The Puzzle of Ligand Binding to Corynebacterium ammoniagenes FAD Synthetase*

In bacteria, riboflavin phosphorylation and subsequent conversion of FMN into FAD are carried out by FAD synthetase, a single bifunctional enzyme. Both reactions require ATP and Mg2+. The N-terminal domain of FAD synthetase appears to be responsible for the adenylyltransferase activity, whereas the C-terminal domain would be in charge of the kinase activity. Binding to Corynebacterium ammoniagenes FAD synthetase of its products and substrates, as well as of several analogues, is analyzed. Binding parameters for adenine nucleotides to each one of the two adenine nucleotide sites are reported. In addition, it is demonstrated for the first time that the enzyme presents two independent flavin sites, each one related with one of the enzymatic activities. The binding parameters of flavins to these sites are also provided. The presence of Mg2+ and of both adenine nucleotides and flavins cooperatively modulates the interaction parameters for the other ligands. Our data also suggest that during its double catalytic cycle FAD synthetase must suffer conformational changes induced by adenine nucleotide-Mg2+ or flavin binding. They might include not only rearrangement of the different protein loops but also alternative conformations between domains.

The two catalytic cycles of FADS involve the binding of two ATP, one RF, and one FMN molecules as substrates and the production of one ADP, one PP i , one FMN, and one FAD. The proposed pathway for the phosphorylation reaction would be for RF to bind before ATP, whereas ADP releases prior FMN (10). In the adenylylation process, FMN is proposed to bind after ATP and the PP i to be released preceding FAD (10). The two enzymatic activities differ in their specificity for divalent cations, optimal pH, and temperature (8,11,12). The presence of Mg 2ϩ improves the turnover of both processes but, although low concentrations (Ͻ1 mM) enhance the kinase activity, much larger concentrations (ϳ10 mM) are required for maximal FAD production (8). These studies indicated the presence of two independent ATP-binding sites, one at the RF phosphorylation site and one at the FMN-adenylylation site, but a single pocket was proposed to allocate the isoalloxazine-ribityl moieties of both substrates, RF and FMN, in the two reactions (10).
The only structure reported for an FADS is that of Thermotoga maritima (TmFADS), both free and in complex with several substrates (13,14). One of these structures shows simultaneous binding of AMP in the N-terminal domain and ADP and FMN in the C-terminal domain. Thus, the protein folds in two almost independent domains, each one hosting one ATP-binding site, whereas only a flavin-binding site, located in the C-terminal domain, was detected (14). However, in the C-terminal domain the ribityl and phosphate of the flavin are not placed in the putative active site. This is probably due to the disorder in the protein regions hosting the flavin. The C-terminal domain shows structural homology with Homo sapiens and Schizosaccharomyces pombe RFKs (14 -17) and can catalyze the phosphorylation of RF (15). The N-terminal region presents remote similarities with nucleotydyltransferases (18), but it does not appear to be self-sufficient to transform FMN into FAD (15). The TmFADS structure shows a large distance between the reported flavin-binding site and the adenylylation site (14). Although the asymmetric unit is a dimer, it does not appear functional. These observations, together with the fact that the FMN produced in the phosphorylation process has to be released before rebinding as substrate for the second reaction (10), make it logical to propose the presence of a second flavinbinding site, located in the N-terminal domain. The structural model of FADS from Corynebacterium ammoniagenes (CaFADS) shows the main structural arrangements present in TmFADS (13,14), and it also shows the disposition of some loops and a 3 10 -helix at the C-terminal domain that are missed in the TmFADS structure. Additionally, the C-terminal domain of CaFADS showed an insertion around residue 230 with regard to TmFADS. This model also suggests the presence of a novel putative second flavin-binding site in the N-terminal domain (15) (see Fig. 1B). This site would be located in the proximity of the adenine nucleotide-binding site and might bind FMN for adenylylation.
