Modulation of Coiled-Coil Dimer Stability through Surface Residues

May 29, 2017 - Binding affinity, on the other hand, can also be affected by surface residues that face away from the dimerization interface. Here we s...
12 downloads 5 Views 3MB Size
Article pubs.acs.org/JACS

Modulation of Coiled-Coil Dimer Stability through Surface Residues while Preserving Pairing Specificity Igor Drobnak,†,§ Helena Gradišar,†,‡ Ajasja Ljubetič,† Estera Merljak,† and Roman Jerala*,†,‡ †

Department of Synthetic Biology and Immunology, National Institute of Chemistry, Hajdrihova 19, SI-1000 Ljubljana, Slovenia EN-FIST Centre of Excellence, Trg OF 13, SI-1000 Ljubljana, Slovenia



S Supporting Information *

ABSTRACT: The coiled-coil dimer is a widespread protein structural motif and, due to its designability, represents an attractive building block for assembling modular nanostructures. The specificity of coiled-coil dimer pairing is mainly based on hydrophobic and electrostatic interactions between residues at positions a, d, e, and g of the heptad repeat. Binding affinity, on the other hand, can also be affected by surface residues that face away from the dimerization interface. Here we show how design of the local helical propensity of interacting peptides can be used to tune the stabilities of coiled-coil dimers over a wide range. By designing intramolecular charge pairs, regions of high local helical propensity can be engineered to form trigger sequences, and dimer stability is adjusted without changing the peptide length or any of the directly interacting residues. This general principle is demonstrated by a change in thermal stability by more than 30 °C as a result of only two mutations outside the binding interface. The same approach was successfully used to modulate the stabilities in an orthogonal set of coiledcoils without affecting their binding preferences. The stability effects of local helical propensity and peptide charge are well described by a simple linear model, which should help improve current coiled-coil stability prediction algorithms. Our findings enable tuning the stabilities of coiled-coil-based building modules match a diverse range of applications in synthetic biology and nanomaterials.



INTRODUCTION

principles make de novo design of coiled-coil dimers relatively simple.8−14 Designed coiled-coil dimers are of great interest both as inhibitors of native protein−protein interactions15,16 and as building blocks for assembling larger bionanostructures.17−22 In both cases, specificity of coiled-coil pairing is of prime importance: either to inhibit one target among many natural coiled-coils or to bring together the correct parts of a selfassembling nanostructure. This is the motivation for developing orthogonal sets of coiled-coil peptides, where each peptide only binds to its specific partner and not to any other peptide in the set.8,10,11,13 Different applications also require different coiledcoil stabilities, and several approaches have been shown to increase CC stability: extending the length of the coiledcoil,14,23 increasing its overall helical propensity,24 introducing favorable electrostatic interactions,25−27 or covalently linking the peptides.28,29 However, stability is normally coupled to specificity and improving one tends to negatively affect the other.30 This is particularly important in orthogonal CC sets, as any modification to residues involved in the binding interface may change a peptide’s specificity and destroy the orthogonality of the entire set. We therefore sought an alternative means of modulating coiled-coil stability without affecting its length or specificity and without the need for covalent modifications.

Coiled-coil (CC) dimers are widely distributed structural motifs in natural proteins, with different functional roles, such as the regulation of the assembly of transcription factors,1 regulation of vesicular trafficking,2 and formation of scaffold/structural elements.2 The coiled-coil dimer as a protein structural element is quite well understood [reviewed at length in refs 3−5]. Two or more α-helices coil around each other so that their sidechains pack together much like a zipper. Supercoiling changes the pitch of the α-helices from 3.6 to 3.5 residues per turn, resulting in a repeat unit of 7 residues (one heptad). Residues are labeled a−g according to their position in the heptad. The characteristic coiled-coil motif consists of hydrophobic residues at positions a and d that face the opposing helix; these are flanked by charged residues at positions e and g. Binding stability and specificity are mainly determined by the burial of hydrophobic side chains at positions a and d, and by complementary electrostatic interactions between residues at positions e and g. This general CC pattern is compatible with a multitude of possible stoichiometries and relative chain orientations.6 Even minor differences in the binding interface (e.g., β-branching side chains or Asn at position a or d) can therefore dramatically affect the CC structure.7 On the other hand, residues at positions b, c, and f face away from the opposing helix and do not form interchain interactions (Figure 1a); they are therefore designated as surface residues. These © 2017 American Chemical Society

