2 ′-Deoxyribosyltransferase from Bacillus psychrosaccharolyticus : A Mesophilic-Like Biocatalyst for the Synthesis of Modified Nucleosides from a Psychrotolerant Bacterium

Structure-function relationships of a novel 2′-deoxyribosyltransferase from the psychrotolerant bacterium Bacillus psychrosaccharolyticus (BpNDT) have been exhaustively studied by biochemical and high resolution crystallographic analyses. Despite BpNDT exhibiting some structural features characteristic of cold-adapted enzymes such as localized flexibility in critical loops, its biochemical properties are typical of mesophilic enzymes. BpNDT is a highly symmetrical homohexamer with tightly associated subunits that possesses flexible and short loops bordering the active sites. The catalytic center is essentially identical to that of other mesophilic homologues. Moreover, BpNDT shows that it is a mesophilic-like enzyme since it is not heat-labile and exhibits an apparent unfolding temperature (Tm) of 49 ◦C, being active during 96 h at 40 and 50 ◦C. Finally, BpNDT synthesizes natural and modified nucleosides, with preference for purines as acceptors and pyrimidine nucleosides as donors. Remarkably, the synthesis of several therapeutic nucleosides has been efficiently carried out. In this sense, 5-hydroxymethyl-2′-deoxyuridine (5-HMdUrd), 7-deaza-6-hydroxypurine-2′-deoxyriboside (7-DHPdRib) and theophylline-2′-deoxyriboside were synthesized for the first time by an NDT enzyme, showing the biotechnological interest of BpNDT.


Introduction
Cold-adapted microorganisms are a valuable source of enzymes of biotechnological interest, since these biocatalysts are more efficient at low temperatures than their mesophilic and thermophilic counterparts [1,2].Enzymes from psychrophiles present flexible structures, which render them labile at higher temperatures [3,4], but allow high turnover numbers (k cat ) at low-to-moderate temperatures.Psychrophiles are defined as organisms which have an optimal growth temperature at 15 • C or below, a minimal growth temperature at 0 • C or below and a maximum growth temperature at approximately 20 • C [5].Another group of cold-adapted microorganisms display optimal growth at about 20 • C but grow fairly well at temperatures close to 0 • C and are unable to grow at temperatures above 30 • C.These are called psychrotrophs or psychrotolerant organisms.The term psychrotolerant also encompasses mesophilic species which manage to survive at low temperatures [6].An example of psychrotolerant bacteria is Bacillus psychrosaccharolyticus (CECT 4074, ATCC 23296, DSM 6), a facultative anaerobic Gram-positive bacterium that can be found in soil and lowland marshes.In the context of mesophilic Bacillus species, this sugar-digesting and spore-forming rod bacterium most closely resembles Bacillus circulans, although they differ in maximal growth temperature and in cell morphology [7].This bacterium should be considered psychrotolerant because it reaches its optimal growth temperature at 20 • C. In fact, at low temperatures, B. psychrosaccharolyticus expresses cold-induced proteins, the likes of those expressed during the cold shock response in B. subtilis, an observation that points to systematic protein recomposition as the main psychrophilicity mechanism in this microorganism [8].
The genome of B. psychrosaccharolyticus has been published and a gene codifying a putative nucleoside 2 -deoxyribosyltransferase has been identified within it [9].
Nucleoside 2 -deoxyribosyltransferases (NDTs) are a group of enzymes which catalyze exchange of the 2 -deoxyribosyl moiety between 2 -deoxyribonucleosides and nucleobases.Traditionally, nucleoside synthesis has been performed by multistep chemical methods, including several protection-deprotection steps and the use of chemical reagents as well as organic solvents that are expensive and environmentally harmful [10,11].In this sense, enzyme-catalyzed synthesis of natural and non-natural nucleosides in a one-pot, one-step reaction involving NDTs is an interesting alternative to chemical methods [12].Nucleoside analogues are pharmacologically active compounds, which include cytotoxic, antiviral, and immunosuppressive molecules [13,14].As an example, 5-halogenated derivatives from 2 -deoxyuridine are widely used as anticancer agents, inhibiting thymidilate synthase, an important enzyme for DNA synthesis in cell proliferation [15,16].On the other hand, 2,6-diaminopurine nucleosides are used as drugs or prodrugs depending on their susceptibility to adenosine deaminase (ADA) activity in vivo [17].

Biochemical Characterization of Recombinant BpNDT
Optimal conditions of pH (8.0) and temperature (40 °C) were previously determined by our group [25].Here, the effects of several cations and other additives as well as ionic strength (I) on BpNDT activity have been studied.As shown in Table 1, neither monovalent cations nor the presence of Ba 2+ , Ca 2+ and Mg 2+ significantly affected activity, whereas Co 2+ , Cu 2+ , Mn 2+ and Zn 2+ showed deleterious effects.Also, the presence of 2-mercaptoethanol did not interfere with activity as expected for an enzyme lacking cysteines like BpNDT, whereas 5 mM Al 3+ dramatically reduced enzyme activity.The effect of ionic strength (I) on BpNDT activity was studied by adding different concentrations of NaCl to reaction mixtures.Activity was unaffected at concentrations up to 1.0 M NaCl, while activity decreased 25% in the presence of 1.5 M NaCl (data not shown).
Thermal inactivation of BpNDT was analyzed by pre-incubating the enzyme at different temperatures, after which aliquots were withdrawn at the indicated times and tested for activity using 2'-deoxyadenosine synthesis from 2'-deoxycitidine and adenine under standard conditions (Figure 1).Remarkably, BpNDT remained stable at 40 and 50 °C for at least 96 h.Whereas at 60 and 70 °C, deoxyribosyltransferase activity diminishes following a single exponential decay, with a threefold higher half-life at 60 °C than at 70 °C (Table 2).Here, we report on biochemical and high resolution structural characterization of BpNDT, including demonstration of its ability to catalyze the synthesis of several natural and non-natural nucleosides.The high resolution structure of BpNDT revealed some characteristics typical of cold-adapted enzymes although its biochemical behavior is mesophilic.Finally, the enzymatic production of modified nucleosides, including antiviral and antitumoral nucleosides, was successfully carried out.