The present study provides a detailed thermodynamic analysis of the binding of adenine nucleotide and flavin ligands to CaFADS and to its individually cloned C terminus domain. It confirms the presence of two flavin-binding sites in CaFADS and allows assigning affinities to the different flavin and adenine nucleotide-binding sites.

EXPERIMENTAL PROCEDURES
Cloning, Overexpression, and Purification of CaFADS and Its C-terminal Domain-WT CaFADS was cloned, overexpressed, and purified as previously described (15). The separately cloned C-terminal domain (⌬(1-182)FADS) was purified using a similar protocol, replacing the DEAE-cellulose chromatography with a Superose 12 gel filtration (GE Healthcare).
Spectral Analysis-UV-visible spectra were recorded on a Cary-100 spectrophotometer. To determine the extinction coefficient (⑀) of CaFADS in 50 mM Tris/HCl, pH 8.0, and 20 mM sodium phosphate, pH 7.0, the UV-visible absorbance spectrum of the protein was recorded in each buffer. The samples were subsequently diluted with 7.5 M guanidinium hydrochloride in 20 mM sodium phosphate, pH 6.5, to 6 M guanidinium hydrochloride. The absorbance spectra were again recorded. Protein concentration was calculated using the theoretical ⑀ 280 nm for the denatured protein (19). The ⑀ 280 nm in each buffer was obtained from the initial spectra. For ⌬(1-182)FADS, the theoretical ⑀ 280 nm of 14 mM Ϫ1 cm Ϫ1 was used (19).
Difference spectroscopy measurements were carried out in 50 mM Tris/HCl, pH 8.0. The reference cuvette, containing buffer, and the sample cuvette, containing 5-6 M FADS (3 M for titration with RF), were stepwise titrated with aliquots of 1-10 l of the corresponding flavin solutions (ϳ180 M RF, 750 -1500 M FMN, and 450 -600 M FAD). For the measurements in the presence of AMPPNP, a nonhydrolyzable ATP analogue, both cuvettes contained 1 mM AMPPNP and 10 mM MgCl 2 . Dissociation constants and difference extinction coefficients were obtained by nonlinear regression fit of the experimental data to the theoretical equation for a 1:N stoichiometric complex (20), where ⌬⑀ is the change in the flavin extinction coefficient upon ligand binding to FADS, l is the cuvette path length, [FADS] and [L] are the total concentrations of FADS and flavin, respectively, K d is the dissociation constant, and N is the number of flavin-binding sites in FADS. The errors estimated for K d and ⌬⑀ were Ϯ15%.
CD spectra were recorded in a Chirascan spectropolarimeter (Applied Photophysics Ltd.) at 5°C in the far-UV (7 M FADS in 5 mM sodium phosphate, pH 7.0, 0.1-cm cuvette) and in the near-UV (20 M FADS in 20 mM sodium phosphate, pH 7.0, 0.4-cm cuvette). Near-UV CD spectra were also recorded at 20°C with saturating concentrations of ATP, ADP, and/or FMN, at both 0 and 10 mM MgCl 2 .
Fluorescence emission spectra were recorded in an Aminco-Bowman Series 2 spectrometer in 20 mM sodium phosphate, pH 7.0, at 25°C, exciting the protein aromatic residues at 280 nm.
Differential Scanning Calorimetry-The heat capacity of FADS (⌬C P ) was measured as a function of temperature at pH 7.0 with a VP-DSC microcalorimeter (MicroCal LLC). Thermal denaturation scans were performed with 8 and 24 M degassed FADS solutions in 20 mM sodium phosphate, pH 7.0, at a scanning rate of 1°C/min from 10 to 100°C. Reversibility of the unfolding process was checked by sample reheating after cooling inside the calorimetric cell. The unfolding of FADS was not reversible, and therefore only a van't Hoff model-independent analysis was performed (21,22).
