The effect of copper/zinc replacement on the folding free energy of wild type and Cys3Ala/Cys26Ala azurin

The effect of copper/zinc metal ion replacement on the folding free energy of wild type (w.t.) and disulfide bridge depleted (C3A/C26A) azurin has been investigated by differential scanning calorimetry (DSC) and fluorescence techniques. The denaturation experiments have shown that, in both cases, the thermal transitions of the zinc derivative of azurins can be depicted in terms of the classical Lumry–Eyring model, NUF, thus resembling the unfolding path of the two copper proteins. The thermally induced transition of Zn azurin, monitored by fluorescence occurs at lower temperature than the DSC scans indicating that a local conformational rearrangement of the Trp microenvironment, takes place before protein denaturation. For Zn C3A/C26A azurin, the two techniques reveal the same transition temperature. Comparison of the thermodynamic data shows that the presence of Zn in the active site stabilises the three-dimensional structure of azurin only when the disulfide bridge is present. Compared to the copper form of the protein, the unfolding temperature of Zn azurin has increased by 4 °C, while the unfolding free energy, ΔG, is 31 kJ/mol higher. Both enthalpic and entropic factors contribute to the observed ΔG increase. However, the copper/zinc replacement has no effect on the unfolding free energy of C3A/C26A azurin. Taking Cu azurin w.t. as the reference state, for both Cu and Zn C3A/C26A azurin the unfolding free energy is decreased by about 28 kJ/mol, indicating that metal substitution is not able to compensate the destabilising effect induced by the disulfide bridge depletion. It is noteworthy that the thermal denaturation of the Zn derivative, which thermodynamically is the most stable form of azurin, is also characterized by the highest value of the activation energy, E_a, as derived from the kinetic stability analysis.     

Structural heterogeneity of blue copper proteins: an EPR study of amicyanin and of wild-type and Cys3Ala/Cys26Ala mutant azurin

A comparative investigation of the effects of cooling rate and solvent physicochemical properties on the structural heterogeneity of wild-type and disulfide bond depleted azurin (Cys3Ala/Cys26Ala) and of amicyanin has been performed by EPR spectroscopy and computer simulation. By describing the spectral features of the EPR spectra in terms of Gaussian distributions of the components of the g and A tensors of the spin Hamiltonian, we have shown that either the cooling rate or the solvent composition affect the structural heterogeneity of the proteins. Such a heterogeneity has been quantified by the standard deviations sigmag and sigmaA of the parallel components of the axially symmetric tensors. In particular, both parameters become smaller after the slow cooling cycle; such a reduction is more significant when glycerol is added as cosolvent to the protein solutions. The comparison of the deltag and sigmaA values found, for the copper proteins investigated, highlights that the reduction is more marked in the azurins compared to amicyanin and that the Cys3Ala/Cys26Ala azurin mutant has a structural heterogeneity lower than that shown by the wild-type protein. The remarkable similarity of the copper coordination sphere of the proteins suggests a more rigid structure of the azurin protein matrix in the absence of the disulfide bridge compared to wild-type azurin and of amicyanin with respect to both forms of azurin. The former result establishes an important role for the -SS- bond in modulating the flexibility of wild-type azurin.     

Evidence of reduced flexibility in disulfide bridge-depleted azurin: a molecular dynamics simulation study.

Two molecular dynamics simulations have been performed for 2 ns, at room temperature, on fully hydrated wild type and Cys3Ala/Cys26Ala double-mutant azurin, to investigate the role of the unique disulfide bridge on the structure and dynamics of the protein. The results show that the removal of the [bond]SS[bond] bond does not affect the structural features of the protein, whereas alterations of the dynamical properties are observed. The root mean square fluctuations of the atomic positions are, on average, considerably reduced in the azurin mutant with respect to the wild type form. The number of intramolecular hydrogen bonds between protein backbone atoms that are lost during the simulation, with respect to the starting configuration, are reduced in the absence of the disulfide bond. The analysis of the dynamical cross-correlation map, characterising the protein co-ordinated internal motions, demonstrates in the mutated azurin a significant decrease in anti-correlated displacements between protein residues, with the only exception occurring in the region of the mutation sites. The overall findings show a relevant reduction in flexibility as a consequence of the disulfide bridge depletion in azurin, suggesting that the [bond]SS[bond] bond is a structural element which significantly contributes to the dynamic properties of the native protein.     

