The chemoenzymatic synthesis of clofarabine and related 2′-deoxyfluoroarabinosyl nucleosides: the electronic and stereochemical factors determining substrate recognition by E. coli nucleoside phosphorylases

Summary Two approaches to the synthesis of 2-chloro-9-(2-deoxy-2-fluoro-β-D-arabinofuranosyl)adenine (1, clofarabine) were studied. The first approach consists in the chemical synthesis of 2-deoxy-2-fluoro-α-D-arabinofuranose-1-phosphate (12a, 2FAra-1P) via three step conversion of 1,3,5-tri-O-benzoyl-2-deoxy-2-fluoro-α-D-arabinofuranose (9) into the phosphate 12a without isolation of intermediary products. Condensation of 12a with 2-chloroadenine catalyzed by the recombinant E. coli purine nucleoside phosphorylase (PNP) resulted in the formation of clofarabine in 67% yield. The reaction was also studied with a number of purine bases (2-aminoadenine and hypoxanthine), their analogues (5-aza-7-deazaguanine and 8-aza-7-deazahypoxanthine) and thymine. The results were compared with those of a similar reaction with α-D-arabinofuranose-1-phosphate (13a, Ara-1P). Differences of the reactivity of various substrates were analyzed by ab initio calculations in terms of the electronic structure (natural purines vs analogues) and stereochemical features (2FAra-1P vs Ara-1P) of the studied compounds to determine the substrate recognition by E. coli nucleoside phosphorylases. The second approach starts with the cascade one-pot enzymatic transformation of 2-deoxy-2-fluoro-D-arabinose into the phosphate 12a, followed by its condensation with 2-chloroadenine thereby affording clofarabine in ca. 48% yield in 24 h. The following recombinant E. coli enzymes catalyze the sequential conversion of 2-deoxy-2-fluoro-D-arabinose into the phosphate 12a: ribokinase (2-deoxy-2-fluoro-D-arabinofuranose-5-phosphate), phosphopentomutase (PPN; no 1,6-diphosphates of D-hexoses as co-factors required) (12a), and finally PNP. The substrate activities of D-arabinose, D-ribose and D-xylose in the similar cascade syntheses of the relevant 2-chloroadenine nucleosides were studied and compared with the activities of 2-deoxy-2-fluoro-D-arabinose. As expected, D-ribose exhibited the best substrate activity [90% yield of 2-chloroadenosine (8) in 30 min], D-arabinose reached an equilibrium at a concentration of ca. 1:1 of a starting base and the formed 2-chloro-9-(β-D-arabinofuranosyl)adenine (6) in 45 min, the formation of 2-chloro-9-(β-D-xylofuranosyl)adenine (7) proceeded very slowly attaining ca. 8% yield in 48 h.

Despite the detailed analysis and optimization of the clofarabine process [12,13] and the bulk production of the protected nucleoside glycon 9 by chemists at Eli Lilly and Co. [18] this chemical synthesis is connected with the use of great volumes of organic solvents and gives rise to the formation of the undesired α-anomer necessitating the chromatographic purification of the desired β-anomer, and finally affords clofarabine in a low yield.