High Sensitivity Isothermal Titration Calorimetry (ITC)-Measurements were carried out using a high precision VP-ITC system (MicroCal LLC). Typically, a ϳ300 M FAD or FMN solution or a 300 -500 M ADP, ATP, or AMPPNP solution was used to titrate ϳ10 -20 M FADS. Because of the low solubility of Lumiflavin (LF, a RF analogue lacking the ribityl chain) and RF, FADS solutions of ϳ5 and ϳ10 M were titrated with LF and RF ϳ90 and ϳ175 M, respectively. Both the ligand and FADS were dissolved in the same buffer (20 mM phosphate, pH 7.0, with variable MgCl 2 concentrations: 0, 0.8, or 10 mM) and degassed. This buffer was selected because of its low ionization enthalpy, which prevents any buffer influence on the observed enthalpy of binding caused by possible de/protonation events. Titrations of FADS⅐nucleotide or FADS⅐FAD complexes were carried out by stepwise injections of the ligand into a mixture of FADS and a saturating concentration of the adenine nucleotide (400 -500 M) and/or FAD (80 -100 M). Each titration was initiated by a 4-l injection followed by 25-28 stepwise injections of 10 l. The heat evolved after each ligand injection was obtained from the integral of the calorimetric signal. The heat caused by the binding reaction was obtained as the difference between the heat of reaction and the corresponding heat of dilution, the latter estimated as a constant heat throughout the experiment and included as an adjustable parameter in the analysis. The association constant (K a ), the enthalpy change (⌬H), and the stoichiometry (N) were obtained through nonlinear regression of the experimental data to a model for one or two independent binding sites implemented in Origin 7.0. The dissociation constant (K d ), the free energy change (⌬G), and the entropy change (⌬S) were obtained from basic thermodynamic relationships. Usually, when N ϭ 2 there was not enough information in the titrations for using a model with two different binding sites and averaged binding parameters are given. The estimated error was Ϯ15% in K d and Ϯ 0.3 kcal/mol in ⌬H and ϪT⌬S. To

RESULTS
Spectral Properties of CaFADS-CaFADS presented an UVvisible maximum at 279 nm. The ⑀ 279 nm was 28.1 mM Ϫ1 cm Ϫ1 in 50 mM Tris/HCl, pH 8.0, and 25.6 mM Ϫ1 cm Ϫ1 in 20 mM sodium phosphate, pH 7.0. The far-UV CD spectrum of CaFADS showed a negative-positive couplet (208 -191 nm) typical of a secondary ␣-helix structure. The negative band around 222 nm, also typical of ␣-helix, appeared like a shoulder ( Fig. 2A), probably because of an important content of ␤-sheet (Fig. 1B). The near-UV CD spectrum ( Fig. 2A) was sensible to substrate binding. Changes were more evident in the simultaneous presence of FMN, ADP, or ATP, and Mg 2ϩ and suggest that binding takes place by influencing the environment of aromatic residues. Excitation of CaFADS at 280 nm produced a fluorescence emission band centered at 330 nm, indicating that, in general, the Trp residues of the protein are contained in an apolar environment.
Binding of Flavins to FADS Followed by Differential Spectroscopy-Titration of FADS with RF, FMN, and FAD produced the appearance of visible difference spectra (Fig. 2B), indicating changes in the dielectric environment of the isoalloxazine upon interaction with FADS. Although titration with RF or FMN produced similar difference spectra, the spectrum elicited upon interaction with FAD was different (Fig. 2B).
The magnitude of the difference spectra reached saturation (Fig. 2D), allowing the determination of K d and ⌬⑀. Fitting of the data to Equation 1 suggested two independent binding sites for RF with an average K d (K d,av ) of 5.2 M, whereas a single binding site was detected for FMN and FAD with K d values of 13.1 and 2.3 M, respectively. The saturating concentration of AMPPNP-Mg 2ϩ produced a considerable increase in ⌬⑀ when FADS was titrated with RF and, especially, with FMN ( Fig. 2C). AMPPNP-Mg 2ϩ considerably increased the FADS affinity for RF while decreasing that for FMN (K d of 0.6 and 90.5 M, respectively). However, AMPPNP-Mg 2ϩ did not affect the FAD interaction (Fig. 2).