Received: March 9, 2017 Published: May 29, 2017 8229

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society

residue that does not form part of the binding interface.35 This replacement was positioned within the proposed trigger sequence and strongly decreased the predicted local helical propensity by removing an intramolecular salt bridge. Such effects of intramolecular interactions on helical propensity are readily predicted by the Agadir algorithm that incorporates terms for the pairwise (i, i+3) and (i, i+4) interactions.39,40 Given that pairs of surface residues at positions (c, f) and (b, f) are, respectively, 3 and 4 residues apart in the coiled-coil sequence, we can modulate the intrinsic helical propensity of a peptide by adding or removing salt bridges at these positions, leaving the binding interface intact (Figure 1a). Building on these principles, we designed several variants of the prototypical coiled-coil GCN4 peptide with mutations introduced at the b, c, or f sites (Figure 1, Table S1). As our starting point we used the variant GCN4-S that has been successfully used for the design of bionanostructures41 and has a Tm around 25 °C. Replacement of Arg22, which forms a salt bridge with Glu19 as part of the proposed trigger sequence, completely abolished the mutant (GCN4-SD) peptide’s ability to form a coiled-coil dimer, similar to earlier reports.35 Agadirbased prediction indicated that this replacement also strongly decreased the local helical propensity. However, we were able to reintroduce a strong local helical propensity region into the peptide by introducing a residue that is predicted to form an intrachain salt bridge at the opposite end of the peptide. The mutation S11R was designed to introduce a salt bridge between Arg11 and Glu7 (Figure S1), strongly increasing the predicted local helical propensity. Most importantly, this mutation completely recovered the peptide’s ability to form coiledcoils. In fact, the resulting peptide (GCN4-SR) was even more stable than the initial GCN4-S. Static light scattering (SECMALS) confirms that GCN4-SR is largely dimeric while GCN4-SD is monomeric at room temperature (Figure S2). This demonstrates that we have designed a new trigger-like stabilizing sequence using only minimal modifications of the peptide’s surface residues. Moreover, it demonstrates the power of our strategy for the modulation of coiled-coil stability. Engineering Trigger Sequences into Heterodimeric Coiled-Coils. Previous studies of trigger sequences31,33,34 have largely focused on homodimeric parallel coiled-coils, where the position of the local helical propensity regions necessarily coincides on both chains. Heterodimers introduce additional complexity, since the helical propensity of each binding partner can be modified independently. Starting from the P3S:P4S peptide pair from our previously published set of orthogonal coiled-coil heterodimers,10,41 we replaced several residues at positions b, c, and f in order to introduce a higher local helical propensity at the N- or C-terminal end of the peptide (peptides P3SHN and P4SHN, and P3SHC and P4SHC, respectively; Figure 2a; see Table S1 for sequences). SEC-MALS confirms that the new peptides still form heterodimers (Figure S3), while CD spectra and thermal denaturation scans (Figure 2b, Figure S4) demonstrate that the modifications increase the native helical content and stabilize the coiled-coil heterodimer by approximately 10 °C. Coiled-coil stability can, on the other hand, be reduced by introducing like-charged residues at b, c, and f sites (peptides P3SN and P4SN). This suggests that helical propensity is not the only determinant of stability, since the S and SN pairs have a similarly low helical propensity but exhibit different stabilities. Importantly, heterodimeric coiled-coils allow us to mix and match peptides with different helical propensities and charges.

Figure 1. Disruption and reconstitution of the trigger sequence in GCN4 strongly affects the coiled-coil stability. (a) Top-down schematic view of the interchain interactions in a parallel coiled-coil. Only surface residues at positions b, c, and f were modified in this study. An example is also shown of typical salt bridges, we introduce between residues (i, i + 3) or (i, i + 4). (b) Local helical propensity profiles of GCN4 variants predicted by Agadir39,40 with parameters T = 20 °C, pH 7.5, ionic strength = 0.17 M. (c) CD spectra measured at 20 °C and presented as mean residue ellipticity (MRE). (d) Thermal denaturation scans of 40 μM peptides, monitored by the CD signal at 222 nm.

It has been shown that the formation of a number of naturally occurring coiled-coils requires the presence of specific sequence motifs called “trigger sequences”.31,32 Trigger sequences have been defined as local sequence elements featuring a combination of increased helical propensity and a network of inter- and intramolecular salt bridges that enhance the local stability of the coiled-coil.33,34 Mutations of residues within the trigger sequences can strongly impair the formation of coiled-coil dimers.35 On the other hand, computational simulations have suggested that strengthening the trigger sequences should stabilize the coiled-coil36 and the stability of short CC fragments has been improved by optimizing electrostatic interactions.25,27 Coiled-coils have also been stabilized significantly by redesigning only the surface sites b, c, and f.37,38 It has further been shown that redesigning the electrostatic interaction networks can be used to guide supramolecular assembly of coiled-coils into larger structures.18 Based on these findings, we develop here a simple strategy for modulating coiled-coil stability. By introducing salt bridges between residues at b, c, and f sites, we can increase the local helical propensity and stabilize the coiled-coil. Just two mutations at surface positions can dramatically stabilize or destabilize a coiled-coil. By avoiding modifications of the binding interface, we produced coiled-coil pairs that have a wide range of stabilities but maintain their binding specificity and orthogonality. Our results also demonstrate that intrachain electrostatic interactions should be taken into account in order to improve the accuracy of coiled-coil stability prediction.