The investigation of the chemistry of clofarabine and the related nucleosides led us to the conclusion that the search for novel more efficient "green" methods is of reasonable interest. In this context, the study by Yamada et al. on the chemical synthesis of 2-deoxy-2-fluoro-α-D-arabinofuranose-1-phosphate (12a; 2F Ara-1P) and its use in an enzymatic coupling with purine bases is of great interest [19]. 1-(2-Deoxy-2-fluoro-β-D-arabinofuranosyl)thymine (FMAU) [20] was used as a source of the phosphate 12a in the first enzymatic synthesis of purine 2-deoxy-2-fluoro-β-D-arabinofuranosyl nucleosides patented by Krenitsky and co-workers (Wellcome Res. Labs) [21]. The transfer of the pentofuranose residue of FMAU was realized by the concerted action of E. coli thymidine phosphorylase (TP) absorbed on DEAE cellulose for the intermediary generation of the phosphate 12a and E. coli purine nucleoside phosphorylase (PNP) for the condensation of the latter with bases. The two most challenging features of this method are the laborious chemical synthesis of FMAU and the very low substrate activity for E. coli TP. As a consequence, large amounts of the enzymes and a long reaction time are necessary to carry out the enzymatic transfer of the sugar moiety and to produce acceptable yields of desired products. Thus, synthesis of 2,6-diamino-9-(2deoxy-2-fluoro-β-D-arabinofuranosyl)purine (2a) from FMAU (1.2 mmol) was performed in the presence of very large amounts of E. coli TP (160000 IE) and PNP (290000 IE). Therefore, it seems to be reasonable to develop either an efficient chemical synthesis of the phosphate 12a or another route for its generation. Yamada et al. studied in detail the transformation of the fluoride 9 in a 3:1 mixture of the α-and β-anomers of the phosphates 12a and 12b. They employed this mixture in the condensation with adenine and 2,6-diaminopurine catalyzed by PNP from Bacillus stearothermophillus (synthesis of 2b: 500 units of PNP per 1 mmol adenine and ca. 1.2 mmol of the 12a,b (3.2:1) mixture; 2a: 870 units PNP per 1 mmol 2,6-diaminopurine and ca. 0.8 mmol of the 12a,b), and the desired nucleosides 9-(2-deoxy-2-fluoro-β-D-arabinofuranosyl)adenine (2b) and -2-aminoadenine (2a) were obtained in 29 and 39% yield, respectively [19].
Previously, we have applied the MacDonald method for the synthesis of α-D-arabinofuranose-1-phosphate (Ara-1P) and showed that it is a versatile substrate for the enzymatic synthesis of both purine and pyrimidine nucleosides [22,23]. In addition, we demonstrated that D-ribose and 2-deoxy-D-ribose can be converted to nucleosides in the cascade one-pot syn- thesis under the consecutive action of three E. coli enzymes, i.e., ribokinase (RK), phosphopentomutase (PPM) and nucleoside phosphorylases [24][25][26][27] (for a recent review, see [28]). In the present study, we described the synthesis of the phosphate 12a by the modified MacDonald method and investigated its substrate properties for the recombinant E. coli nucleoside phosphorylases. In particular, we focused on reactions with a number of purine bases (2-chloroadenine, 2-aminoadenine and hypoxanthine), their analogues (5-aza-7-deazaguanine and 8-aza-7-deazahypoxanthine) and thymine. The results were compared with those of a similar reaction with α-D-arabinofuranose-1-phosphate (13a; Ara-1P). Differences of the reactivity of various substrates were analyzed by ab initio calculations in terms of the electronic structure (natural purines vs analogues) and stereochemical features ( 2F Ara-1P vs Ara-1P) of the studied compounds to determine the substrate recognition by E. coli nucleoside phosphorylases. Moreover, the cascade one-pot synthesis of clofarabine and related arabino-, xylo-and ribonucleosides (6-8) of 2-chloroadenine starting from 2-deoxy-2fluoro-D-arabinose, D-arabinose, D-ribose and D-xylose was studied ( Figure 1).

Results and Discussion
α-D-Pentofuranose-1-phosphates 12a and 13a as substrates of the E. coli nucleoside phosphorylases. Recently, we have studied the synthesis of purine and pyrimidine β-D-arabinofuranosides by using α-D-arabinofuranose-1-phosphate (Ara-1P) as the glycosylating agent and the respective recombinant E. coli nucleoside phosphorylases as biocatalysts [22,23]. It was thus shown that Ara-1P is a universal glycosylating substrate for the synthesis of both purine and pyrimidine nucleosides. In turn, the MacDonald method, originally proposed by the author for the synthesis of D-hexopyranose-1-phosphates [29][30][31], was used for obtaining Ara-1P. The method comprised the treatment of 1,2,3,5-tetra-O-acetyl-D-arabinofuranose or 1-O-acetyl-2,3,5tri-O-benzoyl-D-arabinofuranose