Screening the Thermotoga maritima genome for new wide-spectrum nucleoside and nucleotide kinases

Enzymes from thermophilic organisms are interesting biocatalysts for a wide variety of applications in organic synthesis, biotechnology, and molecular biology. Next to an increased stability at elevated temperatures, they were described to show a wider substrate spectrum than their mesophilic counterparts. To identify thermostable biocatalysts for the synthesis of nucleotide analogs, we performed a database search on the carbohydrate and nucleotide metabolism of Thermotoga maritima. After expression and purification of 13 enzyme candidates involved in nucleotide synthesis, these enzymes were screened for their substrate scope. We found that the synthesis of 2′-deoxynucleoside 5′-monophosphates (dNMPs) and uridine 5′-monophosphate from nucleosides was catalyzed by the already known wide-spectrum thymidine kinase and the ribokinase. In contrast, no NMP-forming activity was detected for adenosine-specific kinase, uridine kinase, or nucleotidase. The NMP kinases (NMPKs) and the pyruvate-phosphate-dikinase of T. maritima exhibited a rather specific substrate spectrum for the phosphorylation of NMPs, while pyruvate kinase, acetate kinase, and three of the NMPKs showed a broad substrate scope with (2′-deoxy)nucleoside 5′-diphosphates as substrates. Based on these promising results, TmNMPKs were applied in enzymatic cascade reactions for nucleoside 5′-triphosphate synthesis using four modified pyrimidine nucleosides and four purine NMPs as substrates, and we determined that base- and sugar-modified substrates were accepted. In summary, besides the already reported TmTK, NMPKs of T. maritima were identified to be interesting enzyme candidates for the enzymatic production of modified nucleotides.

Natural and modified nucleotides are widely used as drugs to treat cancer or viral infections (1), food additives (2), or as reagents for molecular biology applications (3). Recently, they have contributed significantly to the containment of the Corona pandemic as building blocks of mRNA vaccines (4) or important components of PCR-based diagnostic kits (5).
Nucleotide analogs are mainly synthesized chemically. In contrast to the chemical synthesis routes, biocatalytic approaches offer several advantages: reactions can be performed at milder conditions, the use of harsh solvents is reduced, and the use of protection groups can be avoided due to a high regioselectivity and stereoselectivity (6,7). For the biocatalytic nucleotide synthesis, enzymes of the de novo and salvage nucleotide pathway of various organisms were studied. These include, among others, phosphoribosyltransferases (PRTs), nucleoside kinases (NKs), nucleoside monophosphate kinases (NMPKs) and nucleoside diphosphate kinases (NDPKs). We recently developed a general approach for the one-pot synthesis of nucleoside triphosphates from nucleosides using nucleoside and nucleotide kinases (8). By coupling the phosphorylation reactions with an adenosine triphosphate (ATP) regeneration system, yields were significantly improved (8). Lately, cladribine triphosphate, a drug against leukemia and multiple sclerosis, was synthesized from 2chloroadenine and phosphoribosyl pyrophosphate by an adenine phosphoribosyltransferase, polyphosphate kinase, and ribonucleotide reductase (9). In this cascade reaction, polyphosphates served as a low-cost phosphate donor.
Inspired by the broad substrate scope of TmTK, we studied the enzymes involved in nucleotide synthesis in the hyperthermostable bacterium T. maritima with the aim of identifying additional wide spectrum enzymes for biocatalytic applications. In total, 13 enzymes of T. maritima were expressed, purified, and characterized. The substrate scope of active enzymes was analyzed using a set of 26 natural (deoxy) nucleosides or nucleotides as substrates. Finally, we demonstrated the synthesis of base-and sugar-modified nucleotides in cascade reactions with T. maritima NMPKs. Thus, this study presents a set of thermostable TmNMPKs suitable for the synthesis of natural and modified nucleotides.

