Characterization of the nucleotide-binding domain NsrF from the BceAB-type ABC-transporter NsrFP from the human pathogen Streptococcus agalactiae

Treatment of bacterial infections is a great challenge of our era due to the various resistance mechanisms against antibiotics. Antimicrobial peptides are considered to be potential novel compound as antibiotic treatment. However, some bacteria, especially many human pathogens, are inherently resistant to these compounds, due to the expression of BceAB-type ABC transporters. This rather new transporter family is not very well studied. Here, we report the first full characterization of the nucleotide binding domain of a BceAB type transporter from Streptococcus agalactiae, namely SaNsrF of the transporter SaNsrFP, which confers resistance against nisin and gallidermin. We determined the NTP hydrolysis kinetics and used molecular modeling and simulations in combination with small angle X-ray scattering to obtain structural models of the SaNsrF monomer and dimer. The fact that the SaNsrFH202A variant displayed no ATPase activity was rationalized in terms of changes of the structural dynamics of the dimeric interface. Kinetic data show a clear preference for ATP as a substrate, and the prediction of binding modes allowed us to explain this selectivity over other NTPs.

Therapeutic compounds against bacterial infections are currently one of the biggest needs worldwide. Among antibiotics, antimicrobial peptides (AMP) offer promising potential for the treatment of bacterial infections, alone or in combination with already known molecules 1,2 . An alarming number of pathogenic multidrug resistant strains have evolved under the selective pressure caused by decades of incorrect antibiotic usage. Among them, methicillin-resistant Staphylococcus aureus (MRSA) or vancomycin-resistant Enterococcus (VRE) pose a high risk to therapeutic regimens 3 . To include new classes of antibiotics in therapy, studies were performed with lantibiotics, a class of AMPs. These ribosomally-synthesized peptides exhibit high potency against several human pathogenic bacterial strains [2][3][4] and show high stability to chemical and enzymatic degradation due to multiple intramolecular thioether rings and unsaturated amino acids [4][5][6][7][8] .
Most known lantibiotics act similar in that they inhibit cell wall synthesis 9 . A common target for AMPs is the peptidoglycan layer, which exists in Gram-positive as well as Gram-negative bacteria. It is built up by altering amino sugars such as N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) and stabilized by a cross-linkage of those polymer chains. The inhibition of the cell wall synthesis results in reduced cell growth

Results
Cloning, expression and purification. For substrate transport BceAB-type ABC transporters depend on energy supply generated by ATP hydrolysis, which is mediated by the NBD. Here, we characterized the NBD NsrF of the BceAB-type ABC transporter NsrFP from Streptococcus agalactiae. To heterologously express SaN-srF WT and SaNsrF H202A , we constructed expression vectors using a codon-optimized version of SaNsrF for the heterologous expression in E. coli (Gen Bank accession number: WP_000923537). These constructs expressed a SaNsrF protein with an N-terminal His10-tag attached for purification using Metal Ion Affinity Chromatography. The corresponding SaNsrF constructs were expressed under the control of the plasmid-based T7-promoter via induction with Isopropyl-β-D-thiogalactopyranoside (IPTG). SaNsrF WT was purified to high homogeneity (Fig. 1A), and was examined by Size Exclusion Chromatography coupled to Multiangle Light Scattering (SEC-MALS) 33 , which revealed a molecular mass of 31.9 ± 0.4 kDa for the SaNsrF WT protein (Fig. 1B). This corresponds nicely with the calculated theoretical molecular mass of the recombinant monomer of 30.9 kDa including the His10-tag. Thus, the conducted SEC-MALS analysis revealed that SaNsrF WT exists as a stable monomer in solution, which is in line with previous observations of other NBDs from different ABC transporter families [34][35][36] .
