Glycerol-induced folding of unstructured disulfide-deficient lysozyme into a native-like conformation

Protein Expression and Purification

Details of the construction of the genes for the 1SS-, 2SS-, and 3SS-variants of hen lysozyme are described in the references published previously.10, 11, 18 The expression and purification of lysozyme variants and uniformly ¹⁵N-labeled samples have been described previously.7-9 Concentrations of lysozyme variants were estimated by using A₂₈₀ = 2.64 for 1 mg/ml protein.

Materials

DMSO-d6 and glycerol-d8 were obtained from Cambridge Isotope Laboratories. Formic acid-d2 and dichloroacetic acid-d2 were purchased from Sigma-Aldrich Corporation. D₂O was obtained from Wako Pure Chemical Industries.

CD Measurements

CD spectra of lysozyme variants were obtained on a Jasco J-600 spectropolarimeter (JASCO Corporation) with a thermostatically controlled cell holder. They were measured in 20 mM glycine buffer at pH 3.0 and 4°C by using a cuvette with an optical length of 2 mm or 5 mm for far-UV measurements and 10 mm for near-UV. The cell was defrosted with the flow of dry nitrogen gas below the room temperature. Protein concentrations for near-UV CD spectra were about 25 μM, those for far-UV CD 5 μM.

NMR Experiments

NMR spectra of disulfide-deficient lysozyme were measured in 95% DMSO/5% H₂O (or D₂O) containing 60 mM dichloroacetic acid at pH* 5.5, where pH* is the uncorrected pH electrode reading. The pH* was adjusted by adding DCl and NaOD. Shigemi NMR tubes matched to DMSO-d6 were used in all of the experiments 2D ¹H-¹⁵N HSQC spectra19 and 3D NMR spectra were usually measured at protein concentrations of about 0.2 mM and 1 mM, respectively. Data were recorded at 600.13 MHz on a Bruker Avance 600 DRX spectrometer. Solvent signal suppression was achieved by using the WATERGATE20 scheme. Typically, 2D ¹H-¹⁵N HSQC spectra were collected at 25°C with 1024 × 2048 points in t₁ and t₂ directions, using uniformly ¹⁵N-labeled samples. 3D HSQC-NOESY-HSQC (mixing time of 150 ms) spectra21 were collected with 64 × 64 × 2048 in the t₁, t_2, and t₃ directions. Zero filling was applied before Fourier transformation, and data were processed with a shifted squared sine-bell window function in both dimensions. The DMSO peak was referenced as 2.55 ppm in all spectra.

H/D Exchange of Disulfide-Deficient Variants of Lysozyme

H/D exchange of lysozyme variants was carried out in 20 mM formate buffer at pH* 3.0 and 4°C, and at the protein concentration of 0.5 mg/ml. Freeze-dried proteins were first dissolved in H₂O to make the dissolution complete. Especially, this process was necessary in H/D exchange experiments in the presence of concentrated glycerol, because it was not easy to dissolve the freeze-dried protein directly in 30% glycerol solution. To start H/D exchange, the solution was diluted 10-fold with deuterated formate buffer containing concentrated glycerol-d8 or not. The ratio of x_H = [H₂O]/([H₂O]+[D₂O]) resulted in about 0.1. Strictly speaking, it depends on the final glycerol concentration, because D₂O solution contains concentrated glycerol. As a result, in the mixture of H₂O/D₂O containing 30% glycerol, about 12% of total amide groups remained protonated after the H/D exchange reaction finished. On the other hand, the rate constant of H/D exchange remains unaltered even in the mixture of H₂O and D₂O, because the observed rate, k_obs, is the sum of forward (k_HD) and backward (k_DH) rate constants. That is to say, k_HD and k_DH are proportional to [D₂O] and [H₂O], respectively, so that

H/D exchange reaction was performed for a desired time, and then stopped by freeze-drying partially deuterated proteins. In the presence of glycerol, the reaction mixture was applied to a Waters Sep-Pak cartridge column (reversed phase C8) equilibrated with 20% CH₃CN/80% D₂O mixture and washed with 6 ml of the solution to remove glycerol, and then protein was eluted with 3 ml of 60% CH₃CN/40% D₂O mixture and freeze-dried. The removal of glycerol was carried out as quickly as possible at 4°C and pH* 2.5 to stop further progress of the H/D exchange.

