Thermostable nucleoside phosphorylases are attractive biocatalysts for the synthesis of modified nucleosides. Hence we report on the recombinant expression of three ‘high molecular mass’ purine nucleoside phosphorylases (PNPs) derived from the thermophilic bacteria Deinococcus geothermalis, Geobacillus thermoglucosidasius and from the hyperthermophilic archaeon Aeropyrum pernix (5′–methythioadenosine phosphorylase; ApMTAP). Thermostability studies, kinetic analysis and substrate specificities are reported. The PNPs were stable at their optimal temperatures (DgPNP, 55 °C; GtPNP, 70 °C; ApMTAP, activity rising to 99 °C). Substrate properties were investigated for natural purine nucleosides [adenosine, inosine and their C2′-deoxy counterparts (activity within 50–500 U·mg−1)], analogues with 2′-amino modified 2′-deoxy-adenosine and -inosine (within 0.1–3 U·mg−1) as well as 2′-deoxy-2′-fluoroadenosine (9) and its C2′-arabino diastereomer (10, within 0.01–0.03 U·mg−1). Our results reveal that the structure of the heterocyclic base (e.g. adenine or hypoxanthine) can play a critical role in the phosphorolysis reaction. The implications of this finding may be helpful for reaction mechanism studies or optimization of reaction conditions. Unexpectedly, the diastereomeric 2′-deoxyfluoro adenine ribo- and arabino-nucleosides displayed similar substrate properties. Moreover, cytidine and 2′-deoxycytidine were found to be moderate substrates of the prepared PNPs, with substrate activities in a range similar to those determined for 2′-deoxyfluoro adenine nucleosides 9 and 10. C2′-modified nucleosides are accepted as substrates by all recombinant enzymes studied, making these enzymes promising biocatalysts for the synthesis of modified nucleosides. Indeed, the prepared PNPs performed well in preliminary transglycosylation reactions resulting in the synthesis of 2′-deoxyfluoro adenine ribo- and arabino- nucleosides in moderate yield (24%).
Base and sugar-modified analogues of natural nucleosides are widely used as pharmaceutical agents for the treatment of viral infections and cancer [1-3]. Moreover, the replacement of natural nucleosides by modified species results in improvement in the biochemical properties of synthetic oligonucleotides. For example, C2′–fluorinated nucleosides have been shown to have favorable properties as components of antisense oligonucleotides and small interfering RNAs [4-6]. The chemical synthesis of modified nucleosides is challenged by stereo- and regioselective requirements, as well as the need to protect and deprotect sensitive functional groups. By contrast, enzymes, such as nucleoside phosphorylases (NPs) can be used as biocatalysts for the efficient synthesis of nucleoside analogues under environmentally friendly conditions [7-11].
However, common disadvantages encountered in biocatalysis comprise the instability of many enzymes under harsh reaction conditions and narrow substrate spectra limiting the scope of any potential synthetic reactions. Escherichia coli NP-mediated synthesis of fluorinated nucleosides represents an example of this dilemma . The use of enzymes from thermophiles is a reasonable strategy for overcoming instability problems . Such thermozymes offer the possibility of performing reactions at high temperature, resulting in diminution of the viscosity of the medium and increased substrate diffusion coefficients, and leads to higher overall reaction rates. The final yield may be improved because of the higher solubility of substrates with the increase in temperature . It is noteworthy that thermozymes are often resistant to pressure and organic solvents [15, 16]. They can be expressed in E. coli at high levels and are easily purified by heat treatment [17, 18].
Purine nucleoside phosphorylase (PNP; EC 126.96.36.199) and 5′–methylthioadenosine phosphorylase (MTAP; EC 188.8.131.52) catalyze the reversible phosphorolysis of purine nucleosides, as depicted in Scheme 1 Both belong to the PNP family . According to Shugar and co-workers, PNPs can be grouped into two main classes (for a comprehensive review, see ref. ): PNPs of the first class (‘low molecular mass PNPs’), found mainly in mammalian tissues and in some bacteria, are specific for 6–oxo purine nucleosides only; those of the second class (‘high molecular mass PNPs’) are predominantly found in bacteria and accept 6–oxo (e.g. inosine and guanosine) as well as 6–amino (e.g. adenosine) purine nucleosides as substrates. Because of the broader substrate specificity, the latter group of enzymes is especially interesting for synthetic applications. Although many PNPs have been extensively studied , examples of ‘high molecular mass PNPs’ derived from thermophiles are limited to those from Geobacillus stearothermophilus (GsPNP, punB gene) [9, 20] and Enterobacter aerogenes (EaPNP) [21, 22]; the MTAP from Aeropyrum pernix (ApMTAP) , Sulfolobus solfataricus (SsMTAP) [24, 25] and Pyrococcus furiosus (PfMTAP) , whereby only GsPNP, EaPNP and ApMTAP were considered for synthetic applications. Besides, ApMTAP has been recently applied in synthesis reactions , however, the information about the enzyme's characteristics is not available.
