Abstract
We have characterized the cotranslational folding of two small protein domains of different folds – the α-helical N-terminal domain of HemK and the β-rich FLN5 filamin domain – by measuring the force that the folding protein exerts on the nascent chain when located in different parts of the ribosome exit tunnel (Force-Profile Analysis - FPA), allowing us to compare FPA to three other techniques currently used to study cotranslational folding: real-time FRET, PET, and NMR. We find that FPA identifies the same cotranslational folding transitions as do the other methods, and that these techniques therefore reflect the same basic process of cotranslational folding in similar ways.
In recent years, a number of new experimental methods for analyzing the cotranslational folding of protein domains have been developed. These include real-time FRET analysis [1-4], and methods in which nascent polypeptide chains of defined lengths are arrested in the ribosome and their folding status analyzed by, e.g., cryo-EM [5, 6], protease resistance [4, 7,12], NMR [13-18], photoinduced electron transfer (PET) [3, 4], folding-associated co-translational sequencing [19], optical tweezer pulling [20-23], fluorescence measurements [24], and measuring the force that the folding protein exerts on the nascent chain using a translational arrest peptide (AP) as a force sensor [5, 6, 12, 25-27]. Further, coarse-grained molecular dynamics simulations of various flavors can provide detailed insights into cotranslational folding reactions [28-32], especially when coupled with experimental studies [6, 28, 33].
While these different methods have all been used with success, they have rarely been directly compared on the same target protein. As a first step towards such comparisons, we now report AP-based Force-Profile Analysis (FPA) for two protein domains - the α-helical HemK N-terminal domain (HemK-NTD) and the β-sheet-rich FLN5 filamin domain - that have previously been studied by real-time FRET/PET, and NMR respectively [3, 4, 17]. We find that FPA identifies the same cotranslational folding transitions as do the other methods. We conclude that results obtained with the different methods can be interpreted under a common conceptual framework, albeit at different levels of detail and with different limitations inherent to the analysis.
AP-based force-profiles
FPA is based on the observation that the efficiency of AP-induced translational stalling is reduced when the nascent chain is subject to an external pulling force [23, 34, 35]. As shown in Fig. 1, such a pulling force can be induced by cotranslational protein folding. For constructs where the tether+AP segment is long enough that there is just enough space in the ribosome exit tunnel for the protein to fold if the segment is stretched out from its equilibrium length, some of the free energy gained upon folding will be stored as elastic energy (increased tension) in the nascent chain, reducing stalling and increasing the relative amount of full-length protein (Fig. 1a). By measuring the stalling efficiency (measured as fraction full-length protein, fFL) for a series of constructs of increasing tether length (Fig. 1b, c), a force profile can be generated that shows how the folding force varies with the location of the protein in the exit tunnel [5], and hence when during translation the protein starts to fold. FPA has been used to map the cotranslational folding of more than 10 different proteins so far [12].
Cotranslational folding of the HemK N-terminal domain
In a pioneering study, Holtkamp et al. followed the cotranslational folding of the 73-residue long HemK-NTD (Fig. 2a) in real time using both FRET- and PET-based assays [4]. A subsequent, more detailed analysis [3] led to the identification of four folding intermediates (I-IV) in addition to the native state (V), which were proposed to correspond to the linear, stepwise addition of individual α-helices to a growing, compact core. Furthermore, it was concluded that the folding reaction is rate-limited by translation (which proceeds at a rate of ~4 codons s−1 in the in vitro translation system used [4]). A coarse-grained molecular dynamics study has also provided evidence for the cotranslational appearance of compact folding intermediates composed of the first three (~intermediate II) and first four (~intermediate III) α-helices of HemK-NTD, followed by folding into the native 5-helix bundle (V) at longer tether lengths [28].
