Radical Dehalogenation and Purine Nucleoside Phosphorylase E. Coli: How Does an Admixture of 20,30-Anhydroinosine Hinder 2-ﬂuoro-Cordycepin Synthesis

Article Radical Dehalogenation and Purine Nucleoside Phosphorylase E. coli: How Does an Admixture of 20,30-Anhydroinosine Hinder 2-ﬂuoro-cordycepin Synthesis

Alexey L. Kayushin 1, Julia A. Tokunova 1, Ilja V. Fateev 1 , Alexandra O. Arnautova 1, Maria Ya. Berzina 1, Alexander S. Paramonov 1 , Olga I. Lutonina 1, Elena V. Dorofeeva 1, Konstantin V. Antonov 1, Roman S. Esipov 1, Igor A. Mikhailopulo 2 , Anatoly I. Miroshnikov 1 and Irina D. Konstantinova 1,*

1 Shemyakin and Ovchinnikov Institute of Bioorganic Chemistry RAS, Miklukho-Maklaya 16/10, 117997 GSP, B-437 Moscow, Russia; [email protected] (A.L.K.); [email protected] (J.A.T.); [email protected] (I.V.F.); [email protected] (A.O.A.); [email protected] (M.Y.B.); [email protected] (A.S.P.); [email protected] (O.I.L.); [email protected] (E.V.D.); [email protected] (K.V.A.); [email protected] (R.S.E.); [email protected] (A.I.M.) 2 Institute of Bioorganic Chemistry, National Academy of Sciences, Acad. Kuprevicha 5/2, 220141 Minsk, Belarus; [email protected] * Correspondence: [email protected]; Tel.: +7-905-791-17-19 Abstract: During the preparative synthesis of 2-fluorocordycepin from 2-fluoroadenosine and Citation: Kayushin, A.L.; Tokunova, 30-deoxyinosine catalyzed by E. coli purine nucleoside phosphorylase, a slowdown of the reac- J.A.; Fateev, I.V.; Arnautova, A.O.; tion and decrease of yield down to 5% were encountered. An unknown nucleoside was found Berzina, M.Y.; Paramonov, A.S.; Lutonina, O.I.; Dorofeeva, E.V.; in the reaction mixture and its structure was established. This nucleoside is formed from the 0 0 0 Antonov, K.V.; Esipov, R.S.; et al. admixture of 2 ,3 -anhydroinosine, a byproduct in the preparation of 3- deoxyinosine. More- 0 0 0 0 0 Radical Dehalogenation and Purine over, 2 ,3 -anhydroinosine forms during radical dehalogenation of 9-(2 ,5 -di-O-acetyl-3 -bromo- 0 0 Nucleoside Phosphorylase E. coli: -3 -deoxyxylofuranosyl)hypoxanthine, a precursor of 3 -deoxyinosine in chemical synthesis. The 0 0 How does an Admixture of products of 2 ,3 -anhydroinosine hydrolysis inhibit the formation of 1-phospho-3-deoxyribose during 20,30-Anhydroinosine Hinder the synthesis of 2-fluorocordycepin. The progress of 20,30-anhydroinosine hydrolysis was investi- 2-fluoro-cordycepin Synthesis. gated. The reactions were performed in D2O instead of H2O; this allowed accumulating intermediate Biomolecules 2021, 11, 539. https:// substances in sufficient quantities. Two intermediates were isolated and their structures were con- doi.org/10.3390/biom11040539 0 0 firmed by mass and NMR spectroscopy. A mechanism of 2 ,3 -anhydroinosine hydrolysis in D2O is fully determined for the first time. Academic Editors: Anthony J. Berdis and Jesús Fernandez Lucas Keywords: purine nucleoside phosphorylase; biocatalyze; 30-deoxyinosine; 2-fluorocordycepin; deuterium oxide Received: 14 January 2021 Accepted: 12 March 2021 Published: 7 April 2021

1. Introduction Publisher’s Note: MDPI stays neutral 0 with regard to jurisdictional claims in Among the modiﬁed nucleosides, 3 -deoxyribonucleosides remain very promising 0 published maps and institutional afﬁl- objects to study. The series of their research began with cordycepin—3 -deoxyadenosine iations. (1)—the natural antimetabolite of adenosine (2) (Figure1). Cordycepin was isolated from extracts of Cordyceps millitaris tissues in 1950 year [1,2]. This drug is an antioxidant and has a pronounced immunomodulatory effect. In the medicine of Southeast Asia (China, Japan, and Korea), Cordyceps extract has been used since ancient times as a remedy for tumors,

Copyright: © 2021 by the authors. viral and bacterial infections [3–7]. Currently, cordycepin is produced commercially in Licensee MDPI, Basel, Switzerland. China using the Chinensa militaris [4,8]. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Biomolecules 2021, 11, 539. https://doi.org/10.3390/biom11040539 https://www.mdpi.com/journal/biomolecules Biomolecules 2021, 11, x FOR PEER REVIEW 2 of 17

The introduction of a chlorine or fluorine atom into a heterocycle of a nucleoside cardinally changes the biological properties of the molecule: pharmacokinetic prop- Biomolecules 2021, 11, 539 2 of 17 erties, tissue distribution, metabolic pathway, and rate, as well as pharmacodynamics and cytotoxicity [11–14].

