The expression level for most of the mutants was similar to that of wild-type (WT) CaFADS (2–5 mg protein/g cells). However, H28A CaFADS showed surprisingly low expression (<1 mg protein/g cells), while more than 10 mg protein/g cells were obtained for the S164 variants. In contrast to the WT protein samples, which, contain small amounts of oligomeric species easily separated by gel filtration chromatography [18,20], the H31D variant failed to show a monomeric discrete form.

All the mutants retained the main secondary structural elements of WT CaFADS (Figure S2a). Their near-UV circular dichroism (CD) spectra showed a broad negative band between 260 and 300 nm with two discrete minima at 284 and 291 nm, as for WT CaFADS [21] (Figure S2b), but replacements of H31, N125 and R161 by Ala and of H31 and N125 by Asp decreased the intensity of the signal (Figure S2b). The two tryptophan residues at the FMNAT-module of CaFADS are partially accessible to the solvent. In particular, W181 is relatively close to the putative adenine binding site of ATP, where mutations were introduced (Figure 1b), suggesting they might modify its solvent accessibility.

Addition of saturating concentrations of RF to CaFADS or to the CaFADS:ADP complex for the different variants induced difference spectra in the visible region similar to those for the WT form, indicating interaction of the flavin isoalloxazine ring with the protein (Figure S3a,b). More appreciable changes, both in shape and intensity, were observed upon addition of FMN or FAD to CaFADS (Figure S3c,e), and no difference spectra at all were detected for the R161D variant. Since in the absence of adenine nucleotides both flavins are expected to bind only at the FMNAT-module [20,21], the mutations appear to modulate the isoalloxazine: protein interaction at this module. FMN addition to the preformed CaFADS:ADP complexes increased the difference spectra magnitude (Figure S3f), in agreement with the reported detection of an additional FMN binding site in the RFK-module [20,21]. Mutations also modified the relative intensity of the spectral bands upon addition of FAD (only able to bind at the FMNAT-module) to the preformed CaFADS:ADP complexes with respect to both, the WT CaFADS:ADP complex and the difference spectra in the absence of ADP (Figure S3c,d). Since the mutated positions are, in principle, far from the expected allocation site for the flavin isoalloxazine at the FMNAT-module (see Figures 1 and 5 below), these results suggest that changes in the conformation of the adenine nucleotide binding pocket induce changes either in the isoalloxazine location within the protein or in its binding cavity.

Thermal denaturation of WT CaFADS by differential scanning calorimetry (DSC) showed two overlapping transitions (data not shown) with an apparent unfolding mid-transition temperature (T_m,app) of 42.5 °C (Table S1), suggesting that the unfolding processes of the FMNAT- and RFK-modules are not fully cooperative [21]. Replacement of N125, S164 and T165 by Ala produced a 2–3 °C decrease in the T_m,app, indicating a destabilising effect. However, mutations H28A, H31A, R161A, R161D, S164D and T165D stabilised the enzyme with increases of T_m,app up to 6–7 °C. H28D and N125D had similar T_m,app values to WT CaFADS (Table S1). Therefore, replacement of polar residues in this cavity produces different effects, which apparently depend on the protein environment stabilising the new side chain, on its ability to provide or prevent specific salt-bridges or dipole-dipole contacts with respect to the WT enzyme, as well as on different solvation patterns in the substrate binding cavity. Thus, changes in the polarity of residues at the active site cavity of the FMNAT-module modulate its shape, interacting properties and stability. Therefore, we cannot dismiss replacements, particularly with Asp, resulting in adventitious effects unrelated to the enzyme catalytic mechanism.

The values of k_cat, K_m^FMN, and K_m^ATP for the FMNAT activity of WT CaFADS are 17 min⁻¹, 1.2 μM, and 36 μM, respectively (Table 1), in agreement with data previously reported [9,32]. In general, mutations at H28, H31, N125, R161, S164 and T165 had a deleterious impact on activity.