Results and discussion
Identification of enzymes capable of phosphorylating nucleosides, NMPs, and NDPs in T. maritima To identify enzymes of T. maritima involved in the phosphorylation of nucleosides, we analyzed both the Kyoto Encyclopedia of Genes and Genomes (KEGG) database and the T. maritima genome (Figs. 1 and S1). Next to the known TmTK (18), in the genome an adenosine-specific kinase (TmAsK) and a three-domain uridine kinase (TmUK, annotated as NK in the genome sequence) were found. Adenosinespecific kinases are rarely studied, and the TmUK shows an unusual protein sequence as it consists of a NK superfamily domain and a threonyl-tRNA synthetase domain (Fig. S2A). Neither a deoxycytidine kinase, a (deoxy)adenosine kinase, an inosine kinase, deoxyguanosine kinase, nor a deoxynucleoside kinase was found in T. maritima (Figs. 1 and S1).
As other enzyme classes like nucleotidases (19), acid phosphatases (20), PfkB (15,16,21), or polyphosphate kinases (22) are also known for the phosphorylation of nucleosides or nucleotides, they were also included in this study. Indeed, the KEGG pathway for the purine nucleotide metabolism suggests that the 5 0 /3 0 -nucleotidase SurE (TmNuc, also classified as acid phosphatase) of T. maritima (23) might catalyze the formation of purine NMPs (Fig. 1) which led us to integrate this enzyme into our study. Additionally, the T. maritima genome encodes for six enzymes of the PfkB-family including a ribokinase (24). Ribokinase of T. maritima (TmRK) was described to show a high structural similarity to the NK of Methanocaldococcus janaschii (21) which accepted all natural ribonucleosides (15). Therefore, the TmRK was selected in this study to evaluate whether it can accept nucleosides as substrates. Other PfkBfamily enzymes were excluded from this study as they were shown before to be very specific for their sugar substrate (24).
deoxythymidylate kinase (TmTMPK), and a (deoxy)cytidylate kinase (Tm(d)CMPK) were identified (Fig. 1). Additionally, in the genome, a pyruvate-phosphate-dikinase (TmPPDK) was found. This enzyme was described before to catalyze the phosphorylation of nucleoside monophosphates directly to nucleoside triphosphates using phosphoenolpyruvate and pyrophosphate as substrates (25). The conversion of NDPs to nucleoside 5 0 -triphosphates (NTPs) usually is catalyzed by wide-spectrum NDPKs. Surprisingly, the genome of T. maritima does not encode for a NDPK. Therefore, we included enzymes capable of performing a similar function as NDPKs in this study. We focused on enzymes widely used for ATP regeneration as they are well known to catalyze NTP formation. Furthermore, it was shown that pyruvate kinase is a candidate for a phosphoenolpyruvate (PEP)-dependent NDP kinase activity in Escherichia coli and L. lactis due to a wide substrate scope (26). Studying the T. maritima genome, an acetate kinase (TmAcK) (27) and a pyruvate kinase (TmPK) (28) were identified. A polyphosphate kinase, however, was not identified in the genome of T. maritima (29).
To explore whether the substrate specificities correlate with gene annotation, the 13 described enzymes putatively involved in the phosphorylation of nucleosides, NMPs, or NDPs ( Fig. 1) were expressed, purified, and characterized in a next step.

