Multitask ATPases (NBDs) of bacterial ABC importers type I and their interspecies exchangeability

ATP-binding cassette (ABC) type I importers are widespread in bacteria and play a crucial role in its survival and pathogenesis. They share the same modular architecture comprising two intracellular nucleotide-binding domains (NBDs), two transmembrane domains (TMDs) and a substrate-binding protein. The NBDs bind and hydrolyze ATP, thereby generating conformational changes that are coupled to the TMDs and lead to substrate translocation. A group of multitask NBDs that are able to serve as the cellular motor for multiple sugar importers was recently discovered. To understand why some ABC importers share energy-coupling components, we used the MsmX ATPase from Bacillus subtilis as a model for biological and structural studies. Here we report the first examples of functional hybrid interspecies ABC type I importers in which the NBDs could be exchanged. Furthermore, the first crystal structure of an assigned multitask NBD provides a framework to understand the molecular basis of the broader specificity of interaction with the TMDs.


Results
Interspecies functional exchangeability of multitask ATPases (NBDs) from type I ABC transport systems. To characterize the energy-coupling component of bacterial ABC transporters, and to evaluate their intra-and interspecies exchangeability, we constructed a genetic system in B. subtilis for regulated expression of the msmX allele or its homologs in trans (Fig. 1). The functionality of distinct NBDs alleles was determined by their capacity to substitute MsmX, as energizer of different carbohydrates importers, in a B. subtilis msmX-null mutant. Furthermore, the NBDs were expressed and produced with a His 6 -tag placed at the C-terminus, which allowed the determination of the protein accumulation in B. subtilis cells.
Based on the sequence identity (shown in parenthesis), we selected known and putative NBDs from Grampositive and Gram-negative bacteria: B. subtilis FrlP(YurJ (58%)); B. thuringiensis ABC_Bt (74%); Streptococcus pneumoniae MsmK (64%); Staphylococcus aureus ABC_Sa (66%); Clostridioides difficile ABC_Cd (57%); and Escherichia coli YcjV (64%) and MalK (45%). The analysis of the respective genomes revealed that the putative The sequence of the msmX allele, at the 5′-and 3′-end, was modified to accommodate several features (bottom right). In the C-terminal region of MsmX a LEH 6 -tag was added to enable detection of protein production by western blot. Moreover, two unique restriction sites were introduced in the coding region, NheI (5′-end) and BglII (3′-end), to facilitate sub-cloning of the different NBDs. The functionality of distinct multitask ATPases alleles as well as mutagenized msmX is determined by their ability to complement the role of MsmX as energy generator of distinct sugar importers.
www.nature.com/scientificreports/ NBDs are either orphan or clustered together with other components of an ABC importer ( Supplementary  Fig. 1). Among the targeted NBDs, only MsmK from Strep. pneumoniae 17,18 and MalK 23 from E. coli have been characterized so far. MalK was selected because MalEFGK 2 is the prototype of ABC type I sugar importers and its 3D structure has been determined before 10 . The functionality of these distinct ATPases alleles was determined by their ability to substitute MsmX as the NBD of two different sugar importers-AraNPQ 14,15 and GanSPQ 15,24 in a B. subtilis msmX-null mutant (Fig. 2a). By determining the growth kinetic parameters in the presence of the substrates of each importer as sole carbon and energy source, arabinose-oligomers and galactose-oligomers, respectively, we were able to assess the degree of efficiency of each NBD to substitute MsmX. The results are summarized in Fig. 2b (detailed data in Supplementary Table 1), where a higher doubling time reflects lower functional ability of the tested NBD to energize the respective importer AraNPQ (arabinotriose) or GanSPQ (galactan). This indicates that all NBDs tested were able to complement msmX-null mutation, except YcjV and MalK from E. coli. The ABC_Bt from B. thuringiensis is the more efficient NBD in its ability to substitute MsmX and energize both AraNPQ and GanSPQ, with a doubling time of 113.8 ± 6.2 min (arabinotriose) and 148.6 ± 10.9 min (galactan) when compared with the strain bearing the msmX allele, which has a doubling time of 115.4 ± 9.3 min and 172.4 ± 9.3 min, respectively ( Fig. 2b and Supplementary Table 1). On the other hand, the ABC_Sa from Staph. aureus is the less efficient 373.7 ± 45.8 min (arabinotriose) and 424.1 ± 68.6 min (galactan). YcjV and MalK from E. coli do not support growth in arabinotriose or galactan on a msmX-null mutant genetic background.