Figure 1. Adenosine, cordycepin, and their synthetic analogs. Figure 1. Adenosine, cordycepin, and their synthetic analogs. Synthetic C2-substituted analogs of cordycepin (3, 4) have a very interesting biological activityRecent that may studies be applied have inshown medicine. that One2-fluorocordycepin of the best-known ( analogs3) is more of cordycepin efficient than is 2- fluorocordycepincordycepin, in the (2-F-Cord, treatment3) (Figure of human1)[ 9 ,10African]. This trypanosomiasis compound is currently [9,10]. beingHence, actively creat- studieding effective and tested ways as to an synthesize antitumor and2-fluorocordycepin antiviral drug. The (3) introduction is a good direction of a chlorine of re- or fluorinesearch. atom into a heterocycle of a nucleoside cardinally changes the biological properties of theIt molecule: is known pharmacokinetic that 3′-deoxynucleosides properties, tissue can distribution,be synthesized metabolic using pathway, nucleoside and rate,phosphorylases as well as pharmacodynamics (transglycosylation and reaction). cytotoxicity In [ 11our–14 laboratory,]. an efficient method for obtainingRecent studies of purine have nucleoside shown that phosphorylase 2-fluorocordycepin was (3developed) is more efficient [15], and than we cordy- used cepin,this enzyme in the treatment in transglycosylation of human African reactions. trypanosomiasis [9,10]. Hence, creating effective waysHowever, to synthesize the 2-fluorocordycepin problem of a possible (3) is a goodsour directionce of 3-deoxyribose of research. arises [16,17]. Therefore,It is known it is necessary that 30-deoxynucleosides to determine the can optimal be synthesized 3-deoxyribose using nucleoside donor among phospho- pu- rylasesrine nucleosides. (transglycosylation Recent industrial reaction). In developments our laboratory, have an efficient made cordycepin method for obtainingavailable offor purine use as nucleoside a source of phosphorylase 3-deoxyribose. was However, developed 3′-deoxyinosine [15], and we used (3′-dIno, this enzyme9) is likely in transglycosylationto perform better reactions.as a substrate in the synthesis of new nucleosides. However,To test which the problem compound of a possiblecould be source a better of 3-deoxyribosedonor of 3-deoxyribose, arises [16,17 ].we Therefore, recently itimplemented is necessary to a determinemodified chemical the optimal method 3-deoxyribose for 3′-deoxyinosine donor among synthesis purine nucleosides. from com- Recentmercially industrial available developments inosine (5) have (Scheme made 1) cordycepin [18]. The availabledetails of for this use synthesis as a source will of be 3- 0 0 deoxyribose.discussed later. However, 3 -deoxyinosine (3 -dIno, 9) is likely to perform better as a substrate in the synthesis of new nucleosides. To test which compound could be a better donor of 3-deoxyribose, we recently im- plemented a modified chemical method for 30-deoxyinosine synthesis from commercially available inosine (5) (Scheme1)[18]. The details of this synthesis will be discussed later. The first trial syntheses of 2-fluorocordycepin using purine nucleoside phosphorylase (PNP) (Scheme2) yielded encouraging results: the conversion of 2-fluoroadenosine into the target product reached 67% after 7 days from the start of the reaction. In this synthesis, 2-fluoroadenosine (2-F-Ado) rather than 2-fluoroadenine (2-F-Ade) was used as the base donor. This choice was dictated by the extremely low solubility of 2-fluoroadenine in water (less than 0.5 mM at 50 ◦C) [18]. Biomolecules 2021, 11, x FOR PEER REVIEW 3 of 17

Biomolecules 2021, 11, 539 3 of 17 Biomolecules 2021, 11, x FOR PEER REVIEW 3 of 17

Scheme 1. The synthesis of 3′-deoxyinosine (9).

The first trial syntheses of 2-fluorocordycepin using purine nucleoside phosphorylase (PNP) (Scheme 2) yielded encouraging results: the conversion of 2- fluoroadenosine into the target product reached 67% after 7 days from the start of the reaction. In this synthesis, 2-fluoroadenosine (2-F-Ado) rather than 2-fluoroadenine (2-F-Ade) was used as the base donor. This choice was dictated by the ex- SchemetremelyScheme 1. 1.lowThe The synthesissolubility synthesis of of 30-deoxyinosine 32-fluoroadenine′-deoxyinosine (9). (9 ).in water (less than 0.5 mM at 50° C) [18].

The first trial syntheses of 2-fluorocordycepin using purine nucleoside phosphorylase (PNP) (Scheme 2) yielded encouraging results: the conversion of 2- fluoroadenosine into the target product reached 67% after 7 days from the start of the reaction. In this synthesis, 2-fluoroadenosine (2-F-Ado) rather than 2-fluoroadenine (2-F-Ade) was used as the base donor. This choice was dictated by the extremely low solubility of 2-fluoroadenine in water (less than 0.5 mM at 50° C) [18].

Scheme 2. The synthesis of 2-ﬂuorocordycepin (3). Scheme 2. The synthesis of 2-fluorocordycepin (3). However, when we scaled the synthesis up to hundreds of milligrams of 2-fluorocordycepin, an unexpected slowdown of the reaction and a decrease in the conversion of 2-F-Ado into 2-fluorocordycepin down to 5% were encountered. Although the PNP worked normally, the mixture contained an unexpectedly small quantity of 3-deoxyribose phosphate (judging by the quantity of hypoxanthine, 5.45%), while 2-fluoroadenine (2-F-Ade) was the main product of the reaction (Figure S1). In addition, another nucleoside was found in the reaction mixture (FigureScheme S1, 2. RTThe = synthesis 4.802 min), of [M2-fluorocordycepin + H]+ = 241.0923 (Figure(3). S2). Biomolecules 2021, 11, 539 4 of 17

Therefore, it was crucial to determine what disrupts the normal course of the transglycosylation reaction, what is the nature of the byproduct, and whether it is associated with the work of PNP.

2. Materials and Methods 2.1. General Procedures Unless otherwise noted, the materials were obtained from commercial suppliers and 0 0 0 used without any puriﬁcation. D2O was obtained from Sigma-Aldrich. 2 ,5 -Diacetyl-3 - deoxy-30-xylobromoinosine 7 was synthesized as described in [18]. Recombinant E.coli PNP was obtained as described in [15]. Analytical HPLC was performed on the Waters system (Waters 1525, Waters 2489, Breeze 2), column Nova Pak C18, 4.6 × 150 mm, 5 µm, ﬂow rate 0.5 mL/min. Method (I): gradient H2O → 25% MeOH/H2O, 20 min. Method (II): gradient H2O → 50% MeOH/H2O, 10 min. NMR spectra were recorded on Bruker Avance II 700 spectrometers (Bruker BioSpin, Rheinstetten, Germany) in DMSO-d6 at 30 ◦C. Chemical shifts in ppm (δ) were measured relative to the residual solvent signals as internal standards (2.50). Coupling constants (J) were measured in Hz. Liquid chromatography-mass spectrometry was performed on an Agilent 6210 TOF LC/MS system (Agilent Technologies, Santa Clara, CA, USA). UV spectra were recorded on Hitachi U-2900 spectrophotometer (Tokyo, Japan).

2.2. Synthesis of Nucleosides 2.2.1. Synthesis of 20,50-Di-O-acetyl-30-deoxyinosine (8) A flow of hydrogen was passed through a suspension of Pd/C (2.67 g, 10% wt % loading) in 11 mL of MeOH up to saturation of the catalyst with hydrogen. 5 g of CaCO3 were added; in 1 h, a solution of compound 7 (5 g, 12 mmol) in 20 mL of MeOH was added, and hydrogenation was continued. The reaction progress was monitored by TLC on silica gel (EtOAc–EtOH, 3:2). In 10 h, the reaction mixture was filtered off on Zeolite filter, the precipitate was washed by MeOH (50 mL), the filtrate was concentrated, and the target compound was isolated by chromatography on silica gel (column 2 × 17 cm, elution by a gradient of MeOH in CHCl3 (0% → 8%), 500 mL, flow rate 5 mL/min). Yield 2.80 g (8.3 mmol, 67%), white foam. All the MS and NMR data were identical to those described earlier [18].

