Presence of the cofactor speeds up folding of Desulfovibrio desulfuricans flavodoxin

Abstract

Flavodoxin is an α/β protein with a noncovalently bound flavin-mononucleotide (FMN) cofactor. The apo-protein adopts a structure identical to that of the holo-form, although there is more dynamics in the FMN-binding loops. The equilibrium unfolding processes of Azotobacter vinelandii apo-flavodoxin, and Desulfovibrio desulfuricans ATCC strain 27774 apo- and holo-flavodoxins involve rather stable intermediates. In contrast, we here show that both holo- and apo-forms of flavodoxin from D. desulfuricans ATCC strain 29577 (75% sequence similarity with the strain 27774 protein) unfold in two-state equilibrium processes. Moreover, the FMN cofactor remains bound to the unfolded holo-protein. The folding and unfolding kinetics for holo-flavodoxin exhibit two-state behavior, albeit an additional slower phase is present at very low denaturant concentrations. The extrapolated folding time in water for holo-flavodoxin, ∼280 μsec, is in excellent agreement with that predicted from the protein's native-state topology. Unlike the holo-protein behavior, the folding and unfolding reactions for apo-flavodoxin are best described by two kinetic phases, with rates differing ∼15-fold, suggesting the presence of a kinetic intermediate. Both folding phases for apo-flavodoxin are orders of magnitude slower (40- and 530-fold, respectively) than that for the holo-protein. We conclude that polypeptide–cofactor interactions in the unfolded state of D. desulfuricans strain 29577 flavodoxin alter the kinetic-folding path towards two-state and speed up the folding reaction.

Many proteins in nature require the binding of cofactors to perform their biological activity, and these molecules fold in a cellular environment where their cognate cofactors are present. It has been demonstrated in vitro that many proteins retain the interactions with the cofactors after polypeptide unfolding (Bertini et al. 1997; Robinson et al. 1997; Wittung-Stafshede et al. 1999, 2000; Pozdnyakova et al. 2000). For example, coordination of the hemes in cytochrome b₅₆₂ and myoglobin to the corresponding unfolded polypeptides have been observed (Robinson et al. 1997; Wittung-Stafshede et al. 1999, 1998c). In the case of azurin and the Cu_A domain (Pozdnyakova et al. 2001b; Wittung-Stafshede et al. 1998a, 1998b), the copper ions stay bound to the unfolded polypeptides. Therefore, cofactors may bind to their corresponding polypeptides before folding in vivo. It is possible that local nonrandom structure in an unfolded protein forms due to coordination of a cofactor. Such structural restriction in the unfolded ensemble of molecules may reduce the conformational search for the native state (i.e., the cofactor serves as a nucleation site that directs folding) (Shortle 1996; Luisi et al. 1999). Cofactors have been shown to most often stabilize the native states of the proteins they interact with (Bertini et al. 1997; Leckner et al. 1997; Robinson et al. 1997; Goedken et al. 2000). However, the manner in which cofactors affect the folding pathway remains poorly understood, because kinetic folding studies are frequently conducted in the absence of potentially complicating ligands.

To directly address the role of an organic cofactor in protein folding, we have examined the consequences of cofactor binding on the equilibrium and kinetic folding pathway of the 148-residue, single-domain Desulfovibrio desulfuricans ATCC strain 29577 flavodoxin (Helms and Swenson 1991). The flavodoxin fold consists of one five-stranded parallel β-sheet surrounded by two α-helices on each side, a rather common motif shared by nine protein superfamilies. Flavodoxin participates in photosynthetic electron transfer, and has a redox-active flavin mononucleotide (FMN) cofactor noncovalently bound (Mayhew et al. 1996; Romero et al. 1996; Apiyo et al. 2000). The affinity of FMN to folded D. desulfuricans flavodoxin is extremely strong; the dissociation constant (oxidized FMN) is around 0.1 nM (Caldeira et al. 1994). Residues in two peptide loops (residues 58–66 and 95–102) form the major portion of the FMN binding site, which consist of a combination of aromatic-stacking interactions, an apolar environment, and electrostatic interactions. In particular, the iso-alloxazine ring of the FMN moiety is sandwiched between two aromatic residues (Trp60 and Tyr98), allowing for considerable π-orbital overlap. Upon FMN removal, the apo-protein adopts a structure that is identical to that of the holo-form, except for more dynamics observed by NMR in the FMN binding loop regions (Steensma and van Mierlo 1998).

