The RFK catalytic cycle of the pathogen Streptococcus pneumoniae shows species-specific features in prokaryotic FMN synthesis

Abstract Emergence of multidrug-resistant bacteria forces us to explore new therapeutic strategies, and proteins involved in key metabolic pathways are promising anti-bacterial targets. Bifunctional flavin-adenine dinucleotide (FAD) synthetases (FADS) are prokaryotic enzymes that synthesise the flavin mononucleotide (FMN) and FAD cofactors. The FADS from the human pathogen Streptococcus pneumoniae (SpnFADS)–causative agent of pneumonia in humans − shows relevant catalytic dissimilarities compared to other FADSs. Here, by integrating thermodynamic and kinetic data, we present a global description of the riboflavin kinase activity of SpnFADS, as well as of the inhibition mechanisms regulating this activity. Our data shed light on biophysical determinants that modulate species-specific conformational changes leading to catalytically competent conformations, as well as binding rates and affinities of substrates versus products. This knowledge paves the way for the development of tools − that taking advantage of the regulatory dissimilarities during FMN biosynthesis in different species − might be used in the discovery of specific anti-pneumococcal drugs.


Introduction
Streptococcus pneumoniae is the causative agent of human pneumonia disease 1 , meningitis and bacteremia in children and adults. It is estimated that 1.6 millions of people, including 1 million of children under age five die every year of pneumonia disease 2,3 . The irruption during the last decades of multi-drug resistant pneumococci has revealed the need of finding new drugs, as well as novel drug targets. The bifunctional flavin-adenine dinucleotide (FAD) synthetase (FADS) from S. pneumoniae (SpnFADS) arises as a potential drug target 4,5 , since it synthesises the essential cofactors flavin mononucleotide (FMN) and FAD 6 , involved in a plethora of vital processes as part of flavoproteins and flavoenzymes [7][8][9] . As other bacterial FADSs, SpnFADS produces FMN and FAD through two sequential activities. First, a riboflavin kinase activity (RFK) at its C-terminus module phosphorylates riboflavin (RF) to FMN, and then the adenosine 5 0 -triphosphate (ATP):FMN:adenylyltransferase (FMNAT) activity of the enzyme N-terminus module transforms FMN into FAD [10][11][12] . Two characteristics stand out among the suitable properties of FADSs as drug targets. First, bacterial FADSs differ from the human proteins that synthesise FMN and particularly FAD. Thus, the eukaryotic FMNAT activity is catalysed by an enzyme with a completely different protein folding and chemistry relative to the N-terminus of bacterial FADS [13][14][15][16] . Regarding FMN production, the monofunctional Homo sapiens RFK has been hardly characterised so far. Overall, it is structurally homologous to the RFK module of bacterial enzymes, but only structures containing bound ligands are available 17,18 . Nonetheless, structural data predict differences in conformational changes to achieve the catalytic complex 19,20 , and the scarce biochemical information suggests differences in redox environmental requirements for maximal activity 11,12 . Second, the members of the prokaryotic FADSs family studied up to now differ catalytically among them, which might facilitate the design of new species-specific medicines. Structurally, when comparing SpnFADS with the member of the family so far best characterised-which is that from the organism Corynebacterium ammoniagenes (CaFADS)-it presents a very similar structure, with little differences in the position of some key loops. However, these two proteins only share the 26% of sequence homology 6 . Despite the overall structural similitude among prokaryotic FADSs 6,12,18 , SpnFADS shows three main distinctive functional behaviors; (i) it mainly stabilises monomerswhich are the functional form 6 -or traces of dimers, during catalysis; (ii) its FMNAT activity requires reduced FMN as a substrate; and (iii) its RFK activity is not regulated by the RF substrate 6 .
Here, we focus on the SpnFADS RFK activity, using pre-steady and steady-state biophysical techniques to describe this activity and the inhibition mechanism employed by this enzyme to regulate FMN synthesis. We take a close look at the thermodynamic and kinetic basis that determine the ligand binding order, the cooperativity between ligands and the inhibitory mechanism performed by the reaction products. Furthermore, we compare our results with those obtained for the FADS from C. ammoniagenes to identify key regulatory differences between both proteins.

Cloning, expression, and purification of SpnFADS
SpnFADS was cloned, overexpressed, and purified as previously described 6 . In short, Escherichia coli Bl21 Star TM (DE3) cells were transformed with a pET-15b vector that contains the DNA sequence encoding SpnFADS. Transformed cells were grown and protein expression induced through isopropyl b-D-1-thiogalactopyranoside (IPTG) addition. Cells were harvested and broken by sonication. The supernatant was loaded into a His-Trap affinity column and the protein eluted applying a 10-500 mM imidazole gradient. The His 6 -Tag was removed and then the protein was loaded into HisTrap HP and GSTrap 4B connected columns. The unbound fraction was further purified by size exclusion chromatography. The protein purity was tested and pure protein aliquots were conserved at À80 C.