Biochemical Characterization of Recombinant BpNDT
Optimal conditions of pH (8.0) and temperature (40 • C) were previously determined by our group [25].Here, the effects of several cations and other additives as well as ionic strength (I) on BpNDT activity have been studied.As shown in Table 1, neither monovalent cations nor the presence of Ba 2+ , Ca 2+ and Mg 2+ significantly affected activity, whereas Co 2+ , Cu 2+ , Mn 2+ and Zn 2+ showed deleterious effects.Also, the presence of 2-mercaptoethanol did not interfere with activity as expected for an enzyme lacking cysteines like BpNDT, whereas 5 mM Al 3+ dramatically reduced enzyme activity.The effect of ionic strength (I) on BpNDT activity was studied by adding different concentrations of NaCl to reaction mixtures.Activity was unaffected at concentrations up to 1.0 M NaCl, while activity decreased 25% in the presence of 1.5 M NaCl (data not shown).
Thermal inactivation of BpNDT was analyzed by pre-incubating the enzyme at different temperatures, after which aliquots were withdrawn at the indicated times and tested for activity using 2 -deoxyadenosine synthesis from 2 -deoxycitidine and adenine under standard conditions (Figure 1).Remarkably, BpNDT remained stable at 40 and 50 • C for at least 96 h.Whereas at 60 and 70 • C, deoxyribosyltransferase activity diminishes following a single exponential decay, with a three-fold higher half-life at 60 • C than at 70 • C (Table 2).
Furthermore, in order to characterize the thermal stability of the enzyme, heat denaturation temperature (melting temperature), T m , of BpNDT was determined by fluorescence spectroscopy and differential scanning calorimetry (DSC) experiments (Figure 2).The effect of temperature on the intrinsic fluorescence of BpNDT (excitation at 295 nm; emission from 300 to 420 nm) was evaluated at temperatures ranging from 20 to 80    Furthermore, in order to characterize the thermal stability of the enzyme, heat denaturation temperature (melting temperature), Tm, of BpNDT was determined by fluorescence spectroscopy and differential scanning calorimetry (DSC) experiments (Figure 2).The effect of temperature on the intrinsic fluorescence of BpNDT (excitation at 295 nm; emission from 300 to 420 nm) was evaluated at temperatures ranging from 20 to 80 °C.The values of fluorescence intensity at 335 nm were used to determine the Tm, which was 49 °C.DSC experiments showed a similar Tm value:

Substrate Specificity
A summary of the results obtained from the analysis of the transglycosilation reaction between natural bases and 2′-deoxynuceosides catalyzed by BpNDT is shown in Table 3.According to these results, BpNDT is a type II NDT since it catalyzes the transfer of 2′-deoxyribose between pyrimidine  Furthermore, in order to characterize the thermal stability of the enzyme, heat denaturation temperature (melting temperature), Tm, of BpNDT was determined by fluorescence spectroscopy and differential scanning calorimetry (DSC) experiments (Figure 2).The effect of temperature on the intrinsic fluorescence of BpNDT (excitation at 295 nm; emission from 300 to 420 nm) was evaluated at temperatures ranging from 20 to 80 °C.The values of fluorescence intensity at 335 nm were used to determine the Tm, which was 49 °C.DSC experiments showed a similar Tm value:

Substrate Specificity
), 50   Furthermore, in order to characterize the thermal stability of the enzyme, heat denaturation temperature (melting temperature), Tm, of BpNDT was determined by fluorescence spectroscopy and differential scanning calorimetry (DSC) experiments (Figure 2).The effect of temperature on the intrinsic fluorescence of BpNDT (excitation at 295 nm; emission from 300 to 420 nm) was evaluated at temperatures ranging from 20 to 80 °C.The values of fluorescence intensity at 335 nm were used to determine the Tm, which was 49 °C.DSC experiments showed a similar Tm value:

Substrate Specificity
) and 70 • C ( ).Furthermore, in order to characterize the thermal stability of the enzyme, heat denaturation temperature (melting temperature), Tm, of BpNDT was determined by fluorescence spectroscopy and differential scanning calorimetry (DSC) experiments (Figure 2).The effect of temperature on the intrinsic fluorescence of BpNDT (excitation at 295 nm; emission from 300 to 420 nm) was evaluated at temperatures ranging from 20 to 80 °C.The values of fluorescence intensity at 335 nm were used to determine the Tm, which was 49 °C.DSC experiments showed a similar Tm value: 49.1 °C.

Substrate Specificity
A summary of the results obtained from the analysis of the transglycosilation reaction between natural bases and 2′-deoxynuceosides catalyzed by BpNDT is shown in Table 3.According to these results, BpNDT is a type II NDT since it catalyzes the transfer of 2′-deoxyribose between pyrimidine and purine bases with a marked preference for the latter as acceptors.

Substrate Specificity
A summary of the results obtained from the analysis of the transglycosilation reaction between natural bases and 2 -deoxynuceosides catalyzed by BpNDT is shown in Table 3.According to these results, BpNDT is a type II NDT since it catalyzes the transfer of 2 -deoxyribose between pyrimidine and purine bases with a marked preference for the latter as acceptors.
2 -Deoxycytidine (dCyd) is the best donor, followed by 2 -deoxyuridine (dUrd) and thymidine (dThd).In turn, the most and least preferred acceptors are hypoxanthine (Hyp) and uracil (Ura), respectively.Compared to previously described NDTs [18,19,26,27], BpNDT exhibits higher specific activities, except when compared to LrNDT.The synthesis of dIno from Hyp and dAdo proceeds 5.4 times faster in the presence of BpNDT than with LrNDT or PDT from Lactobacillus helveticus (LhPDT), whereas the transfer between Thd and Ura is twice faster with LrNDT or Lactococcus lactis NDT (LlcNDT) than with BpNDT.Formation of dAdo from Ade and Thd by BpNDT is 1.5-fold and 4.3-fold more efficient than by LrNDT or LhPDT, respectively.Transfer between dCyd and Hyp is catalyzed by BpNDT 12 times better than by NDT from Lactobacillus fermentum (Lf NDT) or LhPDT.