RESULTS Reengineering the Trigger Sequence. It has been previously reported that the Tm of the coiled-coil dimer GCN4-p1 can be decreased by 18 °C by replacing a single 8230

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society

Figure 2. Variation of surface residues modulates stability of coiled-coil heterodimers. (a, c, e) Helical propensity profiles predicted by Agadir.39,40 (b, d, f) Melting temperatures (Tm) of different peptide combinations, at a final concentration of 20 μM for each peptide, determined from the temperature dependence of the CD signal at 222 nm (Figure S4). Combining peptides with different helical propensities produces a range of stabilities spanning 35 °C (d). Mutations at surface sites modulate stability independently of the relative position of high helical propensity regions in P5:P6 variants (f).

increased helical propensity regions of both peptides coincide, show nearly identical stabilities as mixed pairs P5SHC:P6SHN and P5SHN:P6SHC, where the helical propensities do not coincide (Figure 2e,f; Figure S4). This suggests that, at least for CC heterodimers composed of four heptads, regions of high local helical propensity have a strong effect on stability regardless of their relative positions in the two sequences. Modifying Surface Sites Preserves Orthogonality of Binding. Previously we designed a set of orthogonal coiled-coil heterodimers, where specificity was achieved based on the electrostatic complementarity between the e and g sites and matched Asn residues at a positions of the selected heptads on opposing peptide chains.10 We wanted to investigate if variations in the helical propensity that affect the coiled-coil dimer stability affect the orthogonality. We have shown above that we can tune coiled-coil stability by modifying only positions b, c, and f. Since this approach leaves all intermolecular interfaces intact, we anticipated that the orthogonality between peptide pairs should be retained. To test this hypothesis, we introduced modifications into the b, c, and f sites of the two orthogonal coiled-coil pairs, P3:P4 and P5:P6.10 To test how this affects the orthogonality of

In Figure 2c,d we present the results of mixing different variants of peptides based on the previously used P5S:P6S pair.10,41 These include the negatively charged SN variant, SHN with a region of increased helical propensity in its N-terminal half, and SHCQ/SHNR that combine increased helical propensity with a zero net charge (see Table S1 for sequences). As expected, mixing one peptide from a stable pair (SHN or SHNR) with a partner from a less stable pair (S or SN) resulted in a coiled-coil of intermediate stability. A nearly continuous range of stabilities was obtained using this approach, spanning 35 °C between the most stable and the least stable pair. Electrostatic interactions play an important role in these stability shifts, as demonstrated by the effect of salt concentration: P5:P6 variants can be either stabilized or destabilized by the addition of salt, depending on their surface residues (Figure S5). To investigate the effect of the relative positions of high helical propensity regions on coiled-coil dimer formation, we compared the CD spectra and stabilities of P5:P6 variants, where high helical propensity was introduced either into the Nterminal or into the C-terminal heptads (SHN and SHC variants, respectively; Figure 2g; see Table S1 for sequences). Peptide pairs P5SHC:P6SHC and P5SHN:P6SHN, where the 8231

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society

Figure 3. CD spectra of different pairs of peptides from the orthogonal P3:P4, P5:P6 set, with or without a region of increased helical propensity at their N-terminus (SHN and S variants, respectively). The expected (cognate) combinations of P3:P4 and P5:P6 pairs have their spectra highlighted in black. They exhibit higher helical content than mismatched pairs (gray spectra), suggesting that the increase of helical propensity through b, c, f sites does not disrupt orthogonality. Inset shows a heatmap of the CD signal intensity at 222 nm, with larger negative values indicating a higher helical content, characteristic of coiled-coil formation.

S7; peptide sequences in Table S1). We then searched for statistically significant correlations between these variables (Figure 4). We found that once the choice of binding interface is accounted for, the remaining variation in stability can best be explained using a combination of the number of residues above a threshold helical propensity and the product of net peptide charges (proportional to long-range electrostatic repulsion between the peptides). The resulting model predicts the Tm to within ±4.1 °C root-mean-square error (RMSE), estimated using a 5-fold cross-correlation analysis (Figure 4e; see Figure S8 for an overview of the validation procedure). This is compared to a RMSE of 12.0 °C obtained for a negative control model fitted to data with scrambled stability values. The mean and maximum of the helical propensity plots of the constituent peptides also correlate with the coiled-coil stability, but do not provide the same predictive power (Figures S9 and S10). As expected from the results presented in Figure 2e,f, we did not find any significant correlation between coiled-coil stability and the overlap of the helical propensity regions on the two interacting peptides (Figures S9 and S10). The stabilities of our coiled-coil set can also be predicted well using a linear combination of helical propensities, net charges, and CC stability predictions from the sequence-dependent bCIPA algorithm8,42 (Figure 4c,d,e). bCIPA alone predicts the experimental stabilities of the investigated coiled-coil dimers rather poorly (R2 = 0.19, cross-validation RMSE = 9.9 °C compared to 10.6 °C for the scrambled control), but predictions improved considerably when the Agadir-based helical propensity and net charge were taken into account