2.2.2. Synthesis of 30-Deoxyinosine (9) Aq. ammonia (25%; 56 mL) was added to a solution of diacetate 8 (2.8 g, 8.4 mmol) in MeOH (28 mL) and the reaction mixture was stirred at rt. After 3 h the deacetylation was complete [TLC, CHCl3–MeOH, 9:1 (vol)], the mixture was evaporated to dryness and the target product was isolated by a column chromatography on Separon (1.8 × 25 cm, gradient elution with MeOH in water, 0 → 30%, 2.0 L, ﬂow rate 6 mL/min). Yield 1.62 g (6.45 mmol, 77%) as lyophilized powder; purity 99.12% (Rt = 6.947 min, method II). UV −1 −1 (H2O, pH 7.0) λmax, nm (ε,M cm ): 248 (12,100); λmin, nm: 224 (4,500). All the MS and NMR data were identical to those described earlier [18].

2.2.3. Synthesis of 20,30-Anhydroinosine (11) Compound 7 (865 mg, 2.08 mmol) was dissolved in 16 mL of methanol, and 4 mL of 25% aqueous NH4OH was added to the solution. After 48 h incubation at room temperature, the solution was evaporated, and the residue was dried over P2O5. Yield 325 mg. The target compound was puriﬁed by recrystallization from water, yield 250 mg (1.00 mmol, 48%), RT 13.50 min, purity 96.5%. MS- and NMR-spectra data are provided in Supplemen- tary Materials. + + HRMS (ESI ): (m/z): [M + H] calculated for C10H11N4O4 251.0780, found 251.0765, + ◦ 1 [M + Na] calculated for C10H10N4NaO4 273.0600, found 273.0581. Mp 230–232 C. H Biomolecules 2021, 11, 539 5 of 17

NMR (700 MHz, DMSO-d6, 30 ◦C): δ = 12.39 (s, 1H, NH), 8.29 (s, 1H, H-8), 8.09 (s, 1H, H-2), 6.18 (d, J~1.0 Hz, 1H, H-10), 5.03 (br. t, J = 4.85 Hz, 1H, OH-50), 4.46 (d, J = 2.3 Hz, 1H, 0 0 0 13 H-2 ), 4.20 (d, J~2.0 Hz, 1H, H-3 ), 4.19 (t, J~5.0 Hz,1H, H-4 ), 3.53 ppm (m, 2H, CH2). C NMR (176 MHz, DMSO-d6, 30 ◦C): δ = 156.48 (C=O), 147.97 (C-4), 145.64 (C-2), 138.75 (C-8), 124.07 (C-5), 81.73 (C-10), 80.94 (C-40), 60.45 (C-50), 58.16 (C-30), 57.39 ppm (C-20). 15N NMR (71 MHz, DMSO-d6, 30 ◦C): δ = 249.80 (N-7), 214.61 (N-3), 176.46 (N-9), 175.09 ppm (N-1).

2.2.4. Compounds 10, 12, 14

Compound 11 (58 mg, 0.23 mmol) and KH2PO4 (20 mg, 0.15 mmol) were dissolved in ◦ D2O (30 mL) (apparent pH 4.1) and incubated at 50 C. In 96 h the solution was evaporated up to 3 mL, centrifuged, and the desired substances were isolated by HPLC using PerfectSil Target column (MZ-Analysentechnik, Germany), 20 × 250 mm, ﬂow rate 4 mL/min, gradient elution with water-methanol system, 7–15%, 2 h. Yield: 10—10 mg (0.04 mmol, 17%), 12—5 mg (0.02 mmol, 9%), 14—8 mg (0.03 mmol, 13%). The NMR spectrum data are provided in Supplementary Materials.

2.3. Escherichia coli Purine Nucleoside Phosphorylase Inhibition Assay Each reaction mixture (0.1 mL, 50 mM potassium phosphate, pH 7.0) contained 0.5 mM of the tested compound (10, 11, 12, or 14), 0.5 mM nucleoside (inosine, 30-deoxyinosine or 2-ﬂuoroadenosine), and Escherichia coli purine nucleoside phosphorylase (0.0028 µg for inosine, 28 µg for 30-deoxyinosine or 0.28 µg for 2-ﬂuoroadenosine). Reaction mixtures were incubated at room temperature for 1 min. Substrate and product quantities were determined using HPLC (Waters 1525, column Nova-Pak C18, 4 µm, 3.9 × 150 mm, eluent 0.1% aqueous TFA, detection at 254 nm, Waters 2489).

2.4. Calculation of The Reaction Rate Constants and Solvent Kinetic Isotope Effects (KIE)

HPLC data of product 10 syntheses from epoxide 11 in H2O and D2O were used for the calculation of compound 10, 11, 12, and 14 concentrations. Kinetic parameters were determined by nonlinear regression analysis using SciDAVis v1.D013 software. Three consecutive first- or pseudo- first-order reaction models were fitted:

k k k A →1 B →2 C →3 D −k ×t [A] = A0 × e 1 k1 −k1×t −k2×t [B] = A0 × k −k × e − e 2 1 k1×k2 −k ×t k1×k2 −k ×t [C] = A0 × × e 1 + × e 2 (k2−k1)×(k3−k1) (k1−k2)×(k3−k2) k ×k + 1 2 × −k3×t (k −k )×(k −k ) e 1 3 2 3 k2×k3 −k ×t k1×k3 −k ×t [D] = A0 × × 1 − e 1 + × 1 − e 2 (k2−k1)×(k3−k1) (k1−k2)×(k3−k2) k ×k + 1 2 × 1 − e−k3×t (k1−k3)×(k2−k3)