Structural Analysis of BpNDT
BpNDT (10 mg/mL in 20 mM Tris-HCl, 0.1 M NaCl pH 8.0) crystallized in numerous conditions from the commercial screens used in the initial high throughput crystallization trials.High-quality diffraction crystals were prepared manually in 3 M sodium nitrate, 0.1 M sodium acetate trihydrate, pH 4.6 (protein/precipitant drop ratio 1:2).A complete dataset up to 1.9 Å resolution was collected at beamline ID29 at the European Synchrotron Radiation Facility (ESRF; Grenoble, France) and the crystal structure of BpNDT was solved to that resolution by molecular replacement using atomic coordinates of LhPDT as search model (Protein Data Bank (PDB) entry 1S2G).
BpNDT crystallized as a hexamer in the trigonal R3 space group, with two subunits per asymmetric unit (Figure 3A) that superpose almost perfectly (r.m.s.d. for 137 Cα atoms is 0.68 Å).The final model included residues 2-138 (out of 142 amino acids) from both chains (A and B) together with 104 water molecules.The 2F o -F c electron density map showed continuous density for the whole protein excluding some residues coming from the same three regions in both chains: connecting loop β2-α3 (chain A: Pro42; chain B: Gln40, Leu41), connecting loop β3-α4 (chain A: Glu75; chain B: Glu75, Asn76, Tyr77) and connecting loop β4-α5 (chain A and B: Glu104).These loops appear highly flexible, which is highlighted by the high B-factors in these regions of the structure.All residues occupy favorable regions of the Ramachandran plot.Data collection and refinement statistics are shown in Table 4.The architecture of the BpNDT subunit is composed of a central, parallel, five-stranded β-sheet (with 21,345 topology), with helices packed against each of its sides (α1-α4-α5 and α3-α6, respectively).The structure is highly asymmetric due to the orientation with respect to the β-sheet and length of the α3 helix.The overall structural features of the BpNDT subunit are highly similar within the 2'-deoxyribosyltransferases as can be manifested by structural similarity searches carried out either with DALI Lite server v3 [28] or FATCAT [29].The closest structural homologue found in both cases is LhPDT (r.m.s.d. of 1.5 for 130 Cα atoms and a sequence identity: 31%; PDB entry 1S2L), followed by the 2'-deoxyribosyltransferase from Lactobacillus leichmannii (LlNDT) [30] (r.m.s.d. of 2.0 for 136 Cα atoms and a sequence identity: 26%; PDB entry: 1F8X).These two protein homologues are homohexamers endowed with D3 molecular symmetry, similarly to BpNDT (see below), which  The architecture of the BpNDT subunit is composed of a central, parallel, five-stranded β-sheet (with 21,345 topology), with helices packed against each of its sides (α1-α4-α5 and α3-α6, respectively).The structure is highly asymmetric due to the orientation with respect to the β-sheet and length of the α3 helix.The overall structural features of the BpNDT subunit are highly similar within the 2 -deoxyribosyltransferases as can be manifested by structural similarity searches carried out either with DALI Lite server v3 [28] or FATCAT [29].The closest structural homologue found in both cases is LhPDT (r.m.s.d. of 1.5 for 130 Cα atoms and a sequence identity: 31%; PDB entry 1S2L), followed by the 2 -deoxyribosyltransferase from Lactobacillus leichmannii (LlNDT) [30] (r.m.s.d. of 2.0 for 136 Cα atoms and a sequence identity: 26%; PDB entry: 1F8X).These two protein homologues are homohexamers endowed with D3 molecular symmetry, similarly to BpNDT (see below), which resulted from the association of three dimers.In fact, the analysis of the BpNDT crystal protein packing by the PISA [31] server suggested a hexamer as the highest-order stable oligomer formed by three tightly bound dimers.The contacting interface between subunits forming these dimers is large (~2740 Å 2 ) and comparable to those from LhPDT (~3300 Å 2 ) or LlNDT (~3200 Å 2 ), although much smaller than the values observed from the distant, dimeric, eukaryotic homologues from Trypanosoma bruceii (~4950 Å 2 ) or from L. mexicana (~4500 Å 2 ).This contacting interface resulted from the close packing of α4 helices from the two participating subunits and from the fact that the α5 helix of each subunit is sandwiched between helices α3# and α4# (# refers to structural elements from the other subunit).This latter structural feature is crucial within the NDTs since it revealed that each substrate-binding pocket is in fact made up of residues contributed by the two associated subunits, supporting the idea that the dimeric assembly is the minimum catalytic unit required for 2 -deoxyribosyltransferase activity [30].
BpNDT is a homohexamer with D3 molecular symmetry that results from the tight association of three dimers (Figure 3B) with a total contacting interface between them of ~7400 Å 2 .The structural elements that participate in this contacting regions are the α3 and α4 helices and the corresponding α3-β3 and α4-β4 connecting loops.These elements configure a smooth, convex surface that faces the inner, three-fold molecular symmetry axis of the assembly.It is notable that the latter two connecting loops are much shorter (2 residues in both cases) than those from the opposite side of each subunit that protrude towards the bulk solvent and flank the entry to the substrate-binding pockets (loop β1-α1: 5 residues; loop α2-α3: 10 residues; loop β3-α4: 7 residues; loop β4-α5: 8 residues).As expected, whereas the structure of the contacting region with the shortest loops is highly conserved in the three hexameric enzymes BpNDT, LhPDT and LlNDT, the region facing the solvent shows the highest structural variability, which is most probably related to the distinct substrate specificity of each particular enzyme.
Since a structural characteristic of multimeric, cold-adapted enzymes is a comparatively reduced cohesion between monomers than in the mesophilic homologues [32], we analyzed the associative behavior of BpNDT in solution in order to check for the possibility of hexamer dissociation phenomena.With this aim, we studied the average molecular mass of BpNDT by analytical ultracentrifugation assays.Sedimentation velocity experiments showed the existence of one homogeneous species with an experimental sedimentation coefficient of 5.9 S (s 20,w = 6.24), which is compatible with a globular species with a molecular mass of 93 kDa.Conversely, the results from sedimentation equilibrium experiments agree well with the latter ones since they are well described by a unique species with a molecular mass of 91 kDa.Consequently, analytic ultracentrifugation experiments demonstrate that BpNDT in solution is well described as a unique, homogeneous species with a molecular mass that agrees with that of a homohexamer (theoretical molecular mass of BpNDT estimated from its amino acid sequence is 16,398 Da) in agreement with the crystallographic results.