In this study, we report on: (a) the preparation of three recombinant ‘high molecular mass’ PNPs derived from the thermophilic bacteria Deinococcus geothermalis (DgPNP) and Geobacillus thermoglucosidasius 11955 (GtPNP), as well as from the hyperthermophilic archaeon A. pernix (ApMTAP); and (b) the investigation into their physicochemical and substrate specificity. Here, we report on their thermal properties, kinetic parameters and substrate specificities in regard to natural purine substrates (inosine, adenosine and their 2′–deoxy counterparts, 1 and 2, Scheme 2), analogues of natural substrates modified at the C2′ carbon atom [2′–amino-2′–deoxyadenosine (7), 2′–amino-2′–deoxyinosine (8), 2′–deoxy-2′–fluoroadenosine (9) and its C2′–arabino diastereomer 10 (Scheme 2)].
Our results reveal that 2′–modified nucleosides are accepted as substrates by all recombinant enzymes studied, indicating that these enzymes are promising biocatalysts for the synthesis of their respective nucleoside analogues. Indeed, in preliminary experiments for the synthesis of 2′–deoxyfluoro adenine nucleosides 9 and 10, when the prepared PNPs were applied in combination with thermostable pyrimidine nucleosides phosphorylases (PyNP) , using adenine as an acceptor and the respective 2′–deoxy-2′–fluorouridine (dUrd2′F) and 1-(2-deoxy-2-fluoro-β-d–arabinofuranosyl)uracil (dUrd2′F) as the pentofuranose residue donors, the formation of the desired nucleosides was observed with a yield of ~ 24%.
Finally, we tested cytidine (11) and 2′–deoxycytidine (12) as substrates of the PNPs prepared and found that these nucleosides showed moderate substrate activity compared with that of 2′–deoxyfluoro adenine nucleosides 9 and 10.
Cloning, expression and purification
In order to avoid problems arising from stable secondary structures in the 5′–mRNA region , the target gene sequences were cloned with an N–terminal hexahistidine tag.
All three enzymes were functionally expressed in E. coli BL21. Soluble GtPNP was well expressed under standard conditions [TB medium, 37 °C, 100 μm isopropyl β–d–thiogalactoside (IPTG)] as shown in Fig. 1A (lane 1). For the expression of DgPNP (Fig. 1A, lane 3) and ApMTAP (Fig. 1A, lane 5), EnPresso® medium  was used at a lower IPTG concentration (20 μm) to prevent the formation of insoluble protein aggregates. The final cell densities varied from A600 = 19 (ApMTAP) to 25 (DgPNP) and 42 (GtPNP).
In SDS/PAGE analysis, the purified proteins (Fig. 1A, lanes 2, 4 and 6) showed bands corresponding to the theoretically calculated molecular masses of the monomeric subunits (DgPNP, 30.0 kDa; GtPNP, 27.6 kDa; and ApMTAP, 27.0 kDa).
However, in the samples of purified ApMTAP, a second band with a molecular mass of ~ 116 kDa could be seen. It was confirmed by MS analysis that this second band also represents ApMTAP (Tables S1 and S2). Because disulfide bond formation is considered to be a key factor for protein stabilization in thermophilic archaea [29-31], we assumed that the second band related to one of possible oligomerization states of the ApMTAP monomeric subunits connected by the disulfide bonds. This assumption was supported by SDS/PAGE analysis (Fig. 1B, lanes 3 and 6), which revealed that the oligomeric band was predominant when the purified ApMTAP was heat-treated in the absence of dithiothreitol (Fig. 1B, lane 3), whereas the monomeric band was predominant after dithiothreitol treatment (Fig. 1B, lane 6). This phenomenon was not observed for the other proteins. Furthermore, we found that the higher oligomeric state of ApMTAP was catalytically active, whereas the monomeric ApMTAP was inactive (data not shown).
The effect of temperature on the activity of the enzymes was examined between 30 and 99 °C (Fig. 2). DgPNP showed optimal activity at 55 °C (Fig. 2), which is slightly higher than the optimal growth temperature (45–50 °C) of D. geothermalis . The stability half-life at 60 °C was 1.6 h, whereas 80% of the activity was retained at 55 °C for 8 h (Fig. 3A). GtPNP showed an optimal activity at 70 °C (Fig. 2), which is ~ 15 °C higher than the optimal growth temperature (55 °C) of its source microorganism G. thermoglucosidasius  and 10 °C higher than the temperature optimum of GtPyNP, another NP enzyme, from the same microorganism . GtPNP was incubated at 70 °C for 8 h and no significant activity loss could be observed (Fig. 3B), whereas GtPyNP at 70 °C had a half-life of only 1.6 h . The activity of ApMTAP increased exponentially at increased reaction temperatures up to the highest temperature tested (99 °C, Fig. 2), while the source microorganism A. pernix shows optimal growth at temperatures between 90 and 95 °C . ApMTAP is extremely thermostable with an estimated half-life of 69 h at 90 °C (Fig. 3C).
The thermal characteristics of the PNPs studied here along with other thermostable PNPs characterized by a broad substrate specificity (recognition of both 6–amino and 6–oxo purine nucleosides) are summarized in Table 1.
Table 1. Thermal characteristics of PNPs
Thermal stability (t1/2)
Temp. optimum [°C]
Purified enzyme was kept in 30 mm inosine (in 25 mm phosphate buffer) at 60 °C.