with anhydrous phosphoric acid followed by the deprotection with LiOH [22]. A similar approach was studied in the present work for the synthesis of purine and pyrimidine 2-deoxy-2-fluoro-β-D-arabinofuranosides. It was, however, found that the 1-O-benzoate 9 (Scheme 1) was stable under MacDonald reaction conditions in a broad range of temperatures and all attempts to replace the 1-O-benzoate group with a phosphate residue failed. In order to overcome the inertness of the 1-O-benzoate group, we studied the reaction of the 1-O-benzoate 9 with acetyl bromide in anhydrous phosphoric acid at different temperatures and found that under 50 °C for 5 h the starting 9 is transformed into the 1-bromide 10 according to TLC analysis. The reaction mixture containing the intermediate 10 was dissolved in dioxane, cooled to 0 °C, treated with n-Bu 3 N, and stored at room temperature for 12-18 h monitoring the reaction progress by TLC. Then, the reaction mixture was diluted with an equal volume of water, powdered LiOH was gradually added under stirring (pH of 7-8), and the mixture was stirred at room temperature for 1 h. The formed lithium phosphate was filtered off, the water phase was adjusted to pH 11.0 by LiOH (1.0 N aqueous solution), trin-butylamine was extracted by chloroform, and the clear aqueous solution was separated and stored overnight. The formation of the deprotected phosphates 12a,b was monitored by TLC. The pH of the reaction mixture was adjusted to 7.5 by HCl (1.0 N), the mixture was concentrated in vacuo to ca. 10 mL, MeOH and acetone were added, and the mixture was stored at 4 °C for 48 h. The precipitate was removed by centrifugation, washed successively with MeOH, acetone, ether, and dried in vacuo over P 2 O 5 to afford the phosphates 12a,b (the α,β ratio was ca. 1:1 according to the 1 H NMR) as a white powder in 42-50 % yield.
The structures of the α-and β-anomers 12a and 12b were verified by a careful analysis of the 1 H and 13 C NMR spectra {[ 1 H, 1 H] and [ 1 H, 13 C] 2D COSY and NOESY spectra} as well as by the 19 F and 31 P NMR and by comparison with published 1 H and 13 C NMR data for the α-anomer 12a [19] and the closely related D-arabinofuranose-1-phosphates [22,32] (Tables S1 and S2 in Supporting Information File 1). It is noteworthy that 1,2,3,5-tetra-O-acetyl-D-arabinofuranose and 1-O-acetyl-2,3,5-tri-O-benzoyl-D-arabinofuranose were shown to transform in a mixture of 1-phosphates of β-D-arabinopyranose (13b; Ara Pyr -1P; major isomer) and α-D-arabinofuranose (13a; Ara Fur -1P; minor isomer) under MacDonald reaction conditions ( Figure 2) [22]. On the contrary, the α-and β-anomers 12a and 12b are the main products in a ratio of ca In addition to the NMR data, the stereochemistry of both anomers as dilithium salts was analyzed by the ab initio calculations (3-21G; total charge equal to zero; Polak-Ribiere conjugate gradient) employing AMBER force field geometry optimization as a starting approximation. Two structures with the C-3-exo conformation of the pentofuranose ring characterized by the closely related total energy (E T ) values were found for   Table S3 in Supporting Information File 1). Taken together, these data suggest the population of two-three conformers and the experimental couplings are the average values of the corresponding vicinal atoms. It is noteworthy that the vicinal coupling constants H-1/F (9.95 Hz), H-3/F (24.39 Hz) and H-1/P (6.5 Hz) taken from the  A similar analysis of the stereochemistry of the α-D-arabinofuranose-1-phosphate (Ara-1P) revealed a very close resemblance with the deoxyfluoro phosphates 12a (Table S3 in Supporting Information File 1).
As distinct from the α-anomer 12a, the calculated structure of the β-anomer 12b (C-2-exo conformer; E T = −732303.4 kcal/ mol) showed satisfactory correspondence with the NMR data pointing to its conformational rigidity. The 4 J H2,P of 1.28 Hz observed in the 1 H NMR spectrum results from the W-like arrangement of the interacting nuclei and is present in the calculated stereochemistry of the β-anomer 12b ( Figure S1 in Supporting Information File 1). It should be stressed that the 19 F and 31 P NMR data also give strong support to the assignments of the 1 H and 13 C resonances and conformational peculiarities of both anomers. Notably, the difference between the average value of E T for the conformationally mobile anomer 12a and E T of the "rigid" anomer 12b is 2.7 kcal/mol pointing to the similarity of these anomers from a viewpoint of their total energy, which is consistent with the formation in a ca. 1:1 ratio in the synthesis (Table S3 in Supporting Information File 1).