Replacements at any of the two His residues and at N125 prevented transformation of FMN into FAD, indicating that they are crucial for FMNAT catalytic activity. Substitutions at R161 allowed FAD production but a decrease in catalytic efficiency was observed, especially for R161D (Table 1, Figure 2a,b). Mutations in the consensus xSSTxxR motif prevented transformation of FMN when Asp was substituted for S164 or T165, while deleterious effects on the catalytic efficiency were observed for the Ala variants (Table 1, Figure 2a,b). Lack of activity exhibited by some variants might be due to either low k_cat or high K_m values, but the binding studies reported below (Sections 2.6 and 2.7) confirm that the mutations particularly affect the k_cat parameter.

The values of k_cat, K_m^FAD, and K_m^PPi of WT CaFADS for the transformation of FAD and PPi into FMN and ATP are 13 min⁻¹, 1.4 μM, and 114 μM respectively (Table 2, Figure 2c,d), in agreement with data previously reported [32]. Catalytic efficiencies were in a similar range to those for the forward reaction. No reaction was detected in the absence of PPi, indicating that CaFADS has no FAD hydrolase activity. The mutations produced deleterious effects on FADpp activity, preventing in most cases product formation. Activity was only detected for N125D, R161A, R161D, S164A and T165A but, in general, the catalytic efficiency decreased significantly as a consequence of a reduction in k_cat and an increase in K_m^PPi (Table 2, Figure 2c,d).

The RFK activity of WT CaFADS exhibits an inhibition profile on the RF substrate at saturated ATP concentrations that is similarly observed in all the mutants here produced (not shown) [9,20]. A fit of the data to an equation describing the dead-end inhibition by excess of RF, generated the inhibition constant (K_i), K_m^RF, k_cat, and catalytic efficiency values for each variant within a factor of two of those of the WT enzyme (Table S2). Similar profiles and kinetic parameters for RFK activity of WT CaFADS were also observed for all the mutants when assayed at different ATP concentrations while keeping the RF one fixed. Therefore, mutations at H28, H31, N125, R161, S164 and T165 at the FMNAT-module of CaFADS do not produce major deleterious effects on the RFK activity, but only slightly modulate binding and transformation of substrates. Among the residues here mutated only T165 is involved in the interaction with the RFK-module of a neighbouring protomer by establishing one H-bond with D298 (Figure S4). However, the region holding D298 is far from the RFK-module active site and does not involve any of the flexible regions of the RFK-module related with this activity.

ITC analysis of the interaction of WT CaFADS with FMN and FAD was consistent with a single binding site at the FMNAT-module for both ligands, with dissociation constant (K_d) values of 7.8 and 0.7 μM, respectively (Table 3) [9,20,21]. All mutants maintained this single flavin binding site, but binding enthalpies for the interaction of FMN and FAD to H28A, H31A and H31D CaFADS and of FMN to R161D, S164A and S164D CaFADS are significantly reduced (Table S3, Figures 3a,b, 4a,b and S5).

Substitution of histidines in the consensus motif 28-HxGH-31 had minor effects on the CaFADS affinity for FMN. Larger effects were observed on the FAD affinity, with K_d^FAD values up to 65-fold higher than for WT (Table 3). Binding of FMN and FAD to N125D CaFADS was 5- and 40-fold weaker, respectively, with respect to WT (Table 3, Figure 3a,b). Noticeably, when Asp substituted for R161 only a small fraction of the enzyme molecules appeared able to bind both flavins, while replacement with Ala did not significantly affect this parameter. Mutations at R161 produced a 2.5-fold decrease in the affinity for FMN (Table 3). Substitution of S164 resulted in K_d^FMN values only slightly smaller than for WT CaFADS, but the affinity for FAD decreased by 30-fold (Table 3). Finally, replacement of T165 with Ala decreased the CaFADS affinity for FMN and FAD by ~15-fold, while slightly increasing upon replacement with Asp (Table 3).

