The Role of Backbone Hydrogen Bonds in the Transition State for Protein Folding of a PDZ Domain

Backbone hydrogen bonds are important for the structure and stability of proteins. However, since conventional site-directed mutagenesis cannot be applied to perturb the backbone, the contribution of these hydrogen bonds in protein folding and stability has been assessed only for a very limited set of small proteins. We have here investigated effects of five amide-to-ester mutations in the backbone of a PDZ domain, a 90-residue globular protein domain, to probe the influence of hydrogen bonds in a β-sheet for folding and stability. The amide-to-ester mutation removes NH-mediated hydrogen bonds and destabilizes hydrogen bonds formed by the carbonyl oxygen. The overall stability of the PDZ domain generally decreased for all amide-to-ester mutants due to an increase in the unfolding rate constant. For this particular region of the PDZ domain, it is therefore clear that native hydrogen bonds are formed after crossing of the rate-limiting barrier for folding. Moreover, three of the five amide-to-ester mutants displayed an increase in the folding rate constant suggesting that the hydrogen bonds are involved in non-native interactions in the transition state for folding.


Introduction
Protein domains usually fold in a highly co-operative manner with concomitant formation of hundreds of non-covalent bonds. The transition state for a typical protein folding reaction looks like a distorted version of the native state as determined from experimental W value analyses and molecular dynamics simulations [1]. Most experimental analyses of folding transition states use the strategy of truncating side-chains of amino acid residues and probing the effect on protein folding kinetics as well as the stability at equilibrium (W value analysis) [2]. This approach provides information on the relative energetics of the interactions of the side-chains in the transition and ground states. However, hydrogen bonds from backbone amides also play important roles in the folding and stability of proteins although their contribution is difficult to evaluate and their contribution is currently debated [3]. However, by introducing amide-to-ester or amide-to-thioether mutations in the backbone of the protein, it is possible to assess the contribution of these hydrogen bonds to stability [4] and folding kinetics [5,6]. Only very few proteins have been subjected to this approach, for example WW domain variants, for which the role of hydrogen bonds for overall stability as well as folding mechanism was assessed [5,6].
The PSD-95/Discs large/ZO-1 (PDZ) domains make up a family of globular adaptor protein domains, involved in signalling and scaffolding [7]. PDZ domains bind peptide segments of target proteins, often at the C-terminus [8][9][10]. This protein family has also served as a model system for detailed studies of protein folding [11][12][13][14][15]. These studies include detailed models of the transition states of folding characterized by conventional W-value analyses [16][17][18]. In the present paper, we use amide-to-ester mutations to investigate the formation of hydrogen bonds in the transition state for folding of the b-sheet formed by b2 and b3 strands in the second PDZ domain of PSD-95 (PSD-95 PDZ2) (Fig. 1).
This b-sheet is involved in the physiological function of PDZ domains by binding to disordered C-termini of protein ligands though backbone-backbone interactions thus forming an extended b-sheet [19]. We also probed putative backbone hydrogen bonds in the adjacent important carboxylate-binding loop, which is fully preserved among PDZ domains and essential for coordinating the C-terminal carboxylate of the protein ligands. Our data show that backbone hydrogen bonds contribute to the overall stability and form late in the folding reaction. Interestingly, some backbone hydrogen bonds appear to form non-native interactions in the transition state that slow down the folding reaction, suggesting that a certain degree of frustration [20] is present for formation of this b-sheet, which form part of the binding pocket for PDZ domains.

Protein constructs
The PSD-95 PDZ2 construct was similar to the one used in earlier folding studies [13,15], containing residues 155-249 of human PSD-95 including a mutation, Y190W in b3 to probe folding by fluorescence spectroscopy. To make amide-to-ester substitutions in the backbone, an additional mutation, (V178C) was introduced. Thus, the pseudo wild type in the folding experiments in this paper is the double mutant of PSD-95 PDZ2, V178C/Y190W. Semisynthesis of the PSD-95 PDZ2 V178C/ Y190W amide-to-ester mutants was performed as previously described [21].