The α-phosphate 12a (as a mixture of two anomers 12a,b, i.e., the real concentration of the substrate 12a is only half of the indicated concentration) was tested as a substrate of the recombinant E. coli nucleoside phosphorylases [23] in the synthesis of purine and pyrimidine nucleosides (Table 1; Figure 4). It was found that under similar reaction conditions (1.56 units PNP per Table 2: The geometry optimization of the two main tautomeric structures of heterocyclic bases studied by means of the restricted Polak-Ribiere ab initio method (6-31G** level; basic set of parameters; HyperChem 8.1).
The optimization of the clofarabine synthesis showed that at a ratio of 3.3:1 (mol) phosphate to 2-chloroadenine and the use of 1.57 units PNP per 1 µmol of 2-chloroadenine the conversion of the base into the nucleoside at 50 °C was 85% in 24 h, then reduced to 80% and the equilibrium base + 12a nucleoside + inorganic phosphate (P i ) remained constant up to 168 h of reaction time ( Figure 5). These data point to a high stability of the α-phosphate 12a under the studied reaction conditions. On the other hand, the extremely poor solubility of 2-chloroadenine in water (ca. 70 mg in 1.0 L at 50 °C) did not allow to scale up the synthesis, and based on an experiment on a milligram scale clofarabine was isolated from the heterogeneous reaction mixture in 42% yield. The slower conversion of hypoxanthine to the corresponding nucleoside 3a is apparently caused by the electronic structure of the base which differs from those of 2-chloroadenine and 2,6diaminopurine. To investigate this hypothesis we analyzed the electronic structure of 2-chloroadenine, 2-aminoadenine and hypoxanthine by means of the restricted Polak-Ribiere ab initio method (6-31G** level; basic set of parameters; HyperChem 8.10) ( Table 2; see also Table S4 in Supporting Information File 1). According to this analysis, the tautomers of 2-chloroadenine and 2-aminoadenine (DAP) with an sp 3 -hybridized N-9 nitrogen atoms are more populated by 10-11 kcal/mol compared to those with sp 3 -hybridized N-7 nitrogen atoms, whereas both similar tautomers of hypoxanthine with sp 3hybridized N-9 or N-7 atoms are thermodynamically equivalent.
The very important role of the β-carboxy group of Asp204 of the E. coli PNP catalytic site was discussed earlier (see, e.g., [26,28]). It is based on the correct base positioning at the E. coli PNP catalytic site by the protonation of the sp 2 -hybridized nitrogen atom of the imidazole ring, which leads to the enhancement of the nucleophilicity of the second nitrogen atom (activation of base). In the case of 2-chloroadenine and DAP, the N-9 and N-7 sp 2 -hybridized atoms can, in principle, be protonated by the β-carboxy function of Asp204 of the catalytic Scheme 2: Suggested mechanism of purine nucleoside synthesis catalyzed by E. coli purine nucleoside phosphorylase.
site with similar efficiency. However, the equilibrium of two main tautomers is shifted to the N-7 sp 2 -hybridized tautomer, which will be protonated and gives rise to (i) the hydrogen bond connected base, and (ii) the N-9 sp 2 -hybridized tautomers 14 and 15 in the transition state, which are reacting with the electrophilic C-1 atom of the 1-O-phosphate 12a (Scheme 2).