In order to compare FPA to the FRET and PET assays, we recorded a force profile for HemK-NTD at ~5-residue resolution using the Escherichia coli SecM(Ec) AP, Fig. 2b (red curve; in order to be consistent with the previous publications, we measure L from the N-terminus of HemK-NTD to the C-terminal end of the AP in this case, Fig. 1b). Four peaks (A-D) are apparent in the force profile. Peak D (L ≈ 100-110 residues) coincides with the formation of the native state V as seen by FRET, and corresponds to a situation where the C-terminal end of HemK-NTD is 30-35 residues away from the ribosome PTC, close to the exit-port region where other proteins of similar size have been found to fold into their native state [6, 26]. Strikingly, peaks A-C in the force profile nicely match the folding intermediates I-III identified by real-time FRET and PET analysis [3], although peak B is just barely detectable above the background.
To test how mutations in the hydrophobic core of HemK-NTD affect the observed forces, we recorded a force profile for a mutant HemK-NTD in which Leu27 and Leu28 in helix H2 were simultaneously mutated to Ala (Fig. 2a, and blue curve in Fig. 2b). The L28A mutation reduces the in vitro denaturation temperature of HemK-NTD from 50°C to 30°C [4]. The [L27A+ L28A] mutation caused a strong reduction of the amplitude of peak D, but had only marginal effects on peaks A and C. Interestingly, the mutation led to a marked increase the amplitude of peak B. To confirm these effects, we also measured fFL values for a Leu7 to Asn mutation in helix H1 (Fig. 2a) for L values corresponding to peaks A-D. As seen in Fig. 2c, the L7N mutation behaves similarly to the [L27A+ L28A] mutation: fFL is strongly reduced for peak D, not affected for peaks A and C, and increased for peak B.
Finally, we also recorded an in vivo force profile by expression of the HemK-NTD constructs in E. coli cells. An HA tag was added to enable the detection of HemK-NTD translation products by immunoprecipitation. Guided by previous mutagenesis studies of a SecM AP [36], we placed the HA tag overlapping the SecM(Ec) arrest peptide, resulting in an HA-AP sequence that was of the same length and somewhat lower stalling strength than the SecM(Ec) AP used in the in vitro experiments (Supplementary Fig. S1). The HemK-NTD force profile obtained in vivo, Fig. 2b (green curve) is similar to the one obtained in vitro (Fig. 2b, red curve), although apparently lacking the low-amplitude peak B.
We conclude that FRET/PET, MD, and FPA all give similar a similar picture of the cotranslational folding of HemK-NTD, and all detect the formation of at least two folding intermediates as well as the native state. Mutation of residues in the hydrophobic core of HemK-NTD reduce the amplitude of peak D (corresponding to the native state) in the force profile as expected, but have little effect on peaks A and C, and even increase the amplitude of peak B, suggesting that the corresponding folding intermediates have less well-defined tertiary structures than the native fold.
Cotranslational folding of the FLN5 domain
The cotranslational folding of the Ig-like FLN5 filamin domain has previously been studied by NMR measurements on ribosome-attached nascent chains and of purified C-terminally truncated versions of the protein [16, 17]. The main conclusion from these studies is that FLN5 folds only when it has fully cleared the exit tunnel and is some distance away from the ribosome surface, at a tether length of 40-45 residues. This tether length is clearly longer than what is required for folding of a similar-sized Ig-like domain such as the I27 titin domain that folds at a tether length of ~35 residues, as determined by FPA and cryo-EM [6]. The simplest explanation for the long tether length required for cotranslational folding of FLN5 is that the ribosome surface destabilizes the folded state relative to the unfolded state [17], as has been observed for the Top7 protein when tethered to a ribosome [23].