Steady-state RFK activity
The SpnFADS RFK activity was measured at 25 C in 500 mL of 20 mM 1,4-piperazinediethanesulfonic acid (PIPES) and 0.8 mM MgCl 2 , pH 7.0. Reaction samples contained different concentrations of RF (0.5-30 mM) and ATP (10-500 mM), as previously described 21,22 . The inhibitory effect of the products of the reaction was analysed as previously described 23 . In short, the SpnFADS RFK activity was determined at increasing concentrations of FMN, varying the ATP concentration and keeping the RF constant (when studying the inhibitory effect of FMN), and at increasing concentrations of adenosine 5 0 -diphosphate (ADP), varying the RF concentration and keeping the ATP fixed (when studying the inhibitory effect of ADP). The flavin composition of the supernatant was determined using an Alliance high performance liquid chromatography (HPLC) system (Waters, Milford, MA, USA) equipped with a 2707 autosampler and a HSST3 column (4.6 Â 50 mm, 3.5 mm; Waters) preceded by a precolumn (4.6 Â 20 mm, 3.5 mm; Waters) as previously described 21,22 . The FMN concentration was quantified using its standard curve. All the experiments were carried out in triplicate.
Michaelis-Menten (K m ) and catalytic rate (k cat ) constants were obtained by fitting the obtained data to the Michaelis equation 24 . The inhibitory mechanism performed by the products of the RFK reaction À FMN and ADP À was analysed by evaluating their effects on the K m and k cat values, obtained by the individual fitting of data sets to the Michaelis-Menten model. Additionally, the data sets were globally fit utilising the Lineweaver-Burk equations for mixed inhibition 25 (Equation (1)).
where, [S] and [I] are the concentration of substrates and product inhibitor, respectively. K i and K' i are the product inhibition constants 25 . Experiments were performed in triplicate. The estimated errors in k cat , K m, and K i were within ±15% of their values.

Pre-steady-state kinetics
Kinetic experiments in the pre-steady state were registered as previously described 23 , using stopped-flow spectroscopy on an Applied Photophysics SX17. MV spectrophotometer, using the Xscan software (Applied Photophysics Ltd., Leatherhead, UK). Fast kinetic measurements were carried out as previously described 23 , at 25 C in PIPES 20 mM pH 7.0, 0.8 mM MgCl 2 . About 0.2 mM SpnFADS was mixed with reaction samples that contained increasing concentrations of the flavin ligands (FLV, herein indicating RF or FMN), in the presence and in the absence of ADP or ATP (herein referred as ANP). Controls were measured in the same buffer but without MgCl 2 . All concentrations indicated here are the final ones in the reaction cell. The kinetic traces were registered until obtaining three reproducible traces. Kinetic traces were fit to exponential equations (Equation (2)), where each exponential term describes a different process. A linear correction term was added (Equation (3)) when a specific process was not finished within the measuring timeframe.
where, A i and k obs,i are the amplitude and the observed kinetic constant for each process (i) that contributes to the overall timedependent fluorescence change.
The processes whose k obs showed a linear dependency on the flavin concentration were fit to a one-step model that describes the kinetic equilibrium for the formation and the dissociation of enzyme-flavin complex (Equation (4)) where, k on and k off are the complex formation and dissociation kinetic constants, respectively. Experiments were performed in triplicate. The flavin photobleaching was analysed as previously described 23 (not shown).