We recently reported (Apiyo et al. 2000) that the equilibrium-unfolding reaction of D. desulfuricans ATCC strain 27774 showed little (or no) effect of the FMN cofactor. A fluorescence-detected transition was observed prior to secondary-structure disappearance (for both apo- and holo-forms), implying a native-like intermediate present on this flavodoxin's equilibrium-unfolding path. Also, apo-flavodoxin from Azotobacter vinelandii has been shown to adopt a native-like equilibrium-unfolding intermediate (van Mierlo et al. 1998). In contrast to these earlier findings, we here show that the equilibrium-unfolding reactions of both apo- and holo-forms of D. desulfuricans ATCC strain 29577 flavodoxin are two-state processes. Still, flavodoxin from ATCC strain 29577 displays 75.3% amino acid sequence similarity to the protein from strain 27774 (Caldeira et al. 1994), and both proteins adopt the same fold. Apparently the equilibrium-unfolding mechanism is not conserved among the members of the flavodoxin–protein family. Time-resolved kinetic experiments with D. desulfuricans strain 29577 flavodoxin reveal that the kinetic folding and unfolding processes are two-state for the holo-protein, but more complex (suggesting the presence of a kinetic off-pathway intermediate) for the apo-protein. Holo-flavodoxin folds more than one order of magnitude faster than the apo-protein, implying that cofactor interactions in the unfolded state aid in the kinetic polypeptide-folding reaction.

Results and Discussion

Equilibrium unfolding of apo- and holo-flavodoxin

The chemical denaturant guanidine hydrochloride (GuHCl) was used to induce unfolding of D. desulfuricans (ATCC 29577) flavodoxin upon monitoring tryptophan fluorescence and far-UV circular dichroism (CD) signals. Fluorescence reports on tertiary contacts and solvation properties near the tryptophan, whereas far-UV CD monitors the overall secondary structure. In Figure 1 ▶, we show the equilibrium-unfolding curves for holo- (Fig. 1A ▶) and apo- (Fig. 1B ▶) flavodoxin monitored by CD and fluorescence. For both forms, the transitions derived from the two spectroscopic probes coincide, indicative of two-state equilibrium-unfolding processes. The unfolding-transition midpoints occur at 2.1 and 1.7 M GuHCl—corresponding to unfolding-free energies in water, ΔG_U(H₂O), of 22.3 ± 3 kJ/mole and 16.4 ± 2 kJ/mole for holo- and apo-flavodoxin, respectively (Table 1). Unfolding reactions of both forms of flavodoxin are completely reversible.

Fig. 1.

GuHCl-induced equilibrium unfolding transitions for holo- (A) and apo- (B) flavodoxin monitored by tryptophan fluorescence (filled symbols, ex 285 nm) and far-UV CD (open symbols). The dashed lines are two-state fits (see Table 1). Inset: Structure of flavodoxin (PBD F2X) with FMN high-lighted.

Table 1.

Equilibrium and kinetic parameters for folding and unfolding reactions of D. desulfuricans ATCC strain 29577 apo- and holo-flavodoxin; 20°C

	HOLO	APO
Equilibrium data
ΔGU(H2O) (kJ/mol)	22.3 ± 2.8	16.4 ± 2.3
meq (kJ/mol,M)	10.6 ± 1	9.7 ± 1
GuHCl1/2 (M)	2.1 ± 0.2	1.7 ± 0.2
Kinetic data		Fast phase	Slow phase
kF (H2O) (s−1)	3517 ± 1000	90 ± 30	6.6 ± 2.0
mF (kJ/mol,M)	−7.3 ± 0.6	−4.0 ± 1.0	−4.2 ± 1.0
kU (H2O) (s−1)	0.4 ± 0.1	0.28 ± 0.08	0.008 ± 0.002
mU (kJ/mol,M)	2.0 ± 0.3	2.4 ± 0.3	2.4 ± 0.3
mU–mF (kJ/mol,M)	9.3 ± 0.9
RTln[kF(H2O)/kU(H2O)] (kJ/mol)	22.5 ± 1.3
mF/(mF–mU)	0.75