Active Site of BpNDT
As an a priori cold-adapted enzyme, the crystal structure of BpNDT should offer the opportunity to deduce potential structural features of the catalytic machinery of 2 -deoxyribosyltransferases reflecting adaptations to low temperature.In this sense, the superposition of its crystal structure with those from the mesophilic homologues LlNDT and LhPDT (Figure 4A) revealed that the catalytic residues occupy almost identical positions (Figure 4B), a characteristic that has been observed in other cases (see [32] for a review).With the aim to check the participation of such residues in the catalytic mechanism within the context of cold-adaptation, an exhaustive mutagenesis analysis was carried out by preparing a battery of single-point variants of BpNDT (Table 5).
As expected, substitution of the nucleophile Glu85 abolished enzymatic activity of BpNDT.This inactivating effect is obtained either by a shortening of the side chain while maintaining the carboxylate moiety (Glu85Asp variant) or by an exchange of this carboxylate for an amide group (Glu85Gln variant).These results highlight the relevance of the specific orientation and position of the nucleophile side chain in the proposed catalytic mechanism of 2 -deoxyribosyltransferases [20].This can also be deduced from the Tyr5Phe variant where the removal of the hydroxyl group of the Tyr5 side chain fully inactivates BpNDT, which reveals the critical role of the 2.9 Å H-bond between this OH group and the carboxylate OE2 atom of Glu85.On the other hand, the nearby Asp59 side chain (Asp59Asn variant) also resulted essential in the catalytic mechanism, most probably participating in the delocalization of the negative charge in the purine/pyrimidine ring during the cleavage of the glycosidic bond as proposed for LhPDT [20].
Replacement of Gln40 for Glu abolishes catalytic activity of BpNDT.In LlNDT this residue was demonstrated to interact with the O2 and N3 atoms of the pyrimidine ring, therefore being critical for the productive binding of pyrimidine bases [20,30].This residue is substituted with Gly in the type II 2 -deoxyrobosyltransferases LhPDT [20].Remarkably, and in contrast to LlNDT, the Gln40 side chain in BpNDT is not orientated towards the substrate-binding pocket, presumably revealing that dynamic properties of the β2-α3 connecting loop should be important for catalytic activity [20,23].In fact, Gln40 and Leu41 from chain B are ill-defined in the electron density map that together with the high B-factors of this loop is indicative of high flexibility.Probably, this should also be the case for the β3-α4 connecting loop of BpNDT where Asp79 is located, which resulted essential for catalytic activity (Asp79Asn variant) similarly to Asp92 in LlNDT or Asp95 in LhPDT.
Tyr5 side chain fully inactivates BpNDT, which reveals the critical role of the 2.9 Å H-bond between this OH group and the carboxylate OE2 atom of Glu85.On the other hand, the nearby Asp59 side chain (Asp59Asn variant) also resulted essential in the catalytic mechanism, most probably participating in the delocalization of the negative charge in the purine/pyrimidine ring during the cleavage of the glycosidic bond as proposed for LhPDT [20].
Replacement of Gln40 for Glu abolishes catalytic activity of BpNDT.In LlNDT this residue was demonstrated to interact with the O2 and N3 atoms of the pyrimidine ring, therefore being critical for the productive binding of pyrimidine bases [20,30].This residue is substituted with Gly in the type II 2'-deoxyrobosyltransferases LhPDT [20].Remarkably, and in contrast to LlNDT, the Gln40 side chain in BpNDT is not orientated towards the substrate-binding pocket, presumably revealing that dynamic properties of the β2-α3 connecting loop should be important for catalytic activity [20,23].In fact, Gln40 and Leu41 from chain B are ill-defined in the electron density map that together with the high B-factors of this loop is indicative of high flexibility.Probably, this should also be the case for the β3-α4 connecting loop of BpNDT where Asp79 is located, which resulted essential for catalytic activity (Asp79Asn variant) similarly to Asp92 in LlNDT or Asp95 in LhPDT.Finally, the catalytic role of the C-terminal residue of Lys142 is demonstrated in our study with the variants Lys142Tyr and ΔK142.Our results agree with previous ones showing the critical role of the carboxylate group of the C-terminal end [20,30,33].In our case, the Lys142Tyr variant resulted as active as native BpNDT, whereas truncation of Lys142 retained significant activity (Table 5).Unfortunately, the specific location of the Lys142 side chain of BpNDT could not be determined since  Finally, the catalytic role of the C-terminal residue of Lys142 is demonstrated in our study with the variants Lys142Tyr and ∆K142.Our results agree with previous ones showing the critical role of the carboxylate group of the C-terminal end [20,30,33].In our case, the Lys142Tyr variant resulted as active as native BpNDT, whereas truncation of Lys142 retained significant activity (Table 5).Unfortunately, the specific location of the Lys142 side chain of BpNDT could not be determined since the last four amino acids were not interpretable in the 2F o -F c electron density presumably due to intrinsic disorder.
A molecular adaptation to low temperatures observed in some X-ray structures is that catalytic cavities seem to be larger and more accessible to ligands in psychrophilic enzymes than in mesophilic ones [34,35].This can be achieved by a reduction in the length of the loops flanking the active sites, the presence of distinct conformations and/or increases in local flexibility.In this regard, superposition of the structures of BpNDT, LlNDT and LhPDT reveals that the catalytically relevant β2-α3 connecting loop shows a conformation highly variable in the three enzymes (Figure 4C).Also, it is very much shorter in BpNDT (10 residues) than in LlNDT or LhPDT (21 residues), which results in a more open entrance to the binding pocket.Whereas the length and conformation of the rest of the loops around the catalytic cavity (β1-α1, β3-α4 and β4-α5) are similar between BpNDT and LlNDT, the β4-α5 connecting region in LhPDT is much longer than in BpNDT (13 residues versus 7, respectively).Interestingly, these crystal structures also provide indications of dynamic regions through B-factor analysis.As shown in Figure 5, crystallographic B-factors reveal that the β2-α3, β3-α4 and β4-α5 loops of BpNDT are much flexible than in LlNDT or LhPDT indicative of a cold-adaptation for BpNDT.A molecular adaptation to low temperatures observed in some X-ray structures is that catalytic cavities seem to be larger and more accessible to ligands in psychrophilic enzymes than in mesophilic ones ( [34,35].This can be achieved by a reduction in the length of the loops flanking the active sites, the presence of distinct conformations and/or increases in local flexibility.In this regard, superposition of the structures of BpNDT, LlNDT and LhPDT reveals that the catalytically relevant β2-α3 connecting loop shows a conformation highly variable in the three enzymes (Figure 4C).Also, it is very much shorter in BpNDT (10 residues) than in LlNDT or LhPDT (21 residues), which results in a more open entrance to the binding pocket.Whereas the length and conformation of the rest of the loops around the catalytic cavity (β1-α1, β3-α4 and β4-α5) are similar between BpNDT and LlNDT, the β4-α5 connecting region in LhPDT is much longer than in BpNDT (13 residues versus 7, respectively).Interestingly, these crystal structures also provide indications of dynamic regions through B-factor analysis.As shown in Figure 5, crystallographic B-factors reveal that the β2-α3, β3-α4 and β4-α5 loops of BpNDT are much flexible than in LlNDT or LhPDT indicative of a coldadaptation for BpNDT.