Expression and thermal characterization of the T. maritima enzymes
After gene synthesis, all T. maritima enzymes were successfully expressed in E. coli I q cells and purified by affinity chromatography (Fig. S3). Protein yields were in the range of 0.8 to 15 mg g −1 cell pellet wet weight.
One advantage of thermostable enzymes is the possibility to use a heat treatment step to enhance protein purification. For TmTK and TmPK heat steps at 70 to 80 C for 20 to 30 min were already described (18,28). To confirm the thermostability of all expressed enzymes, we analyzed the heat denaturation of the enzymes purified by affinity chromatography in a thermal shift assay at pH values of 7, 8, and 9 in the absence and presence of 1,4-dithiothreitol (DTT). Indeed, no denaturation of the enzymes was detected up to 95 C and during an additional 20 min incubation step at 95 C (Fig. S4). Since the enzymes were shown to be hyperthermostable, they were re-expressed and purified by heat treatment and affinity chromatography.
Activity testing of the T. maritima enzymes at varying pH Having purified enzymes in hand, activity was determined at 37 C using their putative natural nucleoside or nucleotide substrates. Phosphotransferase activity was confirmed for all enzymes except for TmNuc, TmUK, and TmAsK. As nucleosides are not the natural substrates for nucleotidase, additional activity tests were performed with adenosine 5 0 -nucleotides, and phosphatase activity toward adenosine 5 0 -monophosphate (AMP) (23) and adenosine 5 0 -diphosphate (ADP) was shown (data not shown). Hence, TmNuc is not able to phosphorylate nucleosides under the applied conditions. Studies on adenosine-specific kinase (AsK) are rare. The first described enzyme was PAE2307, a protein from the hyperthermophilic archaeon Pyrobaculum aerophilum (30). Structural studies in combination with fluorescence spectroscopic analysis indicated a binding of adenosine and AMP to the adenosine-specific kinase like protein. Furthermore, a phosphorylation of the conserved histidine residue in the putative substrate-binding site was observed. Hence, it was suggested that the enzyme is a new class of adenosine kinase. Therefore, we evaluated if TmAsK can phosphorylate adenosine. However, no AMP-forming activity was observed (data not shown). As biochemical and structural studies on a homologous enzyme of Thermus thermophilus indicated that the enzymatic function of the "adenosine-specific kinase" family is rather an ADP cleavage to AMP (31), we evaluated if the TmAsK shows ADP phosphatase activity. However, no activity was observed using ADP as substrate for TmAsK (data not shown). Additionally, no phosphatase activity was detected with AMP and ATP. Since MgCl 2 was described to inhibit phosphatase activity of AsK (31), reactions were also performed without the addition of MgCl 2 and DTT. However, no activity was detected as well (data not shown). Thus, neither a phosphotransferase nor a phosphatase activity was shown for TmAsK under the conditions tested.
The TmUK was described before to be a result of interkingdom gene fusions, which is shared by Treponema pallidum and T. maritima (32). Using the full-length protein, no UK activity was observed in this study (Fig. S2). However, ATP phosphatase activity was observed in the presence of uridine (Urd). Therefore, we decided to study if truncated versions (TmUKs1-4, Fig. S2) only coding for the UK domain would show Urd phosphorylation activity. Truncated UK variants were cloned, heterologously expressed in E. coli and purified by affinity chromatography. While TmUKs1-3 were produced in substantial amount and with good purity, TmUKs4 could not be produced since only insoluble protein was obtained. Activity testing with Urd as substrate revealed no formation of uridine 5 0 -monophosphate (UMP) with all truncated UK variants. These results together with the observed ATP phosphatase activity in the presence of Urd fit well to the hypothesis that the three-domain UK of T. pallidum and T. maritima are involved in the autoregulation of translation (32).
For all enzymes showing phosphorylating activity toward nucleosides or nucleotides working pH ranges were determined, except for TmTK, TmPK, TmAcK, and TmPPDK. For the latter enzymes, the optimal reaction pH has already been described ( (18,25,27,28), Table 1). Working pH ranges differed depending on the enzyme classes. While TmPK and TmAcK preferred a lower reaction pH, TmTK, TmRK, and nearly all TmNMPKs (except for Tm(d)CMPK) showed a comparable activity in a pH range of 7 to 9 (Table 1 and Fig. S5).

