| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
-barrel metalloprotein
1 Instituto de Biología Molecular y Celular, Universidad Miguel Hernández, 03202-Elche (Alicante), Spain
2 Center of Magnetic Resonance (CERM), University of Florence, 50019 Sesto Fiorentino (Florence), Italy
3 Departamento de Química Inorgánica, Universitat de València, 46100 Burjassot (Valencia), Spain
Reprint requests to: Antonio Donaire, Instituto de Biología Molecular y Celular, Universidad Miguel Hernández, Edificio Torregaitán, Avda. de la Universidad s/n, 03202-Elche (Alicante), Spain; e-mail: adonaire{at}umh.es; fax: +34-96-6658758.
(RECEIVED January 5, 2005; FINAL REVISION March 29, 2005; ACCEPTED April 8, 2005)
 |
Abstract |
|---|
 TOP Abstract IntroductionResultsDiscussionMaterials and methodsReferences |
|---|
-helix) and the last 
-strand (where the other three ligands of the metal ion are located) form the most rigid domain of the protein. The results taken as a whole suggest that the first ligand detaches from the metal ion when the protein unfolds, while the other three ligands remain bound to it. The implications of these findings for the biological folding process of Rc are also discussed. Keywords: rusticyanin; blue copper protein; NMR; protein unfolding; protein dynamics
Abbreviations: Az, azurin • BCP, blue copper protein • GdmCl, guanidinium chloride • NMR, nuclear magnetic resonance • NOE, nuclear Overhauser effect • Rc, rusticyanin
Article and publication are at http://www.proteinscience.org/cgi/doi/10.1110/ps.051337505.
|
Introduction |
|---|
|
TOPAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Metalloproteins, in general, and copper proteins, in particular, are stabilized to a large degree by the prosthetic group (i.e., the metal ion). In principle, this stabilizing effect occurs independently of the role of the metal ion in the function of the macromolecule. Blue copper proteins (BCPs) are relatively small, soluble electron transfer proteins (Malmström and Leckner 1998; Gray et al. 2000; Randall et al. 2000; Vila and Fernandez 2001). They all possess a -barrel structure and are arranged in a so-called Greek key topology, as shown in Figure 1
. The redox potential of the copper ion is higher in BCPs than in aqueous solution, i.e., the reduced form is more stable than the oxidized species. BCPs are highly rigid. In fact, values of the generalized order parameter, S2, obtained from NMR data, range from 0.83 to 0.93 (Kalverda et al. 1999; Thompson et al. 2000; Bertini et al. 2001a; Jiménez et al. 2003a). This rigidity is due to the large network of hydrogen bonds extending between the -strands of these proteins, which in turn is necessary to fulfill their function (electron transport). The active site is particularly rigid. The copper ion is strongly bound to a sulfur atom of a cysteine ligand and to the imidazol nitrogen atoms of two histidine residues (HisN and HisC, respectively, according to their proximity in the sequence to the N- and C-terminal ends). In most of these proteins, an axial sulfur atom of a methionine, weakly bound to the metal, completes the pseudotetrahedral coordination (Randall et al. 2000) (see Fig. 1). This geometry, which remains essentially unaltered upon oxidation or reduction of the copper ion, is imposed by the protein scaffold. This is not the preferred coordination for the copper(II) ion and thus, the metal is found in a strained conformation (entatic/rack mechanism) (Malmström 1994). This characteristic permits a low re-organization energy, which is required for an efficient redox process (Gray et al. 2000; Crane et al. 2001). Traditionally, the strained coordination has also been related to the high redox potential of the BCPs (Malmström 1964, 1994). This concept was reviewed since the redox potential of the BCP Az is higher in the unfolded than in the folded state (Winkler et al. 1997; Wittung-Stafshede et al. 1998; Marks et al. 2004). Recently, an elegant theoretical study factorized the structural determinants of the relative redox potentials of the blue copper sites (Li et al. 2004). In this study, the investigators conclude that the entatic/rack mechanism is one of the major determinants of the relative redox potentials in BCPs. These findings lead us to consider the function of the metal ion in maintaining protein stability and, in turn, the role of protein folding in maintaining each oxidation state of the metal ion (i.e., in governing the redox potentials) in BCPs.
|
We have studied the role of the copper ion in the protein stability and dynamic features of Rc by means of NMR. Rc is unique among the BCPs. In fact, Rc possesses the highest redox potential of the BCP family, 680 mV (Nunzi et al. 1994). In other words, it is particularly efficient in stabilizing the copper(I) ion. Rc is also extremely stable at acidic pH values (<2.5) (Cox and Boxer 1978; Ingledew and Cobley 1980). The existence of a highly hydrophobic core in the molecule could explain, at least partially, these two singularities (Botuyan et al. 1996; Walter et al. 1996; Donaire et al. 2001). Nevertheless, the correlation between the structural and thermodynamic properties of Rc is not well understood and is still a matter of debate (Olsson et al. 2003; Li et al. 2004). To date, the role played by the metal ion and its oxidation state in maintaining protein stability is not well understood either. Hence, Rc represents an extremely interesting case in which to identify the critical properties of BCPs that make these proteins especially efficient as electron mediators.