To measure HSQC spectra of partially deuterated proteins, freeze-dried samples were dissolved in a quenching buffer (95% DMSO-d6/5% D₂O mixture at pH* 5.5 adjusted with dichloroacetate-d2). This process was carefully performed under an atmosphere of dry nitrogen in a polyethylene glove bag. The first HSQC spectrum was obtained about 40 min after dissolving protein in the quenching buffer, so that this is the dead time (τ_d). As mentioned later, H/D exchange rates in the quenching buffer were slower than 1/120 per minute for most residues. The ratio of the peak volume, I_ex(τ_d)/I_ex(0), was about 72–92%, where I_ex(0) and I_ex(τ_d) are a true peak volume and that first observed after τ_d, respectively. Since the ratio remains constant for each residue in a series of H/D exchange reactions, the observed peak volume, I_ex(τ_d), was plotted as a function of H/D exchange time without the correction due to the dead time.

Structures Induced by Glycerol in Disulfide-Deficient Variants of Lysozyme

Effects of glycerol on disulfide-deficient variants of lysozyme which are unstructured in water were studied by measuring their CD spectra. Figure 1A shows far-UV CD spectra of 1SS[64-80] in solutions containing various concentrations of glycerol. The negative ellipticity at 220 nm, [θ]₂₂₀, gradually increased with the increase in glycerol concentration. In a solution of 50% glycerol, the value of [θ]₂₂₀ approached to that of the wild-type lysozyme in water, and exceeded it at 70% glycerol. The dotted line close to the curve d refers to the spectrum of the wild-type lysozyme which was independent of the concentration of glycerol. Far-UV CD spectra of 0SS-variant and 1SS[76-94] were almost identical with those of 1SS[64-80] under all the concentrations of glycerol, and those of 1SS[30-115] were virtually the same as those of 1SS[64-80], although the former was slightly larger than the latter in the negative ellipticity. Figure 1B shows far-UV CD spectra of 2SS[6-127,64-80]. The value of [θ]₂₂₀ in water was somewhat larger than that of 1SS[64-80], but [θ]₂₂₀ approached to the same value at 70% glycerol for both 1SS[64-80] and 2SS[6-127,64-80]. Also, far-UV CD spectra of 1SS[6-127] and 2SS[64-80,76-94] were similar to those of 2SS[6-127,64-80], although shapes of CD spectra were slightly different from each other. These data suggest that the addition of glycerol increases the helical contents of all the disulfide-deficient variants studied, and that the helical contents exceed that of the wild-type at 70% glycerol although the helical structure is non-native.

As shown in Figure 1C, near-UV CD spectra of lysozyme variants except 2SS[6-127,64-80] remained nearly equal to zero even in the presence of concentrated glycerol. 2SS[6-127,64-80] alone could recover the native-like CD spectrum around 280 nm by the addition of glycerol. In Figure 1C, the curve a represents the CD spectrum of the wild-type lysozyme which remained unaltered in water and in 50% glycerol. As shown in Figure 1D, 2SS[6-127,64-80] was completely unstructured in water, but some ordered structure was induced around tryptophan residues by the addition of glycerol, while no definite tertiary structure was observed in other variants such as 0SS, four species of 1SS-variants, and 2SS[64-80,76-94].

To explore the glycerol-induced conformation of 2SS[6-127,64-80] in a residue specific manner, we tried to measure ¹H-¹⁵N HSQC spectra, in 30% glycerol solution, but it was very difficult to assign the cross-peak to an individual residue because of broadening of resonance lines. Therefore, we carried out H/D exchange reactions of amide hydrogens in a solution containing glycerol, and then determined the H/D exchange rates of individual amide hydrogens by measuring HSQC spectra of deuterated protein in the manner described below. From the protection factors of amide hydrogens against the H/D exchange, we deduced the glycerol-induced conformation of 2SS[6-127,64-80].