Evaluation of the substrate scope
To validate the substrate scope of the enzymes phosphorylating either nucleosides or nucleotides, reactions with 26 natural nucleosides or nucleotides were performed and analyzed by either thin-layer chromatography (TLC) or high-performance liquid chromatography (HPLC) after an incubation time of 19 h at 37 C.
TmTK was described to be specific for thymidine and uridine derivatives (18). In our study, TmTK accepted all deoxynucleosides and uridine, with 2 0 -deoxyadenosine being the worst substrate ( Fig. 2A).
The substrate spectrum of TmRK with a set of sugars was studied before, and it was shown that the enzyme is very specific for D-ribose (24). Although TmRK seems to be a rather specific sugar kinase, it phosphorylates the natural nucleosides thymidine and uridine as well as deoxyguanosine, deoxyinosine, and deoxycytidine albeit with lower conversion (Fig. 2A). The preference for pyrimidine nucleosides is in good accordance with Pyrobaculum caldifontis PfkB, which preferred cytidine and uridine (16). Interestingly, while other PfkB members preferred ribonucleosides (15,16), TmRK seems to prefer deoxynucleosides.
As for some reactions using NDPs as substrates and NTPs as phosphate donor, NMP formation was detected, we validated whether NMP formation is the result of the reverse NMPK reaction or based on basal NDP cleaving activity of the enzyme preparation. Reactions were performed with NDP as substrate, but without ATP. Indeed, with some substrates, NMPKs (except for UMPK) were able to catalyze the reaction 2 NDP ↔ NMPþNTP. Accepted substrates correlated well with the NMP preference of the enzymes (Fig. 3). As an example, Tm(d)CMPK converted (d)CDP and UDP to the corresponding NMP and NTP.
TmPPDK only converted AMP and 2 0 -deoxyadenosine 5 0 -monophosphate to the respective NTPs with (d)ADP formed as by-products to a lower percentage (Fig. 2B). Additionally, it was shown that (d)ADP and (d)GDP were substrates of TmPPDK with comparable efficiency (Fig. 3). In good agreement with our results, TmPPDK was described before to be highly specific for AMP when tested with ribonucleotide monophosphates (25). However, little information is available on the substrate scope of PPDKs since mainly reactions with AMP as substrate were reported. For Clostridium symbiosum PPDK, even an inhibitory effect of other NMPs like guanosine 5 0monophosphate (GMP) and inosine 5 0 -monophosphate was shown (36).
For TmAMPK, NTP formation was observed for all AMP analogs (Fig. 4). Interestingly, although CMP and UMP are natural substrates, araUMP was accepted but not araCMP. Furthermore, neither the other UMP analogs nor araGMP were phosphorylated. The preference of AMP analogs is in good agreement with the data for other AMPKs (8,48,49). In contrast to the phosphorylation of deoxycytidine analogs like araCMP (48), the acceptance of araUMP was not described before for AMPKs.
In contrast to the other three TmNMPKs, all uridine analogs were accepted by TmTMPK and TmUMPK (Fig. 4). Notably, for 5-Br-UMP only the diphosphate was detected. This might be explained by the substrate scope observed for both NMPKs. Neither of the two enzymes was able to convert UDP to UTP (Fig. 3). Interestingly, for the other uridine analogs, the triphosphate was formed. Thus, TmAcK very likely accepted the modified uridine diphosphates, except for 5-Br-UDP, as substrates.
In summary, we demonstrated the ability of the T. maritima NMPKs to phosphorylate base-and sugar-modified substrates in enzymatic cascade reactions. Thus, they are interesting thermostable biocatalysts for the in vitro synthesis of modified 5 0 -nucleotides.

Conclusions
In this study, we investigated 13 enzymes of T. maritima for their ability to phosphorylate natural and modified nucleosides or nucleotides. Three of the enzymes, namely TmAsK, TmUK, and TmNuc, did not show phosphotransferase activity with ATP as phosphate donor under the conditions tested. Broad substrate spectra toward natural substrates were shown for TmTK, TmPK, and TmAcK, whereas TmPPDK and TmRK were more specific. While all five TmNMPKs had a narrow substrate scope toward NMPs, we detected a broad NDPK activity for three TmNMPKs. In addition, we demonstrated for the first time the phosphorylation of base-and sugar-modified NMPs by thermostable NMPKs in enzymatic cascade reactions. While we focused on modifications in the 2 0 -position of the ribosyl residue or the 5-position of the base, the synthesis of further modifications can be analyzed in the future. Of particular interest here could be 3 0 -or 4 0 -modified nucleotides or C-nucleotides. Thus, our work lays the foundation for the application of these thermostable enzymes for the biocatalytic production of nucleotide analogs.