FADS Thermal Denaturation Followed by Differential Scanning Calorimetry-FADS thermal denaturation exhibited two partially overlapping transitions (T m,ap ϭ 41.1°C), suggesting that the unfolding processes of the N-and C-terminal domains are not fully cooperative. By direct integration of the signal, a calorimetric ⌬H of 165 kcal/mol and an unfolding ⌬C P of 4.2 kcal/K⅐mol were obtained. A statistical analysis of globular proteins indicates that the unfolding ⌬C P and the unfolding ⌬H at 60°C scale with protein size according to: ⌬C P ϭ 13.9⅐N R cal/(K⅐mol⅐res) and ⌬H(60°C) ϭ 0.698⅐N R kcal/(mol⅐res), where N R is the number of residues in the protein (25). FADS, with 338 residues, has an estimated unfolding ⌬C P of 4.7 kcal/K⅐mol and an unfolding ⌬H(60°C) of 236 kcal/mol. They compare well with the experimental values of 4.2 kcal/K⅐mol and 245 kcal/mol (extrapolation at 60°C). This suggests that the protein is folded in solution with no significant unstructured regions.
A second cycle of thermal denaturation indicated that the unfolding process was not reversible, the less stable domain being responsible for the nonreversibility. The more stable domain showed a reversible thermal denaturation with a T m of 43.8°C. The van't Hoff analysis provides a van't Hoff unfolding ⌬H of 84 kcal/mol. A value of 0.51 for the van't Hoff-calorimetric ⌬H ratio also indicates that the thermal transition is not fully cooperative and that at least two domains unfold independently. Finally, the lack of effect on T m,ap upon increasing protein concentration (from 8 to 24 M) indicated that FADS is in the monomeric state at both concentrations, in agreement with the FADS dilution experiment by ITC.   (Table 1). Similarly, FADS showed a single ADP-binding site in the absence of Mg 2ϩ , the affinity four times lower than that of ATP (Table 1) and two ADPbinding sites in the presence of Mg 2ϩ without altering K d,av ADP . AMPPNP interacted with FADS considerably more weakly than ATP and without major  [Mg 2ϩ ] effects. Therefore, AMPPNP is not a suitable ATP analogue for the purpose of this study.

ITC Analysis of the Interaction of FADS with
Direct titrations also allowed the determination of the enthalpy and entropy components of the interactions (Fig. 4). A favorable enthalpy change drove the interaction of ATP, ADP, and AMPPNP with FADS over the unfavorable entropic contribution, although the magnitude was considerably smaller for the latter. Increasing [Mg 2ϩ ] resulted in an slightly less favorable enthalpic binding contribution and a less unfavorable entropic one (Fig. 4). However, especially for ADP and AMP-PNP, ⌬G remained remarkably insensitive to [Mg 2ϩ ] through entropy/enthalpy compensation.
ITC Analysis of the Binding of Flavins to FADS-The interaction of WT FADS with its RF, FMN, and FAD substrates/products, as well as with LF, was also analyzed (Fig. 5 and Table 1). A single FADS-binding site for FMN and FAD was determined, independently of [Mg 2ϩ ], with K d values in the range of 1-3 M for both flavins. Binding of FMN and FAD was driven by a large enthalpy change but at high cost in entropy (Fig. 6). Increasing [Mg 2ϩ ] hardly influenced the interaction with FMN, but the interaction with FAD resulted in being slightly less favored by the enthalpic contribution and less unfavored by the entropic one (Fig. 6).
Two binding sites were detected when FADS was titrated with RF and LF, both binding only slightly more weakly than FMN or FAD ( Table 1). Binding of RF and LF was also enthalpically driven with a very small opposing entropic contribution that was favorable in the presence of Mg 2ϩ (unlike interactions with FAD and FMN) (Fig. 6). This is consistent with the considerably less polar nature of RF and LF.