Equilibrium and kinetic experiments
All experiments were performed at 25uC in 50 mM potassium phosphate, pH 7.0, and 0.4 M sodium sulphate. Equilibrium denaturation of each of the PSD-95 PDZ2 mutants was performed by monitoring the change in fluorescence at 340 nm (excitation wavelength = 280 nm) at increasing concentration of urea. The data displayed the typical sigmoidal shape of an apparent two-state transition and was fitted to the appropriate equation for solvent denaturation [22]. The most destabilized mutants did not give a well-defined native baseline in the experiments but the mid-point of the transition was well defined and fitting of the data to a constrained m D-N value was used to obtain DG D-N .
Kinetic experiments were performed in a stopped flow instrument (Applied Photophysics SX-17, upgraded to SX-20, Leatherhead, UK). For unfolding experiments, protein-buffer solutions were mixed with urea-buffer at different final concentrations of urea. For refolding experiments, protein-buffer-urea solutions were mixed with buffer-urea. Excitation was at 280 nm and mono-phasic kinetic traces were monitored using a 320 nm cut-off filter. Observed rate constants for (un)folding were measured over a wide range of urea concentration (0-8 M) and analysed as described in the following section.

Analysis of kinetic data
The folding data of the PDZ variants were analysed using an equation, which takes into account a change of rate-limiting step manifested as a kink in the refolding arm of the chevron plot: where RT is 0.59 kcal mol 21 at 25uC and [U] is the concentration of urea. The total kinetic m-value m TOT is related to the sum of the slopes of log k obs versus [U] [23]. For example, in a system where we assume that the denatured state is always in fast equilibrium with the folded state, without accumulation of intermediates (twostate), m TOT is the sum of the slopes of log k F2 and log k U2 versus urea concentration. See Hultqvist et al. [15] and Calosci et al. [16] for details on the equation used in the kinetic analysis.
Two W values could be calculated for the present data set, one for the very early first transition state, TS1: WT /k F1 mut )/DDG D-N , and one for the second transition state, TS2: W TS2 = DDG D-T2 /DDG D-N = 1-DDG T2-N / DDG D-N = 1-RT ln (k U2 mut /k U2 WT )/DDG D-N . DDG D-N is the change upon mutation in free energy between the denatured state and the native state, and was obtained from equilibrium denaturation experiments with a constrained m D-N value. Analysis of kinetic data was performed using Prism (GraphPad Software, Inc.) and Kaleidagraph (Synergy Software).

Results
We designed and synthesized five different amide-to-ester mutants of PSD-95 PDZ2, L170l, G171c, F172Q, I174i and G176c as described [21] (see Powers et al. [24] for nomenclature of amide-to-ester mutations). In general, such backbone mutations remove a hydrogen bond formed by the NH of the amide and destabilize the hydrogen bond formed by the amide carbonyl oxygen [3,24]. In the case of PSD-95 PDZ2 [25,26], the amide NH of Phe172, Ile174 and Gly176 are posed for binding a peptide ligand and not directly involved in intra-domain hydrogen bonds. However, the carbonyl oxygen of these peptide bonds are directly involved in backbone hydrogen bonds with the neighbouring strand, b3 (Fig. 1B). Likewise, the carbonyl oxygen of the Leu170- Gly171 peptide bond forms a hydrogen bond to the backbone of Ala200, which pins the a1 helix to the carboxylate-binding loop (Fig. 1C). The carbonyl oxygen of the Gly169-Leu170 peptide bond appears to form a hydrogen bond to the amide of Gly166, thereby stabilizing the carboxylate-binding loop. Finally, the amide NH of Leu170 and Gly171 might be involved in hydrogen bonding with the backbone carbonyl of Lys168, which would also contribute to the stability of the carboxylate-binding loop (Fig. 1C). Semisynthesis of the PSD-95 PDZ2 amide-to-ester mutants required a Cys residue (Cys178) C-terminal of the backbone mutations [21]. The five amide-to-ester mutants as well as Y190W and the V178C/Y190W double mutant were all subjected to equilibrium and kinetic folding experiments. In our experiments, the V178C/Y190W mutant is considered the pseudo wild type to which the other mutants are compared.