By sequence alignments, His 202 was identified to be the essential residue of the H-loop [37][38][39] . As shown for other NBDs, a point mutation to alanine results in a loss of the ATPase activity of the NBD. We generated this variant of SaNsrF (SaNsrF H202A ), which indeed displayed no NTP hydrolysis (see below). This variant served as a negative control in all our experiments. The lack of NTP hydrolysis for SaNsrF H202A is in line with in vivo studies that show that this variant abolishes the activity of SaNsrFP 8,40 . Activity of Sansrf WT . After successful purification, we functionally characterized SaNsrF WT . To do so, we screened the following parameters for their influence on the ATP hydrolysis velocity: (I) pH, (II) salt concentration, (III) nature of the divalent ion and (IV) temperature (see Supporting Information and Fig. S1). As a result, the optimized conditions were found to be 100 mM HEPES at pH 7 with 0 mM NaCl as an assay buffer. The  Fig. 2A, the SaNsrF WT protein demonstrated a nonlinear dependency of ATPase activity over a range of 0-5 mM ATP. The maximal reaction velocity was calculated to be 190.9 ± 10.0 nmol min −1 mg −1 when using ATP. Moreover, the calculation of the kinetic parameters resulted in a kinetic constant of k half = 0.41 ± 0.05 mM and a Hill coefficient of h = 1.72 ± 0.27 ( Fig. 2A and Table 1). A Hill coefficient > 1 demonstrates a cooperative behaviour, and suggests that SaNsrF WT needs to dimerize to hydrolyze ATP, which is in line with other previously characterized NBDs [41][42][43] . For GTP, the maximal reaction velocity was 221.6 ± 11.1 nmol min −1 mg −1 with a Hill coefficient of h = 1.82 ± 0.27 and a k half value of 0.69 ± 0.07 mM ( Fig. 2B and Table 1). Interestingly, the highest reaction velocity with a value of 339.0 ± 30.4 nmol min −1 mg −1 was reached using CTP as a substrate with the highest measured k half value of 1.23 ± 0.20 mM and a Hill coefficient of 1.63 ± 0.53 ( Fig. 2C and Table 1). The kinetic parameters using UTP as a substrate resulted in comparably high values of v max = 314.8 ± 23.4 nmol min −1 mg −1 , k half = 0.90 ± 0.13 mM and h = 1.55 ± 0.25 ( Fig. 2D and Table 1). The variant SaNsrF H202A displayed no hydrolytic activity for any of the four used NTPs (Fig. 2, dashed lines).
Structural models of SaNsrF monomer and dimer. Since no experimental structure of SaNsrF is available, we generated a structural model of the NBD by comparative modeling. NBDs are the most conserved parts of ABC transporters and in the case of SaNsrF, the templates used for modeling show a sequence identity of ~ 30-40% and a sequence similarity of 84-89% (Table S1). Of these X-ray structures (resolution between 1.7 and 3.4 Å), two constitute NBDs in the functionally active assembly; they were crystallized with the TMD of the macrolide exporter MacAB from Acinetobacter baumannii (PDB ID 5GKO 44 ) and MacAB-like from Streptococcus pneumoniae (PDB ID 5XU1 45 ).
The homology model of SaNsrF WT in the monomeric form is of high quality, given the low overall TopScore 46 (TS) value of 0.24 (Fig. 3A). This superimposition-free score evaluates local distance differences 47 of all atoms in a model, and a value closer to zero indicates higher quality. The regions modeled with lower reliability (TS > 0.5), accounting only for ~ 6% of the total sequence, are located at the β-hairpin (residues [15][16][17][18] and the two C-terminal helices (residues 229-232, 235-236, 246-250). Both substructures can be found in other NDBs, however, indicating the plausibility of the model. For example, when compared to the structure of ComA from Streptococcus mutans (PDB ID 3VX4 48 ), the C-terminal helices have a virtually identical fold, with an RMSD of 0.6 Å for the last 50 residues, based on sequence alignment followed by structural superimposition.
The dimeric SaNsrF WT model is structurally similar to other known structures, given RMSD values of ~ 5 Å or lower (RMSD of 3.5 Å, 4.5 Å and 5.2 Å for PDB IDs 1L2T, 5GKO, and 5XU1, respectively), indicating the suitability of the performed protein-protein docking. The reliability of the model is additionally verified by the presence of conserved motifs ( Fig. 2B and Table S2), such as the phosphate-binding loop (P-loop or Walker A motif), the cofactor-chelating region (Walker B motif), and a short consensus sequence "LSGGQ" (C-loop or ABC signature motif), which signify ABC transporter family membership at the sequence level. Moreover, the α-helical and RecA-like domains are in the canonical head-to-tail arrangement (Fig. 3C). Interestingly, the  www.nature.com/scientificreports/ www.nature.com/scientificreports/ calculated electrostatic potential shows a clear polarization ( Fig. 3D) with positively charged residues (such as R and K) prevalent on the dimer's side oriented towards the membrane (named "top") and negatively charged residues (such as D and E) on the opposite side (named "bottom") in agreement with the expected topology.