Enzymatic Production of Nucleoside Analogues
Non-natural nucleosides were synthesized from specific bases and from the best natural nucleosides, dUrd and dCyd (Table 6).BpNDT was able to catalyze the transfer reaction using most

Enzymatic Production of Nucleoside Analogues
Non-natural nucleosides were synthesized from specific bases and from the best natural nucleosides, dUrd and dCyd (Table 6).BpNDT was able to catalyze the transfer reaction using most of non-natural bases as acceptors, with a slight decrease in yield with respect to the one described for LrNDT [18] when the donor is dUrd.As observed in the case of natural bases, higher activities were obtained when purine bases were used, such as benzimidazole, 2,6-diaminopurine and 2-fluoroadenine.Furthermore, we observed efficient catalysis of the synthesis of four nucleoside analogues: 5-hydroxymethyl-2 -deoxyuridine, 2-fluoro-2 -deoxyadenosine, 7-deaza-6-hydroxypurine-2 -deoxyriboside and theophylline-2 -deoxyriboside.
This result is unprecedented for NDTs.

Discussion
Nowadays, psychrophilic microorganisms are being paid much attention as a source of cold-adapted enzymes with biotechnological interest, since these biocatalysts are more productive at low temperatures than their mesophilic or thermophilic counterparts [1,2].An example of psychrotolerant bacteria is B. psychrosaccharolyticus (CECT 4074, ATCC 23296, DSM 6), which has been used to investigate the mechanism of solvent stress.So, Hsp33, a stress response protein that increases resistance of microorganisms in solvent stress conditions has been identified [36].Likewise, two enzymes with biotechnological application from B. psychrosaccharolyticus have been described: (i) alanine racemase, which showed high catalytic activity even at 0 • C and was extremely labile above 35 • C [37]; and (ii) lactate dehydrogenase, widely used in diagnostics and analytical applications [38].Interestingly, B. psychrosaccharolyticus whole cells have been also reported to display nucleoside 2 -deoxyribosyltransferase activity [24] and the analysis of its genome permitted the identification of the ndt gene that codifies a putative NDT [9].Subsequently, this gene was cloned and overexpressed and the recombinant protein produced and immobilized [25,39].
BpNDT is the first 2 -deoxyribosyltransferase studied among psychrotolerant bacteria.Substrate specificity analyses show that it is a type II NDT since it catalyses the transfer of 2 -deoxyribose between purines and pyrimidines.Remarkably, the dependence of its catalytic activity on temperature reveals that BpNDT is not heat-labile (Figure 1), a characteristic of cold-adapted enzymes that results from catalytic centres more mobile or flexible.Indeed, although we have observed that some loops flanking the substrate cavity of BpNDT are more flexible than in the mesophilic LlNDT or LhPDT (Figures 4 and 5), it is obvious from the thermal inactivation experiments that this structural feature does not result in a heat-labile enzyme: BpNDT exhibits highest activity at 40 • C and pH 8 [25].In fact, it is active under these conditions for a longer period of time than the mesophilic LrNDT: the half-time at 60 • C of BpNDT (21.8 h) is remarkably higher than the one observed for LrNDT (17.9 min) [18].
A consequence of mobile and flexible active sites in psychrophilic enzymes is that they have higher K M values than their mesophilic counterparts [40].In this case, BpNDT also behaves as a mesophilic enzyme since the kinetic parameters we determined using dCyd and Ade as nucleoside donor and base acceptor, respectively (dCyd: K M = 2.0 ± 0.3 mM and k cat (s −1 ) = 24.6 ± 0.1; Ado: K M = 0.4 ± 0.1 mM and k cat (s −1 ) = 28.2 ± 0.1) compare well with published values for other mesophilic homologues [18,27,41,42].
The mesophilic character of BpNDT is also observed when analysing its unfolding behaviour, either by calorimetry or by intrinsic fluorescence spectroscopy.The unfolding temperature value determined in both cases is 49 • C, an unusually high value for a psychrophilic enzyme.Monovalent cations did not affect BpNDT activity, in contrast to other reports [26].It was only diminished by the presence of some divalent cations (Co 2+ , Cu 2+ , Mn 2+ , Zn 2+ ) and Al 3+ .That behaviour could be related to an unspecific effect since BpNDT is not a metalloenzyme, and therefore these ions would create an electrostatic environment where pKa values of functional groups involved in catalysis and/or structure of the enzyme might be altered.Additionally, dramatic reduction of enzyme activity by Al 3+ might be due to the well-known pro-oxidant activity of aluminium.On the other hand, BpNDT has no cysteines in its primary sequence and 2-mercaptoethanol did not interfere with its activity.Scarcity or even absence, as in this case, of disulphide bridges is characteristic of cold-adapted enzymes [1,32].
Regarding substrate specificity, BpNDT accepts different natural and non-natural bases, showing a clear preference for purines as base acceptors (Hyp > Ade > Cyt > Thy ≈ Ura) and pyrimidine nucleosides as donors (dCyd > dUrd ≈ dThd ≈ dAdo > dGuo > dIno).This substrate specificity is similar to other well-known NDTs [12,18,19] except for the strong preference for Hyp as acceptor, only comparable with that of L. lactis subsp.lactis NDT [26].This result suggests that high specificity for Hyp is not exclusive for type I NDTs, traditionally associated with the metabolism of dIno for that reason, among others [19].