In addition, the accumulation of the NBDs expressed in B. subtilis was determined by western blot analysis, using a His 6 -tag antibody (Fig. 2c). Although YcjV and MalK are not able to function as NBDs of the importers AraNPQ and GanSPQ, they both accumulate in the B. subtilis cells, which suggests that protein misfolding/ aggregation or low expression levels can be excluded (Fig. 2c). Furthermore, biochemical characterization shows that YcjV and MalK retain their ATPase activity ( Supplementary Fig. 2). The results indicate functional exchangeability of multitask ATPases among Gram-positive bacteria within the Firmicutes phylum.
A putative signature sequence motif for bacterial multitask ATPases. A multiple sequence alignment of the NBDs examined in this study highlighted amino acids conserved in all sequences from different species (B. subtilis, B. thuringiensis, Strep. pneumoniae, Staph. aureus and C. difficile) capable to achieve the role of MsmX but not present in proteins unable to accomplish its function, as MalK and YcjV from E. coli (among the conserved residues, the only exception is E110 that in MsmK from Strep. pneumoniae is replaced by an aspartate) (Fig. 3a). The majority of these residues is located in the predicted region for the NBD/TMD interaction, between positions 60 and 179 of MsmX. To test their relevance upon complex formation and substrate import, we targeted D77, R104, E110, K154 in MsmX. These amino acids were exchanged by alanines, using site-directed mutagenesis, and their effect in the fitness of each MsmX mutant was analized in vivo, as described above. The results of the growth kinetic parameters revealed a different degree of aptitude of each NBD variant to act as energy motor of the AraNPQ transporter (Fig. 3b and Supplementary Table 2). Regarding single mutations, the variant D77A displayed the highest negative impact in transport/growth (180.9 ± 18.7 min doubling time) when compared with the wild-type (115.4 ± 9.3 min doubling time). In addition, the double-mutations D77A/K154A and R104A/E110A were also tested, revealing a cumulative effect when compared to the single mutants (Fig. 3b), with a more pronounced negative impact in cell growth in the first pair D77A/K154A (251.2 ± 17.8 min doubling time). The latter was overexpressed, purified, and shown to retain ATPase activity ( Supplementary Fig. 2). Since the intracellular accumulation of all MsmX variants is similar, as assessed by western blot (Fig. 3c), data from single and double mutations suggest that all targeted amino acids are important for the NBD-TMD contact and/ or proper conformation of the protein complex required to drive the transport of substrate.

Interspecies functional exchangeability of NBDs beyond the Firmicutes phylum.
To check the presence of putative multitask ATPases in other prokaryotic phyla, a bioinformatic search of genome databases was conducted using the same NBD-TMD interaction region of MsmX (residues 60-179). Among the potential hits beyond the Firmicutes phylum, a putative NBD from an ABC transporter from Synechocystis sp., was selected and integrated into the chromosome of the msmX-null mutant to test its functionality in B. subtilis, as described above. This putative NBD displays sequence identity of 51% with B. subtilis MsmX in the total extension of both proteins, while the identity in the targeted region encompassing potential NBD-TMD interactions is 71% (ABC_Syn; Fig. 3a, bottom). Growth experiments and western blot analysis revealed that ABC_Syn accumulates in B. subtilis cells and is able to substitute MsmX as the NBD of the two B. subtilis sugar importers AraNPQ and GanSPQ (Fig. 4). The doubling time of the strain harboring the hybrid importers is 328.5 ± 7.4 min (arabinotriose) and 366.3 ± 69.3 (galactan) (see Supplementary Table 1), twice the time of the strain bearing the msmX allele (Fig. 2b). This result extends the interspecies exchangeability of NBDs to the Cyanobacteria phylum and to Gram-negative bacteria.