3. Results 3.1. Deciphering the Mechanism of Degradation of Inosine Epoxide in D2O As mentioned earlier, during the scaling up of 2-fluorocordycepin synthesis, we encountered an unexpected slowdown of the reaction and a decrease in the conversion of 2-F-Ado into 2-fluorocordycepin down to 5%. Although the PNP worked normally, the main product of the reaction was 2-fluoroadenine (2-F-Ade, RT 8.655 min) (Figure S1). In addition, another nucleoside was found in the reaction mixture (Figure S1, RT = 4.802 min), [M + H]+ = 241.0923 (Figure S2). Six days after the reaction started the mixture contained 30% of this unknown nucleoside. This percentage did not change during further incubation. The nucleoside was isolated by the column reversed-phase chromatography, yield 10.6 mg (0.04 mmol, 17%). The structural formula of this nucleoside (compound 10) was deduced from NMR data: 1H, 1H-1H-COSY, 1H-13C-HSQC, 1H-13C-HMBC, 1H-15N-HSQC, and 1H-15N-HMBC spectra (Figure2). Biomolecules 2021, 11, x FOR PEER REVIEW 6 of 17

𝑘 ×𝑘 𝑘 ×𝑘 × × 𝐷 = 𝐴 × × 1−𝑒 + × 1−𝑒 𝑘 −𝑘 × 𝑘 −𝑘 𝑘 −𝑘 × 𝑘 −𝑘

𝑘 ×𝑘 + × 1−𝑒× 𝑘 −𝑘 × 𝑘 −𝑘

3. Results

3.1. Deciphering the Mechanism of Degradation of Inosine Epoxide in D2O. As mentioned earlier, during the scaling up of 2-fluorocordycepin synthesis, we encountered an unexpected slowdown of the reaction and a decrease in the conversion of 2-F-Ado into 2-fluorocordycepin down to 5%. Although the PNP worked normally, the main product of the reaction was 2-fluoroadenine (2-F-Ade, RT 8.655 min) (Figure S1). In addition, another nucleoside was found in the reaction mixture (Figure S1, RT = 4.802 min), [M + H]+ = 241.0923 (Figure S2). Six days after the reaction started the mixture contained 30% of this unknown nucleoside. This percentage did not change during further incubation. The nucleoside was isolated by the column reversed-phase chromatography, yield 10.6 mg (0.04 mmol, 17%). The structural formula of this nucleoside (com- Biomolecules 2021, 11, 539 pound 10) was deduced from NMR data: 1H, 1H-1H-COSY, 1H-13C-HSQC,6 of 17 1H-13C- HMBC, 1H-15N-HSQC, and 1H-15N-HMBC spectra (Figure 2).

.5 5 253.5 6 1 .0 76.0 1 1 1 N 6.61, 6.69 br.s H2N 7.12

.3 H 9 3 56.2 1 N 6.38d HN 178.1

HO H1-N15 -HMBC O H1-C13-HMBC

56.2 4.25 m H H 5.57 d 73.1 COSY H OH 3.69 m Figure 2. 2.The The structural structural formula formula of nucleoside of nucleoside10. 10. We supposed that nucleoside 10 is formed from some impurity not detected in our preparationWe supposed of 30-deoxyinosine. that nucleoside More careful10 is formed analysis from using some LC-MS impurity helped us not detect detected in ouran admixture preparation of 20,3 of0-anhydroinosine 3′-deoxyinosine.11 in More the preparation. careful an Thealysis data using of chromate-mass LC-MS helped us detectspectrometry an admixture are shown of in 2 Figure′,3′-anhydroinosine S3. The structure 11 of in compound the preparation.11 was confirmed The data by of chromate-massNMR data (see spectrometry Supplementary are Information). shown in Figure S3. The structure of compound 11 was confirmedWe assumed by NMR that epoxide data (see11 formsSupplementary during dehalogenation Information). of protected nucleoside 7 by tributyltinWe assumed hydride that (Scheme epoxide1, Route11 forms A). To during check thisdehalogenation assumption, we of replaced protected a nucle- osidetributyltin 7 by hydride tributyltin treatment hydride by dehalogenation (Scheme 1, usingRoute H 2A)./Pd To (Scheme check1 ,this Route assumption, B). The we resulted compounds 8 and 9 were identical to those described earlier [18]. No traces of replaced a tributyltin hydride treatment by dehalogenation using H2/Pd (Scheme 1, epoxide 11 were found in nucleoside 9. Compound 9 obtained with the help of catalytic Routedehalogenation B). The wasresulted used in compounds the synthesis 8 of and 2-fluorocordycepin. 9 were identical to those described earlier [18].According No traces to HPLCof epoxide data, the 11 reaction weremixture founddid in notnucleoside contain nucleoside 9. Compound10 (Figure 9 S4). obtained Thewith conversion the help of of9 into catalytic 2-fluorocordycepin dehalogenation3 was 71% was in 16used days. in the synthesis of 2-fluorocordycepin.It is the admixture of compound 11 in the first samples of 30-dIno 9 that reduced the yieldAccording of 2-fluorocordecypin to HPLC in data, the enzymatic the reaction reaction. mixture did not contain nucleoside 10 (FigureWe performedS4). The conversion a series of reactions of 9 into of 2-fluorocordycepin 2-fluorocordecypin synthesis 3 was using71% differentin 16 days. 0 samplesIt is of the 3 -dIno admixture9. The results of compound are shown 11 in Figurein the3 first. samples of 3′-dIno 9 that reduced It can be seen that in cases b and c, the transglycosylation reaction and 2-fluorocordecypin the yield of 2-fluorocordecypin in the enzymatic reaction. synthesis proceed abnormally. At the beginning of the process (in the first two days), there was almost no difference between the three reactions, then the speed of 2-F-Cord 3 accumulation decreased, and conversion did not reach the optimum. Additional portions of the enzyme did not improve the situation—the synthesis of 2-F-Cord 3 would not resume. Moreover, 2-fluoroadenine was formed in the mixture almost exclusively, with only traces of 3-deoxyribose-1-α-phosphate (determined by the amount of hypoxanthine in the mixture). We have never encountered such selective PNP behaviour in nucleoside synthesis. On the other hand, if the slowdown of 2-F-Cord 3 synthesis had begun on the 3rd day, it could be assumed that not epoxide 11 itself, but the products of its degradation violate the normal course of the enzymatic reaction. Perhaps one of those products is a PNP inhibitor, which hinders the normal operation of the enzyme. Therefore, we decided to identify the mechanism of transformation of epoxide 11 into nucleoside 10, isolate the intermediates, and determine their PNP inhibition constants. The formation of nucleoside 10 from 20,30-anhydroinosine was described more than 40 years ago [19]. However, the transformation was performed under relatively harsh conditions (0.1 M NaOH, 80 ◦C, 60 min). In [20], 10 was synthesized from 20,30-anhydroadenosine, an intermediate product was isolated, and a mechanism of 10 formation was suggested. Biomolecules 2021, 11, 539 7 of 17 Biomolecules 2021, 11, x FOR PEER REVIEW 7 of 17

Nonetheless,We performed we believed a series that of more reactions than one of intermediate2-fluorocordecypin might exist, synthesis so we decidedusing dif- to investigateferent samples the reaction of 3′-dIno more 9. closely.The results are shown in Figure 3.