Effects of chaotropic anions on the distribution of conformational substates of amicyanin,wild type and Cys3Ala/Cys26Ala azurin mutant

The effect of azide and thiocyanate on the structure and dynamics of wild type and disulfide bond depleted azurin and of amicyanin has been investigated by electron paramagnetic resonance (EPR) spectroscopy at low temperature. The analysis of the EPR spectra, which can be described in terms of Gaussian distributions of the components of the axial symmetric <--> g and <--> A tensors of the spin-Hamiltonian, has shown that the two small exogenous ligands, known as chaotropic agents, are effective in reducing the structural heterogeneity of the proteins. Such a reduction, quantified by the standard deviations sigma(g axially) and sigma(A axially) and obtained by simulation of the experimental EPR spectra, depends on azide and thiocyanate concentration in solution. In particular, the comparison of the sigma(g axially) and sigma(A axially) values found for the protein samples investigated points out that the lower the protein to anion molar ratios (1:50; 1:100) are, the more marked the reduction in structural heterogeneity is. The thiocyanate effect is stronger than the azide one. Furthermore, the reduction in structural heterogeneity is more marked in the azurins than in amicyanin and the Cys3Ala/Cys26Ala azurin mutant is less flexible compared to the wild-type protein. The effect observed upon N(-)(3) and SCN(-) addition in solution is very similar to that observed when glycerol is added to the solution, suggesting that such perturbing agents behave like cryoprotectors, affecting the protein-solvent interactions in such a way as to suppress the large amplitude motions.     

Thermodynamic analysis of the contributions of the copper ion and the disulfide bridge to azurin stability: synergism among multiple depletions

The stabilizing potential of the copper ion and the disulfide bridge in azurin has been explored with the aim of inspecting the ways in which these two factors influence one another. Specifically, whether copper and disulfide contributions to protein stability are additive has been examined. To this aim, the thermal unfolding of a copper-depleted mutant lacking the disulfide bridge between Cys3 and Cys26 (apo C3A/C26A azurin) was studied by differential scanning calorimetry. A comparison of the unfolding parameters of holo and apo C3A/C26A azurin with the apo C3A/C26A protein has shown that the effects of simultaneous copper and disulfide depletion are additive only at two temperatures: T=15 degrees C and T=67 degrees C. Within this range the presence of the copper ion and the disulfide bridge has a positive synergistic effect on azurin stability. These findings might have implications for the rational use of the stabilizing potential of copper and disulfides in copper protein engineering.     

Effect of an engineered disulfide bond on the folding of T4 lysozyme at low temperatures

Equilibrium and kinetic effects on the folding of T4 lysozyme were investigated by fluorescence emission spectroscopy in cryosolvent. To study the role of disulfide cross-links in stability and folding, a comparison was made with a mutant containing an engineered disulfide bond between Cys-3 (Ile-3 in the wild type) and Cys-97, which links the C-terminal domain to the N terminus of the protein [Perry & Wetzel (1984) Science 226, 555]. In our experimental system, stability toward thermal and denaturant unfolding was increased slightly as a result of the cross-link. The corresponding reduced protein was significantly less stable than the wild type. Unfolding and refolding kinetics were carried out in 35% methanol, pH 6.8 at -15 degrees C, with guanidine hydrochloride as the denaturant. Unfolding/refolding of the wild-type and reduced enzyme showed biphasic kinetics both within and outside the denaturant-induced transition region and were consistent with the presence of a populated intermediate in folding. Double-jump refolding experiments eliminated proline isomerization as a possible cause for the biphasicity. The disulfide mutant protein, however, showed monophasic kinetics in all guanidine concentrations studied.     

Crystal structure of the disulfide bond-deficient azurin mutant C3A/C26A: how important is the S-S bond for folding and stability?