Structural dynamics at the NBD-NBD interface and impact of the SaNsrF H202A substitution. The SaNsrF models were subjected to all-atom MD simulations of in total 10 μs length to investigate the structural dynamics at the NDB-NDB interface and to highlight the impact of the H202A substitution on ATP/ Mg 2+ binding. The RMSD profiles for SaNsrF WT and SaNsrF H202A monomers ( Fig. S2) reach almost immediately a plateau at ~ 4 Å, indicating that the overall structure is mostly invariant over simulation times of 0.5 μs for each replica. Additionally, the low variability of ATP/Mg 2+ coordinates (Fig. S3A,B) suggests that the SaNsrF H202A substitution does not impact ATP/Mg 2+ binding, at least on the timescale of our simulations. The RMSD profile for the SaNsrF WT and SaNsrF H202A dimers is mostly invariant (Fig. S4A) when the structures are superimposed onto the two subunits separately (red and blue lines). However, when the superimposition is done with respect to the least mobile regions in the whole dimer (black line), RMSD values reach ~ 6-9 Å in three out of five replicas for SaNsrF WT , indicating that the arrangement of the two subunits changes during the simulations. In particular, the interface between the subunits partially opens (Fig. S4B) up to ~ 25 Å (Fig. S5). The change of ATP molecule and Mg 2+ ion positions relative to the protein is more marked for SaNsrF WT dimer (Fig. S3). Interestingly, this is not happening in the SaNsrF H202A variant, where the interface seems to be more stable.
In terms of structural mobility, the central region of SaNsrF WT and SaNsrF H202A (residues ~ 50-150) shows a different profile in monomers and dimers (Fig. 4). In monomers (Fig. 4A,B), this region is less mobile than in dimers (Fig. 4C,D), with RMSF values lower than 2 Å and up to 4 Å, respectively. Moreover, in the dimeric SaNsrF H202A variant, this region is slightly less mobile than in SaNsrF WT . The residues of the central region are oriented towards the TM region of the transporter (Fig. 4E,D). In addition, after the alignment of SaNsrF with NBDs of structures containing the TMD (PDB ID 5XU1, Fig. 4G), most of the residues of this central region are located at < 5 Å distance from the coupling helices (CH1, between TM2 and TM3, and C-terminal CH2) of the transporter, suggesting that this central region is involved in NBD-TMD communication (Fig. 4H). A similar result was found for the HlyB transporter 50 , where the X-loop motif (corresponding to residues 137-142 in SaNsrF, located in the central region) has been proposed to be an important part of the NBD-TMD communication. Even though we are considering an ATP-bound pre-hydrolysis state, SaNsrF in the dimer seems to be generally more mobile than in the monomer, in agreement with the idea that a dimeric assembly is needed in order to perform its function.
H-bond analysis in SaNsrF WT and SaNsrF H202A dimers reveals that the number of H-bond interactions between SaNsrF and the ligands (ATP molecules and magnesium ions) is on average higher in the case of the SaNsrF H202A variant (Fig. 5A). This is due to the higher structural stability compared to SaNsrF WT . Besides the three residues used as restraints for protein-protein docking (S43-R152-D176), other residues contribute to the stability of the dimer with H-bond occupancies up to 70%, such as R13, T14, R15, E42, E144, and R178 ( Fig. 5B,C). Surprisingly, the residue-wise H-bond occupancy in SaNsrF WT is significantly higher (p < 0.01) for two specific H-bonds involving both side chains and backbone atoms (D136-R15 and R133-R15), although the interface of the SaNsrF WT dimer is less structurally stable (see above). Indeed, in the initial dimeric model, these interactions are not present, but require the movement of one monomer to the other for them to form.
To conclude, the generated models show a high structural stability over the simulation lengths. In the dimers, the central region is more mobile than in the monomers; in SaNsrF WT , the interface between subunits is structurally less stable than in the SaNsrF H202A variant. Since a shift of one monomer to the other is necessary for NDBs to perform their function, these results together suggest that the mutation SaNsrF H202A impacts the structural dynamics at the SaNsrF interface and not only the catalytic mechanism.