Isothermal titration calorimetry
We performed isothermal titration calorimetry (ITC) assays to elucidate both the ligand binding order and the thermodynamic inhibition produced by the binding of the FLV and ANP ligands. Titrations were performed on an AutoITC200 (MicroCal, Malvern, UK) thermostated at 20 C, as previously described 23 . Approximately 25 mM SpnFADS contained in a 200 mL-cell was titrated with solutions of 180 mM RF, 250 mM FMN, or 350 mM ATP or ADP. Additionally, titrations with ANP ligands into mixtures that contain the protein pre-bound to FLV ligands were performed, as well as titrations with FLV ligands into pre-bound SpnFADS:ANP complexes. The titrations were conducted through 19 stepwise injections of 2 mL of the titrating ligand to the calorimetric cell, as previously described 21,26 . Both the ligands and the protein were dissolved in 20 mM PIPES, pH 7.0, either in presence of 0.8 mM MgCl 2 or in absence of this cation, and degassed before the titration. The enthalpy change (DH), the association constant (K a ), and the binding stoichiometry (N) were obtained through non-linear least squares regression of the data using a homemade fitting routine corresponding to a model for one or two independent binding sites, implemented in Origin 7.0 (OriginLab, Northampton, MA, USA), as previously described 21,26 . The Gibbs free energy (DG), the entropic contribution (ÀTDS), and the dissociation constant (K d ) were obtained using well-known thermodynamic equations.
Cooperativity constants (a) between ANP and FLV ligands were obtained as previously described 23,27,28 . Particular titrations with ANP ligands into mixtures of the protein and 100 mM FLV were fit to a homemade fitting routine corresponding to a model that considers the influence of the FLV ligand in the protein binding affinity for the ANP ligand 27,28 .
Experiments were performed in triplicate. The errors considered in the measured parameters (±15% in K d and K a values, ±0.3 kcal mol À1 in DG, DH, and ÀTDS and ±20% in a) were assumed to be larger than the standard deviation between replicates and the numerical error after fitting analysis.

Results and discussion
The products of the reaction inhibit the RFK activity of SpnFADS Feedback inhibition is a frequent strategy to regulate enzymes involved in key metabolic pathways 29,30 . Some bacteria regulate FMN synthesis through the inhibition of the RFK activity of their bifunctional FADSs by the reaction products-FMN and ADP-and/ or the RF substrate 19,23 , although the inhibition level triggered differs between organisms 6,23 . Here, our Michaelis-Menten plots of the SpnFADS RFK activity, for both reaction substrates-RF and ATP-at increasing concentrations of the FMN and ADP products, reveal that k cat decreases while K m ATP and K m FMN increase ( Figure  1). These data point out to mixed or uncompetitive inhibition mechanisms. Therefore, to discern the inhibition mechanism, we carried out Lineweaver-Burk representations (Figure 1(B,D)), which reveal that both products act as mixed inhibitors ( Figure 1). Thus, FMN and ADP bind to the free enzyme as well as to the SpnFADS:ATP and SpnFADS:RF complexes, respectively. However, their affinities for the free protein and for the complex considerably differ (Table 1), being K i ADP and K i FMN 6.4 and 5.5 times smaller than K i ADP and K i FMN , respectively. That indicates that both reaction products bind preferentially to the free enzyme. Also, the considerably smaller inhibition constants for FMN compared to ADP (K i ADP /K i FMN ¼ 100) reveals the flavin product as a much more potent inhibitor (Table 1). Equivalent data obtained for the so far best characterised FADS-that belonging to C. ammoniagenes (CaFADS)-show that ADP and FMN are, respectively, competitive and uncompetitive inhibitors of its RFK activity 23 . The FMN product appears as a considerable more potent inhibitor of the RFK activity of CaFADS than of SpnFADS, as the K m RF /K i FMN values of 7.1 and 0.8 for CaFADS and SpnFADS, respectively, show. This, together with the fact that RF strongly inhibits the CaFADS RFK activity but not the SpnFADS one 6 , points to species-specific inhibition and activity modulation mechanisms. Physiologically, other preferred regulatory strategies, such as the use of reduced FMN during the FMNAT activity, and the different oligomeric assemblies established along both activities of these enzymes 6,31,32 , might be also behind the distinct requirements for the RFK activity regulation.
The binding of the substrates of the RFK reaction is the fastest and most favored process for SpnFADS Then we used stopped-flow spectrophotometry 23,33 to kinetically identify individual steps during the SpnFADS RFK reaction. Here, we took advantage of two aspects; (i) RF and FMN have the same fluorescence spectra and yields 34 and (ii) under oxidising   23 . This suggests that SpnFADS is not able either to bind oxidised flavins or to internalise their isoalloxazine ring. On the contrary, fast and intense exponential flavin fluorescence variations were observed when mixing SpnFADS simultaneously with ANP (ATP or ADP) and FLV ligands. All samples showed fast initial fluorescence decays, but mixtures containing ATP also presented subsequent fluorescence increases (Figure 2(A,B)). As recently reported for CaFADS 23 , we relate the initial fluorescence decay to FLV binding and internalisation through a conformational change of the loop-L4c, that closes the flavin binding site when ANP is previously bound (Supplementary Figure SP1) 20 . The succeeding fluorescence raise can be similarly related to an ATP-induced conformational change that re-opens the flavin binding site, making the isoalloxazine ring accessible again to the solvent 23 (Figure 2(C, D), schemes). k obs1 and k obs2 represent, respectively, the observed rates for these processes. The linear dependence of k obs1 on FLV concentrations (Figure 2(C)) allows us calculating flavin association and dissociation rate constants (k on and k off , respectively) and the derived dissociation constants (K d ). The binding of the RFK substrates-RF and ATP-is the fastest process (Table 2), and also shows the largest fluorescence decay amplitude (Figure 2(A)). Hence, contrary to that reported for CaFADS 23 , the SpnFADS RFK site preferably binds the RF and ATP substrates over other combinations of ligands, being the binding of the FMN product in presence of the ATP substrate the least favored combination. In addition, k obs2 values show that the conformational rearrangement for flavin release is considerably faster when RF is the flavin initially bound (Figure 2(D)). Moreover, k obs2 shows a biphasic behavior negatively affected by the FMN concentration, indicating that accumulation of FMN hinders its release. Noticeable, k obs1 and k obs2 for processes that involve RF and ATP are in the k cat range, revealing that the reaction steps represented by these parameters are relevant for catalysis. Nevertheless, as the k obs1 /k obs2 ratio indicates, the binding and the internalisation of the reaction substrates is 4 times faster than the subsequent conformational change that releases flavins to the solvent, being this last process the SpnFADS RFK reaction bottleneck.
Considering our results in the context of those reported for the RFK cycle of CaFADS, some key facts are worthy to be highlighted. Both proteins require ANP nucleotides to bind/internalise flavin ligands, and they show the same overall individual processes along the reaction. Additionally, both FADSs are able to bind all  the possible combinations of ANP-FLV ligands, although differences in the relative binding rates and magnitude of the associated spectroscopic changes must be behind the lower inhibition levels shown by SpnFADS. Thus, while binding of the substrates of the RFK reaction-that is RF and ATP-is the fastest and most favored process for SpnFADS (Table 2), it is the slowest one for CaFADS 23 . This fact together with a 10-times higher K i ADP , and a 9-times higher K m RF /K i FMN ratio for CaFADS than for SpnFADS, explains the larger inhibition by the reaction products observed for CaFADS (Table 1) 23 . Comparison of the k obs2 behavior for both enzymes (Figure 2(D)) 23 -k obs2 linearly depends on the RF concentration for SpnFADS while it shows a biphasic behavior for CaFADS-shows that during the CaFADS RFK reaction, high RF concentrations inhibit the ATP-induced conformational change for flavin release, while SpnFADS does not present such inhibition. This might be a determinant of the strong inhibition by RF that CaFADS suffers 21,23 , which is absent in SpnFADS.