In this study we describe first, an NMR titration of apo, Cu(I), and Cu(II)Rc with guanidinium chloride (GdmCl); second, direct evidence of copper coordination (in both oxidation states) to Rc in its completely unfolded state; and third, a study of the dynamic and solvent exchange properties of the protein in an incipient or early step of the unfolding process. The former experiments help to interpret the role of the copper ion in maintaining the protein scaffold and the way in which the protein folds. The latter provide clues as to the secondary structural elements involved in starting the unfolding process. We recently described the formation of aggregated species in the unfolding process of Rc with GdmCl, using several biochemical and spectroscopic methods (Alcaraz and Donaire 2004). Here, the aggregation process is again evidenced. The present results have allowed us to identify the regions that are more resistant to the unfolding process and the structural features responsible for Rc stability.
|
Results |
|---|
|
TOPAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
shows the 1H-15N HSQC spectra of Cu(I)Rc at GdmCl concentrations of 0.0, 2.1, and 6.0 M. The addition of denaturant at low concentrations causes small changes in the 1H and 15N chemical shifts (in the fast exchange regime). This clearly indicates that the essential secondary and tertiary structural elements of the protein are maintained at this GdmCl concentration. In these conditions, the protein scaffold remains essentially unaltered as compared with the completely folded species. This represents a monomeric, essentially folded, species. It is essentially folded because its chemical shifts are substantially coincident with those of the completely folded Rc (see Fig. 2B). It is monomeric since the signals do not present line broadening. On the other hand, the completely unfolded species displays an HSQC spectrum typical of random coil proteins (Fig. 2C).
|
, displays the differences in the 1H and 15N chemical shifts of the folded and incipient unfolded species for each HN pair in the reduced protein (the pattern is analogous for the apo and the oxidized forms; data not shown). The largest variations are observed in Gly28, in the middle (Ile89, Ile90, Asp103, and Val106), and at the end (Gly118) of the long loop L9 (the loop that links strands 8 and 9) (see Fig. 1
); at the beginning of the -strand 9 (Phe120, Gly121, Tyr122, and Thr123); and in some residues of the -strand 8 (Val74, Thr75, and Phe76). Thus, these three secondary structural elements (-strands 8 and 9, and the long loop L9) have a manifest tendency to change when the protein starts to open. It is remarkable that both strands, 8 and 9, face each other in the three-dimensional structure of Rc (see Fig. 1). Other groups that experience changes are those belonging to the end of the strand 4 (Ala42, Ala43, and Leu46). Another three residues (Asp58, His128, and Thr130) also experience observable changes in their chemical shifts. They are close (sequentially and spatially) to the two free (nonligand) histidines of the protein (His57 and His128). The changes in their chemical shifts are probably related to protonation–deprotonation phenomena of both histidines at the working pH value.
|
-barrel structure. Moreover, most of them increase the chemical shifts of their amide protons, indicating an increment in the strength of the hydrogen bonds. Only four of these amide protons (Gly28, Asp103, Gly118, and Gly121) are exposed to the solvent and do not form hydrogen bonds. These changes may be due to direct interactions with molecules of the denaturant agent.
Unfolded protein/slow exchange
Subsequent increments in the denaturant agent concentration produce substantial changes in the intensity of the peaks observed and, at a particular concentration of GdmCl (depending on the form studied), a new set of signals starts to appear in slow exchange regime. As observed in Figure 2C, the low spread of 1H and 15N chemical shifts in the last set of signals (at GdmCl 6.0 M) clearly corresponds to an unfolded species. Complete extinction of the folded species is reached at this GdmCl concentration for Cu(I)Rc. For the other two forms, apo and Cu(II)Rc, essentially the same phenomena are observed, although they occur at different concentrations of the denaturant agent.
Figure 4 (symbol 
) shows the molar fraction of the peak corresponding to the well-resolved signal of Ile155 (see Fig. 2A) as a function of the GdmCl concentration for the apo (Fig. 4A), reduced (Fig. 4B), and oxidized (Fig. 4C) protein. Another 20 well-resolved HN pairs distributed throughout the protein were analyzed and very similar results obtained (data not shown). Thus, this process takes place at the same time throughout the molecule. Three regions can be discerned in each of these curves. First, the peak volumes drastically decrease (until they reach ~55%–60% of the original volume) as the GdmCl concentration increases from 0 to 0.7 M. This last concentration is independent of the form of Rc considered. This decrease is not accompanied by the appearance of any new set of signals. Therefore, it must be related to the formation of a "silent" species (from the NMR point of view). By means of gel filtration and cross-linking experiments, we recently showed that aggregated species are formed when GdmCl is added to Rc (Alcaraz and Donaire 2004). The loss of intensity observed here is related to the presence of these aggregated species, whose proton signals are too broad to be detected. Second, a less pronounced (almost flat) decrease in the peak volumes is observed for GdmCl concentrations between 0.7 and 1.0M for the apo form; 0.7 and 1.7 M for the reduced form; and 0.7 and 2.2 M for the oxidized form. Finally, a sigmoid-like behavior is observed in the last part of each curve. Signals corresponding to the incipient unfolded species completely disappear at GdmCl concentrations of 2.5 M for apoRc and 5.5 M for both holoproteins (Fig. 4, symbol ). The midpoints for the sigmoid curves are 1.7 (apoRc), 3.0 (reduced Rc), and 4.0 M (oxidized Rc). The present data clearly indicate that the folded apoRc is less stable (compared with the unfolded species) than both forms of the holoprotein. In turn, Cu(II) is more efficient than Cu(I) in stabilizing the folded state. The same stabilizing effect of the metal ion has been observed in azurin (Leckner et al. 1997a, b; Marks et al. 2004).