peptide pairs, we measured the CD spectra for a matrix of all combinations of peptides with matching (P3:P4 and P5:P6) or mismatched (P3:P5, P3:P6, P4:P5, P4:P6) binding interfaces and with (SHN) or without (S) a region of high local helical propensity (Figure 3). Only peptide pairs with matching binding interfaces (black traces in Figure 3) produce the characteristic α-helical CD spectra indicative of coiled-coil formation. This demonstrates that modifying helical propensity and coiled-coil stability via b, c, f sites does not affect the orthogonality between the P3:P4 and P5:P6 peptide pairs. Note that although the P5S:P6S pair is only marginally stable at 20 °C, its CD spectrum is still closer to the characteristic coiledcoil compared to the spectra of its neighboring noncognate pairs (P3S:P5S, P3S:P6S, P4S:P5S, P4S:P6S). The differences become more obvious when the spectra are overlaid (Figure S6a). The only mismatched pair that unexpectedly displays some helical structure is P6S:P6SHN. However, its helical content is weaker compared to the corresponding expected pairs, P5S:P6SHN and P5SHN:P6S (overlaid in Figure S6b). Since P6S and P6SHN have the same binding interface, they are also unlikely to be used as part of the same orthogonal set, so their potential to weakly bind to one another should not present many difficulties in practical applications. Predicting the Effect of b, c, f Sites on Coiled-Coil Stability. Finally, we investigated the quantitative correlations between helical propensity and coiled-coil stability. We measured the CD spectra and thermal stabilities of 20 peptide pairs with different binding interfaces (GCN4-S, P3S:P4S, or P5S:P6S), degrees of helical propensity, and net charge (Figure 8232

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society

or denaturing cosolvents. They can be made more stable at the required working conditions or more responsive to environmental changes, depending on the requirements of the application. Here we demonstrate a strategy to modulate the stability of coiled-coil dimers, via design of the local helical propensity by introducing or removing intramolecular (i, i + 3) and (i, i + 4) salt bridges. With this approach, a few point mutations can shift the Tm by at least 35 °C (compare GCN4-SD and GCN4-SR in Figure 1 or P5SN:P6SN and P5SHCQ:P6SHNR in Figure 2). This is comparable to the ΔTm of 36 °C previously achieved by introducing multiple inter- and intramolecular salt bridges through mutations at sites e and f26 and not too far from the ΔTm of 46 °C obtained by adding an entire heptad to the coiled-coil.14 The main advantage of surface redesign is that it can achieve such stability shifts without changing the coiled-coil dimerization interface. Although a previous study has shown that it is possible for b, c, f residues to affect specificity by interacting with e and g sites,38 we have not observed any such effects in our case. The specificity of pairing between different peptides is thus unaffected, allowing us to safely modify the stability of peptide pairs in our orthogonal set of coiled-coils (Figure 3). The energetic reason behind this is that modifying helical propensity simultaneously changes the stabilities of all possible pairs a given peptide can form. Cognate and noncognate CC pairs are (de)stabilized by the same amount, so the difference between them (the stability gap, ΔΔG°) remains the same.45 This means we can shift the stability gap between cognate and noncognate pairs up or down on the temperature scale and align it with the requirements of a particular application, without the need for time-consuming development and testing of new orthogonal peptide sets. Mixing and matching the interacting partners in a heterodimer enables an almost continuous tuning of coiled-coil stability. An intriguing future development could be to introduce metal ion- or pHdependence of coiled-coil stability via b, c, f sites, establishing an allosteric dependency into coiled-coil assemblies and making them responsive to environmental changes without the need to disrupt their carefully designed binding interfaces. The inspiration for our approach came from trigger sequences identified in natural coiled-coils and characterized by a combination of increased local helical propensity and efficient electrostatic interactions.31,33,34 We have shown that it is possible to de novo engineer the stabilizing effect of trigger sequences by introducing intrachain salt bridges (Figure 1). Furthermore, our results show that alignment of trigger sequence-like elements is not required for efficient coiled-coil formation and that the final stability does not depend on their relative positions (Figure 2e,f). This would seem to contradict the previously proposed function of trigger sequences as partially prefolded elements that properly align the two chains of the coiled-coil and nucleate its folding.31 However, we need to point out that both our study and a previous one questioning the necessity of trigger sequences26 were conducted on short coiled-coils whose formation is not kinetically limited. This may not be the case with the much longer natural coiled-coils, where correct in-register alignment of trigger sequences could indeed be important for steering the coiled-coil onto a kinetically efficient path toward its energetic minimum. The mechanism by which surface residues stabilize coiledcoils is likely similar to that observed in other intrinsically disordered proteins that fold upon binding.46,47 At least two