Small angle X-ray scattering. Unfortunately, we were not able to crystalize the SaNsrF protein, although extensively tried. In order to experimentally validate this new model, we choose Small Angle X-Ray Scattering (SAXS) to compare the theoretical model with the experimental scattering (Fig. 6A) measured with the Xenocs Xeuss 2. Based on the experimental data, we calculated an ab initio model for SaNsrF WT with the program GASBOR 51 and obtained a χ 2 value of 0.97. Superimposing the ab initio and the TopModel model reveals that the structure and the envelope obtained by the SAXS experiment overlap, but also a density tail at the C-terminus of SaNsrF WT (Fig. 6B) that is not occupied by the model. Scrutinizing the templates used by TopModel 52 shows that this helical part (Fig. 6B, orange helix) is rather unstructured or even missing. This finding indicated that this region might be highly flexible in solution, thereby covering the available free space in the SAXS envelope (Fig. 6B, red helix). With the program CRYSOL 53 we compared the theoretical scattering curve obtained from the TopModel model against the experimental data. The resulting χ 2 value of 1.16 indicates a good agreement between the prediction and the experiment. We uploaded the SAXS data and the corresponding model of SaNsrF to the Small Angle Scattering Biological Data Bank (SASBDB) 54,55 with the accession code SASDJR3.

Molecular docking of other NTPs.
In order to rationalize the hydrolysis preference for ATP over other NTPs, we predicted the binding mode of these molecules in complex with the SaNsrF WT dimer. Ten different pocket conformations, obtained from five equilibrated structures used also for MD simulations times two pockets each, were considered. When focusing on the configurations with lowest Coulomb (ecoul) and van der Waals (evdw) energies, ATP is slightly enriched compared to the other NTP (3 × ATP, 2 × UTP, 1 × CTP and 1 × GTP), suggesting that ATP binding is preferred due to enthalpic contributions to binding (Fig. 7A). Residues giving rise to this preference are those interacting with the nucleobase, namely F12, T49, A23 of one subunit and F143′ and www.nature.com/scientificreports/ www.nature.com/scientificreports/ E144′ of the other (Fig. 7B). In particular, the phenylalanines are interacting with the nucleobase by π-π stacking interactions, and the amino groups of CTP and GTP form H-bonds with the backbone oxygen of F143′ and the carboxylate group of E144′, respectively. Since in ATP the amino group has the same orientation as in CTP, a similar kind of H-bond pattern can be expected. Over respective pockets 1 or 2, which are not symmetric as described above, ATP shows the largest sums of Coulomb and van der Waals energies compared to the other NTPs (Fig. 7C), indicating strongest binding based on enthalpic components, which is in line with the biochemical data where ATP shows the lowest k half value ( Fig. 2 and Table 1).

Discussion
A rather novel family of ABC-transporters, the Bacitracin efflux (Bce) type transporters, have been identified to confer high-level resistance against bacitracin as well as against lantibiotics such as nisin and gallidermin in Bacillus subtilis, Staphylococcus aureus, and Streptococcus agalactiae 8,14,16,[57][58][59][60] . These transporters have been rudimentarily characterized in vitro. We set out to characterize the NBD of the transporter SaNsrFP; this transporter has been shown to be involved in lantibiotic resistance 8 . www.nature.com/scientificreports/ We have purified and characterized the SaNsrF WT and SaNsrF H202A proteins regarding their ability of ATP hydrolysis. The results revealed that inorganic phosphate is only released in a pH range of 6-8, where an HEPES buffer at pH 7 was found to yield maximal ATPase activity. Interestingly, 20% difference could be found in a TRIS buffer system at the same pH ( Fig S1A). Similar results were obtained by Zaitseva et al. examining the HlyB-NBD 36 . In that study, a correlation between the pH of 6 and the pK a values of the glutamate residue and/ or the γ-phosphate of the nucleotide and between the pH of 8 and the pK a value of the conserved histidine bound in a salt bridge with the γ-phosphate was made. On that basis, the nucleophilic attack on the γ-phosphate is preceded, originating from a hydrolytic water molecule, which results in the cleavage of the γ-phosphate moiety 26,36,61 . Moreover, the importance of the conserved histidine could be confirmed since the SaNsrF H202A variant was shown to be incapable of hydrolysing ATP. Here, the 'linchpin'-role during ATP-hydrolysis is conducted by the H-loop 22,36,38,62 . Also, this allows a possible explanation for the observed decrease of activity with increasing concentrations of NaCl (Fig. S1B). Since the conserved histidine is in contact with the γ-phosphate of the nucleotide by forming a salt bridge, rising salt concentration could disrupt this existing interaction. In contrast, a buffer system containing 300 mM of NaCl was used for protein storage, which indicates an inverse correlation between protein stability and activity at rising NaCl concentrations 63 . The incapability of SaNsrF H202A to hydrolyse ATP supports in vivo studies where a loss of resistance against the lantibiotic nisin was observed when expressed in L. lactis bacterial cells 8 .