Circular dichroism and equilibrium denaturations
The V178C as well as all the amide-to-ester mutations destabilized the protein. To enable a comparison of all mutants and wild type PDZ domain, we performed folding experiments in the presence of 0.4 M sodium sulphate, which stabilized the amide-to-ester mutants such that all were folded in the absence of denaturant, as shown by urea denaturation and far-UV circular dichroism (CD) experiments (Fig. 2). However, three mutants, L170l, F172Q and G176c displayed a CD spectrum with a slightly lower signal at 220 nm. Nevertheless, all mutants bound a peptide ligand in a fluorescence polarization experiment [21], showing that they populate a functional PDZ conformation. The amide-toester mutations destabilized the protein by 0.5-2 kcal mol 21 , with F172Q and I174i being the most destabilizing mutations (Fig. 2). The fluorescence of Trp190 was strongly affected by the G176c mutation, which precluded a quantitative analysis of the equilibrium denaturation experiment of this amide-to-ester mutant. The likely reason for the severe effect on the fluorescence of the G176c mutation is that the Ala175 carbonyl oxygen binds to the backbone amide of Trp190 and its removal changes the local structure and perturbs the fluorescence of Trp190.

Kinetic folding experiments
The pseudo wild type and mutants of PSD-95 PDZ2 were rapidly mixed with urea-buffer solutions using a stopped-flow instrument. The resulting kinetic traces were followed by monitoring the change in Trp fluorescence and were fitted to a single exponential equation to obtain the observed rate constants (k obs ) for unfolding or refolding. The k obs values were plotted versus urea concentration to obtain so-called chevron plots (Fig. 3).
Chevron plots of two-state folders i.e., proteins that populate only two states: the denatured and native states, separated by an energetic barrier, appears as perfectly V-shaped, due to the linear dependence of the logarithm of (un)folding rate constants with denaturant concentration [27]. The chevron plots for both the pseudo wild type and the amide-to-ester mutants displayed a curvature in the refolding arm in agreement with previous folding experiments on PDZ domains [13,15]. This curvature was interpreted as a change in rate-limiting step between two transition states, which are separated by a high-energy intermediate. The intermediate may transiently accumulate, but this cannot be distinguished in a kinetic analysis since different mechanistic scenarios give very similar mathematical solutions [27]. Therefore, only two microscopic rate constants can be appropriately determined from the present dataset, namely, the refolding rate constant k F1 (reflecting the crossing of the first barrier) and the unfolding rate constant k U2 (reflecting the crossing of the second barrier). The refolding rate constant k F2 could also be fitted but, in order not to bias the calculation of W values we chose not to calculate the free energy of unfolding in absence of denaturant, DG D-N , from the kinetic data. For calculations of W values we instead used DDG D-N values obtained from the midpoints of the equilibrium experiments and assuming a similar m D-N value in the curve fitting, for wild type and amide-to-ester mutants (1.0 kcalmol 21 M 21 , from experiments with Y190W) ( Table 1). Fitting of kinetic folding rate constants was done using Eq. 1. Calculated Wvalues were similar whether or not a shared m D-N value was assumed.
Truncation of side-chains in the core of a protein usually results in a decrease of the overall stability at equilibrium, DG D-N . In the case of a two-state folding process, DG D-N is dependent on the forward and reverse rate constants for folding, k F and k U . Usually, mutations of side-chains result in a decrease of k F and/or an increase in k U such that the resulting W values are between 0 and 1 (see methods). None of the amide-to-ester mutations in the present study led to a decrease in refolding rate constant, except possibly for G176c (Fig. 3). In fact, the refolding rate constants increased for three of the mutants as compared to PSD-95 PDZ2 V178C/ Y190W, suggesting that the hydrogen bonds, which are removed by the mutations, are involved in kinetically unfavourable interactions in the transition state for folding. Three backbone mutants, G171c, F172Q and I174i, displayed a significant increase in k F . However, while speeding up the folding reaction, the perturbation of these hydrogen bonds destabilizes the native state as reflected in the increased k U values (Fig. 3, Table 1). Three mutants, L170l, F172Q, I174i, displayed DDG D-N values that were sufficiently large (.0.6 kcal mol 21 ) [2] to calculate W values. Generally, W values were close to zero for the amide-to-ester mutations both for transition state 1 (TS1) and TS2 (Table 1). However, the W value for F172Q changes from only slightly negative in TS1 (20.0760.08) to 20.660.3 in TS2.