25 a b 20 c

10 2-F-Cord content (%)

0 0 50 100 150 200 250 300 350 400 Time (h)

FigureFigure 3. 3.The The dynamics dynamics of of accumulation accumulation (HPLC (HPLC data) data) of 2-F-Cord of 2-F-Cord3 in the 3 in enzymatic the enzymatic reaction reac- using samplestion using of 3 samples0-dIno 9 obtained of 3′-dIno by 9 different obtained methods: by differenta—Catalytic methods: dehalogenation a - Catalytic (Scheme dehalogenation1, route B), b(Scheme—Radical 1,dehalogenation route B), b - Radical (Scheme dehalogenation1, route A), c—Mixture (Scheme of 1,a routeand b A),. Reaction c - Mixture conditions: of a and 1 mM b. Reaction conditions: 1 mM 2-F-Ado, 1.5 mM 3′-dIno, 2 mM KH2PO4, pH 7.0, 50 °C, 12.5 IU 2-F-Ado, 1.5 mM 30-dIno, 2 mM KH PO , pH 7.0, 50 ◦C, 12.5 IU PNP, 1 mL. Epoxide 11 concentration: PNP, 1 mL. Epoxide 11 concentration:2 4 a—0 mM, b—0.24 mM, c—0.12 mM. a—0 mM, b—0.24 mM, c—0.12 mM.

FirstIt can of be all, seen we modifiedthat in cases the synthesisb and c, the of epoxide transglycosylation11. Numerous reaction approaches and 2-fluo- to this synthesisrocordecypin have synthesis been proposed. proceed Most abnormally. commonly, At a xylochloro-the beginning or xylobromo-derivativeof the process (in the offirst a corresponding two days), there ribonucleoside was almost acts no differe as a precursor,nce between and cyclization the three reactions, is performed then using the aspeed strong of base 2-F-Cord (for example, 3 accumulation sodium methoxide decreased, [19 and–21]) conversion followed by did neutralization not reach the of theop- basetimum. or Dowex Additional 1x2 (OH portions−)[22]. of It turnedthe enzy outme that did the not treatment improve of 2the0,50 -diacetyl-3situation—the0-deoxy-3 syn-0- thesis of 2-F-Cord 3 would not resume. xylobromoinosine (it was synthesized earlier; see [18]) by the solution of aqueous NH4OH in methanolMoreover, followed 2-fluoroadenine by simple evaporation was formed and in the crystallization mixture almost gave exclusively, the pure desired with productonly traces11 with of 3-deoxyribose-1- satisfactory yield.α-phosphate (determined by the amount of hypo- xanthineCompound in the mixture).11 was stable We in have pure never water encountered at 50 ◦C during such 72 hours.selective It was PNP stable behaviour under thosein nucleoside conditions synthesis. after the addition of LiCl and NaCl. In 5 mM potassium-phosphate buffer (pH 7.0)On atthe 50 other◦C, the hand, reaction if the proceeds. slowdown The HPLC of 2-F-Cord profile of3 synthesis the reaction had mixture begun is shownon the in3rd Supplementary day, it could be Figure assumed S5. We that detected not epoxide peaks that11 itself, correspond but the to products starting material of its deg-11 (RTradation 13.288 violate min.), nucleoside the normal10 course(RT 7.038 of min.),the enzymatic and several reaction. extra products Perhaps (RT one 4.380, of 7.923,those andproducts 11.188). is Wea PNP supposed inhibitor, that thosewhich peaks hinders (compounds the normal12– 14operation) are peaks of ofthe the enzyme. sought intermediates.Therefore, we However,decided to the identify quantity the of thosemechanism intermediates of transformation was too small of for epoxide isolating 11 theminto nucleoside from the reaction 10, isolate mixture. the Most intermediates, likely, the lifetime and determine of the intermediates their PNP is inhibition too short, andconstants. we had to find a way to slow down the reaction and to accumulate the substances in sufficientThe quantities.formation of nucleoside 10 from 2′,3′-anhydroinosine was described more thanWe 40 decidedyears ago that [19]. the simplestHowever, way the to tr doansformation this is using was D2O performed instead of Hunder2O. HPLC rela- profiletively ofharsh thereaction conditions mixture (0.1 (5 M mM NaOH, potassium-phosphate 80 °C, 60 min). buffer,In [20], appeared 10 waspH synthesized 7.0, 50 ◦C, Dfrom2O) is2′ shown,3′-anhydroadenosine, in Figure S6. an intermediate product was isolated, and a mecha- nismThe of situation10 formation improved was significantly.suggested. Nonetheless, The relative quantities we believed of intermediates that more increased,than one andintermediate we had a chancemight exist, to isolate so we and decided investigate to investigate them. However, the reaction at pH 4.1 more (appeared closely. pH becauseFirst the of reaction all, we wasmodified performed the synthesis in D2O), the of resultsepoxide were 11. evenNumerous better (Figure approaches S7). to this Thesynthesis dynamics have of been nucleosides proposed.10 and Most12 formation commonly, during a xylochloro- hydrolysis or of xylobromo-11 at pH 7.0 andderivative 4.1 in H of2O a and corresponding D2O are shown ribonucleoside in Figures4 and acts5. as a precursor, and cyclization is performed using a strong base (for example, sodium methoxide [19–21]) followed Biomolecules 2021, 11, x FOR PEER REVIEW 8 of 17

by neutralization of the base or Dowex 1x2 (OH−) [22]. It turned out that the treatment of 2′,5′-diacetyl-3′-deoxy-3′-xylobromoinosine (it was synthesized earlier; see [18]) by the solution of aqueous NH4OH in methanol followed by simple evaporation and crystallization gave the pure desired product 11 with satisfactory yield. Compound 11 was stable in pure water at 50 °C during 72 hours. It was stable under those conditions after the addition of LiCl and NaCl. In 5 mM potassium- phosphate buffer (pH 7.0) at 50 °C, the reaction proceeds. The HPLC profile of the reaction mixture is shown in Supplementary Figure S5. We detected peaks that correspond to starting material 11 (RT 13.288 min.), nucleoside 10 (RT 7.038 min.), and several extra products (RT 4.380, 7.923, and 11.188). We supposed that those peaks (compounds 12–14) are peaks of the sought intermediates. However, the quantity of those intermediates was too small for isolating them from the reaction mixture. Most likely, the lifetime of the intermediates is too short, and we had to find a way to slow down the reaction and to accumulate the substances in sufficient quantities. We decided that the simplest way to do this is using D2O instead of H2O. HPLC profile of the reaction mixture (5 mM potassium-phosphate buffer, appeared pH 7.0, 50 °C, D2O) is shown in Figure S6. The situation improved significantly. The relative quantities of intermediates increased, and we had a chance to isolate and investigate them. However, at pH 4.1 (appeared pH because the reaction was performed in D2O), the results were even Biomolecules 2021, 11, 539 better (Figure S7). 8 of 17 The dynamics of nucleosides 10 and 12 formation during hydrolysis of 11 at pH 7.0 and 4.1 in H2O and D2O are shown in Figures 4 and 5.