Azurin has a beta-barrel fold comprising eight beta-strands and one alpha helix. A disulfide bond between residues 3 and 26 connects the N-termini of beta strands beta1 and beta3. Three mutant proteins lacking the disulfide bond were constructed, C3A/C26A, C3A/C26I and a putative salt bridge (SB) in the C3A/S25R/C26A/K27R mutant. All three mutants exhibit spectroscopic properties similar to the wild-type protein. Furthermore, the crystal structure of the C3A/C26A mutant was determined at 2.0 A resolution and, in comparison to the wild-type protein, the only differences are found in the immediate proximity of the mutation. The mutants lose the 628 nm charge-transfer band at a temperature 10-22 degrees C lower than the wild-type protein. The folding of the zinc loaded C3A/C26A mutant was studied by guanidine hydrochloride (GdnHCl) induced denaturation monitored both by fluorescence and CD spectroscopy. The midpoint in the folding equilibrium, at 1.3 M GdnHCl, was observed using both CD and fluorescence spectroscopy. The free energy of folding determined from CD is -24.9 kJ.mol-1, a destabilization of approximately 20 kJ.mol-1 compared to the wild-type Zn2+-protein carrying an intact disulfide bond, indicating that the disulfide bond is important for giving azurin its stable structure. The C3A/C26I mutant is more stable and the SB mutant is less stable than C3A/C26A, both in terms of folding energy and thermal denaturation. The folding intermediate of the wild-type Zn2+-azurin is not observed for the disulfide-deficient C3A/C26A mutant. The rate of unfolding for the C3A/C26A mutant is similar to that of the wild-type protein, suggesting that the site of the mutation is not involved in an early unfolding reaction.     

Thermal stability of wild type and disulfide bridge containing mutant of poplar plastocyanin

A comparative study of the thermal stability of wild type poplar plastocyanin and of a mutant form containing a disulfide bridge between residues 21 and 25 was performed using differential scanning calorimetry and optical spectroscopic techniques. For wild type plastocyanin the transition temperature, determined from the calorimetric profiles, is 62.7 degrees C at the scan rate of 60 degrees C/h, whereas for the mutant it is reduced to 58.0 degrees C. In both cases, the endothermic peak is followed by an exothermic one at higher temperatures. The unfolding process monitored by optical absorption at 596 nm also reveals a reduced thermal stability of the mutated plastocyanin compared to the wild type protein, with transition temperatures of 54.8 and 58.0 degrees C, respectively. For both proteins, the denaturation process was found to be irreversible and dependent on the scan rate preventing the thermodynamic analysis of the unfolding process. In parallel, small conformational changes between wild type and mutant plastocyanin emerge from fluorescence spectroscopy measurements. Here, a difference in the interaction of the two proteins between the microenvironment surrounding the fluorophores and the solvent was proposed. The destabilization observed in the disulfide containing mutant of plastocyanin suggests that the double mutation, Ile21Cys and Glu25Cys, introduces strain into the protein which offsets the stabilizing effect expected from the formation of a covalent crosslink.     

Novel insight into the mechanism of the vitamin K oxidoreductase (VKOR): electron relay through Cys43 and Cys51 reduces VKOR to allow vitamin K reduction and facilitation of vitamin K-dependent protein carboxylation

The vitamin K oxidoreductase (VKOR) reduces vitamin K to support the carboxylation and consequent activation of vitamin K-dependent proteins, but the mechanism of reduction is poorly understood. VKOR is an integral membrane protein that reduces vitamin K using membrane-embedded thiols (Cys-132 and Cys-135), which become oxidized with concomitant VKOR inactivation. VKOR is subsequently reactivated by an unknown redox protein that is currently thought to act directly on the Cys132-Cys135 residues. However, VKOR contains evolutionarily conserved Cys residues (Cys-43 and Cys-51) that reside in a loop outside of the membrane, raising the question of whether they mediate electron transfer from a redox protein to Cys-132/Cys-135. To assess a possible role, the activities of mutants with Ala substituted for Cys (C43A and C51A) were analyzed in intact membranes using reductants that were either membrane-permeable or -impermeable. Both reductants resulted in wild type VKOR reduction of vitamin K epoxide; however, the C43A and C51A mutants only showed activity with the membrane-permeant reductant. We obtained similar results when testing the ability of wild type and mutant VKORs to support carboxylation, using intact membranes from cells coexpressing VKOR and carboxylase. These results indicate a role for Cys-43 and Cys-51 in catalysis, suggesting a relay mechanism in which a redox protein transfers electrons to these loop residues, which in turn reduce the membrane-embedded Cys132-Cys135 disulfide bond to activate VKOR. The results have implications for the mechanism of warfarin resistance, the topology of VKOR in the membrane, and the interaction of VKOR with the carboxylase.     

