-type A19 (15) or trigger factor-deficient A19Δtig::cat strains.

Protein biosynthesis was carried out as described (15). Folinic acid was omitted from the transcription-translation mixture and BODIPY-FL-Met-tRNA_f^Met was added (final concentration ∼3 U/ml) except for fluorescence negative controls, which lacked fluorescent tRNA but included folinic acid.

Protein synthesis (75 μl) was allowed to proceed for 30 min at 37°C. BODIPY-labeled ribosome-released full length apoMb was generated from wild-type RNCs upon addition of 1.5 M NH₂OH in the presence of 0.4 μM GrpE, 0.4 μM DnaJ, 30 μM ATP and 500 μM KCl.

Generation and isolation of RNCs

Ribosome-bound nascent chains were generated by oligodeoxynucleotide-directed mRNA cleavage (Supporting Information). Ribosomal pellets (including ribosome-bound peptides and associated proteins) were isolated upon ultracentrifugation (Beckman TL-100 with TLA 100.1 or equivalent rotor) as follows. Cell-free expression mixtures (75 μl) were loaded onto a 150 μl of sucrose cushion (1.1 M sucrose, 20 mM tris-HCl, 10 mM MgOAc₂, 500 mM NH₄Cl, 0.5 mM EDTA and 1 mM dl-dithiothreitol, pH adjusted to 6.9) (42) and ultracentrifuged at 160,000 × g for 60 min at 4°C. The ribosomal pellet was resuspended in 15 μl (i.e., 1/5 of the original volume) of buffer (10 mM tris-HCl, 10 mM Mg(OAc)₂, 60 mM NH₄Cl, 0.5 mM EDTA, 1 mM dl-dithiothreitol, pH 6.9) by shaking for 60 min on ice in an orbital shaker. The ribosome-bound status of nascent chains was tested by treatment with puromycin (Fig. 1d) added post-translationally (final concentration 0.5 mM), followed by incubation for 15 min at 37°C.

Generation of TF- and DnaK-free samples

In order to analyze nascent proteins in the absence of DnaK and TF, Δtig RNCs were incubated with 2 μM GrpE, 30 μM ATP, and 500 μM KCl (+G conditions). The presence of DnaK in resuspended RNCs was tested under −G and +G conditions by an additional ultracentrifugation/resuspension step.

Determination of DnaK and TF concentrations

Cell-free proteins were separated by SDS-PAGE (8% tris-glycine polyacrylamide gels) (43) and transferred to a PVDF membrane, followed by probing with mouse monoclonal anti-DnaK (Stressgen) or rabbit anti-TF (Genevaux & Hartl) polyclonal antibodies. Goat anti-rabbit or goat anti-mouse alkaline phosphatase conjugates (Novagen) were used as secondary antibodies. Protein bands were detected by chemiluminescence, using CDP-Star^® substrate (Novagen) for TF and Lumi-Phos^® substrate (Pierce) for DnaK. Band intensities were quantified by ImageJ (44) and converted to chaperone concentrations via a calibration curve correlating band intensities with intensities of protein standards (TF or DnaK) of known concentration. TF concentrations of apoMb samples range from 10 to 90 nM. DnaK concentrations of apoMb samples range from 20 to 65 nM in WT and Δtig samples. TF and DnaK concentrations of PIR samples and negative controls (either lacking DNA or reacted with puromycin) were 2-15 nM. To assess chaperone content per nascent chain, chaperone concentrations were divided by the corresponding relative fluorescence intensities evaluated at 123 MHz modulation frequency (Chronos Spectrofluorometer, ISS). Intensities were corrected by background subtraction and then divided by the value obtained for wild-type apoMb₁₅₃ in each experiment, to ensure reliable cross-comparisons.

Fluorescence instrumentation and analysis

All measurements were performed at room temperature. Anisotropy decay data were collected by the multifrequency phase and modulation technique with a Chronos spectrofluorimeter (ISS) equipped with calcite prism polarizers and a laser diode (λ_ex = 473 nm, λ_em > 510 nm with a 51294 filter, Newport Corp.-Oriel). Lifetime measurements were performed on all samples with vertically polarized excitation and 54.7° polarized emission. All frequency-domain fluorescence data were analyzed using the GLOBALS software package (45). A frequency-independent standard deviation of 0.2° for phase and 0.004 for modulation was applied. Lifetime curve fitting required three discrete components, one kept fixed to 0.001 ns to account for minor contributions from light scattering (less than 2% of total intensity). The frequency dependent G-factor was measured independently. Frequency-dependent polynomial corrections were applied to all phase and modulation data to take the G value into account. Depolarization data were fit to multiexponential decay expressions (Figure 3, panel a) including either two or three discrete components (10). In the case of two-components, only two terms, describing the fast (F) and slow (S) components, were included. The fractional amplitudes (F_S and F_F) and apparent rotational correlation time for the faster motion (τ_F) were allowed to vary while the fundamental anisotropy and the slow correlation time were fixed (r(0)=0.37, τ_s=1 μs). In all cases, a three-component expression was also considered, with an added term including F_I and τ_I. The three component analysis was retained only if a significant improvement in fit quality was achieved (Table 1). Spectral simulations were performed with the Vinci software (ISS).

Fluorescence-detected gel electrophoresis

Puromycin-released chains were analyzed by SDS-PAGE with a three layer tris-tricine gel (15, 46). Low pH gels (47) were employed to reveal -/+ puromycin treated peptidyl-tRNAs.