Like many other NBDs, SaNsrF was observed to be strictly dependent on its cofactor Mg 2+39,64,65 , because this is required as a Lewis acid in the catalytic cycle. Mg 2+ is involved in proton abstraction from the nucleotide and the nucleophilic attack of the catalytic water, which results in the hydrolytic cleavage of its γ-phosphate 36 .
Finally, we conducted kinetic measurements including all optimized parameters and the preference of SaN-srF WT and SaNsrF H202A for hydrolysing different NTPs. We propose that the main interaction of the nucleoside triphosphate and the protein occurs by π-π-stacking between the adenine moiety and F12 downstream of the Walker A motif (Fig. 3B,C) as also observed for other NBD's 22,24,25,30 . Also, Mg 2+ , anchored to the protein through Asp and Glu residues of the Walker B motif, interacts with the phosphate region of ATP. The Walker A motif binds to the other side of the phosphate region (Fig. 3B).
Based on a comparison of docked binding poses of other NTPs, additional interacting residues were predicted (Fig. 7B). Amino group-containing NTPs (ATP, CTP and GTP) can form H-bonds with the backbone oxygen of F143′ and the carboxylate group of E144′, whereas purines in ATP and GTP form more extended π-π stacking interactions with F12 and F143′. ATP shows the largest sums of Coulomb and van der Waals energies compared to the other NTPs in either pocket of the NBD, in line with the biochemical data where ATP displayed the lowest k half value ( Fig. 2 and Table 1).
By comparing the measured kinetic parameters of each examined NTP, it becomes obvious that the reactions including UTP or CTP resulted in a significantly higher reaction velocity, respectively, when compared to ATP. Nevertheless, the CTPase and UTPase activities revealed noticeably high kinetic constants (k half ) as well. With regards to the substrate affinity represented by the k half value, a minimum of 0.41 ± 0.05 mM was reached using ATP as a substrate, which signifies ATP as the most favoured of all four tested NTPs for SaNsrF WT 32,[66][67][68] . Considering the physiology of purine (ATP, GTP) and pyrimidine (UTP, CTP) nucleotides, we concluded that the involved aromatic ring systems play a major role concerning the substrate affinity and stability of the protein-substrate-complex. Here, pyrimidine bases exhibit a smaller electron density that can be involved in π-π-stacking. Thus, dissociation of pyrimidine nucleotides from the enzyme occurs faster than purine nucleotides. By contrast, the stabilized protein-purine-complex is less liable www.nature.com/scientificreports/ to dissociation. Together, this may explain the small k half values found for ATP and GTP and the high reaction velocities caused by a high turnover of CTP and UTP. NBDs are assumed to share a large number of properties due to highly conserved sequences and specific motifs (see Fig. 3B,C and Table S2) [22][23][24][25][26]30 . The presence of a certain substrate such as ATP is supposed to induce a dimerization of the two NBD monomers in a typical head-to-tail formation, resulting in two ATP molecules in the dimer interface, sandwiched by the Walker A motif of one monomer and the signature motif of the other one as a cooperative process 22,24,25 .
NBDs hydrolyse ATP, which drives substrate translocation by conformational changes of the TMD. In the case of the BceAB-type ABC transporter SaNsrFP, the energy supply is provided by the BceA-domain SaNsrF 16 . By employing SEC-MALS-coupled analysis we were able to confirm a monomeric state of SaNsrF WT and its variant SaNsrF H202A in solution since the measured molecular masses corresponded with the calculated values for each monomer. This agrees with the oligomeric state of other NBDs from other ABC transporter families in the absence of nucleotide [34][35][36] .