|
the normalized intensity of a well-resolved signal in the unfolded state (peak labeled with the symbol * in Fig. 2C) is also shown (symbol •). All the analyzed signals belonging to the unfolded species behave in a similar way. For the three forms of Rc (apo, reduced, and oxidized), these peaks start to appear at GdmCl concentrations of ~1.3 M, although the pattern of the volumes differs depending on the concentration of GdmCl. For the apoRc (Fig. 4A), a well-delineated sigmoid is observed with the midpoint at 2.0 M, and titration is complete at 2.8 M. In the case of Cu(I)Rc, an ill-defined step is observed (Fig. 4B), and titration is complete at 6.0 M. For the oxidized form (Fig. 4C), titration is almost, but not quite, complete at the last point (higher GdmCl concentrations could not be reached because dilution of the protein prevented HSQC spectra from being obtained). It may be deduced from this figure that there are two steps in this titration: The first step is due to the formation of aggregated species and the second to the unfolding process itself. Probably the same situation is found in the Cu(I)Rc unfolding process (Fig. 4B), although here both processes overlap (and this is why the ill-defined step is observed) (see above). However, in the oxidized state the situation is even more complicated since it is well known that the copper(II) ion reacts slowly with the thiol atom of the cysteine ligand in inorganic complexes or in the unfolded form of Cu(II)Az (Leckner et al. 1997a; Winkler et al. 1997). An analogous reaction could take place in the Rc unfolding process at the final points of the titrations. This slow process could affect the measurement of the two final points of the Cu(II)Rc titration (at very high GdmCl concentrations). Due to the presence of different species (aggregates) in solution, the interpretation of these data in a quantitative way (i.e., to extract thermodynamic data) is not straightforward at this stage and these data can only be analyzed in a qualitative way.
Final points for the apo and holoproteins
The final points for apo, Cu(I), and Cu(II)Rc are shown in Figure 5. The three spectra are superimposable, although a few peaks (labeled with * in Fig. 5B) show a very small intensity in the Cu(II)Rc spectrum or even disappear (signals labeled with lowercase letters). This indicates that the copper(II) ion is bound in the unfolded state. In fact, the signals belonging to the protons close to the metal ion decrease their intensity due to the paramagnetic effect (Bertini et al. 2001b; Donaire et al. 2002). Further evidence of copper(II) coordination in the unfolded state is given by the 1D 1H superweft spectrum of the oxidized Rc in these conditions (Fig. 5C). The two broad signals observed at –22.5 (signal e) and –75 ppm (signal f) can only derive from protons with contact contribution to their isotropic shifts; i.e., they belong to ligands of amino acids covalently bound to the copper(II) ion (Donaire et al. 2002). Coordination of the reduced copper ion is also supported by comparison of the three HSQC spectra. In fact, the signals that disappear in the Cu(II)Rc spectrum (signals a–d in Fig. 5B) are the ones that change their position from the apo to the Cu(I)Rc spectrum (Fig. 5A). This strongly suggests that the copper(I) ion is also bound in the completely unfolded species.
|
Solvent accessibility in the incipient unfolded species
A lyophilized Cu(I)Rc sample was dissolved in D2O in the presence of 2.1 M GdmCl, and then, a set of 1H-15N HSQC experiments were performed consecutively. One hundred and three peaks corresponding to the incipient unfolded species were observed in the first experiment. Figure 3D displays the protection factors of these amide protons. Most of the HN protons with low protection factors belong to loop regions (including the long loop L9). Regarding the secondary structural elements, the -strands 6 and 10 have a small tendency to exchange their amide protons with the solvent and strands 5 and 7, a moderate tendency, while the most protected protons belong to the rest of the -strands.
It is interesting to note that all the residues belonging to the last -strand as well as those from the short -helix 2 (both structures are close to the metal ion and contain the last three ligands of the metal) posses amide protons with medium or elevated protection factors. Thus, the hydrogen bonds of these amino acids are relatively hidden from the solvent; i.e., these two structures are highly robust in this early step of denaturation.
In the same Figure 3D the results of the cleanex (Hwang et al. 1998) and the ePHOGSY (Dalvit and Hommel 1995; Dalvit 1996; Bertini et al. 1997) experiments performed in a reduced sample at 2.1 M GdmCl are also shown. The cleanex experiment (Supplemental Material) essentially displays the same cross peaks as those previously found for the native protein (Jiménez et al. 2003a). The most remarkable feature of this spectrum is the three strong cross peaks corresponding to the amide protons of Ser34 and Gly28 (found in loops L3 and L4, respectively) and to the His143 H
1 proton.
Peaks observed in the ePHOGSY experiment (Supplemental Material) (Dalvit and Hommel 1995; Dalvit 1996) may arise from a direct NOE between protein groups and a water molecule "trapped" within the protein frame. However, as already described, other effects may contribute to ePHOGSY peaks in the absence of a real HN-water NOE (Bertini et al. 1997). These are spin diffusion effects during the mixing time, the unwanted NOEs arising from H protons with the same chemical shifts of bulk water, and exchanged relayed NOEs. When all these effects have been properly taken into account, only the Gly142 and the Val145 amide protons unambiguously arise from direct HN–water NOEs. When we inspect the Rc crystallographic structure, 1rcy file (Walter et al. 1996), only the water molecule numbered 212 forms a hydrogen bond with the amide protons of these two amino acids, as well as with the carbonyl oxygen of Val98. Therefore, this structural water molecule (that is still present in the incipient unfolded state) links with the loop L11 and the beginning of the helix 2 on one side, and with the loop L9 on the other (see Fig. 1).