Thermodynamics explain the modest inhibition of the SpnFADS RFK activity
We next used ITC to determine whether the identified kinetic processes were relevant to reach the thermodynamic equilibrium. We titrated with the substrates and the products of the RFK reaction, both free SpnFADS and its binary mixtures with either ANP or FLV ligands. Since Mg 2þ is necessary for the RFK reaction to occur, we performed all titrations with and without this cation, to determine its role on ligands binding. However, titrations that involve both reaction substrates only were executed without Mg 2þ , since the reaction heat would mask the binding enthalpy (Figure 3(A)). Figure 3 illustrates all the possible binding pathways occurring in the interaction landscape of the SpnFADS RFK module with the substrates and products. As reported for CaFADS 23 , Mg 2þ is key in the formation of SpnFADS:ANP:FLV complexes (Figure 3 and Supplementary Table SP1): (i) by increasing the protein fraction prone to interact with a specific ligand, and consequently, the probability of the pathway, and (ii) by favoring the formation of ternary complexes from the binary ones (compare arrows thickness and DG values in Figure 3(A,B)). The stabilisation of these ternary complexes by Mg 2þ is mainly a consequence of the less unfavorable binding entropy (Supplementary Table SP1), which suggests different conformations in ternary complexes with and without MgCl 2 . It is worthy to highlight that although FLV binding to the free protein is not directly observed by ITC or stopped-flow spectrophotometry, flavins-particularly when Mg 2þ is presenthighly modify the protein affinity for ANP ligands (Figure 3(B), Supplementary Table SP1). This cooperative effect-also observed with CaFADS 23 -evinces that RF and FMN act as slow-binding ligands 35 . However, although their binding is too slow to be measured within the experimental time, it can be indirectly estimated 27,28,36 .
Titrations without MgCl 2 (Figure 3(A)) allow establishing an interaction diagram that includes "pseudo-reactive" pathways with the reaction substrates (SpnFADS:ATP:RF). Comparison of this diagram with that obtained for CaFADS 23 reveals important differences, which might contribute to the regulatory dissimilarities displayed between both proteins. Thus, (i) two alternative  We can extract two main conclusions from these results. First, both proteins achieve the catalytic FADS:RF:ATP complex through different mechanisms; SpnFADS seems to follow a random sequential binding of the RFK substrates, while CaFADS requires its concerted fit 23 . Second, as the RFK reaction progresses and the products accumulate, pathways different from the reactive one are favored in the case of CaFADS, which reduces the efficiency of the overall catalysis 23 . For its part, the pathway that leads to SpnFADS:ATP:RF is the most favored, at least until reaching a high [products]/[substrates] ratio.