The first crystal structure of a multitask ATPase. The MsmX protein from B. subtilis was recombinantly overexpressed for structural characterization. During purification, MsmX wild-type showed high propensity to aggregate and high content of nucleic acids. This high level of contamination and aggregation was decreased when using the MsmX K43A variant. K43 is involved in ATP coordination and catalysis, interacting with the β-and γ-phosphate groups. Its replacement by an alanine resulted in 80% decreased ATPase activity ( Supplementary Fig. 3), impairing the normal function of MsmX to energize the AraNPQ transporter in B. subtilis (Supplementary Table 2). This mutant crystallized in I222 space group, with 1 molecule in the asymmetric unit. The crystal structure was solved at 1.67 Å resolution and the final electron density map allowed to model a detailed refined structure, which comprises all residues of MsmX K43A, except the N-terminus methionine.  Supplementary Table 1). At least three independent experiments were performed in each condition and error bars indicate the standard deviation of the mean. Statistical significance between doubling time of each strain bearing a MsmX homolog and the strain ISN10 (msmX_Bs) is indicated (*p < 0.05; ns: not significant) and was obtained using the R software version 3.6.2 (https ://www.r-proje ct.org/). www.nature.com/scientificreports/ Interestingly, most crystal structures of NBDs from ABC transporters show a dimer in the asymmetric unit, promoted either by the C-terminal domains or the presence of ATP (or analogs). This is not the case for B. subtilis MsmX, that crystallized as a monomer with one molecule in the asymmetric unit, and very few contacts www.nature.com/scientificreports/ with symmetry related molecules, as indicated by analyzing the protein interface using PISA server 25 . In fact, MsmX is eluted as a monomer in the last step of purification by size exclusion chromatography, suggesting that the protein is also in the monomeric form during purification. MsmX K43A shares the fold of well-studied NBDs of ABC type I importers ( Fig. 5a,b) 8 . As shown in Fig. 5b, MsmX N-terminal domain presents the general α/β-type ATPase domain fold, which comprises the RecA-like domain (residues 2-87 and 155-235, in blue) and the α-helical domain (residues 88-154, in yellow). The regulatory C-terminal domain (residues 236-365, in red) forms a mixed barrel with 3 α-helices and 11 β-strands (Figs. 5b, 6). MsmX K43A has a high structural similarity with the MalK protein from E. coli and Pyrococcus horikoshii, as derived by PDBeFold 26 (Supplementary Table 3). These proteins have been crystallized in the open, semi-open and closed conformations, corresponding to the ligand-free, ADP + Mg 2+ or ATP bound forms, respectively 23,[27][28][29][30][31] . The presence of ATP is mandatory to achieve the functional closed conformation of the MalK dimer; the MalFGK 2 complex achieves the pre-translocation state in the presence of the Maltose Binding Protein and maltose, while the outward open form when, additionally, it is complexed with ATP. The coupled opening and closure of the NBDs interface has been described as dependent on rigid-body rotations of the TMDs 27 . Since MsmX was crystallized in the absence of ligands, the structure here described very likely corresponds to the open conformation of this multitask NBD protein that should form a dimeric structure upon ATP binding.
The overall RMSD between B. subtilis MsmX and E. coli MalK in its ligand-free form (PDB ID 1Q1E) is 1.84 Å for 337 superimposed C α atoms (Supplementary Table 3). However, the superposition of both structures clearly shows higher structural conservation at the N-terminal domains of both proteins in comparison to the C-terminal domains (Fig. 7): the RMSD value is 1.41 Å for the superposition of 221 C α atoms of the N-terminal domain, and 2.02 Å for 115 C α atoms of the C-terminal domain. This result is not surprising, since the C-terminal domain of NBDs is involved in regulatory roles, while the N-terminal one is mostly involved in the catalytic activity and interaction with TMDs, and hence, more conserved 8 (see below).