Mobility studies
Mobility studies of the protein backbone in the incipient unfolded state (GdmCl 2.1M) were performed (see Fig. 2B). Measurements of the 15N relaxation rates and 1H-15N NOE values were carried out in a reduced Rc sample. Figure 6 shows the experimental data. The relaxation values of 100 of the 140 HN pairs of this protein (Rc possesses 155 amino acids, 14 of which are proline residues) were analyzed. Average values for the longitudinal and transversal relaxation rates (eliminating mobile residues as indicated in Materials and Methods) were 1.6 ± 0.2 and 10.5 ± 1.5 sec–1, respectively (R2/R1 ratio of 6.5 ± 1.7 sec–1). The average value of the 1H-15N NOEs was 0.74 ± 0.14. The signals belonging to the residues Thr2, Leu3, Ala44, Thr79, and Gly108 showed R1 values significantly lower than the average (Fig. 6A). The residues Thr2, Leu3, Thr11, Lys36, Gly69, Thr79, Ala97, and Gly108 displayed low R2 values (Fig. 6B). Of these, residues Thr2, Leu3, Thr79, and Gly108 also gave negative NOE values (Fig. 6C). These regions experience motions on the pico- to nanosecond timescale. On the contrary, the residues His57, Asp58, Leu64, Glu65, Ala70, and Trp127 displayed R2 values higher than the average value (Fig. 6B); i.e., they experience conformational exchange on the micro- to millisecond timescale. His57 has been shown to undergo a protonation/deprotonation process at the working pH value in the folded state (Jiménez et al. 2003a). Leu64 and Glu65 belong to the short -strand 7, while Ala70 is located on the short loop L8 (connecting strands 7 and 8) (see Fig. 1). Consequently, the -strand 7 is one of the domains with a propensity to change to an open conformation.
|
The relaxation parameters were fitted to the dynamic isotropic model (Mandel et al. 1995). No statistical improvement was observed when more independent variables were added (axially symmetric and totally anisotropic models). Once the isotropic model for the global protein was assumed, the relaxation data for each individual HN peak were fitted to the five models previously described (Mandel et al. 1995). The values of the generalized order parameter (S2) per residue are shown in Figure 7A. The average value (discarding those whose standard deviation was >2
) was 0.83 ± 0.07, slightly but significantly slower than that reported by us for the native protein, 0.93 ± 0.03 (Jiménez et al. 2004). Fifty-seven of the 100 residues were fitted according to model 1 and 28 were fitted to model 2 (i.e., they present mobility on the subnanosecond timescale). Thirteen residues presented conformational exchange (they belong to model 3) and two of them (Thr79 and Tyr135) experienced both kinds of motions (model 4). Figure 7B displays the values of the correlation times of the internal motions, 
e, and the exchange rates, Rex, for these residues (Mandel et al. 1995).
|
), with values of S2 <0.4. It is interesting to note that Thr79 is at the beginning of the long loop L9 (see Fig. 1), which lacks a defined secondary structure. It is also noticeable that His85, the first ligand of the copper ion in the sequence, is sequentially and spatially close to this region. Ala97 and Ile102 (Fig. 7A) are another two amino acids with small order parameters (0.68 and 0.70, respectively), while Val101 experiences conformational exchange (see Fig. 7B). All these residues belong to the same long loop L9 and are close to a region rich in proline residues (residues 94, 95, 100, and 104 are prolines). These data taken as a whole indicate that the surroundings of this region have a propensity to experience mobility on a large range of timescales (some on the pico- or nanosecond timescale, others on the micro- to millisecond timescale) (see Fig. 7B). Taking into account the fast exchange properties of the loop L9 (see Fig. 3D), we conclude that (excluding the N-terminal end) the region rich in proline residues is the most flexible region of the protein.
Other residues with S2 values lower than the average are Ala44, Leu46, and Phe51 (Fig. 7A). They are either at the end of the -strand 4 or on the following loop (L5). These are two secondary structural elements with HN protons that exchange easily (see Fig. 3D). This loop contains three prolines (amino acid numbers 47, 50, and 52) and a glycine (Gly48), which are typical residues with a predisposition to undergo conformational movements. Finally, Thr130 also displays a generalized order parameter (0.69) lower than the average, which could be related to the protonation/deprotonation equilibrium of the nearby His128.
|
Discussion |
|---|
|
TOPAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
When GdmCl is added to a solution containing Rc, the first observation is a drastic decrease in the volume of the peaks (see Fig. 4A–C). Integrals of these volumes (corrected by the dilution factors) are very different from those of the initial signals. Thus, nondetected species (i.e., with fast transversal relaxation rates) must be responsible for the percentage of the signal lost. By means of gel filtration chromatography and glutaraldehyde cross-linking experiments, we previously demonstrated the existence of aggregated species in Rc when GdmCl is present (Alcaraz and Donaire 2004). We conclude that these aggregates are present in solution in our NMR titrations. It should be pointed out that these aggregates are in a slow regime exchange with other species, since no changes either in the line broadening or in the relaxation properties of the signals are observed. This suggests that such an aggregation process, affecting the molecule as a whole, must be relatively slow on the NMR timescale. In turn, these aggregates must be made up of at least four monomers per molecule (since simpler structures would be observed). Taking into account these considerations, expression 2 (see Materials and Methods) is sufficiently justified. From this equation, the molar fraction of all species (including aggregated species) can be deduced. In Figure 4, the percentages of the aggregated species are also displayed.