Furthermore, this is in line with our SAXS data that allowed the construction of a volumetric envelope of the SaNsrF WT monomer. The experimental structure of SaNsrF has not been published yet. Here, we generated a structural model using TopModel 52 based on five main templates 1F3O_A, 5XU1_B, 2PCL_A, 5GKO_A, 2OLJ_A (Fig. 3A, 6B). We compared this model with the volumetric envelope obtained from SAXS data, showing high www.nature.com/scientificreports/ reliability and agreement with experimental data. It is striking that the density of the protein model is partly not occupied. A flexible C-terminus could be the reason, which would make a temporary fit of the versatile C-terminal helix to the proposed model possible. As for well-studied NBDs such as HisP, the modeled SaNsrF dimer exhibits the typical head-to-tail formation including two sandwiched ATP molecules in the dimer interface between the Walker A motif of the first monomer and the C-loop of the second one 22,24,25,30 . Therefore, the SaNsrF protein shares many structural similarities with other known NBDs. As the γ-phosphate moiety of ATP was predicted to be in close proximity of the conserved histidine (H-loop) and the cofactor Mg 2+ , one can deduce a consensus with the hypothesis of the H-loop acting as a sensor, whereas the cofactor is involved in hydrolytic cleavage while being coordinated by the Walker B motif (Fig. 3B,C) 22,23,26,28 . Furthermore, in SaNsrF WT , the interface between subunits is structurally less stable than in SaNsrF H202A . Since a shift of one monomer to the other is necessary for NDBs to perform their function, these results suggest that the substitution SaNsrF H202A impacts the structural dynamics at the SaNsrF interface and not only the catalytic mechanism. Clearly, the SaNsrF protein represent an isolated NBD and we do not know if the kinetic correspond to the ATP hydrolysis that will occur in the presence of the transmembrane protein SaNsrP. However, when comparing the data with known NBDs which has been described before in the presence and absence of the transmembrane segment it can be observed that v max might be changed, the k m values however remains very similar. For example the ATP hydrolysis kinetics have been described for the HlyB NBD as well as for the purified full length transporter in detergent solution 26,36,38,43,69 . Here the NBD showed a v max of 200 nmol min −1 mg −1 with a k m value of 0.31 where as the full length transporter displayed a lower v max of 8.1 nmol min −1 mg −1 with a k m value of 0.36. This reduction is likely due to the detergent, which is present to keep the HlyB transporter in solution. Important, however is that in both cases the kinetic displayed cooperativity (Hill coefficient > 1) as in the case of SaNsrF and the corresponding histidine mutation also resulted in an inactive protein. This shows that our NTP analysis of the SaNsrF will likely be similar even when the TMD SaNsrP is present. The same observations were found for the nisin transporter NisT from L. lactis 70 and the nukacin ISK-1 transporter NukT from Staphylococcus arneri ISK-1 71 albeit in detergent solution.
In summary, the experiments revealed the first detailed insights into biochemical properties of the BceA domain of the BceAB-type ABC transporter SaNsrFP. We showed that SaNsrF WT and its variant SaNsrF H202A exist as monomers in solution and determined several physiological and structural properties of the protein by evaluating its ATPase activity in comprehensive in vitro studies and molecular modelling and simulations. Hence, this study contributes to the mechanistic and structural understanding of the BceAB-type ABC transporter family, which opens up the possibility to pharmacologically target this family in order to combat multidrug-resistant species in the long run. It further confirms in vivo data where the H202A variant of SaNsrF displayed a loss in the activity, which now can be pinpointed to a lack of ATP hydrolysis, and shows that this variant can well serve as a negative control in studies concerning BceAB type transporters since the histidine is conserved throughout the sequence of this family.

Materials and methods
Expression of SaNsrF WT and Sansrf H202A . E. coli BL21 (DE3) strains were transformed via heat shock method 72 with pET-16b-NHis 10 -SaNsrF WT or pET-16b-NHis 10 -SaNsrF H202A , respectively. Precultures were selectively grown with 20 µg mL −1 ampicillin at 37 °C and 180 rpm overnight. Lysogeny Broth (LB) medium was pre-incubated with 20 µg mL −1 ampicillin and inoculated with the respective preculture to an OD 600 of 0.1. The cultures were grown to an OD 600 of 0.4 at 37 °C and 180 rpm whereupon the temperature was reduced to 18 °C. Protein expression was induced by the addition of 1 mM IPTG at an OD 600 of 0.8 and the cultures were further grown overnight.