Both main tautomers of hypoxanthine are thermodynamically equivalent and two types of binding and activation of the base such as 14, 15 and 16 might be expected. Indeed, we have earlier shown that the trans-2′-deoxyribosylation of N 2 -acetylguanine, the electronic structure of which is very similar to that of hypoxanthine (Table 2), resulted in the quick formation of the N-7 nucleoside as a kinetic product, which is gradually rearranged into the thermodynamically more stable N-9 nucleoside [33]. We hypothesize that the formation of N 2 -acetyl-7-(2deoxy-β-D-ribofuranosyl)guanine is realized through a transition state structure such as 16 (Scheme 2). Both N-7 and N-9 regioisomeric nucleosides were isolated in pure state and the isomerization of the former into N 2 -acetyl-2′-deoxyguanosine in phosphate buffer in the presence of E. coli PNP was proved. However, in the case of the similar trans-2-deoxyribosylation of hypoxanthine and the condensation of hypoxanthine with the phosphate 12a the formation of the corresponding N-7 nucleo-sides was not observed by using HPLC analysis of the reaction mixtures. The formation of 2′-deoxyinosine proceeded very quickly, the synthesis of 9-(2-deoxy-2-fluoro-β-D-arabinofuranosyl)hypoxathine (3a) proceeded slowly compared with the synthesis of nucleosides 1 and 2a under similar reaction conditions. It appears to be reasonable that the stereoelectronic interaction of the C-6 carbonyl group with arabino-fluorine atom and/or the α-1-O-phosphate function hinders the nucleophilic attack on the C-1 electrophilic carbon atom of 12a. In addition, the stereochemistry of α-1-O-phosphates of 2-deoxy-D-ribofuranose and 2-deoxy-2-fluoro-D-arabinofuranose differs (vide supra), which may contribute to the interaction with the C-6 carbonyl group of the base hindering the formation of a productive transition state of the base and 12a.
The good substrate properties of the phosphate 12a prompted us to test it in the enzymatic synthesis of base-modified nucleosides 9-(2-deoxy-2-fluoro-β-D-arabinofuranosyl)-5-aza-7deazaguanine (4a; purine numbering) and 9-(2-deoxy-2-fluoroβ-D-arabinofuranosyl)-8-aza-7-deazahypoxanthine (5a; purine numbering) and compare it with the synthesis of the related arabinosides 4b and 5b under analogous reaction conditions. Note that 5-aza-7-deazaguanine (17) was earlier shown to be a good substrate in the enzymatic synthesis of its N 9 -2′-deoxy-β-D-ribofuranoside by using 2-deoxy-α-D-ribofuranose-1-phosphate as a co-substrate and PNP from bovine spleen as a biocatalyst [34]. Recently, an efficient transformation of 8-aza-7deazahypoxanthine (allopurinol) into its N 9 -2′-deoxy-β-D-ribofuranoside has also been demonstrated in the transglycosylation reaction and in the cascade one-pot synthesis by using E. coli ribokinase (RK), phosphopentomutase (PPM) and nucleoside phosphorylases as biocatalysts [26]. It was found that the irreversible conversion of 5-aza-7-deazaguanine (17) into its arabinoside 4b proceeds smoothly achieving 50% yield after 24 h and remains at this point for over two weeks. The formation of the 2′-deoxyfluoro counterpart 4a was followed upon a similar path albeit with lower efficiency. The rate of the allopurinol condensation with Ara-1P and 2F Ara-1P was very slow, especially in the case of the latter phosphate ( Figure 6). Remarkably, the high substrate activity of 5-aza-7-deazaguanine and the very low activity of allopurinol is well-correlated with the electronic structure of these bases. In particular, there is a strong correlation with the partial negative charges of the corresponding N-9 sp 2 -hybridized nitrogen atoms (Table 2). It was found that nucleosides 4b and 5b are not substrates for E. coli PNP and moderately inhibit the synthetic reaction (work in progress). Data for the conversion of hypoxanthine and guanine in the corresponding arabinosides 3b and 3c under similar reaction conditions are included in this work to allow for a comparison. It is noteworthy that the guanine + Ara-1P ara-G + inorganic phosphate (P i ) equilibrium is strongly shifted toward the product formation under used reaction conditions, whereas in the case of hypoxanthine a similar equilibrium is established at ca. 65% concentration of 9-(β-D-arabinofuranosyl)hypoxanthine (3b; ara-Hyp). The high substrate activity of 5-aza-7-deazaguanine in the enzymatic synthesis of its arabinoside 4b catalyzed by E. coli PNP is rather unexpected. Owing to the absence of the N-7 nitrogen atom the C-6 carbonyl function of 5-aza-7-deazaguanine represents the only possibility for the correct binding at the catalytic site of PNP. On the other hand, the 17a 17b equilibrium, which is slightly biased toward the N-9 sp 3 tautomer 17a (Figure 7) and the high partial charge of the nucleophilic sp 2 hybridized N-9 nitrogen atom of the tautomer 17b may be responsible for the high substrate activity of this base towards E. coli PNP. The chemical synthesis of 4a has recently been published [35]. The condensation of a N-2 isobutyryl-protected base with bromide 10 led to the formation of a ca. 1:1 mixture of the α-and β-anomers, but the deprotection and separation of these into individual compounds represented a serious problem and both nucleosides were obtained in very low yields. These data motivated us to study the enzymatic synthesis of the β-anomer 4a. The reaction of 5-aza-7-deazaguanine (17) and the phosphate 12a in water solution (pH 7.0) at 52 °C in the presence of E. coli PNP (1.58 units per 1 µmol of base) proceeded very slowly. The HPLC analysis of the reaction mixture showed that after 16 days ca. 35% of the starting base was transformed into nucleoside. The reaction mixture was concentrated and the residue placed on the RP silica gel column that was eluted first with water to recover the base (62%) and then with water/ ethanol (99:1, v/v) to afford the desired nucleoside 4a in 81% yield (calculated for the consumed base).