Unpaired cysteine-54 interferes with the ability of an engineered disulfide to stabilize T4 lysozyme

We have introduced an intramolecular disulfide bond into T4 lysozyme and have shown this molecule to be significantly more stable than the wild-type molecule to irreversible thermal inactivation [Perry, L.J., & Wetzel, R. (1984) Science (Washington, D.C.) 226, 555-557]. Wild-type T4 lysozyme contains two free cysteines, at positions 54 and 97, and no disulfide bonds. By directed mutagenesis of the cloned T4 lysozyme gene, we replaced Ile-3 with Cys. Oxidation in vitro generated an intramolecular disulfide bond; proteolytic mapping showed this bond to connect Cys-3 to Cys-97. While this molecule exhibited substantially more stability against thermal inactivation than wild type, its stability was further enhanced by additional modification with thiol-specific reagents. This and other evidence suggest that at basic pH and elevated temperatures Cys-54 is involved in intermolecular thiol/disulfide interchange with the engineered disulfide, leading to inactive oligomers. Mutagenic replacement of Cys-54 with Thr or Val in the disulfide-cross-linked variant generated lysozymes exhibiting greatly enhanced stability toward irreversible thermal inactivation.     

Crystal structure of the double azurin mutant Cys3Ser/Ser100Pro from Pseudomonas aeruginosa at 1.8 A resolution: its folding-unfolding energy and unfolding kinetics

Azurin is a cupredoxin, which functions as an electron carrier. Its fold is dominated by a beta-sheet structure. In the present study, azurin serves as a model system to investigate the importance of a conserved disulphide bond for protein stability and folding/unfolding. For this purpose, we have examined two azurin mutants, the single mutant Cys3Ser, which disrupts azurin's conserved disulphide bond, and the double mutant Cys3Ser/Ser100Pro, which contains an additional mutation at a site distant from the conserved disulphide. The crystal structure of the azurin double mutant has been determined to 1.8 A resolution(2), with a crystallographic R-factor of 17.5% (R(free)=20.8%). A comparison with the wild-type structure reveals that structural differences are limited to the sites of the mutations. Also, the rates of folding and unfolding as determined by CD and fluorescence spectroscopy are almost unchanged. The main difference to wild-type azurin is a destabilisation by approximately 20 kJ x mol(-1), constituting half the total folding energy of the wild-type protein. Thus, the disulphide bond constitutes a vital component in giving azurin its stable fold.     

The influence of active site loop mutations on the thermal stability of azurin from Pseudomonas aeruginosa

The copper site and overall structures of azurin (AZ) variants in which the amicyanin (AMI) and plastocyanin (PC) metal binding loops have been introduced, AZAMI and AZPC, respectively, are similar to that of AZ, whereas the loop conformations resemble those in the native proteins. To assess the influence of these loop mutations on stability, the thermal unfolding of AZAMI and AZPC has been investigated by differential scanning calorimetry, absorption and fluorescence spectroscopy. The calorimetric profiles of both variants exhibit a complex shape consisting of two endothermic peaks and an exothermic peak. The temperature of the maximum heat of absorption for the single endothermic peak is 82.7°C for AZ, whereas for AZAMI and AZPC the most intense endothermic peaks are at 74.9 and 68.1°C comparable to values for AMI and PC, respectively. Denaturation investigated using the temperature dependence of the absorbance at ～600nm and Trp emission, also demonstrates decreased stability for both loop mutants. The thermal transition between the native and the denaturated states is irreversible, scan rate dependent and consistent with the two-state irreversible model. The structure of the active-site loop has a dramatic effect on the kinetic stability and the unfolding pathway of cupredoxins.     

Heat stability of the potato tuber ADP-glucose pyrophosphorylase: role of Cys residue 12 in the small subunit

Most of the ADP-glucose pyrophosphorylases from different sources are stable to a heat treatment. We found that in the potato (Solanum tuberosum L.) tuber enzyme, the intermolecular disulfide bridge located between Cys12 of the small subunits is responsible for the stability at 60 degrees C. When this unique disulfide bond is cleaved the enzyme is stable up to 40 degrees C. Mutation of Cys12 in the small subunit into either Ala or Ser yielded enzymes with stability similar to the reduced form of the wild type. Concurrently, the enzyme with a truncated small subunit on the N-terminal was stable only up to 40 degrees C. Thus, the N-terminal is important for the stability of the enzyme because of the presence of a disulfide bond.     