The substrate specificities of the enzymes prepared were studied using a number of natural purine nucleosides [adenosine (Ado), inosine (Ino), 2′–deoxyadenosine (dAdo), 2′–deoxyinosine (dIno)] and 2′-modified analogues (dAdo2′NH2 and dIno2′NH2, dAdo2′F and dAdo2′F).
Phosphorolysis of purine nucleosides
The results show that all three enzymes recognize both 6–oxopurine (e.g. Ino, dIno, dIno2′NH2) and 6–aminopurine nucleosides (e.g. Ado, dAdo, dAdo2′NH2) (Fig. 4). Although specific activities for the same substrate are distinct for each enzyme, it can be generally stated that natural substrates are significantly better accepted than 2′–amino modified substrates (50–500 versus 0.05–3 U·mg−1), whereas the latter are significantly better accepted than the 2′–fluoro substituted compounds (0.01–0.03 U·mg−1). Moreover, it is noteworthy that ApMTAP is characterized by relatively low specific activities with natural substrates, but by a relatively broad substrate spectrum at the same time, including dAdo2′NH2 and dIno2′NH2, dAdo2′F and dAdo2′F; whereas GtPNP is highly specific for natural substrates but also accepts modified substrates. DgPNP was the best candidate for phosphorolysis of the 2′–fluorinated purine nucleosides (Fig. 4). Interestingly, GtPNP and ApMTAP displayed similar levels of activity with dAdo2′NH2. However, the replacement of the heterocyclic base adenine of dAdo2′NH2 with hypoxanthine (resulting in dIno2′NH2) led to a significant decrease in substrate activity in the case of GtPNP, whereas the substrate activity towards ApMTAP was slightly enhanced relative to dAdo2′NH2 (Fig. 4). Notably, a similar characteristic of the GtPNP activity is observed in the case of reactions with Ado and Ino, whereas dAdo and dIno showed reversed dependence of activities.
Phosphorolysis of pyrimidine nucleosides
Araki et al. patented the method for producing cytidine compounds using recombinant E. coli PNP as biocatalyst and 2-deoxy-α-d–ribofuranose-1-phosphate and cytosine as substrates in 100 mm Tris/HCl buffer . Recently, Stepchenko et al. described the phosphorolysis of 2′–deoxycytidine (dCyd), 2′–deoxyuridine (dUrd) and 2′–deoxythymidine (dThd) under the conditions of recombinant E. coli PNP in 80 mm phosphate buffer (pH 7.0) at 15 °C. It was shown that dCyd and dUrd are good substrates and ~ 50% was phosphorolyzed after 24 h, whereas substrate activity of dThd was unexpectedly lower . These data prompted us to test a number of pyrimidine nucleosides as substrates for the PNPs under investigation. It was found that substrate efficiency of cytidine (Cyd) is increasing in the following order: DgPNP (12 mU·mg−1) < GtPNP (14 mU·mg−1) < ApMTAP (36 mU·mg−1). Interestingly, in the case of dCyd, the reverse decrease in substrate activity was observed, viz. DgPNP (34 mU·mg−1) > GtPNP (17 mU·mg−1) > ApMTAP (10 mU·mg−1). The activities of DgPNP, GtPNP and ApMTAP with Urd (0.97, 0.83 and 4.21 mU·mg−1, respectively) and thymidine (< 0.05, 0.26 and 6.83 mU·mg−1, respectively) were found to be rather low.
The kinetic parameters of the thermostable PNPs are summarized in Table 2, along with other literature data. Km and Vmax were measured in 50 mm potassium phosphate buffer and typical Michaelis–Menten kinetics were observed. Note that TtPNP, EcPNP, SsMTAP and PfMTAP are hexamers, whereas TtPNPII is a monomer.
Table 2. Kinetic parameters of PNPs on inosine and adenosine. NA, data not available
The results show that the enzymes have a higher affinity (low Km values) towards adenosine than to inosine, e.g. 5–8 times higher in the cases of DgPNP and GtPNP. Similar remarkable differences in the substrate affinity were also described for PNPs from T. thermophilus  and PfMTAP . The kcat values of the nucleoside phosphorylases assessed for the natural substrates dIno and dAdo point to a very high catalytic activity of GtPNP (cf. the relevant data in Fig. 4), provoking reasonable interest to its use in biotechnological transformations.
Multiple sequence alignment
DgPNP, GtPNP and ApMTAP share a sequence identity with EcPNP of 54, 55 and 30%, whereas GtPNP shows 86% identity with GsPNP; ApMTAP shows 45% identity with SsMTAP.
The amino acid sequence alignment of ‘high molecular mass PNPs’ from different thermopiles with E. coli is shown in Fig. 5. The thermal characteristics and substrate specificities of these proteins were described above.
The alignment shows that the amino acid residues known to be involved in the active site of E. coli PNP and S. solfataricus MTAP [19, 24, 39] are generally conserved, with the exception of Met64, Ser90, Cys91, Ala156, Pro162 and Phe167 in EcPNP which have variations in the thermostable PNPs (indicated in boxes in Fig. 5).