From a chemical viewpoint, the formation of the glycosidic bond results from the nucleophilic attack of the sp 2 -hybridized N-9 nitrogen atom of a heterocyclic base on the electrophilic C-1 carbon atom of α-D-pentofuranose-1-phosphates. The observed differences in the rates of enzymatic reactions of the phosphate 12a and its arabino-counterpart 13a may be partly explained by differences in the electrophilicity of the C-1 carbon atoms of the aforementioned phosphates, which can be assessed by comparing the partial positive charges of these carbon atoms. Indeed, the ab initio calculations yielded the  following average values for the three most populated conformers of 2F Ara-1P and Ara-1P 0.405 e and 0.434 e, respectively (Table 3). Data for the 1-phosphates of α-D-ribofuranose and 2-deoxy-α-D-ribofuranose are included in Table 3 to allow for a comparison.
A comparison of the partial positive charges of the C-1 carbon atoms of 2F Ara-1P and Ara-1P pointed to a greater reactivity of the latter and it was experimentally validated in the synthesis of base-modified nucleosides 4a,b and 5a,b (vide supra). However, from our experience in the synthesis of various nucleosides, we know that the enzymatic synthesis of purine and pyrimidine 2′-deoxy-β-D-ribofuranosides is usually most effective, though it does not follow from the data of Table 3 for dRibo-1P. Obviously, the effectiveness of nucleoside synthesis is controlled by several factors, and the role of each of these factors depends on the substrate structure.
The most unexpected finding is that E. coli UP and TP are not able to catalyze the condensation of uracil and thymine with the phosphate 12a (cf. [21]), whereas Ara-1P was shown to be a good substrate for the recombinant E. coli UP and PNP nucleoside phosphorylases [22]. It is noteworthy that Ara-1P synthesized by Wright and Khorana was found to be inactive as a substrate for E. coli TP [36]. One can reasonable assume that the phosphate 12a is incapable to adopt a pentofuranose-ring conformation at the catalytic sites of E. coli UP and TP, which would be compatible with the coupling to pyrimidine bases. There are convincing arguments that the transition states of the phosphorolysis of pyrimidine [37] and purine [38] nucleosides are characterized by one common feature, i.e., the strongly unfavorable pentofuranose ring flattening which is accompanied by the population of an unusual C-4′-endo conformation. It is very likely that the reaction of the synthesis of both types of nucleosides is also realized through a flattened C-4-endo conformation of a furanose sugar ring of the relevant 1-O-phosphates in the transition state. The high conformational mobility of the pentofuranose rings of Ara-1P and 2F Ara-1P as derived from the NMR data is an important prerequisite for the requried flattening. However, such a flattening results in a virtual eclipse arrangement of the substituents at the C-3-C-2-C-1 fragment of both phosphates, and the energy of the eclipse H-3-C-2F(arabino)-H-1 repulsion is higher than the energy of the eclipse H-3-C-2OH(arabino)-H-1 [39][40][41], i.e., the C-2 fluorine atom of 2F Ara-1P exerts a much higher energy barrier for a flattening compared to that of the C-2 hydroxy group of Ara-1P. Moreover, the absence of substrate activity of 2F Ara-1P for E. coli UP and TP points to the different requirements of E. coli nucleoside phosphorylases to the degree of the α-D-pentofuranose-1-phosphate flattening at the catalytic centers of UP and TP, on the one hand, and PNP, on the other hand. In addition, it appears to be obvious that TP imposes the most stringent requirement for the spatial organization of the α-D-pentofuranose-1-phosphates in the transition state of the synthetic reaction.