Binding of both flavins to WT CaFADS was driven by a large enthalpic change with a high cost in entropy (Figure 4a,b, Table S3). Three of the four histidine replacements at 28-HxGH-31 altered the profile for FMN binding, now driven by a very low enthalpic contribution and a favourable entropic driving force. The same behavior was observed for R161D and mutants at S164 (Figure 4a). The thermodynamic profile for the interaction of FAD was similar for all the variants. The only exceptions were H28A, H31A and H31D, which showed a considerably decreased enthalpic contribution while the entropy became favourable (Figure 4b). In addition, changes in the magnitude of both contributions are observed for variants R161D, and, especially, S164D (Figure 4b). It is remarkable that in S164D, the effect on the thermodynamic parameters was opposite for the interaction with either FMN or FAD, suggesting that in this variant, the overall mode of interaction and the residues involved in their binding might differ. Particularly, the small values of the enthalpic contributions to binding suggest that the mutations prevent the formation of electrostatic interactions and H-bonds.

Two independent ATP binding sites were detected for WT CaFADS in the presence of 10 mM Mg²⁺ with an average K_d^ATP value (K_d,av^ATP:Mg) of 30 μM (Table 3) [9,20,21]. In the absence of magnesium, only one binding site, identified in the FMNAT-module, was detected with a K_d^ATP of 5.5 μM (Table 3) [9,20]. Two ATP:Mg²⁺ binding sites were also detected for all the CaFADS variants, but in general mutations produced deleterious effects in the affinity. Only H31A and R161A showed K_d,av^ATP values in the WT range. In general, K_d,av^ATP were more than 3-fold higher than for WT, and up to 20-fold in H28A (Table 3, Figure 3c).

In the absence of magnesium the interaction of variants at H28 and H31, if any, is particularly weak. Only H28D showed appreciable binding, although the affinity considerably decreased with respect to WT (K_d^ATP ~7-fold higher). The rest of the variants kept the ability to bind one ATP molecule at the FMNAT site, but with the only exception of the N125A and T165D mutants, the affinity for the adenine nucleotide was particularly weak (close to 20-fold for N125D, R161D, S164A and S164D) (Table 3, Figure 3d). These observations, together with the considerable decrease in the magnitude of the enthalpic and entropic contributions for ATP binding to the variants in the HxGH motif (Figure 4c), can be explained by the ATP binding site in the FMNAT-module being significantly altered by the mutations. Moreover, no good correlation was observed between K_d^ATP and K_m^ATP for active variants (with K_m^ATP values lower than K_d^ATP values (Tables 1 and 3)), suggesting that in the CaFADS mutants, reactions might proceed through a different mechanism from WT. Once again these results indicate the importance of polarity in the cavity of the catalytic site for both binding of ligands and catalysis.

Catalysis by enzymes of the α/β phosphodiesterase NT family is proposed to achieve nucleotidylation of substrates without the direct involvement of active site residues in covalent or acid-base catalysis [27,30,33,34]. Therefore, in FADS the adenylylation reaction is expected to occur via a nucleophilic attack by the 5′-P of the FMN on the α-P of ATP, releasing FAD and PPi [31,35]. Superposition of the FMNAT-module of CaFADS with some NTs in complex with substrates allowed tentative positioning of ATP and FMN (Figure 5a) [18].