Protein purification. SaNsrF WT and SaNsrF H202A were purified using Immobilized Metal Ion Chromatography (IMAC). Therefore, a 5 mL HiTrap Chelating HP column, loaded with Zn 2+ , was equilibrated with low IMAC-buffer (100 mM HEPES at pH 8, 300 mM NaCl, 20% glycerol). Protein elution was undertaken with the high IMAC-buffer (low IMAC-buffer plus 125 mM histidine). A washing step of 40-percent high IMAC-buffer was introduced before. The concentrated eluted proteins were then injected onto a Superdex 75 16/60 size exclusion column at a flow rate of 0.5 mL min −1 , pre-equilibrated with SEC buffer (100 mM HEPES at pH 8, 300 mM NaCl, 20% glycerol). Protein eluates were collected and stored at 4 °C.
ATPase activity assay. The ATPase activity of SaNsrF WT and SaNsrF H202A (diluted in 100 mM HEPES at pH 8, 100 mM NaCl) was examined by the Malachite Green Phosphate Assay at a protein concentration of 0.1 mg mL −1 that was initially undertaken at room temperature (20 °C). Several parameters were screened to determine the optimal buffer and temperature conditions for the protein activity (see Supplementary Information).
Kinetic measurements for SaNsrF WT and SaNsrF H202A were performed under the influence of NTP (ATP, GTP, CTP, UTP) with concentrations ranging from 0 to 5 mM.
Therefore, the kinetics were fitted using the Hill equation: www.nature.com/scientificreports/ All shown data are representing the average of a triple evaluation at least, with the standard deviation reported as errors.
Small angle X-ray scattering (SAXS). We collected all SAXS data on our Xeuss 2.0 Q-Xoom system from Xenocs, equipped with a PILATUS 3 R 300 K detector (Dectris) and a GENIX 3D CU Ultra Low Divergence x-ray beam delivery system (Xenocs). The chosen sample to detector distance for the experiment was 0.55 m, results in an achievable q-range of 0.18-6 nm −1 . All measurements were performed at 15 °C with protein concentrations between 0.5 and 4.2 mg mL −1 . Samples were injected in the Low Noise Flow Cell (Xenocs) via autosampler. For each sample, twelve frames with an exposer time of ten minutes were collected. By comparing these frames, we excluded the possibility of aggregation and radiation damage during the measurement. Data were scaled to absolute intensity against water. All used programs for data processing were part of the ATSAS Software package (Version 3.0.1), available on the EMBL website 73 . Primary data reduction was performed with the program PRIMUS 74 . With the Guinier approximation we determined the forward scattering I(0) and the radius of gyration (Rg) 75 . The program GNOM was used to estimate the maximum particle dimension (D max ) with the pair distribution function p(r) 76 . Low resolution ab initio models were calculated using GASBOR 51 . The superposition of a predicted SaNsrF model (see below) was done using the program SUPCOMB 56 .
Structural models of SaNsrF complexes. As an experimental SaNsrF structure is not available, a homology model was constructed using the template-based protein structure prediction program TopModel 52 and the SaNsrF WT sequence as input (NCBI Reference Sequence: WP_000923535.1). In order to build a SaNsrF model arranged in a dimeric assembly with substrate (ATP) and cofactor (Mg 2+ ) bound, starting from the SaN-srF WT monomer in the apo state, a search for sequence similarity and structural properties was performed on the Protein Data Bank. The results were filtered according to the following criteria: sequence identity ≥ 33% and E-value cutoff 0.001 as determined by BLAST 77 ; oligomeric state equals 2; sequence length of 250 ± 50 residues; resolution ≤ 2 Å. Out of six results, only one (PDB ID: 1L2T 28 ) is crystallized as a functionally active "ATP sandwich" symmetrical dimer and was therefore used as a reference. Since ATP is bound at the interface of the dimer and its binding is influenced by both protein subunits, both protein-ligand and protein-protein docking would be particularly challenging in this case. Hence, we constructed first the SaNsrF WT dimer in the apo form and the ATP/Mg 2+ -bound form subsequently.
To do so, protein-protein docking was performed with the program HADDOCK 78,79 , using distances between respective three residues that bridge the two subunits together with H-bond interactions as restraints (S40/S43, R153/R152 and D177/D176, for PDB ID 1L2T/SaNsrF WT sequences, respectively). The most similar docking solution to the reference PDB ID 1L2T was used for further modeling steps.