ITC Analysis of Flavin Binding to Preformed FADS⅐Nucleotide Complexes-A single FAD-binding site was unequivocally determined in the presence of saturating concentrations of ADP or AMPPNP (Table 2 and Fig. 7). AMPPNP had minor effects on the binding parameters, but the presence of ADP considerably reduced the FADS affinity for FAD at low [Mg 2ϩ ] ( Table 2). This reduction in the affinity is the result of considerably less favorable enthalpic and less unfavorable entropic binding contributions (Figs. 6 and 8). However, at high [Mg 2ϩ ], binding affinity of FAD to FADS⅐ADP was similar to that to free FADS because of the less favorable enthalpic contribution being compensated with a favorable binding entropy (Figs. 6 and 8).
The preformed FADS⅐ADP complex was able to bind two FMN molecules. Therefore, the presence of ADP promoted the appearance of a second FMN-binding site (Tables 2 and  3). The presence of Mg 2ϩ enhanced the FADS⅐ADP complex affinity for FMN (Table 2). At the highest [Mg 2ϩ ] assayed, the independent K d values for the two FMN-binding sites could be determined. Additionally, the simultaneous presence of Mg 2ϩ and ADP had an important effect in the enthalpic and entropic binding contributions (Figs. 6 and 8). Binding of FMN to the preformed FADS⅐ADP complex was enthalpically driven with an opposing entropic contribution in the absence of Mg 2ϩ . However, its presence promoted the enthalpic contribution to become less favorable and transformed the unfavorable entropy into a favorable one, making the interaction stronger. Binding of FMN to the FADS⅐AMPPNP complex showed similar affinity values to those for the FADS⅐ADP complex, but a second FMN-binding site was not observed ( Table 2).
Binding of RF to FADS preloaded with ADP pointed out the presence of a single RF-binding site that turned into two sites in the presence of Mg 2ϩ ( Table 2). The cation also causes a slight increase in the affinity for RF. Two binding sites to FADS⅐AMPPNP were also detected. The favorable enthalpic and, especially, the unfavorable entropic contributions became considerably less intense upon increasing [Mg 2ϩ ] (Fig. 8).

TABLE 1 Stoichiometry and interaction parameters for the FADS-nucleotide and FADS-flavin interaction determined by ITC
The data were obtained at 25°C in 20 mM sodium phosphate, pH 7.0.

ITC Analysis of Competitive Ligand Binding to Preformed FADS Complexes-A set of experiments in the presence of an
excess of FAD as competitive ligand was carried out to establish the binding domain of ligands with a single binding site ( Table  2). According to previous modeling studies (15) and to our own data, FAD must bind at the N-terminal domain, blocking both the adenine nucleotide and the putative flavin nucleotide-binding sites of this domain. Under FAD saturation and lacking Mg 2ϩ , FMN binding was hardly detected (Table 2). Because in the presence of ADP and 10 mM MgCl 2 FADS binds two FMN molecules (Table 2), FADS was simultaneously saturated with ADP and FAD at 10 mM MgCl 2 and titrated with FMN. Two FMN-binding sites were also detected, but the FMN interaction was weaker than in the single ADP presence (Table 2). Two RF-binding sites were detected in the FAD⅐FADS complex with a K d,av RF of 13.5 M (Table 2). Therefore, FAD makes the interaction of FADS with one of the RF molecules slightly weaker.
In the FADS⅐FAD complex, a single binding site for ATP or ADP was again observed (Tables 1 and 2). ATP affinity was not affected, but the presence of FAD considerably hindered ADP binding ( Table 2).