The amino acid residues from 62 to 76 of EcPNP are considered to be highly conserved sequence motifs  involved in ribose binding . In the thermozymes, especially in ApMTAP (Fig. 5), smaller residues such as Ile66*, Gly68, Ala71, Ala72, Val73* and Val74* replace Met64*, Ile66, Cys69, Ser70* and Ile71* of EcPNP, respectively (residues highlighted in pink of Fig. 5 are designated by * in the text). Moreover, in or close to the phosphate- and base-binding sites, similar phenomena can be observed: smaller residues in ApMTAP (or SsMTAP) Gly25*, Ala44*, Ala164*, Ala167, Ala172 and Ala173* replace Leu23*, Val42*, Leu158*, Ser161, Met166 and Phe167* of EcPNP, respectively. On the one hand, the smaller residues provide more space for the modified substrate , which may explain the relatively broad substrate spectrum of ApMTAP. On the other hand, the amino acid composition of thermophilic proteins is thought to be correlated with their thermostability . Indeed, compared with EcPNP, ApMTAP contains more hydrophobic and small residues such as Val (10.4 versus 8.4%), Leu (12.3 versus 6.3%) and Gly (13.9 versus 9.6%). Furthermore, it is known that the substitution of Lys by Arg residues is advantageous at high temperatures because the Arg δ–guanidino moiety has less reactivity than the Lys amino group due to its high pKa and resonance stabilization . It can be observed that the Arg content increases from EcPNP (5.0%), GtPNP (6.0%), DgPNP (6.9%) to ApMTAP (8.2%), whereas the Lys content decreases from EcPNP (5.9%), GtPNP (3.4%), ApMTAP (1.6%) to DgPNP (0.8%).
Among the eight PNPs shown in Fig. 5, only TtPNPI  cannot accept Ado as a substrate, supposedly because of its residues substitutions in the base-binding sites: Glu156, Phe178 and Asn204 instead of Ala156, Val178 and Asp204 in EcPNP. Note that these residues are highly conserved in other PNPs (Fig. 5). The Asp204 (EcPNP) is considered as the key base-binding amino acid residue by the protonation of the purine base at N7 accompanied by hydrogen bonding with the C6 amino group . According to Mikleušević et al. , mutation of Arg204 to Asn in EcPNP results in a 1000–fold decrease in the enzymatic activity, which further provides evidence for the role of Asp204 in the PNP specificity of Ado. Asp204 (corresponding to Asp223 in DgPNP, Asp 203 in GtPNP and Asp210 in ApMTAP) was conserved in the PNPs studied here and as a consequence they can well accept Ado as substrate. However, to explain the difference in the reaction rate and substrate preference between the studied PNPs and EcPNP (see Table 2), profound structure studies of the catalytic site are needed.
Synthesis of fluorinated purine nucleosides
The aforementioned results of this study and data from previous studies  prompted us to investigate the synthesis of adenine 2′-deoxyfluoro-ribo- and -arabino-nucleosides, dAdo2′F (17) and dAdo2′F (18), respectively, using adenine as an acceptor and the respective 2′-fluorinated uracil nucleosides, dUrd2′F (13) and dUrd2′F(14), as donors of the pentofuranose residues (Scheme 3). Preliminary experiments showed the formation of: (a) dAdo2′F using TtPyNP and DgPNP as biocatalysts (55 °C, 24 h) with a yield of 23%, and (b) dAdo2′F employing a TtPyNP and ApMTAP combination of enzymes (80 °C, 24 h) with a yield of 24%. These data point to the possibility of developing a practical synthesis of both 2′-fluorinated adenine nucleosides, although an optimization of the reaction conditions is necessary to increase the yield.
Pyrimidine and purine nucleosides modified at the C2′ carbon atom of the pentofuranose ring are of great importance as constituents of artificial oligonucleotides of medicinal potential [41, 42]. Among them, 2′-amino-2′-deoxy- and 2′-deoxy-2′-fluoro-β-d–ribo-nucleosides attracted much attention during the last two decades and a number of oligonucleotides containing these modified nucleosides at diverse position of the oligonucleotide chain have been synthesized and their properties investigated [43-45]. More recently, 2′-deoxy-2′-fluoro-β-d–arabino-nucleosides have been shown to be of great importance for antisense and low molecular mass oligonucleotides as potential anticancer drugs [6, 46, 47]. Unfortunately, chemical synthesis of pyrimidine and especially purine 2′-deoxyfluoro-ribo- and -arabino-nucleosides, as well as their 2′-aminodeoxy counterparts suffers from many drawbacks despite of enormous progress achieved in this field of research. The search for enzymatic transformations that can efficiently replace some of the chemical steps in the production of these nucleosides using the principles of ‘green chemistry’ is therefore of great importance.
In our previous work, we described the recombinant expression and biocatalytic characterization of two thermostable PyNPs, isolated from G. thermoglucosidasius (GtPyNP) and T. thermophilus (TtPyNP) . It was found that GtPyNP displays very low phosphorolytic activity towards dUrd2′F and dUrd2′F, whereas TtPyNP shows rather satisfactory activity with both 2′-deoxyfluoro-ribo- and -arabino-nucleosides. This observation pointed to the possibility of employing TtPyNP as a biocatalyst in the transglycosylation reaction with adenine as the pentofuranosyl acceptor and dUrd2′F or dUrd2′F as donors of the corresponding ribo- and arabino-pentofuranose.