The unfolding processes of other BCPs have been studied (Leckner et al. 1997b; Capaldi et al. 1999; Bai et al. 2001; Mizuguchi et al. 2003; Pozdnyakova and Wittung-Stafshede 2003; Sandberg et al. 2003) and no aggregation forms were found. Nevertheless, it is well known that similar aggregates have been identified in -barrel proteins (Kobayashi et al. 2000; Kjellsson et al. 2003). The presence of these states in Rc is probably related either to the high content in hydrophobic residues of this BCP or to the fact that this is the largest BCP. Indeed, the protein size is a crucial parameter to discern between proteins that unfold in a two-step process from those that unfold via intermediates (Dobson 2003).
Hints from the unfolding process
Figure 8 classifies the secondary structural elements of Rc according to the parameters derived from our NMR experiments. High P-factors and high S2 values indicate residues hidden from the solvent and/or belonging to highly rigid substructures of the protein. In contrast, low protection factors and low generalized order parameters, as well as the existence of cleanex cross peaks, designate the regions exposed to the solvent and lacking a robust structure. The information that the variations in the chemical shifts provides is more ambiguous. Large differences in this parameter derive from chemical changes in the surroundings of the amide protons. On the one hand, amide protons exposed to the solvent (such as those belonging to loop regions or to glycine residues) can modify their chemical shifts simply because of the change in the solvent polarity when passing from water to GdmCl. This would be the case of the amino acids close to the regions rich in proline residues (loops L5 and L9). On the other hand, amide protons belonging to the strands 5, 8, and 9 (see Fig. 8) probably increase the strength of the hydrogen bonds in which they participate. In effect, the average chemical shift variations found here are due, in most cases, to an increment toward the downfield regions of the proton chemical shifts. Then, these chemical shift variations reflect -strands that interact more strongly with each other when the protein starts to unfold. In turn, the large mobility that the loops L5 and especially L9 experience in the incipient unfolded state probably permits the interaction between these three chains to increase.
|
) unequivocally indicate that two domains are the first to undergo the unfolding processes: the N-terminal end and the loop L9 (mainly in the region rich in prolines). Another two regions have a similar, although not so pronounced, behavior: On the one hand, the loop L5, which also possesses a large degree of mobility on the pico- to nanosecond timescale (see Fig. 7B); and, on the other hand, the loop L8, close to Ala70, which experiences mobility on the micro- to millisecond timescale (see Fig. 7B).
In contrast to these mobile, accessible regions, the most rigid, buried structures of the protein correspond to the -strands (especially -strands 9 and 11) and to the short -helix 2. Two of the ligands of the metal ion (His143 and Met148) are located in these secondary structural elements. This means that the metal ion interactions are essential in maintaining this structure (as is also deduced from the corresponding titrations) (see Fig. 4).
Role of the copper ion
NMR titrations unequivocally demonstrate the stabilizing effect of the metal ion in Rc unfolding. Formation of the completely unfolded state is dependent on the metal ion and its oxidation state (see Fig. 4A–C). This confirms that not only the global tertiary conformation of the protein but also the metal ion architecture itself is modified in this step. In the case of the apo form, complete formation of the unfolded species is reached at 2.8 M GdmCl (see Fig. 4A), while a 6.0 M concentration is necessary to obtain the same result for Cu(I)Rc. With the oxidized protein, some nonnegligible aggregation exists (Fig. 4C), even at the last point of the titration (6.0 M). The midpoints of the titrations follow the order Cu(II)>Cu(I)>apo (Fig. 4, symbol •). Consequently, the same order is maintained for the stabilizing effect in the folded as compared with the unfolded forms. In other words, the copper ion provides Rc with an excess of thermodynamic stability compared with the apo form. In this respect, the oxidized state seems to be more effective than the reduced one.
The stabilizing effect of the metal ion is also corroborated by the fact that the domain containing the metal ion (i.e., the -helix 2 and the strands 10 and 11) is one of the most rigid regions of the protein. Also, the hydrogen bonds in this domain present slow exchange. Therefore, these secondary structural elements are strongly bound to each other.
The thermodynamic cycles of the unfolding process for the apo, Cu(I), and Cu(II) forms of Az have been studied in detail (Marks et al. 2004; Wittung-Stafshede 2004). In Az, copper(II) stabilizes the folded form more efficiently than the reduced species does. At first glance, this result seems contradictory to the fact that the azurin redox potential is higher than that of the Cu(II)/Cu(I) pair in aqueous solution; i.e., the reduced state is stabilized to a greater extent than the oxidized state in the folded form (Vila and Fernandez 2001). The apparent paradox vanishes if we take into account that the reduced and the oxidized forms of the completely unfolded conformations of Az are different (Pozdnyakova et al. 2001; Marks et al. 2004), so no direct comparison between the two unfolded titrations can be made (both their starting and final points are different). Our present results for Rc are analogous to those obtained for Az.