ITC Analysis of Flavin and Adenine Nucleotide Binding to the C-terminal Domain of CaFADS (⌬(1-182)FADS)-
The ability of ⌬(1-182)FADS to interact with flavins and adenine nucleotides was also studied (Table 3 and Fig. 9). ⌬(1-182)FADS was able to bind a single ATP molecule, but its affinity was weaker than that of WT FADS (Table 3). Despite their smaller magnitude, the enthalpic and entropic contributions at 10 mM MgCl 2 showed a profile similar to that of WT FADS (Figs. 4 and 9). Noticeably, in the absence of Mg 2ϩ , ATP binding to ⌬(1-182)FADS was driven by a favorable entropic contribution, whereas the favorable enthalpic one was considerably smaller than for binding to WT FADS. Binding of ADP to ⌬(1-182)FADS was only detected in the presence of Mg 2ϩ , with K d and enthalpic and entropic binding contributions similar to those for WT FADS (Table 3 and Figs. 4 and 9).
No calorimetric profiles were detected upon titration of ⌬(1-182)FADS with FAD. Titration with FMN only produced a residual profile that could not be related to a real interaction (Fig. 9, left column). However, RF was shown to bind at a single site with K d of 1.8 and 7.2 M at 0 and 10 mM MgCl 2 , respectively (Table 3). Binding was entropically driven in the absence of Mg 2ϩ but became enthalpically driven, with an entropic binding contribution still favorable, in its presence (Fig.  9). ADP increased the affinity for RF and, particularly, for FMN (Table 3) (Fig. 9, left column). ADP turned the entropy into an unfavorable contribution for the RF interaction, whereas the enthalpic one became much more favorable. The FMN interaction was also enthalpically driven, but with a favorable entropic binding contribution of almost the same magnitude (Fig. 9).

DISCUSSION
This study provides a thermodynamic characterization of substrate/product binding to CaFADS and of its individually cloned C terminus domain, analogous in sequence and structure to the eukaryotic RFKs (15).
Our data confirm that CaFADS holds two different adenine nucleotide-binding sites, each of them involved in each one of the enzyme catalytic activities (10). In the absence of Mg 2ϩ , the adenine nucleotide binds to a single site ( Table 3). Comparison of data for the interaction of the adenine nucleotides with CaF-ADS, both free and saturated with FAD, and with ⌬(1-182)FADS (Tables 1-3) allows us to distinguish between both sites. Thus, in the absence of Mg 2ϩ , ATP only binds to the phosphorylation site at the C-terminal domain, whereas ADP only binds to the adenylylation site at the N-terminal domain.
According to the structural model for CaFADS in complex with its ligands (Fig. 1B) (15), an electrostatic repulsion between the highly conserved Asp 25 and the ␥-phosphate of ATP could prevent ATP binding in the absence of Mg 2ϩ , ultimately preventing the adenylyltranferase reaction in the absence of the cation (8). In the C-terminal domain, specificity for ATP versus ADP in the absence of Mg 2ϩ might relate with the localization of the phosphates, particularly of their acidic groups; when binding ATP, the ␥-phosphate acidic groups might be stabilized by the positively charged 195-202 loop, whereas in the case of ADP the negative charges, located on the ␤-phosphate, would be surrounded by the negatively charged 268 -277 loop.
The presence of Mg 2ϩ induces the appearance of a second adenine nucleotide-binding site ( Table 1), suggesting that either Mg 2ϩ contributes to reduce the mentioned electrostatic repulsions by participating in the bonding network of the binding pocket or it induces conformational changes in the protein.
Our experimental data do not allow determination of the parameters for each one of the two ATP-or ADP-binding sites, suggesting that the type of interactions with the protein must not differ widely. Particularly, polar contacts are mostly responsible for the protein-nucleotide interaction in both sites (Fig. 4), according with the nature of the protein regions where they must bind (15). Nevertheless, the results suggest particular differences. This is in agreement with the selective binding in each of the sites and with the functional and structural differences expected for both sites (10,14,15).