NMR Spectra in DMSO/D₂O Solution

To determine H/D exchange rates of amide hydrogens, H/D exchange reaction must be virtually stopped during the acquisition of NMR signals (about 15 min). In the conventional H/D exchange NMR method, therefore, partially deuterated proteins are rapidly refolded immediately after the H/D exchange reaction to stop the progress of the reaction, then the HSQC spectra of refolded proteins are measured. Since disulfide-deficient variants of lysozyme are inherently unstructured in water, H/D exchange time-constants of almost all amide hydrogens are much faster than 15 min. Therefore, it was impossible to measure their H/D exchange rates by the conventional method. To measure the NMR spectra of partially deuterated proteins, we adopted the mixture of 95% DMSO and 5% D₂O at pH* 5.5 adjusted with dichloroacetate-d2 as a quenching buffer, because the minimum exchange rate was found to be about 100-fold reduced in the quenching buffer relative to that in water.16 This solvent system had another advantage that it significantly suppressed broadening of the resonance lines of lysozyme variants and enabled us to obtain well-separated NMR spectra as shown in Figure 2. The peak assignment was readily carried out using 3D HSQC-NOESY-HSQC spectra. Almost all residues were successfully identified on the HSQC spectrum of 2SS[6-127,64-80] except K1, W63, P70, P79, R112, and L129. In addition, the HSQC spectrum of 0SS variant was measured in DMSO/H₂O solution. The assigned chemical shifts for lysozyme variant, 0SS and 2SS[6-127,64-80], have been deposited in the BMRB database with accession numbers of 11051 and 11052, respectively. Figure 3 shows the chemical shift differences of amide protons of 2SS[6-127,64-80] from those of 0SS. Actually, large differences were observed near the disulfide bridges existing in 2SS[6-127,64-80].

H/D Exchange Rates of 2SS[6-127,64-80] in DMSO/D₂O Solution During the Acquisition of NMR Spectra

As mentioned earlier, the HSQC spectra of partially deuterated protein after the H/D exchange were measured in the quenching buffer (mixture of 95% DMSO/5% D₂O) at 25°C. To see to what extent the H/D exchange progresses during the acquisition of NMR spectra, we actually observed the time course of HSQC spectra in this solvent. The time constants of H/D exchange in the quenching buffer were determined for most amide hydrogens of 2SS[6-127,64-80], and are summarized in Table I. Generally speaking, time constants of most Gly, Ser, and Asn were fast (about 45 to 120 min.), those of Ala were around 150 min, and those of hydrophobic residues were slow ranging from 500 to 1000 min, although the values were dependent on neighboring residues. As a result, H/D exchange time-constants of most residues were longer than 120 min except R5, C6, H15, G16, D18, D48, G49, D52, D87, D101, G102, and D119 with too fast exchange rates.

H/D Exchange Reaction of Disulfide-Deficient Variants of Lysozyme in the Absence of Glycerol

H/D exchange reactions of 0SS- and 2SS[6-127,64-80]-variants were carried out at 4°C and pH* 3.0 in the absence of glycerol. After various time of 5, 10, 30, 60, 120, and 480 min, the exchange reaction was stopped by freeze-drying the protein, then the partially deuterated protein was dissolved in the quenching buffer for the measurements of HSQC spectra. Figure 4a shows some examples of H/D exchange reactions of amide hydrogens. H/D exchange time-constants were determined as described in Methods. In Figure 4a, the best-fitting curves for F34, V92, I55, and W108 had time-constants of 34, 50, 66, and 163 min, respectively.

However, the time-constants of some residues could not be precisely determined for some reasons. Amide hydrogens of R5, S6 (in 0SS), H15, G16, D18, D48, G49, D52, D87, D101, G102, and D119 exchanged so fast even in the quenching buffer that the calculation of their peak volumes was difficult. In the case of residues located near the disulfide bridges in 2SS[6-127,64-80], the peak-heights remarkably diminished due to the peak broadening as shown in Figure 2, so that we could not determine the exchange rates of C6, E7, W63, C64, N65, D66, G67, C80, and S81. Further, time-constants of some residues could not be precisely determined because of an overlap of cross-peaks. For example in 2SS[6-127,64-80], Y20, L83, and I124 overlapped each other, and the cross-peaks of A10 and A76, L17 and N93, N19 and N74, A95 and Q57, K96 and I55, could not be separated from each other because of an insufficient resolution of cross-peaks, which was inevitable to shorten the acquisition time of NMR spectra.