The functional motifs conserved within NBD proteins are present exclusively in the N-terminal domain and were assigned based on E. coli MalK protein structures in its ligand-free (PDB ID 1Q1E) and ATP-bound forms (PDB ID 1Q12) (Figs. 5, 6, 7). The A-loop (residues [11][12][13][14][15][16][17][18][19], located between the antiparallel β1 and β2 strands www.nature.com/scientificreports/ of NBDs, is the least conserved motif between the two proteins, with different conformations depending on the presence or absence of ATP (Fig. 8a). In the case of MalK, the β-strands are much longer in the ligand-free form than in the ATP-bound form, suggesting that substrate binding gives rise to a longer and more flexible loop. In the ligand-free MsmX K43A structure here reported, the A-loop length is similar to the one in ATP-bound MalK, already in a conformation that favours nucleotide binding. Also contributing to nucleotide accommodation is the hydrophobic residue Y13, present in the A-loop and responsible for ATP positioning: in MalK, it is W13 located next to β1 strand; in MsmX the corresponding Y13 is located in the middle of the A-loop, possibly conferring higher degree of flexibility to bind ATP (Fig. 8a). The Walker A motif (residues 37-44) is fully conserved in MsmX K43A and MalK (Figs. 6 and 8b), and, as expected, the MsmX A43 residue superposes with MalK K42.
In the MsmX Q-loop (residues 81-87), Q83 structurally aligns with the conserved glutamine (Q82) involved in the hydrolysis cycle of ATP in the ligand-free and ATP-bound MalK structures (Fig. 8b). Walker B comprises residues 155-160, which form strand β7. The sequence is conserved in MsmX and MalK (Fig. 6), except for L157 of MalK that is replaced by M158 in MsmX. This small difference must not affect significantly the interaction with strand β8 since the hydrophobic nature of the residue is maintained. Walker B MalK D158 and E159, responsible for Mg 2+ coordination and water polarization during ATP hydrolysis, respectively, are in the same position as the corresponding residues in MsmX (D159 and E160) 8 . The H-loop (residues 192-194) is well conserved between MsmX and MalK, both at the sequence (Fig. 6) and structural levels (Fig. 8c), suggesting that MsmX H193 might be involved in the interaction with the γ-phosphate of the ATP substrate. In our MsmX K43A structure we observe the H193 imidazole ring is occupying the ATP site, at 4 Å from a sulphate anion, probably from the crystallization solution. The sulphate anion atoms are at the same position as the α-phosphate of the ATP molecule in the MalK ATP-bound form.
The D-loop (residues 163-166) is located downstream of Walker B (Fig. 6). The structural alignment of the two proteins, however, shows that MsmX D-loop and the neighbouring residues (160-170) superpose poorly with MalK either in the ATP-bound or in the ligand-free forms (Fig. 8d). MsmX D166 is at 7.3 Å from the corresponding residue of MalK ligand free, and at 8.9 Å from ATP-bound form, occupying the ATP binding site. N163, L164 and D165 side chains of ligand-free MalK are facing the dimerization interface, while in MsmX, these are pointing inwards (Fig. 8d). Next to the D-loop, K168 of MsmX affects the electrostatic potential of this region, that is more hydrophobic in MalK (with A167 in the corresponding position). Together, these differences suggest that this region is likely to go through conformational changes in order to allow NBD dimerization and substrate binding. The ABC signature (ABC motif in Fig. 6, residues 135-139), characteristic of the ABC superfamily of P-loop NTPases 8 , is strictly conserved between MsmX and MalK protein sequences (Fig. 6). In MsmX crystal structure, this motif is located between α-helices 5 and 6, while in MalK, this region forms a single α-helix and has a 4.7 Å shift upon ATP binding (Fig. 8d).
The regulatory C-terminal domain is the most diverse region within NBDs of ABC transporters, since it is strictly related with regulatory mechanisms, such as carbohydrate preference (Fig. 7) 32 . At secondary structure level, MsmX presents a unique helix α11, between β13 and β14, while E. coli MalK possesses an extended loop    Fig. 1). On the other hand, the less efficient energy-coupling component in the assembled hybrid transporters, the NBD from Staph. aureus appears to belong to an operon encoding a complete ABC carbohydrate uptake system ( Supplementary Fig. 1). Intra-species exchangeability was previously observed between FrlP (YurJ) from B. subtilis and MsmX in the hybrid transporter AraNPQ when the gene frlP (yurJ) was ectopically expressed 15 . Here we show that FrlP (YurJ), like MsmX, is also able to energize the GanSPQ transporter ( Fig. 2 Table 1). FrlP (YurJ) is encoded by a gene adjacent to an operon comprising other components of a system for uptake of fructosamines 38,39 . Functional exchangeability of two distinct and apparently specific NBDs, encoded in operons together, or adjacent, to the other components of a complete ABC type I, have been described before 40 . The UgpC and MalK of Escherichia coli, are NBDs of ABC type I transporters comprised in operons with their cognate TMDs and SBP, transporting sn-glycerol-3-phosphate and maltose, respectively. Using genetic complementation experiments UgpC and MalK were shown to be functionally exchangeable but the hybrid transporters were less efficient than the wild-type 40 . Similarly, two ABC ATPases of Streptococcus mutans, MsmK and MalK, encoded by genes located in the vicinity of the remaining components of their partner ABC transporter, can substitute each other and energize both transporters for raffinose/stachyose and for maltodextrins, respectively 41 . Our results reinforce the observation that functional exchangeability of NBDs has been exclusively reported in bacterial ABC sugar importers of the subfamily CUT1 that transport di-, tri-and higher oligosaccharides. Sharing the ATPase may allow for additional levels of regulation over the ABC sugar importers. B. subtilis MsmX has a similar C-terminal domain to MalK from E. coli, which has been shown to contribute to carbohydrate preference 32 , and thus a similar role is plausible.