General information
All chemicals and solvents were of analytical grade or higher and purchased from Sigma-Aldrich, Carl Roth, TCI Deutschland, Carbosynth, or VWR. Nucleosides and nucleotides were acquired from Alfa Aesar, Carl Roth, Sigma-Aldrich, Carbosynth Limited, and TCI. Stock solutions with concentrations of 10 and 50 mM were prepared in deionized water, and aliquots were stored at −20 C.

Identification of enzymes involved in the nucleotide salvage pathway of T. maritima
The KEGG pathway of T. maritima in comparison to the reference pathway was analyzed (purine metabolism: tma00230 and pyrimidine metabolism: tma00240). All enzymes involved in the nucleotide pathway were chosen for gene synthesis. Furthermore, the genome of T. maritima (accession number: NC_023151.1) was analyzed for proteins being able to phosphorylate either nucleosides, NMPs, or NDPs.

Expression and purification of the T. maritima enzymes
Genes of interest were synthesized by GeneArt (Thermo-Fisher). Enzymes were expressed in E. coli I q (NEB) using EnPresso B medium (Enpresso) based on the manufacturer's recommendations. Briefly, a 50 ml main culture was inoculated either from a fresh transformation plate or an LB preculture. Protein expression was induced by the addition of 200 μM IPTG. Cells were harvested by centrifugation and stored at −20 C until further use. For purification, the cell pellet was resuspended in binding buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, pH 8), including 0.1 mM phenylmethylsulfonyl fluoride, 1 mg ml −1 lysozyme, 1 mM MgCl 2 , and 0.6 mg ml −1 DNase. After incubation for 30 min at room temperature, cells were disrupted by sonification and prepurified by heat treatment for 20 min at 80 C. After centrifugation, the clear supernatant was loaded onto 0.5 to 2 ml Ni-NTA agarose (Jena Bioscience) columns. The columns were washed four times with 2 column volumes of washing buffer (binding buffer with 20 mM imidazole) and six times with 0.5 column volumes of elution buffer (binding buffer with 250 mM imidazole). The buffer of the elution fractions was exchanged by dialysis to 50 mM Tris-HCl pH 7.6, and enzymes were stored in 50% (v/v) glycerol at −20 C.
The purity of the enzyme preparations was analyzed by SDS-PAGE. Protein concentration was determined by A280 measurements at a ThermoFisher Scientific NanoDrop One using the molar extinction coefficient E1% predicted by Protparam.
Expression and purification of the D. melanogaster deoxynucleoside kinase The D. melanogaster deoxynucleoside kinase (DmdNK) plasmid was kindly provided by Prof. Birgitte Munch-Petersen (Roskilde University). The GST-fusion protein was produced as described before (8,56) in E. coli BL21. The purified and tag-free enzyme was stored in 50% glycerol, 1% Triton X-100 and 1 mM DTT at −20 C.

Thermal shift assay
The melting point was analyzed as described before (57). In a total volume of 25 μl, 0.2 g l −1 enzyme and 5× SYPRO orange were incubated in 70 mM Tris-HCl pH 7, pH 8, or pH 9 with 0 or 5 mM DTT. The sealed PCR plate was incubated at 50 C for 30 s and then heated in steps of 0.5 C per 5 s to a final temperature of 95 C in a Bio-Rad CFX96 Real-Time system. Fluorescence was measured at λex = 470 nm and λem = 570 nm. To estimate protein stability at elevated temperatures, proteins were further incubated at 95 C for 20 min. Fluorescence was measured again afterward. The melting point was determined as the inflection point of the fluorescence intensity over the temperature.
To analyze the nucleotide phosphatase activity of AsK and Nuc, the nucleotides AMP, ADP, and ATP were applied as substrates in the standard reaction mixture without an additional phosphate donor. AsK reactions were also performed without the addition of MgCl 2 and/or DTT.
Reactions were incubated for 19 h at 37 C and validated by TLC or HPLC.