Here, we studied this reaction testing DgPNP, GtPNP and ApMTAP as biocatalysts for the coupling of the intermediates 2′-deoxyfluoro-ribo- and arabino-1-phosphates with adenine. It was found that DgPNP catalyzes the formation of the desired dAdo2′F under very mild conditions (55 °C, 24 h) with a yield of 24%. It is noteworthy that the undesired chemical transformation of dUrd2′F into O2,2′–anhydro-1-(β-d–arabinofuranosyl)uracil was very low (2%) at 55 °C . The use of TtPyNP and ApMTAP as a combination of enzymes, and dUrd2′F and adenine as substrates (80 °C, 24 h) resulted in the formation of dAdo2′F with a yield of 24%; TtPyNP and GtPNP as a combination of enzymes for the same synthesis resulted in the formation of dAdo2′F with a yield of 14%. It is obvious that these results are very promising from the viewpoint of their application in the development of practical syntheses of dAdo2′F and dAdo2′F.
Phosphorolysis of purine nucleosides catalyzed by GtPyNP and TtPyNP
During the course of the abovementioned studies, we observed that the thermostable PyNPs, GtPyNP and TtPyNP, phosphorolytically cleaved not only pyrimidine nucleosides , but also various purine nucleosides. To investigate this phenomenon further, we determined the temperature dependence of PyNP activity with inosine in order to exclude the possibility that the PNP activity comes from E. coli PNP. It was found that temperature dependence of GtPyNP and TtPyNP activity with inosine (Fig. 6) is similar to that obtained in experiments with one of the natural substrates, uridine (see Fig. 4A in Ref. ). Hence, we assume that the PNP activity is inherent to the PyNPs and not the result of contamination with E. coli PNP.
Furthermore, the activities of GtPyNP (at 60 °C) and TtPyNP (at 80 °C) with the other purine nucleosides were examined and the results are summarized in Table 3. It is obvious that both PyNPs display considerable activities with Ado and Ino. In fact, TtPyNP accepts the purine significantly better than 2′–fluorinated pyrimidine nucleosides. By contrast, the activity with the natural pyrimidine nucleoside cytidine was extremely low (TtPyNP) or was not detectable (GtPyNP).
Table 3. Activity of GtPyNP and TtPyNP with natural and non-natural nucleosides. Reaction conditions: 1 mm substrate in 50 mm phosphate buffer at 60 °C (GtPyNP) or 80 °C (TtPyNP)
In this study, we prepared three recombinant PNPs from thermophiles and studied their properties and substrate specificity for natural substrates (Ado, dAdo, Ino, dIno) and three sets of analogues, viz., (a) dAdo2′NH2 and dIno2′NH2, (b) dAdo2′F and dAdo2′F, (c) Cyd and dCyd. The reasons for selection of the analogues are as follows: the first have closest structural relation of the relevant natural purine nucleosides; the diastereomeric deoxyfluoro adenine nucleosides also structurally mimic dAdo. However, it is well known that the replacement of a hydrogen atom or hydroxyl group in the sugar moiety of nucleosides gives rise to derivatives with unexpected physico-chemical and biological properties. Finally, pyrimidine nucleosides are not typical substrates of PNP, however, Cyd and dCyd mimic the pyrimidine fragment of adenine base of N3–(β-d–ribofuranosyl)adenine that was shown to be the substrate of PNP [48-50] and a good substrate activity of dCyd and 2′-deoxyuridine for E. coli PNP was also described [36, 37].
The role of the vicinal C2′ and C3′ hydroxyl groups of natural purine nucleosides in the binding site of PNPs has been discussed in a number of publications [38, 51, 52]. The PNPs studied here and E. coli PNP contain the same cluster of amino acids (Val–Glu–Met–Glu). In this cluster, the last residue Glu181 of E. coli PNP was suggested to play an important role in the binding of ribose via the vicinal C2′ and C3′ hydroxyl groups and in facilitating substrate activation by flattering the pentofuranose ring . In a similar way, one can suggest that the relevant Glu residues of the enzymes under study [DgPNP (Glu200), GtPNP (Glu180) and ApMTAP (Glu187); (Fig. 5)] are involved in the ribose binding site. It was previously shown that E. coli PNP accepts the natural purine ribo- and 2′-deoxyribo-nucleosides with similar efficiency. The results of this study also showed a similar trend (Fig. 4), indicating that the C2′ hydroxyl groups of Ado and Ino are not essential for the enzymatic reaction. Note that there is a lot of theoretical and experimental data convincingly demonstrating that natural ribo- and 2′-deoxyribo-nucleosides have rather high conformational flexibility, viz., S ↔ N conformation of the pentofuranose ring and base rotation around the glycosidic bond. These conformational peculiarities allow PNPs adopting the required spatial structure of a natural nucleoside substrate in the catalytic site, which is necessary to form the productive enzyme-substrate complex.