Figure 4. Prediction of coiled-coil melting temperatures (Tm) is significantly improved when local helical propensity and net charge are taken into account. (a−d) Correlations between model-predicted and experimentally measured Tm values for all 20 coiled-coils of our data set. All models predict the Tm as a linear combination of several parameters of each coiled-coil: interface type, type of dimerization interface (GCN4-S, P3S:P4S, or P5S:P6S); helicity, number of residues with an Agadir-predicted helical propensity above the threshold level of 11%; charge, product of net charges of the two peptides; bCIPA, Tm predicted by the bCIPA algorithm.8,42 R2 is the coefficient of determination. (e) Root mean squared errors in the Tm predictions of each model, calculated using 5-fold cross-validation. The calculation was performed using a range of different helical propensity thresholds for the helicity parameter, shown on the x-axis.

(R2 = 0.88, cross-validation RMSE = 4.2 °C compared to 11.6 °C for the scrambled control). While bCIPA already makes use of the information on the helical propensity of individual residues, our results demonstrate that the prediction can be improved further by taking into account (i, i + 3) and (i, i + 4) interactions that strongly affect the helical propensity of peptides, as well as net peptide charges that can affect peptide association.



DISCUSSION Coiled-coils are of great interest to bionanotechnology as the simplest and best-understood protein dimerization domains. They are commonly used in facilitating specific noncovalent association between proteins1 and have been applied to scaffolding of enzymes,43 assembly of multimeric complexes,17,18,20,44 and construction of novel protein folds composed of concatenated coiled-coil dimer-forming segments.21 Modulation of coiled-coil stability is very important for extending this range of applications even further. By modulating the stabilities of their constituent coiled-coils, entire assemblies can be made more or less sensitive to temperature 8233

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society

equipped with a Peltier thermal control block (Melcor, NJ, now part of Laird Technologies). Final concentrations were 20 μM for each peptide in heterodimer pairs and for individual peptides. In the case of homodimerizing peptides, the final concentration was 40 μM monomer, which corresponds to 20 μM dimer. All experiments were performed in a quartz cuvette with a 1 mm path length. CD spectra were measured every 1 nm from 200 to 260 nm with a bandwidth of 1 nm and 0.5 s integration time. The reported spectra are averages of three replicates measured on the same sample. Temperature denaturation scans were performed by measuring the CD signal at 222 nm. Temperature was varied by setting the thermal control block to heat from 0 to 100 °C in 1 °C increments at a rate of 1 °C/min. The actual sample temperatures reported here were measured directly in the cuvette using a temperature probe. At the end of the thermal denaturation scan, another CD spectrum was measured at the final temperature. Finally, the sample was cooled back down to 20 °C rapidly, and another CD spectrum was measured. For all peptides, the spectra after refolding matched the ones taken before the temperature scan very well, indicating that the coiled-coils can refold quickly after thermal denaturation. Size-exclusion chromatography coupled to static light scattering (SEC-MALS). SEC-MALS experiments were performed on the Waters e2695 HPLC system, coupled with a 2489 UV detector (Waters, MA) and a Dawn8+ multiple-angle light scattering detector (Wyatt, CA). Peptide samples of 100−200 μM concentrations were filtered using Durapore 0.1 μm centrifuge filters (Merck Millipore, Ireland) before being injected onto the Biosep SEC-S2000 column (Phenomenex, CA). Injection volume was 50 μL, and the flow rate was 0.5 mL/min. Sample run time was 50 min with typical retention times around 20−35 min. Data analysis was carried out using Astra 7.0 software (Wyatt, CA). Model Fitting and Cross-Validation. Experimental thermal denaturation curves were fitted with a two-state equilibrium model, AB ↔ A + B for heterodimers and N2 ↔ 2D for homodimers. The fitting procedure was based on minimizing χ2 (the sum of squared deviations of model from experimental data) using the Nelder−Mead algorithm,51 implemented in a custom-written C++ program. The procedure has been described in detail previously.52 We tested multiple linear models for predicting the experimental Tm from the sequence of its constituent peptides. All tested models are listed in Table S2. Each model was first fitted to the entire experimental data set and the coefficient of determination (R2) was calculated as the fraction of the variance in experimental Tm values that is explained by the linear model. As a more rigorous test of the model’s predictive power, we performed 5-fold cross-correlation analysis. The data set was split into five segments of four data points (four CC dimers). One segment at a time was left out of the data set, and the model was trained using the remaining 16 points. Then the errors in the predicted Tm values for the left-out CC dimers were calculated. This was repeated for each segment of data, so that prediction errors were calculated for all points in the data set; the root-mean-square of these errors (RMSE) is the measure of how well the model can predict the Tm values of peptides that had not been used to train the model (a lower score is better). The cross-validation procedure was also repeated on 20 scrambled data sets, where the experimental Tm values were randomly permuted between the CC pairs. The mean crossvalidation RMSE of these 20 scrambled data sets represents a negative control: the kind of RMSE we can expect for fitting each model to data that has no real underlying relationship. The entire procedure is summarized in Figure S6. The linear model fitting and cross-validation were implemented as a custom-written Python script, making use of the scikit-learn package version 0.16,53 specifically the predictor class linear_model.LinearRegression and the function cross_validation.cross_val_score().