Chemicals
Culture media were purchased from Becton Dickinson (Franklin Lakes, NJ, USA).Substrates (nucleosides, 2 -deoxynucleosides and bases) and all other chemical reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Cloning, Sequencing and Tagging of BpNDT
The gene encoding nucleoside 2 -deoxyribosyltransferase from B. psychrosaccharolyticus CECT 4074 was amplified by the Polymerase Chain Reaction (PCR) with a 6-His tag coding sequence added to its 5 end using pET28a(+)-BpNDT [25] as template.DNA amplification was performed under standard conditions in a Mastercycler (Eppendorf, Hamburg, Germany) thermocycler using Pfu DNA polymerase.The amplified 0.43-kb product was inserted into a pET28a(+) vector, purified with the High Pure Plasmid Isolation Kit (Roche Diagnostics, Basel, Switzerland) and sequenced to confirm the absence of mutations.This recombinant plasmid (pET28BpndtHis) was used to transform competent E. coli BL21 (DE3) cells before growing them at 37 • C on LB (Luria Bertani) medium with kanamycin (50 µg/mL).
DNA manipulation and transformations were carried out according to standard methods [45].DNA sequencing was performed by the dideoxy chain termination method [46] with an automated sequencer, 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA).

Production and Purification of Recombinant His-Tag BpNDT
E. coli BL21 (DE3) cells harboring pET28BpndtHis were grown at 37 • C in LB medium supplemented with kanamycin (50 µg/mL).When cultures reached an optical density of 0.6 at 600 nm, expression of BpndtHis was induced with 0.5 mM IPTG for 2.5 h at the same temperature.Cells were then harvested by centrifugation at 3500× g for 10 min, resuspended in buffer A (20 mM sodium phosphate, 500 mM NaCl, 10 mM imidazole buffer, pH 7.5) and disrupted by sonication on ice employing a Digital Sonifier 450 (Branson Ultrasonics Corporation, Danbury, CT, USA).The resulting cell extract was applied onto a 1 mL Nickel Rapid Run Cartridge (ABT, Torrejon de Ardoz, Spain) equilibrated with buffer A and washed at a flow rate of 1 mL/min until the eluate contained no protein.Adsorbed protein was then eluted using buffer B (20 mM sodium phosphate, 500 mM NaCl, 500 mM imidazole buffer at pH 7.5).Purified BpNDT-His was analyzed by SDS-PAGE using a gel containing 15% acrylamide [47].Protein concentration was measured by the Bradford method [48].

N-Deoxyribosyltransferase Assay
Reaction mixtures contained 0.40 µg of electrophoretically pure enzyme, 10 mM 2 -deoxyuridine and 10 mM adenine in 50 mM HEPES buffer, pH 8.0 in a final volume of 40 µL.Reactions were conducted at 40 • C with shaking (30 r.p.m.) for 5 min and were stopped by addition of 40 µL of cold methanol and heating for 5 min at 95 • C.After centrifugation at 9000× g for 2 min, supernatants were half-diluted with water and analyzed by HPLC using an Agilent 1100 Series system (Agilent, Santa Clara, CA, USA) equipped with an ACE ® C18-PFP column (dimensions: 250 × 46 mm; particle size: 5 µm) (Advanced Chromatography Technologies Ltd., Aberdeen, UK) at a flow rate of 0.980 mL/min.The nucleoside product was eluted into the diode array detector for quantification at 254 nm by applying two successive gradients: 100 to 90% trimethyl ammonium acetate and 0 to 10% acetonitrile in ten minutes followed by 90 to 100% trimethyl ammonium acetate and 10 to 0% acetonitrile in ten minutes.One unit of enzyme activity was defined as the amount of enzyme required to produce 1 µmol of product per minute under these conditions.
Synthesis of natural nucleosides was studied using 10 mM substrates (2 -deoxyribonucleosides and bases) while 2 mM was used for non-natural nucleosides.The buffer was 50 mM HEPES at pH 8.0 in both cases.
Retention times for reference compounds were as follows:

Thermal Inactivation Studies
Thermal inactivation kinetics were studied by incubating enzyme aliquots for different times (2 to 96 h) at several temperatures ranging from 40 to 70 • C.After thermal treatment, aliquots were put on ice for 5 minutes and the remaining activity was determined by incubating 5 µL of the treated enzyme with 10 mM 2 -deoxycytidine and 10 mM adenine and quantifying synthesis of 2 -deoxyadenosine at 40 • C in 50 mM HEPES buffer at pH 8.0.
Data were fit to single exponential decays assuming first-order, unimolecular and irreversible reactions involving only two different enzymatic states (E active → E inactive ).This is algebraically described by Equation (1): where A and A 0 are the residual and initial activities, respectively, for a given inactivation time (t) and k inac is the first-order inactivation rate constant.

Analytical Ultracentrifugation Analysis
Sedimentation velocity and sedimentation equilibrium experiments for BpNDT were performed in 50 mM potassium phosphate buffer, pH 7.0 at 50,000× g using an Optima XL-1 analytical ultracentrifuge (Beckman-Coulter Life Sciences, Indianapolis, IN, USA), equipped with absorbance optics, an An-60 Ti rotor and standard (12-mm optical path) double-sector center pieces of charcoal-filled Epon.Baseline offsets were determined at 200,000× g.The apparent sedimentation coefficient distribution, c(s), and sedimentation coefficient s were calculated from sedimentation velocity data using program SEDFIT 12.52 [49].The whole-cell weight-average bM w (buoyant molar mass) values were obtained by fitting experimental data to the equation for the radial concentration distribution of an ideal solute at sedimentation equilibrium, using program HETEROANALYSIS 1.1.44[50].The corresponding apparent weight-average molar masses (M w ) were determined from the buoyant masses, taking into account the partial specific volumes of the protein (0.738 mL/g) obtained from the amino acid composition using program SEDNTERP version 201220111 BETA [51].