Compounds of the first and second sets are functionally competent, i.e. contain all the functionality of inherent natural substrate, purine nucleosides and, a priori, one can expect their substrate activity for PNPs to be similar to those of natural substrates. Indeed, the amino function of dAdo2′NH2 and dIno2′NH2 is isosteric for the hydroxyl group and can functionally replace the hydroxyl group. The only important peculiarity of the amino function vs hydroxyl group consists in that the former can exist in the C2′-ammonium cation ↔ C2′-amino equilibrium implying the pH dependence of the substrate activity, on the one hand (Scheme 4), and the ionic interaction of the phosphate anion and ammonium cation impeding the correct spatial arrangement of the productive nucleoside/phosphate/enzyme complex, on the other.
The observed dramatic decrease in substrate activity by going from natural substrates to dAdo2′NH2 (7) and dIno2′NH2 (8) (2–3 orders of magnitude) prompted us to: (a) test the substrate activity of both aminodeoxy nucleosides at different pH values of the reaction mixtures (Table S3, Fig. S1), and (b) analyze the C2′-ammonium cation and C2′-amino forms of the former by the restricted Hartree–Fock (RHF) ab initio method using a basis set of STO-3G (hyper-chem, 8.1 release) and the PM3 geometry optimization as starting approximation for the ab initio calculations compared to similar data for Ado (Table S4).
Preliminary experiments on the pH dependence on the substrate properties gave a rather variable picture of activities (Fig. S1) and pointed to the need to perform more detailed studies to obtain meaningful results. For example, a plot of kcat/Km as a function of pH is needed to reflect the essential ionizing groups of the free enzyme that play a role in both substrate binding and catalytic processing .
Analysis of the amino and ammonium forms of dAdo2′NH2 by the ab initio method revealed the essential differences in the base orientation about the glycosyl bond as well as the ribofuranose rings of two species, on the one hand, and some difference of both forms from the natural substrate, on the other hand. Conformational analysis of dAdo2′NH2 (7) in water by NMR spectroscopy clearly pointed to its high S ↔ N and anti ↔ syn conformational flexibility . Taking into account that both functionally competent forms are in equilibrium under optimal pH values for PNP functioning, an enzyme can accept the most suitable conformer for binding. On the whole, these considerations allow us to suggest the involvement of the C2′-ammonium cation in an interaction with phosphate anion as a likely reason for the very low substrate activity of dAdo2′NH2 (7) and dIno2′NH2 (8).
In the case of diastereomeric adenine deoxyfluoro nucleosides dAdo2′F (9) and dAdo2′F (10), the most unexpected finding is that both functionally competent analogues show very similar extremely low (3–4 orders of magnitude compared with that of Ado) substrate activity for the all investigated enzymes. The H and F atoms have similar van der Waals' radii and thus do not create any steric hindrances to the substrate binding in the catalytic center of the investigated enzymes. However, a fluorine atom, as an isostere for the hydrogen atom , differs dramatically from its isostere owing to the strongest electronegative character exerting diverse influences on electronic properties of proximal atoms as well as on the stereochemistry of the pentofuranose ring.
An ab initio analysis of dAdo2′F and dAdo2′F and comparison of the results obtained with those of X–ray crystallographic data for Ado and dAdo2′F (Table 4), as well as the data of the conformational NMR analysis of dAdo2′F  and dAdo2′F  unexpectedly revealed a rather broad stereochemical and electronic similarity of the 2′-deoxyfluoro-ribo- and -arabino-nucleosides that is in harmony with the activity data (Fig. 4). The only reasonable explanation of the observed very low substrate activity of dAdo2′F and dAdo2′F may be connected to the structural peculiarities of the productive complex giving rise to the glycoside bond cleavage. Bennet et al. suggested flattering of the pentofuranose ring of the substrate (the C4′–O4′–C1′–C2′ torsion angle strives to 0°) as one of the most important features of the reaction . As distinct from the Ado/ara-adenine (Ade) pair, the C2′ fluorine atom of dAdo2′F and dAdo2′F in all likelihood exerts a much higher energy barrier than that of Ado and dAdo for flattering of the pentofuranose ring, which is unexpectedly similar to and independent of the fluorine ribo- or arabino- configuration.
Table 4. Oligonucleotide primers and GenBank number of PNPs in this study
Here, we described the preparation of three PNPs from the thermophilic microorganisms D. geothermalis, G. thermoglucosidasius and A. pernix, aiming at the search for new efficient biocatalysts for the synthesis of modified nucleosides. The recombinant expression in E. coli resulted in a high yield of biologically active enzymes. The PNPs were found to be stable at their optimal temperatures (DgPNP: optimal activity at 55 °C retaining thereupon 80% for 8 h; GtPNP: optimal activity at 70 °C without losing activity for 8 h; ApMTAP: activity exponentially rising up to 99 °C). Natural substrates (Ado, Ino and their 2′-deoxy counterparts) are effectively phosphorolyzed by all three investigated enzymes (activity level within 50–500U·mg−1). The activity of the enzymes for dAdo2′NH2 and dIno2′NH2 was found to be approximately two orders of magnitude lower (within 0.01–3.25 U·mg−1) than their activity for natural substrates. ApMTAP showed the highest level of enzymatic activity for dAdo2′NH2 and dIno2′NH2. A similar further decrease in substrate activity was found for dAdo2′F and dAdo2′F (within 0.01–0.03 U·mg−1). DgPNP showed the highest activity for this type of substrate among the enzymes tested here. It should be stressed that the substrate activities of dAdo2′F and dAdo2′F are unexpectedly similar and do not depend on the fluorine atom configuration. Interestingly, cytidine was phosphorolyzed by the all studied PNPs with an efficiency similar to that of adenine 2′-deoxyfluoro nucleosides. Furthermore, uridine and thymidine were also phosphorolyzed by the studied PNPs (except for DgPNP, which cannot accept thymidine as substrate), but the activity was 6–30 times lower than that for cytidine.