effects can contribute to the observed stabilization. The higher helical propensity can serve to partially preform the peptide’s structure in its monomeric state. This way, the peptide loses less conformational freedom upon binding, which results in a lower entropic penalty and stronger binding. The second effect is simply the result of additional favorable interactions, such as salt bridges, that stabilize the final structure but are not present in the unbound state: they are not sufficiently strong to constrain the protein backbone themselves, but they do contribute once the protein is folded. The CD spectra of free peptides (gray spectra in Figure 2b,e,h; diagonal in Figure 3) are characteristic of largely unstructured chains, suggesting that the second mechanism is likely to play an important role. Nevertheless, peptides with an increased helical propensity also display some helical character in their monomeric forms, which could contribute through the first mechanism. Our results are also relevant to improving predictions of coiled-coil stability based on their amino acid sequence. Current CC stability prediction algorithms understandably focus mostly on residues at the binding interface.8,12,42,45,48 While the helical propensity of individual amino acid residues has also been taken into account, 30,45 intramolecular interactions featuring the b, c, and f sites have so far been largely neglected. Yet, it is known that (i, i + 3) and (i, i + 4) interactions are important to a peptide’s helical propensity40 and to coiled-coil stability.25,26,35,38 Figure 4 demonstrates that we can make good use of the calculated helical propensity to predict the differences in the stability between coiled-coils with the same binding interface. Based on the current peptide set, the best way to utilize this information is to consider the number of residues with high helical propensity. Additionally, results in Figure 2 and Figure 4 suggest that the net charge attraction or repulsion is also a relevant secondary descriptor of coiled-coil stability. This comes with the caveat that the data set in this study is quite limited when compared to the full range of possible coiled-coil sequences. A truly general predictor of CC stabilities will need to be parametrized and validated on a much larger data set, containing both CCs that have different binding interfaces and those with the same interface but differing at the b, c, f sites. In summary, we provide a minimally disruptive strategy for tuning the stability of coiled-coil dimers that is compatible with existing orthogonal coiled-coil dimer sets. This should facilitate the design of new coiled-coil-based assemblies with stabilities adapted to the requirements of specific applications. This study also underscores the importance of considering surface residues in making theoretical predictions of coiled-coil stabilities.



METHODS

Reagents and Buffers. The peptides used in this study were purchased from Proteogenix (France). The N-termini of the peptides were protected by acetylation and the C-termini by amidation. The peptides were dissolved to a stock concentration of approximately 5 mg/mL in deionized water. NH4Cl was added (at a final concentration of 1 mg/mL) to improve solubility if needed. The working buffer for most experiments was 20 mM Tris, pH 7.5, with 150 mM NaCl and 1 mM TCEP as reducing agent for peptides that contain cysteines. The exception is GCN4 variant peptides that displayed higher stability in the absence of NaCl and do not contain cysteines; they were measured in 20 mM Tris, pH 7.5 without additives. Peptide concentrations were determined from the measured absorbance at 280 nm and extinction coefficients calculated using the ProtParam web tool.49,50 Circular Dichroism (CD) Spectroscopy. CD measurements were performed on the ChiraScan instrument (Applied Photophysics, UK),