Enzyme Crystallization and Data Collection
Crystallization of BpNDT was performed at 291 K by the sitting-drop vapour diffusion method with Innovaplate SD-2 96-well plates using a Nanodrop Innovadyne robot.Each drop contained 250 nL of protein (9 mg/mL) in Tris-HCl buffer (20 mM Tris-HCl pH 8.0 containing 0.1 M NaCl) and 250 nL of reservoir solution.Drops were equilibrated against 65 µL reservoir solution.Crystals were observed in numerous conditions from all tested screens.After scaling and optimisation of preliminary crystallization conditions, high-quality diffraction crystals were prepared in 3 M sodium nitrate, 0.1 M sodium acetate trihydrate, pH 4.6 (protein/precipitant drop ratio 1:2).
For diffraction data collection, BpNDT crystals were transferred to an optimized cryoprotectant solution consisting of mother liquor plus 10% (v/v) 2-methyl-2,4-pentanediol before being cooled to 100 K in a cold nitrogen-gas stream.Diffraction data were recorded on a Pilatus 6M pixel detector (Dectris LTD) at beamline ID29 at the European Synchrotron Radiation Facility (ESRF) (Grenoble, France).A total of 1800 images were collected with a 0.1 • oscillation angle.Diffraction images were processed with XDS [52] and the space group examination was performed with POINTLESS from the Collaborative Computational Project Nº 4 (CCP4) software package [53].Crystals of BpNDT belonged to the trigonal space group R3, with two molecules in the asymmetric unit and 41% solvent content within the unit cell.A summary of data collection statistics is provided in Table 4.

Structure Solution and Refinement
The structure of BpNDT was solved by molecular replacement using phenix.phaser [54].The atomic coordinates of nucleoside 2 -deoxyribosyltransferase from Lactobacillus helveticus were used as search model (LhPDT; PDB entry 1S2G).Model rebuilding was performed manually using COOT [55] and refinement was carried out with phenix.refine[56] in PHENIX [57].Refinement steps included xyz refinement, TLS (Translation, Libration, Screw), individual atomic displacement parameters (ADPs), addition of ligands and solvent molecules.The refined structure has a final R-factor of 18.5% (R free = 22.3%) for data up to 1.90 Å. Analysis of the interfacial surfaces was done with the PISA server [31].Analysis of the secondary structure was done with the DSSP (Dictionary of Secondary Structure of Proteins) server [58].Stereochemistry validation was done with the Phenix MolProbity tool plus de wwPDB Deposition server.PyMOL [59] was used for structure visualization and figure preparation.Data collection and refinement statistics are listed in Table 1.

Spectroscopic Studies
The molar extinction coefficient of the native enzyme (ε nat ) was determined using an accurate method which includes both measured and calculated properties [58] and uses Equation (2): where A nat was the absorbance at 280 nm of BpNDT (70 µg/mL) in 10 mM potassium phosphate buffer, pH 7.0; Aunf is that for unfolded protein (in presence of 6 M guanidine hydrochloride) and ε unf is the molar extinction coefficient of the unfolded protein, which was calculated to be 30,940 M −1 •cm −1 from the amino acid composition of the enzyme using ProtParam program (http://web.expasy.org/protparam).The effect of temperature on fluorescence spectra of pure BpNDT was evaluated from 20 to 80 • C increasing 20 • C/h, using an excitation wavelength of 295 nm.Fluorescence emission spectra were recorded using an SLM-Aminco•Bowman Series 2 spectrofluorometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with thermostatted cell holder with 0.4 cm and 1 cm path-lengths for excitation and emission, respectively.Both excitation and emission slit widths were 5 nm.The scan rates were 60 nm/min.Protein concentration was 0.1 mg/mL in 50 mM potassium phosphate buffer, pH 7.0.

Differential Scanning Calorimetry Studies
Differential scanning calorimetry (DSC) studies were performed using a Microcalorimeter VP-DSC calorimeter (Malvern Instruments, Malvern, UK).Five scans of pure BpNDT (0.36 mg/mL in 50 mM potassium phosphate buffer, pH 7.0) were recorded from 15 to 90 • C at a scan rate of 20 • C/h.

Kinetic Studies
Kinetic parameters were calculated for the synthesis of 2 -deoxyadenosine from 2 -deoxycytidine and adenine with one substrate fixed at different concentrations (1, 2.5, 5, 7.5, 10, 15 mM) while varying the other one from 0.25 to 60 mM.K M and k cat were determined by fitting velocity data to the Michaelis-Menten model using non-linear regression analysis.

Accession Number
The atomic coordinates and structure factors have been deposited in the Protein Data Bank with the accession code 6EVS.

Conclusions
The present work on BpNDT has provided structural and biochemical results that are apparently inconsistent regarding the adaptation of this enzyme to work at low temperatures.Hence, the thermal inactivation and unfolding experiments on BpNDT showed that this enzyme is not heat labile, presenting unfolding curves typical for a mesophilic enzyme.Conversely, the structural analyses do not reveal clear-cut features characteristic of cold-adapted enzymes.In particular, the catalytic machinery of BpNDT is essentially identical to that of mesophilic homologues and a great number of stabilizing interactions involved in the association of subunits within the homohexamer can be identified.In this regard, it should be remarked that most cellular adaptations to low temperatures and the underlying molecular mechanisms are not fully understood and although general trends at the molecular level are expected to exist explaining cold-adaptation, it cannot be discarded individual strategies for each protein based on specific combinations of structural alterations [32].In the case of BpNDT, we believe that features derived from its multimeric nature such as the specific topological features of this oligomer in which active sites face the solvent in an equatorial arrangement with loops at their entrances endowed with local flexibility may be critical factors underpinning its adaptation to low temperatures.Yet, we believe that our results highlight the necessity to integrate molecular information coming from functional and structural analyses of individual macromolecules into the high level cellular processes where they participate to fully explain cold-adaptation.
Finally, in this work, we have shown that one-pot, one-step nucleoside synthesis catalyzed by BpNDT represents a valuable alternative to chemical methods and nucleoside phosphorylases [60].Furthermore, BpNDT is an interesting biocatalyst from an industrial point of view, since it can perform at 40 • C but also at low temperatures, minimizing undesirable chemical reactions that can occur at higher temperatures and protecting heat-labile substrates [61].