Equilibrium unfolding of Bombyx mori glycyl-tRNA synthetase

Unfolding of Bombyx mori glycyl-tRNA synthetase was examined by multiple spectroscopic techniques. Tryptophan fluorescence of wild type enzyme and an N-terminally truncated form (N55) increased at low concentrations of urea or guanidine-HCl followed by a reduction in intensity at intermediate denaturant concentrations; a transition at higher denaturant was detected as decreased fluorescence intensity and a red-shifted emission. Solute quenching of fluorescence indicated that tryptophans become progressively solvent-exposed during unfolding. Wild type enzyme had stronger negative CD bands between 220 and 230 nm than the mutant, indicative of greater alpha-helical content. Urea or guanidine-HCl caused a reduction in ellipticity at 222 nm at low denaturant concentration with the wild type enzyme, a transition that is absent in the mutant; both enzymes exhibited a cooperative transition at higher denaturant concentrations. Both enzymes dissociate to monomers in 1.5 m urea. Unfolding of wild type enzyme is described by a multistate unfolding and a parallel two state unfolding; the two-state component is absent in the mutant. Changes in spectral properties associated with unfolding were largely reversible after dilution to low denaturant. Unfolding of glycyl-tRNA synthetase is complex with a native state, a native-like monomer, partially unfolded states, and the unfolded state.     

Design and synthesis of the pseudo-EF hand in calbindin D9K: effect of amino acid substitutions in the alpha-helical regions

A series of 37-residue analogues of the pseudo-EF hand in bovine calbindin D9K has been synthesized by the solid phase method. In the presence of calcium an alpha-helical induction of up to 44% was observed for the peptide with the native sequence with a Kd for calcium binding of 0.35 mM. A number of amino acid substitutions have been carried out to study the packing of the two alpha-helices based on the crystal structure of the entire protein. Three strategies were employed: (1) replacement of the Leu residues, which in the crystal structure do not contribute to the hydrophobic interaction between the two helices, by Gln or Ala in order to control the orientation of the helix packing, (2) stabilization of the individual helix by introducing a Glu-...Lys+ salt bridge or by changing the N-terminal charge to compensate for the helix dipole moment, and (3) introduction of a disulfide bond between the two helices to help the packing of the helices. The mutants with the substitution of (Leu-30, Leu-32) to (Gln-30, Gln-32), (Gln-30, Ala-32), and (Ala-30,Ala-32) designed based on the strategy 1 do not show any affinity for calcium and have low alpha-helicity. The Leu-30 to Lys-30 mutant designed to form a salt bridge between the side chains of Glu-26 and Lys-30 has an apparent Kd for calcium of 6.8 mM. Kd of the N-terminal acetylated and succinylated mutants are 0.41 and 0.45 mM, respectively, and no increase in the alpha-helix content relative to that of the natural sequence peptide is observed. The disulfide containing mutants, namely Tyr-13, Leu-31 to Cys-13, Cys-31 and Tyr-13, Leu-31 to Cys-13, hCys-31, show apparent Kd values of 0.93 and 2.1 mM, respectively. The former mutant shows the highest alpha-helix content among the peptides studied in the presence and absence of calcium. While it is difficult to construct an isolated and rigid helix-loop-helix motif with peptides of this size, introduction of a disulfide bond proved to be effective for this purpose.     

Dramatic stabilization of the native state of human carbonic anhydrase II by an engineered disulfide bond

To find a disulfide pair that could stabilize the enzyme human carbonic anhydrase II (HCA II), we grafted the disulfide bridge from the related and unusually stable carbonic anhydrase form from Neisseria gonorrhoeae (NGCA) into the human enzyme. Thus, the two Cys residues at positions 23 and 203 were engineered into a pseudo-wild-type form of HCA II (C206S), giving the mutant C206S/A23C/L203C. The disulfide bond was not formed spontaneously. The native state of the reduced form of the mutant was markedly destabilized (2.9 kcal/mol) compared to that of HCA II. Formation of a disulfide bridge was achieved by treatment by oxidized glutathione. This led to a significant stabilization of the native conformation. Compared to HCA II the unfolding midpoint for the variant was increased from 0.9 to 1.7 M guanidine HCl, corresponding to a stabilization of 3.7 kcal/mol. This makes the human enzyme almost as stable as the model protein NGCA, for which the unfolding of the native state has a midpoint at 2.1 M guanidine HCl. The stabilized protein underwent, contrary to all other investigated variants of HCA II, an apparent two-state unfolding transition, as judged from intrinsic Trp fluorescence measurements. A molten-globule intermediate is nevertheless formed but is suppressed because of the high denaturant pressure it faces upon rupture of the native state.     