and Supplementary
In order to identify potential signature motifs related with the multitasking role of this family of proteins, we analyzed the crystal structure of MalK, inspecting the residues involved in the interface with the MalF and MalG proteins. Since this interface is conserved among the different catalytic conformations of the MalFGK 2 complex during maltose transport, we used a MalFGK 2 structure (PDB ID 3FH6:A) with the MalK dimer in the open conformation and no ligands bound 27 , for the comparison with MsmX K43A crystal structure (RMSD of 2.12 Å for 310 C α atoms superposed). The 'EAA' motif 42 of each TMD contains the conserved coupling helix 43 that www.nature.com/scientificreports/ interacts with MalK through a surface cleft involving the Q-loop and a hydrophobic pocket. The MalK F81-Y87 region accommodates the MalF/G helix, providing an hydrophobic environment; in MsmX, the corresponding phenylalanine (F82) is flipped, causing the side chains of Q83 and N84 to point towards the complex interface, which decreases the hydrophobic character of the cleft ( Fig. 9a and b). Interestingly, N84 is conserved in all NBDs capable of substituting MsmX energizing function, while MalK possesses a serine residue at this position (S83) (Fig. 3a); this difference might affect the range of possible interactions with the coupling helix of TMDs. Furthermore, in the hydrophobic pocket involved at the interface with the MalF/G, MalK A73 is at 3 Å from the MalF helix, while in MsmX this residue is replaced by K74, which affects considerably the electrostatic potential, the volume and the hydrophobicity of the corresponding pocket (Fig. 9a). The diverse features of the MsmX pocket, in particular its larger surface area, may potentially contribute to the interaction with multiple TMDs. The sequence alignment presented in Fig. 3a shows that the charged residues D77, R104, E110 and K154, conserved among NBDs able to functionally replace MsmX, are exchanged by hydrophobic or neutral residues in E. coli MalK (G76, A103, V109 and S153) and YcjV. The structural comparison of MsmX K43A with MalFGK 2 indicates that these residues, although not located at the complex interface with the TMDs, may contribute to a charged environment in all the multitask proteins (Fig. 10). D77 and K154 in the MsmX K43A structure are not forming a salt bridge but are H-bonded through the main chain atoms (3 Å between the peptide carbonyl of D77 and NH of K154). These two residues are located just upstream of strands β6 and β7, respectively, which flank the α-helical domain (Fig. 6). In addition, D77 might have an additional role considering its proximity to the hydrophobic pocket residues. This analysis supports the observation that D77A and D77A/K154A variants had a pronounced effect in impairing cell growth and sugar transport energization (Fig. 3b). Given the role of the helical domain in the NBD-TMD interaction in ABC transporters 8 , residues D77 and K154 may be involved in conformational rearrangements involving salt bridges/H-bonds needed to select diverse TMDs for complex assembling. R104 and E110 also belong to the α-helical domain of MsmX, with R104 sitting in a loop between α3 and α4, and E110 part of α4 (Fig. 6). The two charged residues (R104, E110) in MsmX K43A are not interacting (nearest distance of the side chains is 6.9 Å) but again are contributing to the charged environment of this region, compared to MalK (Fig. 10). In addition to the putative role of the mutated residues in NBD-TMD interaction, the possibility of their involvement in the correct conformation of the assembled transporter, required to drive the translocation of the substrate across the membrane, should also be considered.