The model suggests that H28 and H31 might contribute to allocate the α-P of ATP in the catalytically competent complex (Figure 5b), in agreement with the lack of FMNAT and FADpp activities, when both histidines, particularly H31, are replaced (Tables 1 and 2). Binding characterisation additionally supports this model, since replacement of both histidines alters the enthalpic/entropic ATP binding profiles (Figure 4c,d). The isoalloxazine and ribityl of FMN or FAD are proposed to bind in a cavity formed by several flexible protein loops (Figures 1a and 5a) [18]. In silico models suggest that H28, H31, N125 and R161 might orientate the phosphate of FMN towards the α-P of ATP during the FMNAT activity (Figure 5b), and the PPi portion of FAD for the FADpp one. Removal of H31 will prevent a favourable interaction, which apparently is critical to allocate the phosphate of FMN at the catalytic site. This is further supported by the, in general, low FMN and FAD binding enthalpy in these mutants (Figure 4a,b) while difference spectra confirm that the isoalloxazine portion is still able to interact with the FMNAT-module (Figure S3). Therefore, in WT CaFADS, the enthalpic contribution to FMN binding is mainly provided by interaction of the HxGH motif, particularly H31, with the phosphates. This is in agreement with the fact that removal of H31 has a similar deleterious effect on the enthalpic contribution as removal of the phosphate of FMN for RF binding [21]. On the other hand, the entropy changes additionally suggest a more dynamic interaction, in agreement with conformational changes in the environment of some aromatic side chains at the FMNAT-module modifying the binding organisation of the substrates (Figure S2). Modeling of the position of ATP suggests that its adenine ring would be inserted at the bottom of the cavity establishing H-bonds with main chain atoms of some hydrophobic residues [18], while the oxygen connecting the α-P and β-P of ATP might be situated at the interacting distance of the Nɛ of H31 (Figure 5b).

Mutational studies on the equivalent histidines in CTP-glycerol-3-phosphate and CTP-phosphocholine synthase from Bacillus subtillis, as well as in human and murine 3′-phosphoadenosine 5′-phosphosulfate synthases, reported deleterious effects on k_cat, with minor changes in K_m[22,23]. Alanine substitutions in Methanobacterium thermoautotropicum nicotinamide mononucleotide adenylyltransferase (NMNAT) produced variants with some activity, and the comparative analysis suggested that the second histidine was more relevant as the catalytic group [26,28,29]. Our data indicate that, in CaFADS, this motif is critical both for the catalytically competent ternary complex to be formed and for the catalysis itself, suggesting both histidines might be involved in the stabilisation of the pentacoordinate transition state arising from nucleophilic attack at the α-P of ATP [22,34]. The lack of activity made it impossible to determine, from the kinetic parameters, which of them is more relevant. However, binding parameters and the structural model suggest a more important contribution for H31, while H28 would stabilise the position of H31 (through stacking and H-bond interactions, as reported for other proteins of the NT superfamily) [26,29].

The model for the complex of the FMNAT-module with ATP and FMN, would putatively locate N125 at an adequate distance to interact both with the α-P of ATP and with the phosphate of FMN (equivalent position to the PPi in the crystallographic structure) through H-bonds (Figure 5b). Either replacement of N125 by Ala or Asp produces variants unable to transform FMN into FAD and, with a very low efficiency in the FADpp activity (Tables 1 and 2). The interaction of N125D with flavins and ATP is clearly altered (Table 3, Figure 3). By contrast, although the presence of an Ala at this position still allows binding it cannot stabilise the phosphates, and substrates are not able to achieve a catalytically competent interaction. Since the FADS sequence alignments show that in general either Asn or Asp can be found in this position [17], the properties of this position might be key for each particular FADS species. Therefore, both our structural models and biochemical data suggest that N125 contributes in locating and orientating both ligands as a previous step for catalysis, and envisages a possible involvement in the stabilisation of the transition state.