In the case of the holo-protein, FMN remains coordinated to the unfolded polypeptide; this was also reported previously for D. desulfuricans ATCC strain 27774 flavodoxin (Apiyo et al. 2000). The conclusion that FMN stays bound to the unfolded protein is based on several observations (Fig. 2 ▶). Both dialysis and gel-filtration experiments with the unfolded holo-protein show that the FMN stays with the polypeptide fraction. Moreover, the FMN emission from the unfolded holo-protein is quenched compared to that of free FMN, indicating protein–FMN interactions (Fig. 2A ▶). That FMN can associate with unfolded flavodoxin in a 1:1 ratio is also clear from direct binding experiments between free FMN and apo-flavodoxin at denaturing conditions (using isothermal titration calorimetry and fluorescence quenching; Fig. 2B ▶). Finally, equilibrium unfolding curves are identical for different flavodoxin concentrations (5–100 μM tested); if FMN dissociated from the unfolded polypeptide, a protein–concentration dependence of the stability would be observed. Using the K_D for binding of FMN to native flavodoxin (0.1 nM; Caldeira et al. 1994) and the derived unfolding-free energies for holo- and apo-flavodoxin (Table 1), the K_D for FMN interacting with the unfolded polypeptide is estimated to be 1.2 nM. It is not known if the FMN binding site in the unfolded protein is similar to that in the folded protein. The high predicted affinity, however, suggests the presence of at least some native-like interactions in the unfolded state.

Fig. 2.

(A) Fluorescence from unfolded holo-flavodoxin (solid line), free FMN (dashed line), and unfolded apo-flavodoxin (dotted line); all in the presence of 3 M GuHCl, excitation at 285 nm. (B) FMN-fluorescence decrease (shown as relative emission change at 525 nm, excitation at 285 nm) upon mixing FMN and unfolded apo-flavodoxin in 1:10 ratio (10 μM FMN; 100 μM apo-protein; 3 M GuHCl). Pseudofirst-order rate constant: 0.3 sec⁻¹. Inset: Isothermal titration calorimetry data for FMN added to a solution of unfolded apo-flavodoxin (100 aμM apo-flavodoxin; 3 M GuHCl).

The equilibrium unfolding results reported here for D. desulfuricans ATCC strain 29577 flavodoxin should be compared to our previous study of the D. desulfuricans ATCC strain 27774 protein. The two proteins share 75% amino acid sequence similarity; thus, their behavior with respect to folding and unfolding is expected to be rather similar. In case of the protein from ATCC 27774, equilibrium unfolding (as monitored by far-UV CD) results in unfolding-free energies of 20 ± 2 and 19 ± 1 kJ/mole, and transition midpoints at 1.8 and 1.6 M GuHCl, for holo- and apo-forms, respectively (Apiyo et al. 2000). The absolute stability values for the two species of flavodoxin thus match well with each other; however, the difference in stability between holo- and apo-forms is larger for flavodoxin from ATCC 29577 than for the protein from ATCC 27774 (6 ± 5 kJ/mole versus 1 ± 3 kJ/mole). Nevertheless, both these differences in energy are small; cofactors may stabilize their corresponding protein much more; for example, the copper in azurin stabilizes the protein by 23 kJ/mole (Pozdnyakova et al. 2001a).