Besides 2SS[6-127,64-80], H/D exchange time-constants for the residues of 0SS-variant were determined and are shown in Figure 5 together with those of 2SS[6-127,64-80]. The exchange rates of the same residue in the two variants were quite similar to each other. In Figure 6, exchange time constants of 0SS-variant measured in 20 mM formate buffer are compared with those expected in a random coil model, which are predicted according to Bai et al.22 The results indicate that time constants in the formate buffer, τ_for, are significantly faster than Bai's predicted values, τ_pre. Bai's experimental data were obtained in 0.5M KCl to shield possible charge effects, whereas our experiments had to be performed in 20 mM formate buffer to avoid the aggregation of disulfide-deficient variants. Since the difference in salt concentrations seems the cause of the disagreement,23 H/D exchange reactions of 0SS-variant were measured in a solution containing 4.0M GuHCl. As shown in Figure 6, the time constants in 4.0M GuHCl, τ_gdm, were nearly equal to Bai's predicted values. These data suggest that neither 0SS nor 2SS[6-127,64-80] has any structure preventing amide hydrogens from being exchanged. Although the helical contents of 2SS[6-127,64-80] were somewhat larger than those of 0SS-variant judged from their far-UV CD spectra shown in Figures 1A and 1B, in the absence of glycerol, two disulfide bridges existing in 2SS[6-127,64-80] do not appear to promote the formation of partially folded structures.

H/D Exchange Reaction of Disulfide-Deficient Variants of Lysozyme in the Presence of Glycerol

H/D exchange reactions of 2SS[6-127,64-80] were performed in solutions containing 10, 20, and 30% glycerol. Figures 4b, 4c, and 4d show the H/D exchange reactions of F34, I55, and V92, respectively, in the presence of glycerol. In a solution of 10% glycerol, H/D exchange reactions were carried out for 30 min, 1.0, 8.0, and 30 h at pH* 3.0 and 4.0°C. Exchange reaction was somewhat delayed compared with that in water, but it finished within 30 h for almost all residues. When the concentration of glycerol increased from 10 to 20%, exchange reaction definitely slowed down in the regions of L8-R14, W28-F34, F38-N39, Q41, Y53-R61, L83-S85, T89, S91-S100, W108, W111, and W123-R125. In the presence of 30% glycerol, HSQC spectra of partially deuterated samples were measured after each H/D exchange reaction was carried out for 1.0, 8.0, 32, 72, 168, 336, and 1008 h. The peak volume of 42 residues diminished fast, but more than 30 cross-peaks still remained definitely even after 168 h. In the case of 30% glycerol, time-constants of H/D exchange were determined only for residues whose peak volume was enough detectable after 8.0 h. For other residues whose exchange reaction almost finished in 8.0 h, time-constants were not estimated anymore.

Protection factors against the H/D exchange reaction in a glycerol solution were determined for individual amide hydrogens as the ratio of τ_obs to τ_for, where τ_obs and τ_for are time constants of exchange reaction measured in a glycerol solution and in a formate buffer solution, respectively. Figure 7 shows the protection factors of each residue in 10, 20, and 30% glycerol solutions. In a solution of 10% glycerol, protection factors uniformly increased to about 5 to 10 at most residues. It is likely that non-native secondary structures are induced anywhere along the polypeptide chain, but the conformation appears non-persistent, that is, to be easily broken and rebuilt again. With the increasing concentration of glycerol, some selected regions grow the protection factors, but other regions do not any more. In highly protected regions, secondary structure may be frozen by the tertiary structure induced by glycerol. Most highly protected regions are V9-M12, V29-F34, I55-I58, L83-L84, T89, and V92-V99, and their protection factors reached about 1000 in 30% glycerol solution. Strictly speaking, the cross-peaks of K96 and I55, those of A95 and Q57, and those of I124 and L83 overlapped each other, so we could not determine their time constants separately. A common value was assigned as the protection factors for these residues.

In H/D exchange reactions of L8, W28, F38, N39, Q41, Y53, G54, N59, S60, S85, S91, N93, A94, and S100, a fast phase of exchange reaction was observed within an hour and followed by a slow phase. The signal amplitude of the fast phase was close to a half of total one for these residues. That is to say, such a residue was weakly protected in some ensemble of structures, while it was strongly protected in another one. Tertiary structures induced in concentrated glycerol solution might be inhomogeneous with a very slow inter-conversion rate between ensembles. These residues were placed on the periphery of highly protected regions. In general, such a fast phase was also observed for highly protected residues mentioned earlier, but the amplitude of the fast phase was relatively small. In Figure 7, the protection factors of the slow phase are plotted for all residues.