With regard to the 161-xxSSTxxR-168 motif, replacements at S164 and T165 also produce important deleterious effects in the FMNAT and FADpp activities of CaFADS (Tables 1 and 2). Mutations at S164, despite allowing FMN binding, appear to produce a different protein-ligand organisation in the cavity (Table 3, Figure 4). Moreover, interactions with ATP and, especially, with FAD were clearly altered, indicating that this serine must stabilise ligand binding, probably by H-bond contact. In fact, introduction of a negatively charge residue produces a variant completely inactive to both activities, but able to bind flavins and ATP (Tables 1–3). Such behavior might be explained by an electrostatic repulsion between the introduced aspartate and the phosphates that avoids their proper position in the active site. The low catalytic efficiencies observed when an Ala replaced S164 supports that this serine might provide polarity for transition state stabilisation. Replacement of T165 by an Ala produced deleterious effects in the interaction with FMN and FAD (Table 3). Despite our kinetic parameters indicating that T165 is critical for substrate binding and catalysis (Tables 1 and 2, Figure 2), the structural model does not envisage any apparent interaction with ligands. Therefore, the interaction of the FMNAT-module with its substrates might include conformational changes to allow T165 to play a role in the catalytically competent complex or, alternatively, T165 might be involved in a conformational change necessary for binding and catalysis. Finally, the position of R161 is usually occupied by a positively charged side chain, Arg or Lys, in the FADS family. However, non-polar residues, as Val and Ile, can be also found in FADSs and NTs [17]. Replacements at R161 in CaFADS produce deleterious effects on k_cat (Tables 1 and 2, Figure 2), and only a few molecules of the R161D variant appear active to bind FAD, FMN and ATP. All together these observations might be explained by a folding difference inducing a significant energetic barrier for ligand binding that produces an effect unrelated to the enzyme catalytic mechanism itself. This is consistent with the absence of difference spectra for the interaction of this mutant with either FMN or FAD. These observations suggest a role for R161 in ATP binding and probably also in the stabilisation of the transient state during the reaction, as also proposed for the equivalent residue in NMNAT [28]. The structural model analysis does not envisage any direct interaction of this residue with ligands, suggesting conformational changes in the protein module upon ATP interaction.

Mutations were introduced into the pET28a-CaFADS plasmid [17] by using the QuikChange kit (Stratagene, La Jolla, CA, USA) in combination with adequate synthetic oligonucleotides (Supplementary Material), and verified by DNA sequence analysis. WT and mutated CaFADS were over-expressed in E. coli and purified as described [17]. Samples were dialysed in 20 mM PIPES, 10 mM MgCl₂, pH 7.0 and stored at −80 °C. A gel filtration chromatography with HiPrep 26/60 Sephacryl S-200 HR column (GE Healthcare, Uppsala, Sweden) was used to separate the monomeric fraction [18,20].

CD spectra were recorded in a Chirascan spectropolarimeter (Applied Photophysics Ltd., Leatherhead, Surrey, UK) at 25 °C. 5 μM CaFADS in 5 mM PIPES, 10 mM MgCl2, pH 7.0 and 20 μM CaFADS in 20 mM PIPES, 10 mM MgCl₂, pH 7.0 were used in the far-UV (cuvette path length, 0.1 cm) and near-UV CD (0.4 cm), respectively. Difference spectra (Figure S3) were obtained upon addition of saturated concentrations for each one of the different ligands (50–100 μM FAD or FMN and 20–40 μM RF) in 20 mM PIPES, 10 mM MgCl₂, pH 7.0 to samples of either CaFADS (3–4 μM) or the CaFADS:ADP (3–4 μM with final ADP concentration of ~450 μM) complexes for the different variants.