Discussion
Mutations of side-chains have in combination with simulation shaped our knowledge about the protein folding reaction [1,[28][29][30]. On the other hand, there are only very few studies employing backbone mutations to study protein folding, following the first papers more than 10 years ago [4][5][6]. The primary reason is that amide-to-ester mutations and similar modifications of the back- bone are still far from trivial to introduce in proteins. Yet they provide the possibility to probe the energetics of a fundamental aspect of protein structure, the hydrogen bond.
We have removed or destabilized hydrogen bonds from a bsheet and from the connecting loop in PSD-95 PDZ2 by introduction of backbone amide-to-ester mutations (Fig. 1) using expressed protein ligation [21,31]. The backbone hydrogen bonds in the b-sheet contribute significantly to the global stability of this protein domain (Table 1) and/or the local structure in case of the carbonyl group of Ala175 (see Results section). Moreover, the L170l mutation in the loop resulted in global destabilization of close to one kcal mol 21 suggesting that at least one of the two putative hydrogen bonds depicted in Fig. 1C contributes to the stability of the folded state. The G171c mutation also potentially targets two hydrogen bonds (see Results section and Fig. 1C). While the small value of DDG D-N suggests a minor contribution to overall stability, the large error precludes a quantitative assessment. However, the more accurately determined increase in the k U2 value upon G171c mutation suggests that the putative hydrogen bond(s) contribute to the stability of the domain. Polar groups are solvated in the denatured state. In order to achieve a stable fold, it is important to replace each desolvated hydrogen bond with an internal hydrogen bond in the folded state. Thus, if we count the hydrogen bonds (hydrogen bond inventory) we expect to have a similar or even identical number on each side of the folding reaction, or the reaction will become energetically unfavourable. The details of such energetics can be rather complex [32] and the actual contribution of a particular hydrogen bond to overall stability of a protein is therefore complex to deduce and highly context-dependent [3]. Nevertheless, our results for PSD-95 PDZ2 (observed DDG D-N around 1 kcal mol 21 ) agree well with those of previous studies using backbone modifications.
In examining the folding kinetics we make the interesting observation that mutation of three of the backbone peptide bonds resulted in a slightly increased folding rate constant (Table 1, Fig. 3). Such kinetics result in W values ,0 and has been observed for mutations involving truncations of side-chains (For example refs. [33,34]) but also for a thioether backbone mutation in the YAP WW domain probing formation of a b-hairpin [5]. One likely interpretation of an increase in k F1 is that non-native interactions involving the targeted hydrogen bond donors and/or acceptors are formed in the initial transition state for the folding reaction of the wild type protein [35]. Removal of the non-native interaction speeds up folding. Such non-native interactions might thus reflect a frustrated energy landscape and could involve misaligned b-strands as suggested for a circularly permuted PDZ domain [18]. In particular, the large negative W value resulting  from the F172Q mutation suggests a selective stabilization of TS2, which is not present in the ground states. The structures of two consecutive rate-limiting barriers were previously deduced based on W-value analyses of two different PDZ domains, PTP-BL PDZ2 [17] and PSD-95 PDZ3 [16]. These studies show that the late transition state (TS3 in Fig. 3D) has native-like side-chain interactions in large parts of the PDZ domain structure. Folding nuclei were identified in the strands b1, b4 and b6 for PTP-BL PDZ2 and in a2, b5 and b6 for PSD-95 PDZ3. We later identified a very early transition state (TS1) in the folding of PDZ domains [15] and redefined the numbering of transition states according to Fig. 3D. The compactness in terms of b T value and the W value analyses of the later transition states (TS2 and TS3) all suggest that the structure of the first transition state TS1 is highly heterogeneous. The current study addresses formation of backbone hydrogen bonds in TS1 and TS2, in the strands b2 and b3, as well as the loop connecting b1 and b2 and the helix a1 (Fig. 4). We find very low W values for the backbone hydrogen bonds in this region for the two early transition states (TS1 and TS2), in agreement with the previous studies [16,17]. Due to the relatively low thermodynamic stability of the amide-to-ester mutants of PSD-95 PDZ2, we could not measure W-values for the late, more native-like third transition state (TS3).
The amide-to-ester mutations are all in the ligand-binding groove. Mutations in this region have shaped the broad but overlapping ligand specificity in the PDZ family. It is reasonable that the ligand-binding groove of PDZ domains, which is optimized for function, should not affect the folding pathway, since this could lead to misfolding. Thus, while other parts of the PDZ domain govern early events in the folding reaction of PDZ domains [16][17][18], the ligand-binding groove may accept thermodynamically unfavourable amino acid side-chains, such as the conserved His in a2 [16,36], allowing for evolution of new specificities [37,38].