In the native structure of the wild-type lysozyme, A-helix including A9-M12, B-helix including V29-F34, and C-helix including T89-V99 enclose the side chains of Ile 55 and Leu 56. Also, Q57 and I58 are located in the interface between two domains in the native structure. Therefore, the amino acid residues whose protection factors reach 1000 appear to form a native-like rigid structure with centering on Ile 55 and Leu 56. In Figure 8, the highly protected region is represented by sticks-and-balls on the ribbon diagram of the native structure of lysozyme. This region looks like a pivot of lysozyme folding with three helices and three β-strands anchored together. Also, it is suggestive that protection factors of F38, N39, and L83-S85 which are located near the pivot in the native structure also approach 1000. On the other hand, protection factors increased to 100, but did not exceed it in the regions of D-helix (residues 108–115) and C-terminal 3₁₀-helix (residues 120–124). Protection factors were limited to 10 in the region of β1- and β2-strands (residues 41–54), and other regions corresponding to irregular structures of the native lysozyme. The preferential hydration seems insufficient to freeze the structural freedom of these regions because of the lack of cooperativity in interactions among residues.

Preferential Hydration by Glycerol

The effect of solvent additives on protein stability can be interpreted in terms of the solvation properties of protein. In general, large cosolvent molecules are excluded from the solvation shell near the protein's surface because of volume exclusion, and protein is said to be preferentially hydrated.13, 14 It was experimentally shown that glycerol resulted excluded from the surface of wild-type lysozyme in water/glycerol mixtures, confirming that lysozyme was preferentially hydrated.24 The protein surface area in contact with the solvent tends to be minimized by negative binding of glycerol. In most of the previous studies on the effects of glycerol, proteins were originally in the native state and the preferential hydration further stabilized them preventing the exposure of buried groups to the solvent.18, 25, 26

In the present study, the disulfide-deficient variant, 2SS[6-127,64-80], is inherently unstructured in pure water. Preferential hydration by glycerol induces 2SS[6-127,64-80] to recover a native-like tertiary structure in a part of molecule. [θ]₂₂₀ of 2SS[6-127,64-80] was relatively large at low concentrations of glycerol compared with that of 0SS- or 1SS-variants, but it only increased gradually in the range of 20 to 70% glycerol. At the early stage below 10% glycerol, the helical contents of 2SS[6-127,64-80] may increase easily at most of residues, but they will soon reach saturation. In such an early stage, it seems likely that the helical contents are relatively large but the protection factors remain low. Protection factors of about 10 mean that the protected form of a given amide merely persists nine times longer than the exposed one. Suppose each residue independently changes its conformation, the probability that several consecutive residues are in a protected form is low. As a result, a helical conformation is not persistent and short secondary structures only float around along the polypeptide chain. In the case of 0SS- or 1SS-variants of lysozyme, [θ]₂₂₀ increased with the increase in concentration of glycerol, and it approached to the same value as in 2SS[6-127,64-80] at 70% glycerol, but the protection factors for H/D exchange of 1SS[64-80] and/or 1SS[76-94] variants remained low (around 10) even in 70% glycerol solution. These observations are similar to those for 2SS[6-127,64-80] in 10% glycerol solution. The conformation of 2SS[6-127,64-80] in this stage looks like a disordered one rich in helices.

With increasing concentration of glycerol, the preferential hydration seems to induce weakly protected segments in 2SS[6-127,64-80] to associate with each others through hydrophobic interactions. If the cooperativity gets to work among some structural elements, a rigid tertiary structure may emerge and protection factors rapidly increase to reach 1000 in the part of molecule. In 20 to 30% glycerol solution, the cooperativity seems to work among residues centering on Ile 55 and Leu 56 surrounded by A-, B-, and C-helices, while β1- and β2-strands, D-helix and C-terminal 3₁₀-helix appear to be independent from the former region.