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b01690. 8234

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society



(15) Strauss, H. M.; Keller, S. In Protein-Protein Interactions as New Drug Targets; Klussmann, E., Scott, J., Eds.; Springer: Berlin, Heidelberg, 2008; Vol. 186, pp 461−482. (16) Rezaei Araghi, R.; Keating, A. E. Curr. Opin. Struct. Biol. 2016, 39, 27−38. (17) Wagner, D. E.; Phillips, C. L.; Ali, W. M.; Nybakken, G. E.; Crawford, E. D.; Schwab, A. D.; Smith, W. F.; Fairman, R. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 12656−12661. (18) Ryadnov, M. G. Angew. Chem., Int. Ed. 2007, 46, 969−972. (19) Apostolovic, B.; Danial, M.; Klok, H.-A. Chem. Soc. Rev. 2010, 39, 3541−3575. (20) Fletcher, J. M.; Harniman, R. L.; Barnes, F. R. H.; Boyle, A. L.; Collins, A.; Mantell, J.; Sharp, T. H.; Antognozzi, M.; Booth, P. J.; Linden, N.; Miles, M. J.; Sessions, R. B.; Verkade, P.; Woolfson, D. N. Science 2013, 340, 595−599. (21) Gradišar, H.; Božič, S.; Doles, T.; Vengust, D.; HafnerBratkovič, I.; Mertelj, A.; Webb, B.; Šali, A.; Klavžar, S.; Jerala, R. Nat. Chem. Biol. 2013, 9, 362−366. (22) Thomson, A. R.; Wood, C. W.; Burton, A. J.; Bartlett, G. J.; Sessions, R. B.; Brady, R. L.; Woolfson, D. N. Science 2014, 346, 485− 488. (23) Litowski, J. R.; Hodges, R. S. J. Pept. Res. 2001, 58, 477−492. (24) Zitzewitz, J. A.; Ibarra-Molero, B.; Fishel, D. R.; Terry, K. L.; Matthews, C. R. J. Mol. Biol. 2000, 296, 1105−1116. (25) Burkhard, P.; Ivaninskii, S.; Lustig, A. J. Mol. Biol. 2002, 318, 901−910. (26) Lee, D. L.; Lavigne, P.; Hodges, R. S. J. Mol. Biol. 2001, 306, 539−553. (27) Burkhard, P.; Meier, M.; Lustig, A. Protein Sci. 2000, 9, 2294− 2301. (28) Moran, L. B.; Schneider, J. P.; Kentsis, A.; Reddy, G. A.; Sosnick, T. R. Proc. Natl. Acad. Sci. U. S. A. 1999, 96, 10699−10704. (29) Wuo, M. G.; Mahon, A. B.; Arora, P. S. J. Am. Chem. Soc. 2015, 137, 11618−11621. (30) Grigoryan, G.; Reinke, A. W.; Keating, A. E. Nature 2009, 458, 859−864. (31) Kammerer, R. A.; Schulthess, T.; Landwehr, R.; Lustig, A.; Engel, J.; Aebi, U.; Steinmetz, M. O. Proc. Natl. Acad. Sci. U. S. A. 1998, 95, 13419−13424. (32) Frank, S.; Lustig, A.; Schulthess, T.; Engel, J.; Kammerer, R. A. J. Biol. Chem. 2000, 275, 11672−11677. (33) Burkhard, P.; Kammerer, R. A.; Steinmetz, M. O.; Bourenkov, G. P.; Aebi, U. Structure 2000, 8, 223−230. (34) Steinmetz, M. O.; Jelesarov, I.; Matousek, W. M.; Honnappa, S.; Jahnke, W.; Missimer, J. H.; Frank, S.; Alexandrescu, A. T.; Kammerer, R. A. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 7062−7067. (35) Kammerer, R. A.; Jaravine, V. A.; Frank, S.; Schulthess, T.; Landwehr, R.; Lustig, A.; García-Echeverría, C.; Alexandrescu, A. T.; Engel, J.; Steinmetz, M. O. J. Biol. Chem. 2001, 276, 13685−13688. (36) Chapagain, P. P.; Liu, Y.; Gerstman, B. S. J. Chem. Phys. 2008, 129, 175103. (37) Dahiyat, B. I.; Benjamin Gordon, D.; Mayo, S. L. Protein Sci. 1997, 6, 1333−1337. (38) Kaplan, J. B.; Reinke, A. W.; Keating, A. E. Protein Sci. 2014, 23, 940−953. (39) Lacroix, E.; Viguera, A. R.; Serrano, L. J. Mol. Biol. 1998, 284, 173−191. (40) Muñoz, V.; Serrano, L. Nat. Struct. Biol. 1994, 1, 399−409. (41) Ljubetič, A.; Lapenta, F.; Gradišar, H.; Drobnak, I.; Aupič, J.; Strmšek, Ž .; Lainšcě k, D.; Hafner-Bratkovič, I.; Majerle, A.; Benčina, M.; Butina, K.; Pisanski, T.; Ć irković Veličković, T. A. R.; DjinovićCarugo, K.; Jerala, R. 2017, submitted for publication. (42) Hagemann, U. B.; Mason, J. M.; Müller, K. M.; Arndt, K. M. J. Mol. Biol. 2008, 381, 73−88. (43) Minten, I. J.; Claessen, V. I.; Blank, K.; Rowan, A. E.; Nolte, R. J. M.; Cornelissen, J. J. L. M. Chem. Sci. 2011, 2, 358−362. (44) Noble, J. E.; De Santis, E.; Ravi, J.; Lamarre, B.; Castelletto, V.; Mantell, J.; Ray, S.; Ryadnov, M. G. J. Am. Chem. Soc. 2016, 138, 12202−12210.