100

1 60 2 3 4 40

20 Nucleoside content (%) content Nucleoside

0 0 50 100 150 200 250 300 350 400 Time (h)

Biomolecules 2021, 11, x FOR PEER REVIEW 9 of 17 FigureFigure 4. 4.The The percentage percentage of of nucleosides nucleosides10 and10 and12 in 12 the in reactionthe reaction mixtures mixtures during during hydrolysis hydroly- of 11 sis of 11 (pH 7.0) in H2O (curves 1 and 3) and in D2O (curves 2 and 4). 1, 2—nucleoside 10, (pH 7.0) in H2O (curves 1 and 3) and in D2O (curves 2 and 4). 1, 2—nucleoside 10, 3, 4—nucleoside 12. 3, 4—nucleoside 12.

100

1 60 2 3 4 40

20 Nucleoside content (%)

0 0 50 100 150 200 250 300 350 400 Time (h)

FigureFigure 5. 5.The The percentage percentage of of nucleosides nucleosides10 and10 and12 in 12 the in reactionthe reaction mixtures mixtures during during hydrolysis hydroly- of 11 sis of 11 (pH 4.1) in H2O (curves 1 and 3) and in D2O (curves 2 and 4). 1, 2—nucleoside 10, (pH 4.1) in H2O (curves 1 and 3) and in D2O (curves 2 and 4). 1, 2—nucleoside 10, 3, 4—nucleoside 12. 3, 4—nucleoside 12.

It can be seen that the optimal conditions for obtaining of 12 are D2O at appeared pH 4.1.It can So, be we seen performed that the hydrolysis optimal conditions of 11 in D2 Ofor at obtaining appeared of pH 12 4.1 are and D2O isolated at ap- intermediatespeared pH 4.112. So,and we14 (Seeperformed Supplementary hydrolysis Information). of 11 in D Then2O at our appeared goal was pH to identify4.1 and theisolated chain intermediates of transformation—which 12 and 14 (See of intermediates Supplementary forms Information). from 11 primarily Then and our which goal onewas is to an identify immediate the precursorchain of transforma of 10. tion—which of intermediates forms from 11 primarilyWe put and three which reactions: one is hydrolysis an immediate of 11, precursor hydrolysis of of 1012., and hydrolysis of 14, and measuredWe put the three UV spectra reactions: of the hydrolysis reaction mixtures of 11, hydrolysis overtime. of Individual 12, and hydrolysis spectra of 10 of, 1114,, 12and, and measured14 are shown the UV in Figurespectra6. of the reaction mixtures overtime. Individual spectra of 10, 11, 12, and 14 are shown in Figure 6.

2.0 215 247 254 276 310 Abs

1.5 10 11 12 1.0 14

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

Figure 6. The UV spectra of compounds 10, 11, 12, and 14 (D2O, appeared pH 4.1).

The set of UV spectra for hydrolysis of 11 is shown in Figure 7. The picture is somewhat complicated due to the overlay of spectra of individual components. Biomolecules 2021, 11, x FOR PEER REVIEW 9 of 17

100

1 60 2 3 4 40

20 Nucleoside content (%)

0 0 50 100 150 200 250 300 350 400 Time (h)

Figure 5. The percentage of nucleosides 10 and 12 in the reaction mixtures during hydrolysis of 11 (pH 4.1) in H2O (curves 1 and 3) and in D2O (curves 2 and 4). 1, 2—nucleoside 10, 3, 4—nucleoside 12.

It can be seen that the optimal conditions for obtaining of 12 are D2O at appeared pH 4.1. So, we performed hydrolysis of 11 in D2O at appeared pH 4.1 and isolated intermediates 12 and 14 (See Supplementary Information). Then our goal was to identify the chain of transformation—which of intermediates forms from 11 primarily and which one is an immediate precursor of 10. Biomolecules 2021, 11, 539 We put three reactions: hydrolysis of 11, hydrolysis of 12, and hydrolysis 9of of 14, 17 and measured the UV spectra of the reaction mixtures overtime. Individual spectra of 10, 11, 12, and 14 are shown in Figure 6.

2.0 215 247 254 276 310 Abs

1.5 10 11 12 1.0 14

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

2 FigureFigure 6. 6. The TheUV UV spectra spectra of of compounds compounds10, 1110,, 1211,, and 12, 14and(D 142O, (D appearedO, appeared pH 4.1). pH 4.1). Biomolecules 2021, 11, x FOR PEER REVIEW 10 of 17

TheThe set set of of UV UV spectra spectra for hydrolysisfor hydrolysis of 11 isof shown 11 is shown in Figure in7. Figure The picture 7. The is somewhat picture is complicatedsomewhat complicated due to the overlay due to of spectrathe overlay of individual of spectra components. of individual However, components. it can be seenHowever, that, in it 48 can h (magentabe seen that, line), in some 48 h quantity(magenta of line),12 (shoulder some quantity at 270 nm) of 12 and (shoulder14 (λmax 310at 270 nm) nm) is formed. and 14 (λmax 310 nm) is formed.

2.0 247 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

FigureFigure 7. 7.The The UV UV spectra spectra of of the the reaction reaction mixtures mixtures obtained obtained during during hydrolysis hydrolysis of 11 (D of2O, 11 appeared (D2O, pHappeared 4.1). pH 4.1).

TheThe set set of of spectra spectra obtained obtained during during hydrolysis hydrolysis of 12 ofis 12 much is much more more useful useful (Figure (Figure8). 8). In 24 h after reaction started, noticeable amounts of 14 (λmax 310 nm) are formed (blue line). The absorption at 254 nm (λmax of 12) decreases, and an additional absorption maximum at 274 nm (λmax of 10) is appears. In 48 h (magenta line), the amounts of both 14 and 10 increase, and the amount of 12 is negligible. In 96 h (navy line), the amount of 14 decreases, while the amount of 10 continues increasing. In 144 h (violet line), the amount of 14 is minimal—almost all 14 is converted into 10. These data allowed us to assume, that 11 is converted into 12, then 12 is converted into 14, and, finally, 14 is converted into 10.