Overall, the structural superposition of MsmX and MalFGK 2 at the NBD-TMD interface is poor, suggesting some degree of flexibility to account for putative rearrangements upon complex formation. Additionally, all targeted residues may influence the electrostatic potential of this region in the multitask ATPases, with a putative role in the promiscuous interaction with TMDs from different transporters. Interestingly, ABC_Syn from Synechocystis sp. complemented MsmX in vivo, but when compared to MsmX, D77 and E110 are conserved in the cyanobacteria protein, while R104 and K154 are replaced by Q103 and S153 (ABC_Syn numbering) (Fig. 3a). These observations support that multitasking activity is not dependent on individual mutations but on the overall structure and electrostatic potential of the interaction region that might vary between phylum. Further structural investigation of these multitask ATPases in the context of the multisubunit complex transporter will clarify with greater detail the sequence and structural signatures associated with the multitask role of these ATPases.
Over the last decade, multitask ATPases have been shown to play a central role in carbohydrate uptake in Firmicutes and its inactivation reduces colonization and attenuates virulence in pathogenic species 15,17,44 . To our knowledge this study is the first to show that these NBDs of ABC type I importers are functionally exchangeable between species among the Firmicutes. The results provide evidences that multitask ATPases are present in other bacteria phyla and may be widespread in bacteria. Furthermore, the first high resolution crystal structure of an assigned multitask NBD provides a framework to understand the basis of the broader specificity of interaction with the TMDs. It remains to be seen if ABC transporters could be targets for antibacterial therapy development 45,46 but these specific multitask ATPases, which are not present in humans, may represent a new target to disrupt carbohydrate uptake.

Methods
DNA manipulation and sequencing. DNA manipulations were carried out as described previously by Sambrook et al. 47 . Restriction enzymes, T4 ligase and FastAP Thermosensitive Alkaline Phosphatase used in DNA cloning were purchased from Thermo Fisher Scientific and were used in accordance with the manufacturer's instructions. DNA ligations were performed using T4 DNA Ligase. For PCR reactions Phusion High-Fidelity DNA Polymerase, also from Thermo Fisher Scientific, was used. Clean-up of DNA from agarose gels and restriction and PCR reactions was performed with GFX Gel Band Purification kit (GE Healthcare Life Sciences). Plasmid DNA extraction and clean-up was done using the NZYMiniprep kit from NZYTech. Plasmid DNA and PCR product Sanger sequencing was performed at STAB VIDA (https ://www.stabv ida.com/). All primers used in this work are listed in Supplementary Table 4.

Construction of vectors for expression of wild-type and mutagenized MsmX in E. coli. All vec-
tors used in this work are described in Supplementary Table 5. pMJ22 15 was used as a vector for the expression of wild-type MsmX in E. coli BL21(DE3)pLysS. Mutagenic primers ARA757 and ARA758 introduced a mutation in codon AAA (Lys) to GCA (Ala), substituting the lysine at position 43 of MsmX for an alanine, creating the expression vector pAM11.  Protein extracts of B. subtilis and Western Blot analysis. B. subtilis strains were grown in CSK minimal medium supplemented with 0.1% (w/v) of arabinose, as previously described for growth kinetics parameters and doubling time determination 14 . Total protein content and its quantification were performed as described previously 15 . 