Sequences and net charges of tested peptides, linear Tm prediction models, molecular models of GCN-S, GCNSD, and GCN-SR, multimerization state of GCN4 variants and heteromeric coiled-coil pairs with increase helical propensity, raw data showing that variation of surface residues modulates stability of coiled-coil heterodimers, effect of salt on the stabilities of several P5:P6 variants, CD spectra at 20 °C for related peptide pairs, helical propensities, CD spectra, and thermal denaturation profiles for all tested peptide pairs, and comparison of different Tm prediction models (PDF)

AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Roman Jerala: 0000-0002-6337-5251 Present Address §

I.D.: Lek d. d., Kolodvorska 27, SI-1234 Mengeš, Slovenia.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This project was financed by Slovenian Research Agency (program no. P4-0176, projects N4-0037, L4-6812, J4-5528), by a grant from the ICGEB (CRP/SLO14-03), and by the Slovenian Ministry of Education, Science, and Sport through ERANET SynBio project Bioorigami (ERASYNBIO1-006). Contribution of the COST actions CM-1306 and CM-1304 is acknowledged.



REFERENCES

(1) Robson Marsden, H.; Kros, A. Angew. Chem., Int. Ed. 2010, 49, 2988−3005. (2) Truebestein, L.; Leonard, T. A. BioEssays 2016, 38, 903−916. (3) Lupas, A. N.; Gruber, M. In Fibrous Proteins: Coiled-Coils, Collagen and Elastomers; Parry, D., Squire, J., Eds.; Advances in Protein Chemistry; Academic Press: San Diego, CA, 2005; Vol. 70, pp 37−78. (4) Mason, J. M.; Müller, K. M.; Arndt, K. M. In Protein Engineering Protocols; Arndt, K. M., Müller, K. M., Eds.; Methods in Molecular Biology; Humana Press: Totowa, NJ, 2007; Vol. 352, pp 35−70. (5) Woolfson, D. N. In Fibrous Proteins: Coiled-Coils, Collagen and Elastomers; Parry, D., Squire, J., Eds.; Advances in Protein Chemistry; Academic Press: San Diego, CA, 2005; Vol. 70, pp 79−112. (6) Moutevelis, E.; Woolfson, D. N. J. Mol. Biol. 2009, 385, 726−732. (7) Harbury, P.; Zhang, T.; Kim, P.; Alber, T. Science 1993, 262, 1401−1407. (8) Crooks, R. O.; Baxter, D.; Panek, A. S.; Lubben, A. T.; Mason, J. M. J. Mol. Biol. 2016, 428, 385−398. (9) Fletcher, J. M.; Boyle, A. L.; Bruning, M.; Bartlett, G. J.; Vincent, T. L.; Zaccai, N. R.; Armstrong, C. T.; Bromley, E. H. C.; Booth, P. J.; Brady, R. L.; Thomson, A. R.; Woolfson, D. N. ACS Synth. Biol. 2012, 1, 240−250. (10) Gradišar, H.; Jerala, R. J. Pept. Sci. 2011, 17 (2), 100−106. (11) Negron, C.; Keating, A. E. J. Am. Chem. Soc. 2014, 136, 16544− 16556. (12) Potapov, V.; Kaplan, J. B.; Keating, A. E. PLoS Comput. Biol. 2015, 11, e1004046. (13) Reinke, A. W.; Grant, R. A.; Keating, A. E. J. Am. Chem. Soc. 2010, 132, 6025−6031. (14) Thomas, F.; Boyle, A. L.; Burton, A. J.; Woolfson, D. N. J. Am. Chem. Soc. 2013, 135, 5161−5166. 8235

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236

Article

Journal of the American Chemical Society (45) Mason, J. M.; Schmitz, M. A.; Müller, K. M.; Arndt, K. M. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 8989−8994. (46) Higo, J.; Nishimura, Y.; Nakamura, H. J. Am. Chem. Soc. 2011, 133, 10448−10458. (47) Wright, P. E.; Dyson, H. J. Curr. Opin. Struct. Biol. 2009, 19, 31− 38. (48) Grigoryan, G.; Keating, A. E. J. Mol. Biol. 2006, 355, 1125− 1142. (49) Gasteiger, E.; Hoogland, C.; Gattiker, A.; Duvaud, S.; Wilkins, M. R.; Appel, R. D.; Bairoch, A. In The proteomics protocols handbook; Walker, J. M., Ed.; Humana Press: Totowa, NJ, 2005; pp 571−607. (50) Pace, C. N.; Vajdos, F.; Fee, L.; Grimsley, G.; Gray, T. Protein Sci. 1995, 4, 2411−2423. (51) Press, W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P. Numerical Recipes in C++: The Art of Scientific Computing, 2nd ed.; Cambridge University Press: New York, NY, 2002. (52) Drobnak, I.; Vesnaver, G.; Lah, J. J. Phys. Chem. B 2010, 114, 8713−8722. (53) Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V.; Vanderplas, J.; Passos, A.; Cournapeau, D.; Brucher, M.; Perrot, M.; Duchesnay, É. J. Mach. Learn. Res. 2011, 12, 2825−2830.

8236

DOI: 10.1021/jacs.7b01690 J. Am. Chem. Soc. 2017, 139, 8229−8236