The significant difference between the results on the two species is that equilibrium unfolding of flavodoxin from ATCC 27774 proceeds through a native-like intermediate, with altered tryptophan emission but native far-UV CD (Apiyo et al. 2000), whereas equilibrium unfolding of flavodoxin from ATCC 29577 occurs in a two-state process. This discrepancy may be explained by the native-like intermediate having a structural perturbation that is caused by some of the residues (25% of total) not present in the ATCC 29577 protein. The equilibrium unfolding reactions of two other flavodoxins have also been reported. A. vinelandii apo-flavodoxin unfolds via a relatively stable equilibrium unfolding intermediate (van Mierlo et al. 1998), whereas the equilibrium unfolding reaction of Anabaena PCC 7119 apo-flavodoxin is two state (Genzor et al. 1996). Thus, in contrast to many other protein families, small differences in amino acid sequence appear sufficient to alter the equilibrium folding mechanism among the members of the flavodoxin–protein family.

Folding dynamics for holo-flavodoxin

Stopped-flow mixing, while monitoring changes in tryptophan emission, was used to probe the unfolding and refolding kinetics of flavodoxin with and without the cofactor. The time-resolved traces collected for holo-flavodoxin unfolding and refolding are best fit by mono-exponential decay equations. The kinetic results for holo-flavodoxin are shown in a semilogarithmic plot of ln k versus GuHCl concentration in Figure 3A ▶. The logarithm of the unfolding-rate constants increases linearly with denaturant concentration, whereas the logarithm of the speed of refolding increases linearly, going to lower denaturant concentration; this typical V-shape behavior is expected for two-state transitions (Fersht 1997). However, at the lowest GuHCl concentrations studied in the refolding experiments, the kinetics are biphasic. There is first a rapid phase that follows the GuHCl dependence observed for the rates at higher denaturant concentrations, followed by a slower phase. The amplitude of the slower phase increases when the denaturant concentration is lowered. We speculate that upon refolding into conditions strongly favoring the native state, refolding is so rapid that some misfolding occurs, which subsequently is resolved in the slower phase. That the slower phase at low GuHCl concentration is due to apo-molecules (folding slower, see below) is excluded, because the experiments are performed at concentrations well above FMN's K_D to the unfolded polypeptide. The parameters obtained from fitting the kinetic data in Figure 3A ▶ are listed in Table 1; they agree well with the equilibrium unfolding parameters, in further support of an apparent two-state mechanism. The folding transition state for holo-flavodoxin appears rather native-like: m_F/(m_U − m_F) = 0.75, in accord with corresponding data on other β-sheet proteins (Fersht 1999).

Fig. 3.

Natural logarithm of folding and unfolding rate constants for holo- (A) and apo- (B) flavodoxin as a function of GuHCl concentration. In A, the open symbols represent an additional slower phase observed only at the lower GuHCl concentrations (corresponding to 50% of amplitude at 0.5 M GuHCl and 40% of amplitude at 0.7 M GuHCl). In B, filled symbols represent the fast folding and unfolding phases; the open symbols represent the slow phases. Parameters from fits (dashed lines) are listed in Table 1.

The extrapolated folding time for holo-flavodoxin in the absence of a denaturant is 284 ± 60 μsec (neglecting the additional slow phase observed at low denaturant concentrations). A correlation between folding speed and native-state topology (described by the parameter relative contact order) has been reported for a large number of small proteins folding by two-state kinetics (Plaxco et al. 1998). Proteins with mostly local contacts (such as helical proteins) fold more rapidly than proteins with mostly long-range contacts (such as β-sheet proteins) in their native states. This finding was later explained in terms of an extended nucleus with native-like topology in the transition-state for folding (Fersht 2000). Based on the relationship between contact order and folding speed, flavodoxin is predicted to have a folding time of 100 μsec (K. Plaxco, pers. comm.). Thus, the prediction is in very close agreement with our experimental result. Holo-flavodoxin is one of very few proteins with more than 100 residues (Fersht 1999)—perhaps the largest—that folds with simple two-state kinetics (and obeys the contact–order correlation).