Enthalpic destabilization of a mutant human lysozyme lacking a disulfide bridge between cysteine-77 and cysteine-95.

To understand the role of disulfide bridges in protein stability, the thermodynamic changes in the denaturation of two mutant human lysozymes lacking a disulfide bridge between Cys-77 and Cys-95 (C77A and C77/95A) were analyzed using differential scanning calorimetry (DSC). At pH 3.0 and 57 degrees C, the stabilities of both the C77A and C77/95A mutants were decreased about 4.6 kcal.mol-1 in Gibbs free energy change. Under the same conditions, the enthalpy changes (delta H) were 94.8 and 90.8 kcal.mol-1, respectively, which were smaller than that of the wild type (100.8 kcal.mol-1). The destabilization of the mutants was caused by enthalpic factors. Although X-ray crystallography indicated that the mutants preserve the wild-type tertiary structure, removal of the disulfide bridge increased the flexibility of the native state of the mutants. This was indicated both by an increase in the crystallographic thermal factors (B-factors) and by a decrease in the affinity of N-acetylglucosamine trimer [(NAG)3] observed using isothermal titration calorimetry (DTC) due to entropic effects. Thus, the effect of cross-linking on the stability of a protein is not solely explained by the entropy change in denaturation.     

Tyrosine emission in the tryptophanless azurin from Pseudomonas fluorescens.

A strain of Pseudomonas fluorescens contains an azurin with no tryptophan and two tyrosines. This protein is interesting because it allows one to study both the structure of azurin and the emission of tyrosines in proteins. Comprehensive measurements were carried out including spectrophotometric and fluorimetric titration, fluorescence quantum yield, fluorescence polarization, and I- quenching. In the copper-containing protein, almost independent of the copper ion oxidation, the fluorescence quantum yield is approximately 60% of that of the apoprotein. The latter has the remarkable property that its quantum yield is even greater than free tyrosine. The two tyrosines in the metalloprotein have different pKa's, 10.75 and 12.78, but there is only one average pKa, 10.9 in the apoprotein. The polarization of the fluorescence at 310 nm (290-nm excitation) is 0.32 for the metalloproteins and 0.34 for the apoprotein. I- hardly quenches the fluorescence. The conclusion is that the two tyrosines are inaccesible to the solvent, located in nonpolar environments, larger than or equal to 20 A apart, and not adjacent to the disulfide bridge.     

Hydrophobic core variant ubiquitin forms a molten globule conformation at acidic pH

The conformational properties of hydrophobic core variant ubiquitin (Val26 to Ala mutation) in an acidic solution were studied. The intrinsic tryptophan fluorescence emission spectrum, far-UV and near-UV circular dichroic spectra, the fluorescence emission spectrum of 8-anilinonaphthalene-1-sulfonic acid in the presence of V26A ubiquitin, and urea-induced unfolding measurements indicate this variant ubiquitin to be in the partially folded molten globule conformation in solution at pH 2. The folding kinetics from molten globule to the native state was nearly identical to those from the unfolded state to the native state. This observation suggests that the equilibrium molten globule state of hydrophobic core variant ubiquitin is an on-pathway folding intermediate.     

毛细管电泳测定牛红细胞超氧化物歧化酶的巯基位点

本文介绍了毛细管电泳分析蛋白质酶解产物中含巯基多肽的方法。还原的及天然的牛红细胞超氧化物歧化酶(SOD)经4-乙烯吡啶修饰后,由TPCK-胰蛋白酶水解,在254nm检测到还原的SOD水解物中含3个巯基多肽,天然的SOD为1个疏基多肽且其毛细管电泳行为与上述3个多肽之一相一致。分析它们的氨基酸顺序,证实Cys-6为游离的巯基,Cys-55和Cys~(-144)形成二硫键。