It was found that the previously described thermostable pyrimidine nucleoside phosphorylase TtPyNP  and the present described DgPNP and ApMTAP can be applied for the transglycosylation reaction of adenine in the presence of dUrd2′F and dUrd2′F with the formation of the corresponding adenine nucleosides dAdo2′F and dAdo2′F at a yield of ~ 24%.
The investigation of the underlying enzymatic mechanisms of the prepared thermostable PNPs and their employment for the synthesis of purine nucleosides of biological interest are in progress.
Materials and methods
Bacterial strains, cloning, expression and purification
Escherichia coli TOP10 (Invitrogen, Carlsbad, CA, USA) and BL21 (Novagen, Darmstadt, Germany) strains were used for gene cloning and protein expression, respectively.
The genomic DNA of T. thermophilus HB27 and G. thermoglucosidasius 11955 were isolated as previously reported . The genomic DNA of D. geothermalis and A. pernix were isolated by Marco Casteleijn (Bioprocess Engineering Laboratory of the University of Oulu, Finland). The target gene sequences were amplified from genomic DNA by PCR using Pfu DNA polymerase (Fermentas, Vilnius, Lithuania). The used primer pairs are listed in Table 4.
The PNP encoding genes were cloned via BamHI/HindIII digestion (FastDigest restriction endonucleases; Fermentas) and subsequent ligation (T4 DNA Ligase; Roche, Mannheim, Germany) into a modified pCTUT7 vector . In brief, this vector includes an IPTG inducible lac promoter and a hexahistidine coding sequence connected to the 5′–end of the target gene.
DgPNP and ApMTAP were expressed in E. coli BL21 in EnPresso® medium (BioSilta, Oulu, Finland) by the use of the Enbase® technology  using Ultra Yield Flasks™ and AirOTop™ seals (both from BioSilta, Oulu, Finland). The expression of DgPNP and ApMTAP was induced after overnight cultivation at 37 or 30 °C, respectively, by the addition of 20 μm IPTG. GtPNP was expressed in E. coli BL21 cells in Terrific Broth (TB) medium  by the induction of 100 μm IPTG to the culture growing at 37 °C for 2.5 h. Cells were harvested by centrifugation (16 000 g, 5 min, 4 °C) after either 24 h (for EnPresso® medium) or 3.5 h (for TB medium) after induction, respectively.
Cell disruption and protein purification were carried out as described previously . Briefly, cells were broken by ultrasonic treatment (UP200S sonicator; Hielscher Ultrasonics GmbH, Teltow, Germany), then the lysate was heat-treated at 50 °C (DgPNP), 65 °C (GtPNP) or 85 °C (ApMTAP) for 15 min, and after centrifugation (16 000 g, 20 min, 4 °C) the supernatant containing the target protein was further purified by metal ion-affinity chromatography (Ni-NTA Superflow cartridge; Qiagen, Hilden, Germany) and the HiPrep 26/10 desalting column on an Äkta FPLC system (GE Healthcare, Munich, Germany).
The protein purity was checked by SDS/PAGE according to a standard protocol . To analyze whether the disulfide bridge exists in the enzymes, the purified enzymes (0.5 mg·mL−1) were treated with or without the reducing agent, i.e. 0.1 m dithiothreitol, in the incubation step of SDS/PAGE analysis. The molecular mass marker (Sm0431) was purchased from Fermentas. The purified protein concentration was determined by using a NanoDrop1000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, ME, USA)  and applying E280 nm1% = 6.3 (for DgPNP), or 8.8 (for GtPNP and ApMTAP), respectively. The absorption coefficients were theoretically calculated from the amino acid sequence (vector nti software; Invitrogen).
Enzyme activity assay
All the reactions were performed within 200–1000 μL scale in Eppendorf tubes, which were incubated in a thermomixer (Biozym, Hessisch Oldendorf, Germany) at 300 rpm. The standard activity assay (phosphorolysis) was carried out in potassium phosphate buffer (50 mm; pH 7.0) containing 1 mm purine nucleoside. After 2 min preheating at the corresponding temperature, a certain amount of purified enzyme (~ 0.5 μg·mL−1 for natural substrates; 0.1 mg·mL−1 for modified substrates) was added into the mixture and the reaction was stopped (addition of 1/2 vol of 10% trichloroacetic acid or methanol) after a defined time interval so that < 10% of the substrate was converted to the product. Under these conditions, the reaction rate was linear as a function of time and enzyme concentration. After centrifugation (20 000 g, 15 °C, 25 min) the samples were stored at −20 °C for following analyses. Negative controls (without enzyme addition) were performed in parallel. The reaction mixtures were analyzed by HPLC as described in . The substrate conversion was calculated from the HPLC chromatograms:
One unit (U) of enzyme activity was defined as the amount of the enzyme catalyzing the conversion of 1 μmol of substrate per minute under the respective assay conditions.