In the oxidized form of the unfolded Az, it is known that the copper ion is coordinated to a cysteine, a methionine, and a histidine ligand (HisC) (Pozdnyakova et al. 2001; Marks et al. 2004). These three ligands are on the same loop of the metal ion for all the BCPs (Buning et al. 2000; Donaire et al. 2002) (see Fig. 1). Figure 5 clearly shows that the copper ion (in both oxidation states) is coordinated to the unfolded Rc. From these spectra it is not possible to make deductions concerning ligands of the metal ion. However, other data here reported provide us with clues to the identity of such possible ligands. In fact, the last loop and the last -strand are precisely the least disordered regions of Rc in the incipient unfolded state (Fig. 7). Moreover, their amide protons exchange very slowly with the solvent (Fig. 8B). These two data taken together indicate that this is the most rigid domain of the protein. In turn, since the distinctive feature of this region is its coordination to the metal ion, and since the metal ion clearly maintains the stability of the protein (see Fig. 8), we suggest that these three amino acids (Cys138, His143, and Met148) are probably coordinated to the metal ion in the unfolded state of Rc.
It has been shown that HisN is not coordinated to the metal ion in the unfolded state of Az (Romero et al. 1998; Pozdnyakova et al. 2001). HisN in Rc, as in all BCPs (Buning et al. 2000), is located in another domain of the protein (see Fig. 1). Our relaxation data show that the domain where this ligand is located (the loop L9) is one of the most mobile regions of the protein. It is thus likely that this histidine is not coordinated to the metal ion in the completely unfolded form; i.e., Rc shows the same pattern as that observed in Az (Romero et al. 1998; Pozdnyakova et al. 2001). Moreover, the equivalent region to this long loop in the folded forms of Az (Kalverda et al. 1999), pseudoazurin (Thompson et al. 2000) and plastocyanin (Bertini et al. 2001a) also present a relatively high degree of mobility. Then this could be a general process in BCPs: The ligands close together in the amino acid sequence (HisC, and the cysteine and methionine residues, all sited on the last loop) remain bound to the metal ion when the protein unfolds, while the ligand far from the others in the primary structure (HisN) (see Fig. 1) disrupts its interaction in the unfolding process.
Relevance in the biological folding process
Two features can be deduced from the mechanism of Rc folding. The first is related to the role of the proline regions. The two most mobile regions of Rc (L5 and L9) posses a high content of proline residues. The correct cis–trans conformations of proline residues are decisive in the speed of the folding of the BCP plastocyanin (Dyson et al. 1992; Bai et al. 2001). In fact, an erroneous conformation of the proline residues gives rise to non-active structures in the protein folding process. The high flexibility of the proline regions in Rc results in a low energy barrier to correct bad conformations if such misfolding processes occur.
The second characteristic relates to the kinetics of the process and the uptake of the metal ion by the protein. In the case of the BCP Az, the speed of formation of the three-dimensional protein structure is dependent on the tertiary contacts of the -barrel (Pozdnyakova and Wittung-Stafshede 2002, 2003). It is also independent of the metal ion (Pozdnyakova and Wittung-Stafshede 2001b). However, formation of the active protein (the holoprotein) is closely related to the sequence of necessary events (protein folding and copper uptake). While in the case of Az, the binding of the copper is fast (of the order of milliseconds) for the unfolded protein, it is quite slow (from minutes to hours) for the folded protein (Pozdnyakova and Wittung-Stafshede 2001b). We obtained similar results with Rc by stopped-flow measurements (L. Alcaraz and A. Donaire, unpubl.). Thus, it was concluded that the copper ion bound to the protein prior to its folding (Wittung-Stafshede 2004). With our present data we can conclude that a similar behavior is likely to happen with Cu(II)Rc.
A doubt remains about the oxidation state of the copper ion in this process. In fact, in Az most of the evidence of copper coordination is related to the oxidized state. Nevertheless, copper(II) is toxic in the cell, and only the reduced form is found (O’Halloran and Culotta 2000). Our data (Fig. 5B) indicate that copper(I) is also bound in the unfolded state of Rc, thus making the mechanism proposed above possible.
|
Materials and methods |
|---|
|
TOPAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Atomic absorption
Quantitative analysis of copper in unfolded Rc was performed using a SpectraA-220FS (Varian) atomic absorption spectrometer equipped with an automated sampling system. Two Rc samples were metallated: One of them was oxidized and the other reduced. Excess metal and EDTA was eliminated from both as described above. Then, they were incubated with 6 M GdmCl for 1 h and dialyzed three times against the same buffer (3 h each). The final protein concentration of these samples was 7.0 µM. Atomic absorption spectroscopy was carried out with these two samples. The absorbance versus concentration (µg/L) calibration curve was previously determined following standard procedures.
NMR measurements
NMR experiments were performed at 296K in a Bruker Avance 500 spectrometer operating in a magnetic field of 11.7 Tesla. 1H and 15N frequencies were 500.13 and 50.68 MHz, respectively. The assignments of the peaks at a concentration of GdmCl of 2.1 M were achieved by following the chemical shifts from the native protein up to the incipient unfolded state present at that concentration of denaturant agent. A 3D 1H-15N HSQC-NOESY experiment (80 msec of mixing time) and a 3D 1H-15N HSQC-TOCSY (with 50 msec of spin lock) were also performed to assign some peaks in the most overcrowded regions. The carrier signal was set to the H2O signal in all the experiments.