The CaFADS RFK activity was measured at a final volume of 500 μL in 20 mM PIPES, 0.8 mM MgCl₂, pH 7.0 containing 0.5–45 μM RF, 10–400 μM ATP. The reaction mixture was pre-incubated at 37 °C and the reaction was initiated by addition of 20–50 nM of the enzyme. The mixture was incubated at 37 °C for periods ranging from 0.5–2 min and, finally, the reaction was stopped by heating at 100 °C for 5 min [20,21]. The protein was removed from the reaction mixture by centrifugation, and the supernatant composition was analysed using an Alliance HPLC system (Waters, Milford, MA, USA) equipped with a 2707 autosampler. An aliquot of 20 μL of the filtered solution was applied to a HSST3 column (4.6 × 150 mm, 3.5 μm, Waters), preceded by a pre-column (4.6 × 20 mm, 3.5 μm, Waters) of the same material. The chromatography was developed at 1 mL/min with a 6-min isocratic program of methanol 40% (v/v) in 5 mM ammonium acetate (pH 6.0). Detection of flavins was carried out using a 2475 Multi λ fluorescence detector (Waters, Milford, MA, USA), an excitation wavelength of 470 nm and emission of 530 nm. Under these conditions the retention times for the flavins were 2.25 min for FAD, 3.43 min for FMN, and 5.50 min for RF. The amounts of produced FMN or/and FAD were determined from their corresponding standard curves obtained under the same conditions. Kinetic data obtained for one substrate at saturated concentrations of the second one (nanomol of flavin transformed per minute) were interpreted using the Michaelis-Menten kinetic model, obtaining k_cat and K_m with errors of ±10%. A model describing the substrate inhibition effect produced in a bi-substrate enzyme kinetic was used to interpret the kinetic evolution of the RFK activity on RF concentration (see [20] and Supplementary Material) [36]. In these situations, errors in the apparent K_m and k_cat (^appK_m and ^appk_cat) increased with K_i getting closer to K_m^S[20].

The CaFADS FMNAT and FADpp activities were fluorimetrically measured using a continuous assay. Time dependent measurements were performed at a final volume of 1 mL in 20 mM PIPES, 10 mM MgCl₂, pH 7.0, containing 3–8 μM FMN/FAD, 10–400 μM ATP/PPi and 20–40 nM of enzyme at 37 °C (higher concentrations for variants with low activity). Measurements were performed in a Cary Eclipse spectrophotofluorimeter with excitation and emission wavelengths at 420 nm and 530 nm, respectively. FAD and FMN fluorescence were individually calibrated using standard solutions. The rate of FAD formation was calculated from the rate of fluorescence decrease (ΔF/Δt), measured as the tangent to the initial part of the experimental curve by applying the equation:

where ΔF is the decrease in the value of the fluorescence expressed in arbitrary units, Δt is the measurement of reaction time expressed in per minute, and K_FMN and K_FAD are the FMN and FAD fluorescent rate constants expressed in per micromolar [37]. The kinetic data obtained for one substrate at saturated concentrations of the second one were interpreted using the Michaelis-Menten kinetic model. Data are media of at least three different experiments allowing for estimation of k_cat and K_m errors of ±10%.

Thermal stability was determined with a VP-DSC microcalorimeter (MicroCal LLC, Northampton, MA, USA). Protein samples and reference solutions were degassed and carefully loaded into the cells to avoid bubble formation. The baseline of the instrument was routinely recorded before the experiments with both cells filled with buffer. Thermal denaturation scans were performed with 20 μM FADS solutions in 20 mM PIPES, 10 mM MgCl₂, pH 7.0, at a scanning rate of 1 °C/min from 10 °C to 80 °C. Errors in the T_m,app data are within ±0.3–0.5 °C.

Measurements were carried out using a high precision VP-ITC (MicroCal LLC, Northampton, MA, USA) following the procedure described previously [20,21]. Typically, 200 μM FAD or FMN and 300 μM ATP solutions were used to titrate ~20 μM CaFADS. Ligand and CaFADS were dissolved in 20 mM PIPES, pH 7.0, at the indicated MgCl₂ concentration. Control titrations were performed by injecting ligand into buffer solution. The association constant, the enthalpy change (ΔH) and the stoichiometry, or their average values, were obtained through non-linear regression of the experimental data to a home derived model for one or two independent binding sites implemented in Origin 7.0 (OriginLab Corporation, Northampton, MA, 2002) (see Supplementary Material). The K_d, the free energy change (ΔG), and the entropy change (ΔS) were obtained from basic thermodynamic relationships. Experiments were performed in duplicate or triplicate. The errors considered in the measured parameters (±15% in K_d and ±0.3 kcal/mol in ΔH and −TΔS) were taken as larger than the standard deviation between replicates and the numerical error after fitting analysis.