Temperature optimum and thermal stability measurements
In order to determine temperature optima of the enzymes, the reaction mixture (1 mm inosine in 50 mm potassium phosphate buffer, pH 7.0) was preheated for 2 min at different temperatures (from 30 to 99 °C), suitably diluted enzyme solution was added, and the reaction was stopped by the addition of trichloroacetic acid after 3 min.
Thermal stability was analyzed by incubating purified PNP aliquots (13–55 μg·mL−1) in 50 mm phosphate buffer (pH 7.0) on a PCR machine (Eppendorf, Hamburg, Germany) at the respective temperatures. After defined time intervals, tubes were withdrawn and cooled on ice. The residual activity of the incubated enzyme was determined at 50 °C for DgPNP, 60 °C for GtPNP and 80 °C for ApMTAP using inosine as a substrate under standard assay conditions. In order to determine the half-life, the residual activity data over time were fitted to the exponential decay equation a = a0·e(−λ ·t) (sigmaplot 11.0; Systat Software GmbH, Erkrath, Germany), where λ is the first-order deactivation coefficient, a is the residual activity and a0 is the original activity before treatment. Finally, the half-life was obtained by t1/2= ln 2·λ−1.
Kinetic parameters determination
Reactions were performed in triplicate at least to obtain the initial velocity at a fixed phosphate concentration (50 mm, pH 7.0) and the substrate concentration varied over a 20-fold range (spanning 0.25–5 times the Michaelis–Menten constant) . Reaction temperatures were: 55 °C for DgPNP, 70 °C for GtPNP and 80 °C for ApMTAP. Km and Vmax were determined by nonlinear regression based on the Michaelis–Menten equation (applied in sigmaplot 11.0).
Multiple sequence alignment
Protein identities were assessed with the protein basic local alignment tool (BLAST) of the NCBI web server . The multiple alignments were constructed using Expresso mode of T-Coffee server , which implemented an automated identification of suitable structural templates via a BLAST search against the PDB database . The corresponding UniProt Protein accession numbers are P0ABP8 (EcPNP), G0E416 (EaPNP), Q1IY92 (DgPNP), F8CYG4 (GtPNP), P77835 (GsPNP), Q72IR2 (TtPNPI), Q9YDC0 (ApMTAP) and P50389 (SsMTAP). The figure of alignment was created using Jalview2 .
Synthesis of fluorinated purine nucleosides
The syntheses of adenine 2′–deoxyfluoro-ribo- and -arabino-nucleosides (dAdo2′F and dAdo2′F, respectively) were performed in phosphate buffer (2 mm, pH 6.5) containing 2 mm dUrd2′F or dUrd2′F and 1 mm Ade. The enzyme loading was 0.1 mg·mL−1 for each nucleoside phosphorylase, viz., GtPyNP or TtPyNP and PNP. The reaction was conducted at defined temperatures for up to 24 h. The synthesized nucleosides, dAdo2′F and dAdo2′F were identified by comparison of their retention times and UV spectra with those of authentic samples by HPLC (Fig. S2). The calculated yields are based on the amount of adenine applied into the reaction.
Preparation of thermostable PyNP, isolated from the thermophilic microorganisms Geobacillus thermoglucosidasius (GtPyNP) and Thermus thermophilus (TtPyNP) was described in .
Nucleosides and bases
Ade, Ado and Ino were purchased from Carl Roth (Karlsruhe, Germany); hypoxanthine, dAdo, dIno, Cyt and dCyt were purchased from Sigma-Aldrich (Steinheim, Germany); dAdo2′F was purchased from Metkinen Chemistry (Kuusisto, Finland); dAdo2′NH2, dIno2′NH2 and dAdo2′F were kindly provided by Prof. Alex Azhayev (Metkinen Chemistry).
We sincerely thank Dr Marco G. Casteleijn (Bioprocess Engineering Laboratory, University of Oulu, Finland) for providing us with the genomic DNA of D. geothermalis and A. pernix. We further thank Prof. Alex Azhayev (Metkinen Chemistry) for the donation of the substrates of 1-(2-deoxy-2-fluoro-β-d–arabinofuranosyl)uracil, 2′–amino-2′–deoxyadenosine and 2′–amino-2′–deoxyinosine. We are thankful to the Alexander von Humboldt Foundation (Bonn–Bad–Godesberg, Germany) for computer and program facilities used here. We are also indebted to Prof. Steven E. Ealick (Baker Laboratory, Cornell University) and Dr William Bill Parker (Southern Research Institute, Birmingham) for kindly providing the Vmax data for E. coli PNP. This work is part of the Cluster of Excellence ‘Unifying Concepts in Catalysis’ coordinated by the Technische Universität Berlin. Financial support by the Deutsche Forschungsgemeinschaft (DFG) within the framework of the German Initiative for Excellence is gratefully acknowledged (EXC 314).