Folding dynamics for apo-flavodoxin

The kinetics for apo-flavodoxin folding and unfolding differs from the data on the holo-protein. In Figure 3B ▶, we show that both unfolding and refolding processes are biphasic throughout the denaturant-concentration range, with approximately 50% of the amplitude change in each phase. Explanations for the bi-exponential behavior may be sequential (presence of an intermediate) or parallel (two populations of molecules) reaction mechanisms (Fersht 1999). If the kinetics are bi-exponential due to the presence of two parallel pathways there must be two populations of apo-molecules in both the folded and unfolded states. The apo-protein, however, appears homogeneous according to native-gel electrophoresis and mass-spectroscopy analysis, eliminating the presence of two chemically different fractions of apo-molecules. In addition, the two-state equilibrium unfolding behavior indicates only one population of apo-molecules. It is, therefore, probable that apo-flavodoxin unfolding and refolding reactions involve a kinetic intermediate. However, the intermediate cannot be on-pathway, because the thermodynamic stability calculated from the observed rate constants, assuming an on-pathway mechanism, is too large (30 kJ/mole) compared to the equilibrium data (16 kJ/mole). Moreover, a linear off-pathway mechanism does not support bi-phasic unfolding kinetics.

In contrast, a triangular three-species mechanism can explain the observed biphasic behavior in the folding and unfolding reactions, assuming the intermediate and unfolded forms to have similar fluorescence properties. Upon unfolding, a fraction of apo-molecules unfolds directly to the unfolded state (with the fast unfolding rate), whereas the rest of the molecules unfold to the intermediate state (with the slow unfolding rate), which then rapidly converts to the unfolded species (invisible to spectroscopic detection). Upon refolding, a fraction of molecules folds directly to the native state (with the fast folding rate), whereas the rest of the molecules fold to the intermediate conformation (invisible to detection). The latter molecules then fold directly to the native state, or unfold back before reaching the native state, with the slower folding rate. The experimental rate constants are reasonable only if the intermediate is in essence off-pathway (i.e., most molecules adopting the intermediate conformation during folding must unfold back before reaching the native state). The faster folding and unfolding rate constants, associated with the unfolded-to-folded transition in the triangular mechanism, corresponds to a stability of ∼14 kJ/mole, which is in good agreement with the equilibrium unfolding result (Table 1). The triangular mechanism is supported by a recent report on the kinetic folding behavior of apo-flavodoxin from Anabaena PCC 7119 (Fernandez-Recio et al. 2001). Two kinetic phases were also resolved during Anabaena apo-flavodoxin folding and unfolding reactions. The authors concluded, through rigorous modeling, that the Anabaena protein, like we suggest for D. desulfuricans apo-flavodoxin, folds via an essentially off-pathway intermediate in a triangular mechanism (Fernandez-Recio et al. 2001).

The chemotactic (apo) protein CheY is a sequentially unrelated protein that shares the flavodoxin-like fold. This protein, like D. desulfuricans and Anabaena apo-flavodoxins, displays a kinetic folding mechanism that involves a misfolded intermediate (Lopez-Hernandez and Serrano 1996).

Concluding remarks

Only a few other studies targeting the kinetic effect of cofactors on folding have been reported. There are no other kinetic folding studies addressing the effect of the cofactor in flavodoxin proteins. α-Lactalbumin was concluded to refold more quickly in the presence of metals, but the metals had no effect on the unfolding speed (Kuwajima et al. 1989). In contrast, calcium ions were shown to stabilize RNase and staphylococcal nuclease proteins by a mechanism based on decreasing the unfolding speed (Sugawara et al. 1991; Goedken et al. 2000). Also, in the case of azurin, the copper ion stabilizes the native state by decreasing the unfolding speed (Pozdnyakova et al. 2001a). For azurin, the folding rate predicted from the native-state contact order agrees with the folding speeds for both apo- and holo-forms, suggesting that interactions with copper during the folding process do not affect this polypeptide's folding transition state.