Conformational differences explain the dissimilar regulation of prokaryotic RFK cycles
Structurally, the different mechanisms by which RF and ATP get allocated in the RFK active site might be related to the simpler conformational changes required by SpnFADS. Figure 4 shows the relative conformation of the catalytic PTAN motif in SpnFADS regarding CaFADS (Figure 4(B)) and CaFADS:FMN:ADP structures (Figure 4(C)). Remarkably, the conformation of this motif in SpnFADS resembles that of CaFADS in a ternary complex with the products of the RFK reaction. So, the ligand-induced conformational change of this motif, which is crucial for the CaFADS RFK cycle 20 , is not necessary for SpnFADS. This allows flexibility in the allocation order of substrates in the SpnFADS RFK active site. Figure 3 also reveals highly different DG values for titrations of either free SpnFADS or of its binary complexes with the same ligand, as well as that flavins are able to bind the enzyme in the presence of ANP ligands but not in its absence. These effects can be explained considering binding cooperative effects between ANP and FLV ligands (Table 3), and slow binding of flavins to free SpnFADS. Therefore, we carried out a co-operativity study that again came out with fundamental differences regarding heterotropic cooperativity in the CaFADS RFK module. Our data show that for SpnFADS the RF and ATP substrates always show positive cooperativity (Table 3). On the contrary, in the case of CaFADS low RF concentrations facilitate ATP binding, while increasing RF amounts hinder it 23 . This different cooperative behavior of the substrates explains the inhibition by the RF substrate that CaFADS shows but not SpnFADS. Structurally, such effect might be also related to the different conformation of the PTAN motif. In CaFADS, the occupation of the ANP binding site by RF -when it is in excess-might prevent the ligand-induced conformational change of this motif, which is necessary for ATP binding ( Figure  4) 23 . This structural rearrangement is not necessary for SpnFADS (Figure 4(B)), and therefore, the excess of RF does not hinder ATP binding.
Collectively, our results shed light on the kinetic and thermodynamic basis behind the substrates and products inhibition regulatory differences in the RFK cycles of SpnFADS and CaFADS. These essential enzymes differ in the substrates binding order, as well as in the inhibitory potency of the products and their action mechanism. The lesser inhibition level presented by SpnFADSregarding CaFADS-is because of the binding of the reaction substrates is the fastest and thermodynamically most favored process, regardless of products accumulation. The structural origin of such divergences is the dissimilar conformation that the key PTAN motif shows in both FADSs, which points to species-specific conformational changes during the RFK activity. This fact, together with other differential regulatory strategies, such as the use of reduced substrates, provide us with a broader frame in which we could work in the development of new anti-microbials specific for S. pneumoniae. That way, it might be feasible to envisage drugs binding to a particular SpnFADS conformation, being unable to interact-due to the structural constraints and functional differences-with other FADSs.

Conclusions
In conclusion, here we present a complete description of the RFK catalytic cycle of SpnFADS, integrating thermodynamic and kinetic data, both in the pre-steady and in the steady-state, obtained using different biophysical and biochemical tools. We consider our research highly relevant in a double way. On one hand, combining physico-chemical tools commonly used in the biophysical characterisation of proteins, we have designed a strategy to go beyond the information directly extracted from this kind of techniques to offer a model that contributes to explain the dynamics during the SpnFADS RFK catalytic cycle. On the other hand, the results derived from our study are also relevant for the scientific community. Thus, we have found important differences between the RFK catalytic cycles of two structurally similar essential proteins, i.e. CaFADS and SpnFADS, shedding light on the thermodynamic and kinetic determinants that led to these inhibitory differences. Taking into account the pathogenic character of S. pneumoniae, the essentiality of its FADS and the last tendencies in discovering species-specific drugs, which reduce the resistances emergence, the knowledge here presented might facilitate the development of drugs able to bind a specific conformation of SpnFADS, having no effect on other members of the FADS family.

Disclosure statement
No potential conflict of interest was reported by the authors.  a is the cooperativity coefficient, N the fraction of total protein able to bind the titrating ligand, and Dh the enthalpy change associated to each process.