20 μg of total protein from each extract and 0.25 μg of purified MsmX-His 6 were loaded in a 12.5% SDS-PAGE handcast gel and run at constant electrical current (30 mA) for 50 min. Transferred proteins were visualized in the membranes with Ponceau Red. The membranes were blocked with powdered milk solution www.nature.com/scientificreports/ in TBS-Tween (5% w/v), washed and then blotted overnight with a mix of anti-His 1:2,000 (6xHis Epitope TAG, PA1-983B, Thermo Fisher Scientific) and anti-flagellin 1:10,000 (polyclonal antibody raised in rabbit-a gift from Paulo Tavares, I2BC, Gif-sur-Yvette), followed by incubation with HRP-conjugated goat anti-rabbit IgG antibody 1:10,000 (Thermo Scientific, Pierce Antibody Products). Flagellin detection constitutes the loading control. Signal was detected by enhanced chemiluminescence using SuperSignal West Pico PLUS (Thermo Fisher Scientific); Amersham Hyperfilm plates (GE Healthcare Life Sciences) were exposed to luminescence inside a closed Hypercassette Autoradiography Cassette (GE Healthcare Life Sciences). Frozen or fresh pellets were resuspended in buffer A (10 mM PBS pH 7.4, 500 mM NaCl, 10% glycerol and 10 mM Imidazole) supplemented with 10 μg/mL DNaseI and 10 mM MgCl 2 . Cell lysis was performed by sonication and the resulting supernatant applied onto a 5 mL HisTrap HP Column (GE Healthcare Life Sciences). Proteins retained in the column were eluted using a linear gradient of imidazole. Fractions containing MsmX protein were pooled and desalted to buffer B (10 mM Tris-HCl pH 7.4 @ 20 °C, 300 mM NaCl, 10 mM MgCl 2 ) using 5 mL HiTrap Desalting columns (GE Healthcare Life Sciences). Protein fractions were pooled and concentrated with Vivaspin Turbo concentrators with 10 kDa molecular weight cut-off (Sartorius), and the resulting sample applied onto a Superdex 75 10/300 GL (GE Healthcare Life Sciences) pre-equilibrated with buffer B. In the case of MsmX wild-type, purification was optimized by including in the buffer A benzonase 5 mU/mL (instead of DNAseI) and ½ tablet of cOmplete ULTRA Tablets, Mini, EDTA-free, EASYpack Protease Inhibitor Cocktail (Roche); and using buffer B with slightly different composition (50 mM Tris-HCl pH 6.8 @ 20 °C, 300 mM NaCl, 10 mM MgCl 2 , 10% glycerol and 5 mM 2-mercaptoethanol). The final yield of MsmX wild-type and K43A mutant was 20 and 40 mg/L of culture, respectively. Crystallization, data collection and structure determination of MsmX K43A. Crystallization of wild-type MsmX, in spite of extensive attempts resulted only in poorly diffracting crystals. Therefore, further crystallization trials were pursued with MsmX K43A variant. These were performed at 20 °C with the crystallization robot Oryx 8 (Douglas Instruments Limited) using the sitting drop vapour-diffusion method, with 0.5 μL of protein at 30 mg/mL and 0.5 μL of reservoir solution containing 100 mM (NH 4 ) 2 SO 4 , 100 mM HEPES pH 7.5 and 30% (w/v) polyethylene glycol 400 (from the MemStart + MemSys HT-96 screen, MD1-33, Molecular Dimensions). Crystals with maximum dimensions of 0.1 mm × 0.3 mm × 0.06 mm were formed within four days. The diffracting properties of the MsmX K43A mutant crystals decreased during handling, harvesting and cryoprotection with different solutions. The best diffracting data was obtained when the crystals were flash cooled in liquid nitrogen directly from the crystallization drop, taking advantage of the cryoprotectant properties of polyethylene glycol 400.

Construction of vectors for the expression of wild-type and mutagenized
A complete dataset was collected from a crystal diffracting up to 1.67 Å on ID23 beamline of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). X-ray diffraction data reduction was processed using the software package XDS 54 . Anisotropic data correction and cut-off of merged intensities were performed using STARANISO web server 55 followed by scaling with Aimless 56 software from CCP4 57 program suite. The number of molecules per asymmetric unit was estimated based on Matthews coefficient 58 . Structure determination was achieved by molecular replacement using Phaser software 59 and an ensemble of structures-PDB ID 1V43 and 1Q12-as a search model; the N-terminal and C-terminal domains were searched separately to achieve a partial and a final solution, respectively. Arp/wARP 60 was used for partial model building with subsequent manual inspection using Coot 61 . Structure refinement was performed interactively using REFMAC5 62 from CCP4 suite, Phenix.refine 63 and Coot. Final R work and R free values are 0.20 and 0.25, respectively. The quality of the model was assessed by MolProbity 64 . Statistics of data collection and structure refinement are shown in Supplementary  Table 7. The model coordinates and structural factors were deposited in PDB under the accession code 6YIR. Secondary structure elements were assigned using STRIDE software 35