The unfolding process was monitored by NMR. 1H-15N HSQC experiments (Bodenhausen and Ruben 1980) were recorded after addition of the corresponding aliquots of a solution of 8.0 M GdmCl to the Rc sample. For the oxidized protein, a superweft 1H experiment (Inubushi and Becker 1983) was also recorded at each point. In the first steps of the titrations an incipient unfolded state appears in fast exchange regime. Thus, the weighted average of the 1H and 15N chemical shifts can be calculated from the equation (Garrett et al. 1997):
 | (1) |
where the differences 
HN and N refer to changes in the chemical shifts of the amide protons and the nitrogen atoms, respectively, between the incipient unfolded and completely folded species.
For higher concentrations of the denaturant agent, the incipient unfolded and completely unfolded species appear in a slow exchange regime. The molar fractions of both species, 
F and U, were determined from the ratios VF/V0 and VU/V0, respectively. Here, V0 is the volume of the signal analyzed at the initial point of the titration (GdmCl 0.0 M) and VF and VU are the measured volumes of the cross peaks (corrected by the dilution factors) of the incipient and completely unfolded species, respectively.
The molar fraction of the aggregated species, A (see Discussion) was calculated from the expression:
 | (2) |
where the total molar fraction, 0, equals unity.
For the H2O/D2O exchange experiments, a sample of GdmCl was lyophilized in D2O. This deuterated GdmCl was used to prepare a stock solution of 2.1 M GdmCl in D2O. A sample of Rc, previously lyophilized, was dissolved in the GdmCl solution and immediately introduced in an NMR tube. Then, a 1H-15N HSQC spectrum of this freshly prepared sample was recorded (0 h). HSQC experiments in analogous conditions were recorded every 15 min for the first 8 h, every 2 h the following 2 d, and at increasing intervals over 2 wk.
The decay of the individual peak volume was fitted to the single exponential decay curve, I(t)=I(0) exp(Rexcht), where I(t) and I(0) are the volumes at a given time t and t=0, respectively, and Rexch is the rate constant of the H2O/D2O exchange reaction. The protection factors, P, were calculated as logP=log(Rint/Rexch), where Rint is the intrinsic exchange rate for the proton (Hvidt and Nielsen 1966). In turn, Rint rates were calculated using the program SPHERE (http://www.fccc.edu/research/labs/roder/sphere).
Experiments to determine amide proton in exchange with the bulk solvent, 15N-(CLEANEX-PM)-FHSQC (cleanex experiment) (Hwang et al. 1998), and with water molecules residing in the protein for longer than the correlation time (ePHOGSY) (Dalvit and Hommel 1995; Dalvit 1996) were also performed in a reduced sample (2.1 M GdmCl, acetate buffer 100 mM [pH 5.5]) by applying the previously reported pulse sequences.
The dynamic features of Rc in the reduced state in the presence of GdmCl 2.1 M were investigated by means of heteronuclear NMR. 15N longitudinal, R1 (Peng and Wagner 1994), and transversal, R2 (Kay et al. 1992; Peng and Wagner 1994) relaxation rates, and the 1H-15N NOE (Grzesiek and Bax 1993) values were measured using the pulse sequences previously reported. For R1 measurements, eight experiments with 15N recovery delays of 10, 60, 130, 150, 250, 290, 500, and 800 msec were performed. The relaxation delay after the acquisition time was 3 sec. Another eight experiments with 15N recovery delays of 6.9, 13.8, 27.6, 48.3, 75.9, 110.4, 151.8, and 207 msec during the transversal evolution of the 15N nucleus were carried out to determine the R2 values. Water suppression was achieved by an echo–anti-echo scheme (Sklenar and Bax 1987). For the determination of 1H-15N NOE values, three experiments with 1H presaturation for 2.5 sec and one without were performed. In this last case, the strong solvent signal was partially eliminated by a flip-back approach (Grzesiek and Bax 1993). For R2 and 1H-15N NOE experiments, 3.2 sec of recycle time were employed to ensure the complete relaxation of the nuclei.
For all NMR experiments, FIDs were apodized to a final data matrix of 2048 x 1024 points, zero filled, weighted by Gaussian and sine square functions (shifted 60°) in acquisition and evolution dimensions, respectively, and Fourier transformed. Only the downfield part of the 1H spectra, containing the HN connectivities, was kept for the data analysis. All the NMR spectra were processed with the XWINNMR program (Bruker) and examined with the Sparky software (http://www.cgl.ucsf.edu/home/sparky/).
The mobility data were analyzed as previously described (Jiménez et al. 2003a, 2004). The overall tumbling of Rc was calculated from the R2/R1 ratios of each HN pair using the program Quadric Diffusion 1.11 (Lee et al. 1997). The average structure (previously minimized) of the family of Rc structures (1cur file [Botuyan et al. 1996] from the Protein Data Bank) was used in these calculations. R2/R1 values larger than twice the standard deviation () of the average value were eliminated (Tjandra et al. 1995). This process was repeated until all values fell within the average value ±2. Throughout this article the same criterion has been taken for discerning between data that do and do not deviate from the average value. Relaxation rates that deviate >2 are considered to deviate significantly from the average. The diffusion parameters corresponding to the isotropic, axially symmetric, and fully anisotropic models were tested (Palmer et al. 1991; Mandel et al. 1995). The F factor (Bevington and Robinson 1992) was used to check the statistical improvement of the fitting to each model.