Given that the NMR data indicate that FLN5 folds at a considerably longer tether length than I27, we were curious to see if this is also reflected in its force profile. The FLN5 force profile obtained with a mutant AP, SecM(Ec, S→K) (see Supplement 1 Materials and Methods), that is somewhat more resistant to pulling force than is SecM(Ec) is shown in Fig. 3, together with a curve derived from the NMR data showing averaged relative intensities of three 15N amide resonances arising from the unfolded FLN5 domain as a function of L [17]. In order to be consistent with the previous publications [16, 17], we measure L from the C-terminus of FLN5 to the C-terminal end of the AP (Fig. 1b). The force profile has two peaks: one at L = 35-47 residues that coincides with the main folding transition detected by NMR, and one at L ≈ 20 residues. The latter corresponds roughly to an initial drop in the NMR profile that has been suggested to reflect the formation of a marginally stable folding intermediate in which the C-terminal β-strand is disordered [16]. There is an apparently significant dip both in the force profile and the NMR profile at L = 43 residues; at present, we have no explanation for this precisely localized feature in the two profiles.
To validate that the main peak at L = 37-47 residues in the force profile reflects folding into the native state, we also analyzed an unstable mutant of FLN5, Y719E, using the weaker SecM(Ec) AP. As expected, the main peak is of significantly lower amplitude in this mutant; notably, however, the peak at L ≈ 20 residues is unaffected (Supplementary Fig. S2).
In conclusion, the main cotranslational folding transition of FLN5 to the native state is detected both by NMR and FPA, and a putative folding intermediate is also apparent in both data sets.
Summary
We have characterized the cotranslational folding of two small protein domains of different folds – the α-helical HemK-NTD and the β-rich FLN5 – by FPA, allowing us to compare FPA to three other techniques currently used to study cotranslational folding: real-time FRET, PET, and NMR. The results show broad agreement between FPA on the one hand, and the three biophysical techniques on the other. Previously, we have shown that FPA also agrees with results from on-ribosome pulse-proteolysis [12] and coarse-grained MD simulations [5, 6, 37].
It thus appears that all these techniques reflect the same basic process of cotranslational folding in similar ways. They should therefore be possible to use interchangeably, at least for proteins that fold fast relative to the rate of translation. For proteins that fold more slowly, methods such as NMR, PET, pulse-proteolysis and FPA that are based on an analysis of nascent chains stably tethered to the ribosome may indicate folding at somewhat shorter tether lengths than will methods such as real-time FRET, in which the nascent chain is being continuously lengthened during the assay.
Techniques that use stable tethering can potentially yield folding profiles with single-residue resolution, since the nascent chain can be lengthened one residue at a time. Real-time techniques, in contrast, average over a population of elongating ribosome-nascent chain complexes and further depend on an estimate of the local translation rate to convert from time to nascent-chain length, and therefore cannot achieve single-residue resolution. The different techniques thus have somewhat different limitations.
It is particularly interesting to note that all the techniques appear able to detect intermediates on the cotranslational folding pathway, not just the formation of the native state. In-depth analysis of cotranslational folding intermediates may shed further light on how the presence of the ribosome affects protein folding.
Supplementary Table 1
In vitro expression constructs of HemK: [HemK-amino terminal domain]-[HemK linker domain]-[Arrest Peptide]-[LepB derived C-terminal tail]. For in vivo expression the Arrest Peptide sequence is substituted to be GSYPYDVPDYAPIRAGP. Sequences where the Leu27,28Ala substitutions were made are underlined and those where the Leu7Asn was made are highlighted.
List of FLN5-FLN6 constructs [1]. All constructs were modified to include a LepB derived C-terminal tail subsequent to the critical proline residue in the SecM(Ec) AP. The ‘length’ corresponds to the number of residues between the C terminus of FLN5 to the critical proline residue at the C-terminal end of the SecM(Ec) AP. Non-folding controls with the Y719E mutation [1] were arrested with the SecM(Ec) AP, and were included for certain critical lengths used in the NMR study. Constructs 16, 11, and 1 were generated with only the stronger SecM(Ec, S→K) AP.
Acknowledgements
FLN5 constructs previously used for NMR studies were kindly provided by Lisa Cabrita and John Christodoulou, University College London. This work was supported by grants from the Knut and Alice Wallenberg Foundation, the Swedish Cancer Foundation, and the Swedish Research Council to GvH.