An Early Stage of Lysozyme Folding and the Role of Disulfide Bridges

Although the helical contents of 0SS- or 1SS-variants increased due to the preferential hydration by glycerol, the effect of glycerol alone was insufficient for these variants to fold successfully. For example, 1SS[64-80] with a single disulfide bridge Cys64-Cys80 was rich in helices but still remained non-persistent judged from the protection factors. The stability of glycerol-induced structure in 2SS[6-127,64-80] suggests that Cys6-Cys127 plays an important role for the formation of the stable α-domain structure, because Cys64-Cys80 is far from the α-domain itself. However, 1SS[6-127] containing only one disulfide bridge Cys6-Cys127 could not recover the tertiary structure even in 70% glycerol as well as 1SS[64-80]. Besides Cys6-Cys127, either Cys30-Cys115 or Cys64-Cys80 is necessary to keep the α-domain stable.

When both Cys6-Cys127 and Cys30-Cys115 were present, 2SS[6-127,30-115] maintained the native-like structure even in the absence of glycerol.9 As pointed out in our previous article, many side-chains are closely packed within the hydrophobic box in the α-domain, and many long-range NOE contacts were detected between side-chain and main-chain protons. The α-domain structure was never like a molten globule, rather close to the native one. On the other hand, when Cys64-Cys80 coexists with Cys6-Cys127, the tertiary structure of the α-domain was lost in pure water, but recovered in 30% glycerol solution. The preferential hydration by glycerol seems to reinforce the hydrophobic box enclosed by A-, B-, and C-helices. What is the role of Cys64-Cys80 in 2SS[6-127,64-80]? Besides the preferential hydration by glycerol, the constraint on the backbone conformation by Cys64-Cys80 is necessary for 2SS[6-127,64-80] to stabilize the glycerol-induced structure. Our previous studies on 3SS-variants of lysozyme showed that Cys64-Cys80 made a large contribution toward maintaining the β-sheet structure.8 Suppose Cys64-Cys80 is removed from 2SS[6-127,64-80], it may lead β-sheet to fluctuate significantly, so that the cooperativity between two domains may weaken, and the side-chains of Ile 55 and Leu 56 may come out the pocket surrounded by A-, B-, and C-helices. As a result, the glycerol-induced structure in the α-domain may be disrupted simultaneously. Probably Cys64-Cys80 is necessary for side-chains of Ile 55 and Leu 56 to anchor at the hydrophobic box in the α-domain.

The glycerol-induced structure in 2SS[6-127,64-80] is rather close to a molten globule. At relatively low concentration of glycerol, it looked like a disordered structure rich in helices. With increasing glycerol concentration, a rigid tertiary structure emerged but long-range NOE contacts were not detected because of the resonance-line broadening. These are the characteristics of the molten globule. Further, it is intriguing that the glycerol-induced structure in 2SS[6-127, 64-80] is similar to that of 2SS[6-127,30-115]. In particular, residues centering on Ile 55 and Leu 56 were highly protected against the H/D exchange in both 2SS-variants. The difference between them is that the hydrophobic box looks like a molten globule in the former, and a crystalline structure in the latter, and that the C-terminal region including D-helix and 3₁₀-helix is flexible in the former, and rigid in the latter. It implies that the structural motif common to both 2SS[6-127,64-80] and 2SS[6-127,30-115], in which residues centering on Ile 55 and Leu 56 form a native-like persistent structure through hydrophobic interaction as shown in Figure 8, is crucial to the lysozyme folding.

The wild-type lysozyme is a protein for which folding has been studied in great detail by combination of different techniques.5 When the folding was initiated by dilution of denaturants from 6M GuHCl, it was complex with a variety of distinct steps. At the intermediate stage of folding, most highly populated species was the α-domain folding intermediate, in which the α-domain formed a highly stable structure with a hydrophobic box, while the β-domain was still unstable and rapidly fluctuating.6 The glycerol-induced structure in 2SS[6-127,64-80] is close to the α-domain folding intermediate of the wild-type in most of characteristics. Thus, it seems a molten globule at equilibrium. It is likely that a molten globule grows from a disordered form to an ordered one during the folding. At the early stage of molten globule such as in 10% glycerol, it may be a disordered conformation rich in helices, while a stable native-like structure may emerge in a part of molecule at the late stage such as in 30% glycerol. If that is the case, the late molten globule is still distinguished definitely from the native state by the global cooperative transition between them.