Both, SaNsrF WT monomer and dimer structures were preprocessed with the Protein Preparation Wizard 80 of Schrödinger's Maestro Suite. Since the residues at the binding sites are highly conserved, ATP and Mg 2+ are considered to bind in a very similar way as in PDB ID 1L2T. Thus, their coordinates were copied from the reference into the SaNsrF WT model after alignment to one protein subunit. Residues located ≤ 5 Å away from the ATP molecules were energy-minimized using the OPLS 2005 force field 81 with standard cutoff values for van der Waals, electrostatic, and H-bond interactions, until the average RMSD of non-hydrogen atoms reaches 0.30 Å. Bond orders as well as missing hydrogen atoms were assigned, and the H-bond network was optimized. Finally, residue 202 was substituted to construct the SaNsrF H202A variant of the monomer and dimer.

Molecular dynamics simulations.
In order to validate the modeled protein-protein interface and the ATP binding mode, and to investigate the impact of the SaNsrF H202A substitution on structural dynamics, a set of MD simulations was performed using Amber 2019 82 . Four different ATP/Mg 2+ -bound SaNsrF systems were prepared for this with the LEaP program 83 : monomer and dimer, both for SaNsrF WT and SaNsrF H202A .
After establishing charge neutrality by adding sodium counter ions, each system was placed in a truncated octahedral box of TIP3P 84 water with a distance of the nearest atom to the border of the box of ≥ 11 Å. Structural relaxation, thermalization, and production runs of MD simulations were conducted with pmemd.cuda 85 using the ff14SB force field 86 for the protein, Joung-Cheatham parameters 87 for ions, and available ATP parameters 88 . For each starting complex, five independent replicas of 500 ns length each were performed, resulting in a cumulative simulation time of 10 µs. In order to set up independent replicas and obtain slightly different starting structures, the target temperature was set to different values during thermalization (299.8 K, 299.9 K, 300.0 K, 300.1 K, 300.2 K and 300.3 K). A detailed description of the thermalization protocol can be found elsewhere 89 . The analysis of the MD trajectories was carried out with cpptraj 90 on snapshots extracted every 1 ns. All the MD-generated conformations were clustered applying a hierarchical agglomerative approach and an RMSD cutoff value of 4 Å. The representative structure of the SaNsrF WT monomer was compared to the experimentally determined SAXS density.
The representative structure of the most populated cluster for the SaNsrF WT dimer was used to calculate the electrostatic potential with the Adaptive Poisson-Boltzmann Solver (APBS) software package 49 as implemented in PyMOL 91 . Dielectric constants (ε) of 2.0 and 78.0 were used, respectively, for the protein and for water, and the concentration of monovalent cations and anions was set to 0.15 M.
To measure structural mobility, we computed the residue-wise root-mean-square fluctuation (RMSF) of backbone atoms. Structural changes over time, both for the apo SaNsrF proteins and the ATP/Mg 2+ -bound form, were detected calculating the root-mean-square deviation of atomic positions (RMSD) compared to the initial structure. To describe the changes occurring at the level of the interface, we performed two analyses: (I) measurement of the distance between the center of mass of two residues located in opposite subunits at the center of the interface (S43 and S146); (II) H-bond analysis (i) in terms of the total number of interactions between Scientific RepoRtS | (2020) 10:15208 | https://doi.org/10.1038/s41598-020-72237-7 www.nature.com/scientificreports/ two subunits (SaNsrF A -SaNsrF B ) and between protein and ligands (SaNsrF-(ATP-Mg 2+ )) and ii) residue-wise H-bound occupancy between residues of the two subunits (SaNsrF A -SaNsrF B ), allowing to identify which residues perform more frequent H-bonds throughout the simulations. For this analysis, only H-bonds with the following criteria were considered: occupancy between specific donor and acceptor > 1%; H-bond present in at least two replicas of the same system; H-bonds between two residues with residue-wise occupancy > 10% in at least one system.

Molecular docking of other NTPs. To predict the binding mode of other NTPs in complex with the
SaNsrF WT dimer, molecular docking was performed. The starting points for these calculations were the five structures resulting from thermalization and equilibration steps, then used also for independent MD simulations replicas (production). First, for each binding site a cubic grid of 20 Å length centered on the respective ATP molecule was built in the Maestro platform 92 , for a total of 10 different grids. Then, starting from the ATP structures, other NTPs were built (GTP, CTP and UTP) by modifying the nucleobase. The generated conformations were refined and scored with the Glide-Extra precision (XP) mode of Glide 93 . Only the best solution for each NTP in each grid was considered. The Coulomb interaction energy (ecoul) and the van der Waals energy (evdw), components of the XP GlideScore scoring function, were computed, and used to describe the enthalpic contribution of binding.