Determination of the reaction pH
The influence of the reaction pH was evaluated by applying 70 mM Tris-HCl pH 7, 8, and 9. The following substrates were used: Urd: RK, 2 0 -deoxyadenosine 5 0 -monophosphate: AMPK, CMP: (d)CMPK, GMP: GMPK, TMP: TMPK, and UMP: UMPK. The reactions were preheated to 37 C and started by the addition of diluted enzyme stock solution. After 5 min, a 30 μl sample was stopped with 780 μl cold deionized water and analyzed by the luminescent assay.

Thin-layer chromatography (TLC)
After the reactions were stopped by freezing, 5 μl were spotted on a TLC plate. The running buffer consisted of deionized water, 25% ammonia, isopropanol and dioxan in a ratio of 1:4:5:0 (screening) or 4:3:2:4 (cascade reactions). The reaction compounds were identified by comparison to authentic standards under UV-light.

High-performance liquid chromatography
For HPLC analysis, samples were stopped with ice-cooled deionized water and freeze-thawed. After centrifugation (21,500g, 4 C, 15 min), samples were analyzed with a KNA-UER Azura or Agilent 1200 system using a Phenomenex (Aschaffenburg) reversed phase Kinetex EVO C18 column (250 × 4.6 mm) as previously described (8,58). Briefly, samples were measured at 260 nm and 34 C with a flow rate of 1 ml min −1 . Isocratic elution was performed using 80% A (0.1 M KH 2 PO 4 /K 2 HPO 4 , 8 mM tetrabutylammonium bisulfate, pH ca. 5.4) and 20% B (70% A, 30% MeOH) for 4 min followed by a gradient to 40% A and 60% B over 10 min, to 38% A and 62% B in 12 min, and back to 80% A and 20% B in 0.5 min. Initial conditions were maintained for 2.5 min.
Enzyme cascade reactions were analyzed using an adapted gradient: after an isocratic step of 80% A and 20% B for 4 min, 100% B was obtained in 16 min and maintained for 10 min.
Then, initial conditions were restored in 0.5 min and maintained for 9.5 min.
Typical retention times were as follows: AMP: 7.5 min, ADP: 14.6 min, ATP: 19.6 min. Substrates and products were identified based on their retention time and UV absorption spectra.
Conversions were calculated according to (1) as the ratio between the peak area of compound X (P X ) and the sum of all peak areas of the substrate and product(s) in the reaction (P total ).

Luminescent assay
The luminescent assay was performed as previously described (58). Briefly, the luminescence of 90 μl sample and 10 μl Kinase-Glo Max reagent (Promega) was measured by a Tecan infinite M1000 plate reader. The product formation was calculated from the remaining ATP in the reaction according to formula (2), where Lum R is the average luminescence signal of the reaction sample, and Lum NC is the average luminescence signal of the negative control.

Data availability
All data depicted visually in the items and in the main text as well as in the Supplementary Material are available on request from the corresponding author.
Acknowledgments-We are grateful to the BioNukleo GmbH members Julia Schollmeyer and Sarah Westarp for their support in the lab. We thank Birgitte Munch-Petersen for suppling Drosophila melanogaster deoxynucleoside kinase plasmid. We acknowledge support by the Open Access Publication Fund of TU Berlin.
Funding and additional information-K. F. W. was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), grant number 392246628. This Project is supported by the Federal Ministry for Economic Affairs and Climate Action (BMWK) on the basis of a decision by the German Bundestag, project number ZF4184202SK9.
Conflict of interest-A. K. is CEO of the biotech company Bio-Nukleo GmbH. M. F. and C. M. are scientists at BioNukleo GmbH, and P. N. is a member of the advisory board. The authors declare no conflict of interest.