2.0 254 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

Figure 8. The UV spectra of the reaction mixtures obtained during hydrolysis of 12 (D2O, appeared pH 4.1). Biomolecules 2021, 11, x FOR PEER REVIEW 10 of 17

However, it can be seen that, in 48 h (magenta line), some quantity of 12 (shoulder at 270 nm) and 14 (λmax 310 nm) is formed.

2.0 247 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

Figure 7. The UV spectra of the reaction mixtures obtained during hydrolysis of 11 (D2O, appeared pH 4.1).

The set of spectra obtained during hydrolysis of 12 is much more useful (Figure 8). In 24 h after reaction started, noticeable amounts of 14 (λmax 310 nm) are formed (blue line). The absorption at 254 nm (λmax of 12) decreases, and an additional absorption maximum at 274 nm (λmax of 10) is appears. In 48 h (magenta line), the amounts of both 14 and 10 increase, and the amount of 12 is negligible. In 96 h (navy line), the amount of 14 decreases, while the amount of 10 continues increasing. In Biomolecules 2021, 11, 539 144 h (violet line), the amount of 14 is minimal—almost all 14 is converted into10 of 10 17. These data allowed us to assume, that 11 is converted into 12, then 12 is converted into 14, and, finally, 14 is converted into 10.

2.0 254 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

FigureFigure 8. 8.The The UV UV spectra spectra of theof the reaction reaction mixtures mixtures obtained obtained during during hydrolysis hydrolysis of 12 (D of2O, 12 appeared (D2O, pHappeared 4.1). pH 4.1).

In 24 h after reaction started, noticeable amounts of 14 (λmax 310 nm) are formed (blue line). The absorption at 254 nm (λmax of 12) decreases, and an additional absorption maximum at 274 nm (λmax of 10) is appears. In 48 h (magenta line), the amounts of both 14 Biomolecules 2021, 11, x FOR PEER REVIEW 11 of 17 and 10 increase, and the amount of 12 is negligible. In 96 h (navy line), the amount of 14 decreases, while the amount of 10 continues increasing. In 144 h (violet line), the amount of 14 is minimal—almost all 14 is converted into 10. These data allowed us to assume, that 11 is convertedThe last set into of12 spectra, then 12 wasis converted obtained into during14, and, hydrolysis ﬁnally, 14 ofis converted14 (Figure into 9).10 The. absorptionThe last at set 310 of spectra nm is waspermanently obtained during decreased hydrolysis while of absorption14 (Figure9). at The 274 absorption nm is in- atcreased. 310 nm This is permanently confirms, decreased that 14 is while a direct absorption precursor at 274 of nm 10, is and increased. a chain This of conﬁrms,transfor- thatmation14 is is a 11 direct–12– precursor14–10. of 10, and a chain of transformation is 11–12–14–10.

2.0 215 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

FigureFigure 9. 9. TheThe UVUV spectra spectra of of reaction reaction mixtures mixtures obtained obtained during during hydrolysis hydrolysis of 14 of(D 142 O,(D2 appearedO, ap- pHpeared 4.1). pH 4.1).

The structure of compounds 12 and 14 was deduced from NMR data: 1H- and 1H-1H-COSY, 1H-13C-HSQC, 1H-13C-HMBC, 1H-15N-HSQC, and 1H-15N-HMBC. As- signments of main signals for 12 are shown in Figure 10.

Figure 10. Assignments of main NMR signals for compound 12.

Integrals of C2 and C8 protons (8.16 ppm and 7.97 ppm) in 1H-NMR spectra of 12 proved to be 0.66 and 0.17 correspondingly (integral 1.00 is assigned to anomer proton - 6.08 ppm) (Figure S8). This can be explained by the H - D exchange in those positions. In 2H-NMR spectra of 12, two peaks at 8.16 and 7.97 ppm with corresponding integrals are observed (Figure S9). Biomolecules 2021, 11, x FOR PEER REVIEW 11 of 17

The last set of spectra was obtained during hydrolysis of 14 (Figure 9). The absorption at 310 nm is permanently decreased while absorption at 274 nm is increased. This confirms, that 14 is a direct precursor of 10, and a chain of transformation is 11–12–14–10.

2.0 215 276 310 Comp 10 Abs 0 h 24 h 48 h 1.5 72 h 96 h 144 h 192 h 1.0 312 h

0.5

0.0 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm)

Biomolecules 2021, 11, 539 11 of 17 Figure 9. The UV spectra of reaction mixtures obtained during hydrolysis of 14 (D2O, appeared pH 4.1).

The structurestructure of of compounds compounds12 and12 and14 was 14 was deduced deduced from NMRfrom data:NMR1 H-data: and 1H-1H- and1H- COSY,1H-1H-COSY,1H-13C-HSQC, 1H-13C-HSQC,1H-13C-HMBC, 1H-13C-HMBC,1H-15N-HSQC, 1H-15N-HSQC, and 1H-15N-HMBC. and 1H-15 AssignmentsN-HMBC. As- of mainsignments signals of for main12 are signals shown for in 12 Figure are shown10. in Figure 10.

Figure 10. Assignments of main NMR signals for compound 12. Figure 10. Assignments of main NMR signals for compound 12. Biomolecules 2021, 11, x FOR PEER REVIEWIntegrals of C2 and C8 protons (8.16 ppm and 7.97 ppm) in 1H-NMR spectra of 12 12 of 17 Integrals of C2 and C8 protons (8.16 ppm and 7.97 ppm) in 1H-NMR spectra of proved to be 0.66 and 0.17 correspondingly (integral 1.00 is assigned to anomer proton— 6.0812 proved ppm) (Figure to be 0.66 S8). and This 0.17 can becorrespondingly explained by the (integral H—D exchange 1.00 is assigned in those positions. to anomer In 2protonH-NMR - 6.08 spectra ppm) of 12,(Figuretwo peaksS8). This at 8.16 can and be explained 7.97 ppm with by the corresponding H - D exchange integrals in those are 13C-NMR2 data also confirm the presence of deuterium in compound 12 (Figure observedpositions. (Figure In H-NMR S9). spectra of 12, two peaks at 8.16 and 7.97 ppm with corre- S10).sponding13 C-NMR integrals data also are conﬁrmobserved the (Figure presence S9). of deuterium in compound 12 (Figure S10). SignalsSignals of of C2 C2 and and C8 areC8 shiftedare shifted upﬁeld upfield (143.94 →(143.94143.68 → and 143.68 134.39 →and134.16 134.39 ppm) → 134.16 ppm)and are and split are into split triplets. into Signals triplets. in NMR Signals spectra in NMR of nucleoside spectra14 ofwere nucleoside assigned in 14 the were as- signedsame way in (Figure the same 11). way (Figure 11).