The results for ATP and ADP binding (Table 1 and Fig. 4) also indicate that the ␥-phosphate of ATP is actively involved in the binding at both sites and particularly contributes to the enthalpic and entropic parameters. The different enthalpic and entropic contributions in the binding of ATP to the truncated protein compared with WT FADS (Figs. 4 and 9) indicate that the N-terminal region somehow influences nucleotide binding in the C terminus. Earlier studies hypothesized on the presence of a single flavin-binding site shared by both activities (10,13), but the presence of a second site was recently suggested (15). The present work provides experimental evidence of the existence of two independent flavin-binding sites in CaFADS. Two binding sites are observed for RF in FADS, whereas ⌬(1-182)FADS binds one RF molecule by itself (Tables 1 and 3). Therefore, one of these sites is situated in the C-terminal domain. The presence of Mg 2ϩ only slightly modulates the RF affinity. K d RF values (2.8 M at 10 mM Mg 2ϩ ) are considerably larger than those derived from steady state kinetic analysis of CaFADS (79 nM) (10).   (14 -16, 27). A single binding site in CaFADS is observed for both FMN and FAD.
According to the lack of interaction between FAD and ⌬(1-182)FADS (Table 3), FAD must bind at the N-terminal domain, with the isoalloxazine ring putatively allocated by Val 59 , Leu 98 , Tyr 106 , and Phe 128 (15). The FMN preferred binding site in the absence of adenine nucleotides must be the same as FAD.
According to the proposed mechanism, binding of FMN to free FADS would be prevented because of electrostatic repulsion between the FMN phosphate and a hypothetical negative charge around the (single) flavin-binding site (10). Initial ATP-Mg 2ϩ binding would overcome this repulsion and allow subse-quent FMN binding. However, our results invalidate this hypothesis showing that in the absence of ATP-Mg 2ϩ , FMN preferably binds to the adenylylation site. Flavin binding to CaFADS was greatly modulated by the presence of adenine nucleotides (Tables 1-3). Saturation of ⌬(1-182)FADS with ADP-Mg 2ϩ considerably enhanced the affinity of this domain for RF and, especially, for FMN (Fig. 9). The positive cooperativity of ADP-Mg 2ϩ on the binding of RF to ⌬(1-182)FADS is reciprocal (28). Therefore, when any of the two ligands is bound to ⌬(1-182)FADS, there is a 9-fold increase in the binding affinity of the second one. Similarly, ADP-Mg 2ϩ enhances the CaFADS ability to bind RF and promotes the appearance of a second FMN-binding site ( Table 2), that of FMN as a product of the phosphorylation reaction. The fact that FMN only binds to the C-terminal domain in the presence of the second reaction product, ADP, agrees with ADP leaving the phosphorylation site prior to FMN (10,16,27). At the highest [Mg 2ϩ ] assayed, the affinity for FMN at both binding sites differs by 20-fold ( Table 2). Comparison of these K d FMN values with that obtained in the simultaneous excess of ADP and FAD suggests that the highest affinity site must be the one at the N-terminal domain. Thus, ADP allows FMN to bind in the C-terminal region, whereas the affinity in the N-terminal site considerably increases. Therefore, positive cooperativity is also observed in the N-terminal region; when any of the two ligands is bound, the binding affinity of the other one increases 29-fold. Therefore, ADP, Mg 2ϩ , and particularly ADP-Mg 2ϩ produce a synergistic stabilizing effect in the formation of the FADS⅐flavin⅐nucleotide complex. Although the complexes analyzed represent combinations of products or substrate analogues that will never react with each other, similar effects might be expected for the reacting substrates. The positive cooperative binding effect could be a consequence of ADP-Mg 2ϩ mediating additional interactions between the protein and the flavin, as indicated by the favorable enthalpic contribution for the simultaneous binding of RF and ADP-Mg 2ϩ (Fig. 9). On the other hand, binding of one of the ligands could induce a conformational change that would allow the tighter binding or better accommodation of the second ligand in the active site. These possibilities are not mutually exclusive.