Figure 2 .
Figure 2. Determination of apparent melting temperature (Tm) of BpNDT: (a) Effect of temperature on fluorescence emission at 335 nm.(b) Differential scanning calorimetry (DSC) analysis.

Figure 2 .
Figure 2. Determination of apparent melting temperature (Tm) of BpNDT: (a) Effect of temperature on fluorescence emission at 335 nm.(b) Differential scanning calorimetry (DSC) analysis.

Figure 2 .
Figure 2. Determination of apparent melting temperature (Tm) of BpNDT: (a) Effect of temperature on fluorescence emission at 335 nm.(b) Differential scanning calorimetry (DSC) analysis.

Figure 2 .
Figure 2. Determination of apparent melting temperature (Tm) of BpNDT: (a) Effect of temperature on fluorescence emission at 335 nm.(b) Differential scanning calorimetry (DSC) analysis.

Figure 2 .
Figure 2. Determination of apparent melting temperature (T m ) of BpNDT: (a) Effect of temperature on fluorescence emission at 335 nm.(b) Differential scanning calorimetry (DSC) analysis.

a
Values for the highest resolution shell are given in parentheses.

Figure 3 .
Figure 3. Overall structure of BpNDT.(A) The two subunits of BpNDT that form the asymmetric unit of the crystal are shown as ribbon model (blue) and ribbon plus transparent surface (magenta), respectively.The regular secondary structure elements are indicated.(B) Two orientations of the hexameric assembly identified within the BpNDT crystal are shown.Each dimer is depicted in surface model.

Figure 3 .
Figure 3. Overall structure of BpNDT.(A) The two subunits of BpNDT that form the asymmetric unit of the crystal are shown as ribbon model (blue) and ribbon plus transparent surface (magenta), respectively.The regular secondary structure elements are indicated.(B) Two orientations of the hexameric assembly identified within the BpNDT crystal are shown.Each dimer is depicted in surface model.

Figure 4 .
Figure 4. Structural analysis of the BpNDT active site.(A) Multiple sequence alignment of the 2'deoxyribosyltransferases BpNDT, LlNDT and LhPDT.Conserved residues are depicted in white, bold characters within red boxes and highly similar ones in red characters.Catalytic important residues are marked with asterisks.(B) Three dimensional superposition of catalytic residues from BpNDT (yellow, sticks model; PDB code: 6evs), LlNDT (cyan, sticks model; PDB code: 1f8y), and LhPDT (magenta, sticks model; PDB code: 1s2d).Numbering is with respect BpNDT sequence (excluding residues marked with and asterisk).The symbol # indicates residues from the accompanying subunit.The ligand 2-methyl-pseudouridine (2-MPU) is shown to clarify the pyrimidine-binding mode (PDB code: 1f8y).(C) Ribbon models of BpNDT, LlNDT and LhPDT (color code as above) highlighting the different conformation of the catalytically important loop β2-α3.

Figure 4 .
Figure 4. Structural analysis of the BpNDT active site.(A) Multiple sequence alignment of the 2 -deoxyribosyltransferases BpNDT, LlNDT and LhPDT.Conserved residues are depicted in white, bold characters within red boxes and highly similar ones in red characters.Catalytic important residues are marked with asterisks.(B) Three dimensional superposition of catalytic residues from BpNDT (yellow, sticks model; PDB code: 6evs), LlNDT (cyan, sticks model; PDB code: 1f8y), and LhPDT (magenta, sticks model; PDB code: 1s2d).Numbering is with respect BpNDT sequence (excluding residues marked with and asterisk).The symbol # indicates residues from the accompanying subunit.The ligand 2-methyl-pseudouridine (2-MPU) is shown to clarify the pyrimidine-binding mode (PDB code: 1f8y).(C) Ribbon models of BpNDT, LlNDT and LhPDT (color code as above) highlighting the different conformation of the catalytically important loop β2-α3.

Figure 5 .
Figure 5. Analysis of flexibility regions of BpNDT, LlNDT and LhPDT.(A) Crystallographic B-factors are used as probes to reveal flexible regions within the structures the corresponding subunits.These regions correspond to loops at the entrance to the substrate-binding site.B-factors are indicated by color, from blue to red, and via a putty tube representation.Red regions correspond to more flexible regions, which is clearer for BpNDT.(B) The flexible loops of BpNDT within the hexamer are located at the equatorial region of the assembly, facing the bulk solvent.

Figure 5 .
Figure 5. Analysis of flexibility regions of BpNDT, LlNDT and LhPDT.(A) Crystallographic B-factors are used as probes to reveal flexible regions within the structures the corresponding subunits.These regions correspond to loops at the entrance to the substrate-binding site.B-factors are indicated by color, from blue to red, and via a putty tube representation.Red regions correspond to more flexible regions, which is clearer for BpNDT.(B) The flexible loops of BpNDT within the hexamer are located at the equatorial region of the assembly, facing the bulk solvent.

Table 1 .
Effect of several additives on the activity of BpNDT.

Table 1 .
Effect of several additives on the activity of BpNDT.

Table 3 .
Analysis of the substrate specificity of BpNDT in the synthesis of natural nucleosides a .
a Reaction conditions: BpNDT (0.40 µg) was incubated at 40 • C for 5 min with 10 mM substrates in 50 mM HEPES buffer, pH 8 in a final volume of 40 µL.

Table 4 .
Data collection and refinement statistics.
a Values for the highest resolution shell are given in parentheses.

Table 5 .
Impact of different mutations on the activity of BpNDT a .
a Reaction conditions: 0.40 µg of enzyme were incubated at 40 • C for 5 min with 10 mM 2 -dUrd and Ade in HEPES 50 mM buffer pH 8 in a final volume of 40 µL.

Table 5 .
Impact of different mutations on the activity of BpNDT .Reaction conditions: 0.40 μg of enzyme were incubated at 40 °C for 5 min with 10 mM 2′-dUrd and Ade in HEPES 50 mM buffer pH 8 in a final volume of 40 μL. a

Table 6 .
BpNDT-catalyzed synthesis of non-natural nucleosides from natural nucleosides and non-natural bases a .
a Experimental conditions: 0.40 µg of enzyme at 40 • C for 2 h with 1.95 mM substrates in 50 mM HEPES buffer, pH 8 in a final volume of 40 µL.ND, not detected.