One-pot enzymatic transformation of D-pentoses into nucleosides.
In the next series of experiments, we studied the cascade one-pot synthesis of 2-chloroadenine nucleosides in order to estimate the possible applications and limitations of this method [24,27] and to compare the results with those discussed above by using the α-D-pentofuranose-1-phosphates as glycosylating agents. The cascade synthesis involves a sequential conversion of D-pentoses into their 5-monophosphates catalyzed by the recombinant E. coli ribokinase (RK; ATP co-factor) [27], stereospecific isomerization of the 5-phosphates into α-D-pentofuranose-1-phosphates catalyzed by the recombinant E. coli phosphopentomutase (PPM) [24], and the condensation of the latter with 2-chloroadenine catalyzed by PNP, which gives rise to the formation of the desired nucleoside. It should be stressed that 1,6-diphosphates of D-hexoses are not necessary for the transformation of D-2F Ara-5P into α-D-2F Ara-1P as we have showed in our previous works [24][25][26]28]. This transformation was partly optimized by using variable concentrations of ATP, D-pentoses and biocatalysts. The results are shown in Table 4 and Figure 8.   Table 4 and Figure 8.

Conclusion
In summary, the modified MacDonald′ method was employed for the chemical synthesis of 2-deoxy-2-fluoro-α-D-arabinofuranose-1-phosphate (12a), and its substrate properties for the recombinant E. coli nucleoside phosphorylases were studied. The formation of the α-and β-anomers 12a and 12b in a ca. 1:1 ratio as the main products was observed in all preparations. The stereochemistry of the phosphate 12a was investigated by the integrity of NMR methods and ab initio calculations. Results point to the high C-3-exo C-3-endo/C-2-exo ( 3 T 2 ) conformational mobility of the phosphate 12a which resembles that of α-D-arabinofuranose-1-phosphate (Ara-1P) [22]. It is shown that the phosphate 12a (used in a mixture with 12b) is a good substrate for the recombinant E. coli PNP, and the formation of clofarabine (1) and related 2′-deoxyfluoroarabino nucleosides was studied. Analysis of the results suggest that the glycosidic bond formation is strongly dependent on (i) the N-9 base nucleophilicity and the C-1 pentofuranose-1-phosphate elec-trophilicity, and (ii) the ability of the enzyme to force the α-Dpentofuranose-1-phosphate from energetically favorable conformation(s) to the closely planar energetically unfavorable conformation of the furanose ring in the transition state. On the other hand, the requirements of E. coli nucleoside phosphorylases to the degree of the flattening of the furanose ring depend on the type of enzyme and appear to be the most strong in the case of TP, then in the case of UP, and least restrictive in PNP.
The cascade one-pot transformation of 2-chloroadenine into nucleosides by using 2-deoxy-2-fluoro-D-arabinose, D-arabinose, D-xylose and D-ribose as the pentofuranose starting compounds was studied under individually optimized reaction conditions. As expected, D-ribose showed the best substrate activity with 90% yield of 2-chloroadenosine (8) in 30 min. This points to the very efficient three consecutive transformations in D-ribose-5-phosphate catalyzed by RK with ATP as a co-factor, the latter in α-D-ribofuranose-1-phosphate (catalyzed by PPM without any 1,6-diphosphates of D-hexoses as co-factors, (cf. [42]), which finally condensed with 2-chloroadenine catalyzed by PNP. In a similar way, clofarabine (1) and its arabino-counterpart 6 were synthesized in ca. 50% yields in ca. 30 h and 45 min, respectively, whereas the yield of the xylonucleoside 7 did not exceed 8% in 48 h.
The extremely low solubility of 2-chloroadenine is a major challenge in the enzymatic synthesis of its nucleosides. One of the possible solutions to this problem is the use of highly soluble 2-chloroadenosine [43,44] as an in situ donor of the base in reactions catalyzed by PNP [44,45], the results of which are currently investigated in our laboratory.