We have here shown that presence of the organic cofactor FMN in the unfolded state speeds up folding of D. desulfuricans ATCC strain 29577 flavodoxin. The apo-protein refolding rates are decreased ∼40-fold for the faster and ∼530-fold for the slower phase compared to the folding rate for holo-flavodoxin. Interestingly, the presence of FMN in the unfolded state also affects flavodoxin's kinetic folding mechanism. In the presence of FMN, the kinetic folding reaction appears two-state, whereas in the absence of FMN, the reaction becomes more complex (a triangular three-species mechanism fits the experimental data). To our knowledge, there is only one other report of different kinetic folding mechanisms for two homologous proteins. The immunity proteins Im7 and Im9 are 60% identical in sequence and have the same four-helix bundle structure, yet Im7 folds through a kinetic intermediate, whereas Im9 folds by a two-state mechanism (Ferguson et al. 1999). As in the case for apo- and holo-flavodoxin, the less stable form (Im7 and apo-flavodoxin) exhibits the pathway with the kinetic intermediate.

An important issue in modern protein biophysics is whether structurally homologous proteins share common stability and/or folding features. Studies on four different flavodoxins (from D. desulfuricans ATCC strains 29577 and 27774, A. vinelandii and Anabaena) show that the equilibrium unfolding pathway can vary between two- and three-state among the members of the flavodoxin–protein family. However, for each flavodoxin species (D. desulfuricans ATCC strains 29577 and 27774 studied), the equilibrium unfolding mechanism (in sharp contrast to the kinetic folding mechanism) is not affected by the presence, or not, of the FMN cofactor.

Materials and methods

Flavodoxin from D. desulfuricans (ATCC strain 29577) was expressed in Escherichia coli (Helms and Swenson 1991) and purified as described previously (Caldeira et al. 1994; Romero et al. 1996). The FMN was oxidized in all experiments. Apo-flavodoxin was prepared by low-pH dialysis (Apiyo et al. 2000). Fluorescence was monitored on a Varian Eclipse (1-cm × 2-mm cell), CD on an OLIS instrument (1-mm path). All experiments were performed at 20°C in 5 mM phosphate buffer, pH 7.0 using GuHCl of the highest purity (Sigma). There was no protein concentration (in the 5–100-μM range tested) dependence in the equilibrium unfolding transitions and they were fully reversible.

The unfolded state of holo-flavodoxin (in 3 M GuHCl) was subjected to dialysis (10,000 MW cutoff, 1:1000 volume ratio, three volume changes) and gel-filtration experiments, followed by estimation (by protein and FMN absorption as well as FMN emission) of an amount of FMN remaining in the protein sample. FMN emission at 525 nm (ex 285 nm) from unfolded holo-flavodoxin (3 M GuHCl) was compared to that of the same concentration of free FMN at identical solution conditions. To directly verify interactions between the unfolded polypeptide and FMN, binding studies were performed by isothermal titration calorimetry (MicroCal instrument) and by monitoring fluorescence-quenching upon adding unfolded apo-protein to FMN (10 μM FMN, 0–100 μM apo-protein; 3 M GuHCl). Both methods revealed stoichiometric FMN binding to the unfolded polypeptide.

The kinetics of folding and unfolding were followed by tryptophan emission (ex 285 nm; em >305 nm) using an Applied Photophysics SX.18MV stopped-flow (pathlength 2 mm). No amplitude changes occurred in the dead time (<4 msec) of the instrument. Unfolding was measured by mixing five parts denaturant solution (of appropriate concentration) with one part of flavodoxin (60 μM). For refolding, unfolded flavodoxin (60 μM) in GuHCl was mixed 1:5 with appropriate GuHCl/buffer solutions to give desired final conditions.

For each GuHCl concentration, a minimum of five kinetic traces were averaged and fit to monophasic (holo) or biphasic (apo) decay equations using a nonlinear least-squares algorithm supplied by Applied Photophysics. The rate constants for the holo-protein kinetics were fit in KaleidaGraph, assuming standard linear dependence of ln k_F and ln k_U on GuHCl concentration (Fersht 1999):

where m_U is the slope of the unfolding branch and m_F is the slope of the folding branch. k_F(H₂O) and k_U(H₂O) are the folding and unfolding rate constants in absence of GuHCl. For the apo-protein kinetic data, rate constants for each phase were fit separately assuming linear dependences of ln k (k_U^fast, k_U^slow, k_F^fast, or k_F^slow) on GuHCl concentration.