In an attempt to dissect the binding phenomena, the relative enthalpic (⌬⌬H), entropic (⌬⌬S) and free energy changes (⌬⌬G) associated with the binding of flavins to the binary complex FADS⅐ADP versus that to free FADS have been calculated, when possible, according to ⌬⌬G ϭ ⌬G binary Ϫ ⌬G free ; ⌬⌬H ϭ ⌬H binary Ϫ ⌬H free ; and ϪT⌬⌬S ϭ ϪT⌬S binary Ϫ (ϪT⌬S free ) (binary and free refer to the binding of a flavin to the FADS⅐ADP

TABLE 3 Stoichiometry and interaction parameters for the (⌬1-182)FADS-ligand interaction determined by ITC
The data were obtained at 25°C in 20 mM sodium phosphate, pH 7.0.  MARCH 13, 2009 • VOLUME 284 • NUMBER 11 complex and free FADS, respectively). ADP-Mg 2ϩ favored the binding of RF to ⌬(1-182)FADS by Ϫ1.3 kcal/mol. This relates to a much more favorable enthalpic contribution, Ϫ9.9 kcal/ mol, indicating that ADP-Mg 2ϩ promotes the appearance of new favorable polar interactions between ⌬(1-182)FADS and RF. The effect of ADP-Mg 2ϩ on the interaction of FMN with ⌬(1-182)FADS or with the C-terminal of FADS is even more drastic, because flavin binding only occurs when ADP and Mg 2ϩ are simultaneously bound. Moreover, the presence of ADP-Mg 2ϩ promotes not only a favorable enthalpic contribution but also a favorable entropic one, suggesting cooperation of ADP-Mg 2ϩ in the preformation of the FMN-binding site (Figs. 8 and 9). When dissecting the thermodynamics of FMN binding to the N-terminal domain saturated with ADP-Mg 2ϩ versus the free domain, the enhanced interaction (⌬⌬G ϭ Ϫ2.0 kcal/mol) correlated with a considerably unfavorable decrease in the enthalpic contribution and a considerably favorable increase in the entropic one (⌬⌬H ϭ ϩ18.8 kcal/mol and ϪT⌬⌬S ϭ Ϫ20.8 kcal/mol). The presence of ADP-Mg 2ϩ appears to cause a favorable hydrophobic effect for the FMN interaction that was not observed in its absence, and the adenine nucleotide binding "freezes" a protein conformation that favors flavin binding. Although Mg 2ϩ reduces both the enthalpic and entropic contributions in the interaction with substrates, nucleotide binding free energy is employed in reducing the conformational energy penalty occurring in flavin binding. This is in agreement with kinetic analyses suggesting that during the adenylylation cycle the adenine nucleotide binds prior to FMN (10). Despite the observed cooperativity, all of the substrates bind in the absence of the other one. Therefore, FADS appears to show a cooperativity in substrate binding less strict than that reported in other enzymes (29).

Interaction of FADS with Its Ligands
In conclusion, CaFADS holds two independent adenine nucleotide sites and two independent flavin-binding sites, which correlate with the two different reactions catalyzed. Substrates modulate the structural organization of the catalytic sites of the enzyme and the affinity for the rest of the ligands. Additionally, the energetics of the interaction and the protein function itself are related to the structural dynamics of the protein. The lack of some of the loops putatively involved in substrate binding in the crystal structure of TmFADS, together with the in silico models derived for the CaFADS, clearly suggest a high flexibility for these particular loops (13)(14)(15). The structural data reported for monofunctional nucleotydyltransferases and RFKs indicate drastic conformational changes associated with nucleotide and flavin binding, by promoting strong interactions and modifying the coordination between the active site residues and the Mg 2ϩ , ATP and flavin substrates, while simultaneously shielding the active site from the solvent (16, 17, 27, 30 -34). Therefore, it is feasible to expect that during its catalytic cycle FADS will undergo a series of conformational rearrangements to optimally allocate substrates at the active site and to stabilize transition states (34,35).