The relaxation data (R1, R2, and 1H-15N NOE) were analyzed according to the model-free approach of Lipari and Szabo (1982a,b), using the program Modelfree 4.0 (Mandel et al. 1995). The experimental data were adjusted to one of the four different models described (Palmer et al. 1991; Mandel et al. 1995). Here, we follow exactly the same criteria as those previously reported (Jiménez et al. 2004).
|
Footnotes |
|---|
|
Acknowledgments |
|---|
|
References |
|---|
|
TOPAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Bai, Y., Chung, J., Dyson, H.J., and Wright, P.E. 2001. Structural and dynamic characterization of an unfolded state of poplar apo-plastocyanin formed under nondenaturing conditions. Protein Sci. 10: 1056–1066.
Banci, L., Bertini, I., Cramaro, F., Del Conte, R., Rosato, A., and Viezzoli, M.S. 2000. Backbone dynamics of human Cu,Zn superoxide dismutase and of its monomeric F50E/G51E/E133Q mutant: The influence of dimerization on mobility and function. Biochemistry 39: 9108–9118.[CrossRef][Medline]
Bertini, I., Dalvit, C., Luchinat, C., Huber, J.G., and Piccioli, M. 1997. e-PHOGSY experiments on a paramagnetic protein: Location of the catalytic water molecule in the heme crevice of the oxidized form of horse heart cytochrome c. FEBS Lett. 415: 45–48.[CrossRef][Medline]
Bertini, I., Bryant, D.A., Ciurli, S., Dikiy, A., Fernandez, C.O., Luchinat, C., Safarov, N., Vila, A.J., and Zhao, J. 2001a. Backbone dynamics of plastocyanin in both oxidation states. Solution structure of the reduced form and comparison with the oxidized state. J. Biol. Chem. 276: 47217–47226.
Bertini, I., Luchinat, C., and Parigi, G. 2001b. Solution NMR of paramagnetic molecules. Applications to metallobiomolecules and models. Current methods in inorganic chemistry. Elsevier, Amsterdam.
Bevington, P.R. and Robinson, D.K. 1992. Data reduction and error analysis for the physical sciences, pp. 205–209. McGraw-Hill, New York.
Bodenhausen, G. and Ruben, D.J. 1980. Natural abundance nitrogen-15NMR by enhanced heteronuclear spectroscopy. Chem. Phys. Lett. 69: 185–189.[CrossRef]
Botuyan, M.A., Toy-Palmer, A., Chung, J., Blake 2d, R.C., Beroza, P., Case, D.A., and Dyson, H.J. 1996. NMR solution structure of Cu(I) rusticyanin from Thiobacillus ferrooxidans: Structural basis for the extreme acid stability and redox potential. J. Mol. Biol. 263: 752–767.[CrossRef][Medline]
Bruschweiler, R. 2003. New approaches to the dynamic interpretation and prediction of NMR relaxation data from proteins. Curr. Opin. Struct. Biol. 13: 175–183.[CrossRef][Medline]
Buning, C., Canters, G.W., Comba, P., Dennison, C., Jeuken, L., Melter, M., and Sanders-Loehr, J. 2000. Loop-directed mutagenesis of the blue copper protein amicyanin from Paracoccus versutus and its effect on the structure and the activity of the type-1 copper site. J. Am. Chem. Soc. 122: 204–211.[CrossRef]
Campos-Olivas, R. and Summers, M.F. 1999. Backbone dynamics of the N-terminal domain of the HIV-1 capsid protein and comparison with the G94D mutant conferring cyclosporin resistance/dependence. Biochemistry 38: 10262–10271.[CrossRef][Medline]
Capaldi, A.P., Ferguson, S.J., and Radford, S.E. 1999. The Greek key protein apo-pseudoazurin folds through an obligate on-pathway intermediate. J. Mol. Biol. 286: 1621–1632.[CrossRef][Medline]
Cox, J.C. and Boxer, D.H. 1978. The purification and some properties of rusticyanin, a blue copper protein involved in iron(II) oxidation from Thiobacillus ferro-oxidans. Biochem. J. 174: 497–502.[Medline]
Crane, B.R., Di Bilio, A.J., Winkler, J.R., and Gray, H.B. 2001. Electron tunneling in single crystals of Pseudomonas aeruginosa azurins. J. Am. Chem. Soc. 123: 11623–11631.[CrossRef][Medline]
Daggett, V. and Fersht, A.R. 2003. Is there a unifying mechanism for protein folding? Trends Biochem. Sci. 28: 18–25.[CrossRef][Medline]
Dalvit, C. 1996. Homonuclear 1D and 2D NMR experiments for the observation of solvent-solute interactions. J. Magn. Reson. B 112: 282–288.[CrossRef][Medline]
Dalvit, C. and Hommel, U. 1995. Sensitivity-improved detection of protein hydration and its extension to the assignment of fast-exchanging resonances. J. Magn. Reson. B 109: 334–338.[CrossRef]
Dinner, A.R., Sali, A., Smith, L.J., Dobson, C.M., and Karplus, M. 2000. Understanding protein folding via free-energy surfaces from theory and experiment. Trends Biochem. Sci. 25: 331–339.[CrossRef][Medline]
Dobson, C.M. 2003. Protein folding and misfolding. Nature 426: 884–890.[CrossRef][Medline]
Donaire, A., Jiménez, B., Moratal, J.M., Hall, J.F., and Hasnain, S.S. 2001. Electronic characterization of the oxidized state of the blue copper protein rusticyanin by 1HNMR: Is the axialmethionine the dominant influence for the high redox potential? Biochemistry 40: 837–846.[CrossRef]
) were calculated from Equation 2 (see Discussion).