Figure 11. 11.Assignments Assignments of main of main NMR NMR signals signals for compound for compound14. 14.

Compound 13 (see Figures S5 and S6) isolated from the reaction mixtures was found to be a purinyl heterocycle (according to NMR data, see page S-32); λmax 250 nm, λmin 221 nm. The H → D exchange at C8 is not related to hydrolysis of epoxide 11. After incubation of inosine in D2O at 50 °C (pH 4.1) for 4 days, 81% of protons at C8 are changed for deuterium. Thus, we have fully deciphered the structures of 2′,3′-anhydroinosine opening intermediates in a phosphate buffer solution. A full description of the NMR spectra of compounds 10, 12, and 14 is provided in the Supplementary Information.

3.2. Escherichia Coli Purine Nucleoside Phosphorylase Inhibition Assay. After obtaining compounds 10, 12, and 14, a series of experiments were performed to study PNP inhibition. As we mentioned above, the presence of epoxide 11 in the reaction mixture inhibits the synthesis of 2-fluorocordycepin. So, epoxide 11 or one of the products formed during hydrolysis of 11 can be an inhibitor of PNP. In this case, they should inhibit a phosphorolysis of inosine, 3′-deoxyinosine, or 2- fluoroadenosine. The results of phosphorolysis in the presence of compounds 10, 11, 12, and 14 are shown in Figure 12. The reactions conditions are provided in Materials and Methods. Biomolecules 2021, 11, 539 12 of 17

Compound 13 (see Figures S5 and S6) isolated from the reaction mixtures was found to be a purinyl heterocycle (according to NMR data, see page S-32); λmax 250 nm, λmin 221 nm. The H → D exchange at C8 is not related to hydrolysis of epoxide 11. After incubation ◦ of inosine in D2O at 50 C (pH 4.1) for 4 days, 81% of protons at C8 are changed for deuterium. Thus, we have fully deciphered the structures of 20,30-anhydroinosine opening intermediates in a phosphate buffer solution. A full description of the NMR spectra of compounds 10, 12, and 14 is provided in the Supplementary Information.

3.2. Escherichia Coli Purine Nucleoside Phosphorylase Inhibition Assay After obtaining compounds 10, 12, and 14, a series of experiments were performed to study PNP inhibition. As we mentioned above, the presence of epoxide 11 in the reaction mixture inhibits the synthesis of 2-ﬂuorocordycepin. So, epoxide 11 or one of the products formed during hydrolysis of 11 can be an inhibitor of PNP. 0 Biomolecules 2021, 11, x FOR PEER REVIEW In this case, they should inhibit a phosphorolysis of inosine, 3 -deoxyinosine,13 of 17 or 2-ﬂuoroadenosine. The results of phosphorolysis in the presence of compounds 10, 11, 12, and 14 are shown in Figure 12. The reactions conditions are provided in Materials and Methods.

140 Ino 3'-dI 2-F-Ado 120

100

60 PNP activity, % activity, PNP

0 control 10 11 12 14

Figure 12. PNP activity for inosine (green), 30-deoxyinosine (orange), and 2-fluoroadenosine (yellow) phosphorolysis in the presence of compounds 10, 11, 12, and 14. Figure 12. PNP activity for inosine (green), 3′-deoxyinosine (orange), and 2-fluoroadenosine (yellow) phosphorolysisIt can be seen in that the neither presence compound of compounds11 nor 10 compounds, 11, 12, and formed14. from 11 affect E. coli PNP activity. It can be seen that neither compound 11 nor compounds formed from 11 affect E. coli PNP activity.3.3. Calculation of the Reaction Rate Constants and Solvent Kinetic Isotope Effects (KIE) The rate constants and kinetic isotope effects for the synthesis stages of every isolated 3.3. Calculationcompound of werethe Reaction determined Rate Constants (Table1). Threeand Solvent consecutive Kinetic first-Isotope or Effects pseudo (KIE). first-order Thereaction rate constants models were and used. kinetic isotope effects for the synthesis stages of every isolated compound were determined (Table 1). Three consecutive first- or pseudo first-order reaction models were used.

Table 1. Reaction rate constants and solvent kinetic isotope effects for synthesis of 12, 14, and 10.

pH Solvent k1, 1/h k2, 1/h k3, 1/h KIE1 KIE2 KIE3 (pD) H2O 0.011 ± 0.001 0.17 ± 0.02 0.046 ± 0.004 7.0 1.10 ± 0.11 1.31 ± 0.15 0.77 ± 0.07 D2O 0.010 ± 0.001 0.13 ± 0.01 0.060 ± 0.005 H2O 0.0081 ± 0.0006 0.047 ± 0.006 0.015 ± 0.002 4.1 1.14 ± 0.08 1.34 ± 0.17 0.68 ± 0.09 D2O 0.0071 ± 0.0004 0.035 ± 0.003 0.022 ± 0.002

4. Discussion During a series of syntheses of 2-fluorocordycepin from 3′-deoxyinosine catalyzed by E. coli PNP (Scheme 2), we encountered the problem of a decrease in the yield of the target product. At the same time, E. coli PNP worked normally: 2-fluoroadenine was produced, while phosphorolysis of 3′-deoxyinosine did not occur (according to HPLC data, there was practically no hypoxanthine in the reaction mixture). As a result of a thorough study of 3′-deoxyinosine quality, an impurity of Biomolecules 2021, 11, x FOR PEER REVIEW 13 of 17

140 Ino 3'-dI 2-F-Ado 120

100

60 PNP activity, % activity, PNP

0 control 10 11 12 14

Figure 12. PNP activity for inosine (green), 3′-deoxyinosine (orange), and 2-fluoroadenosine (yellow) phosphorolysis in the presence of compounds 10, 11, 12, and 14.

It can be seen that neither compound 11 nor compounds formed from 11 affect E. coli PNP activity.

3.3. Calculation of the Reaction Rate Constants and Solvent Kinetic Isotope Effects (KIE).

Biomolecules 2021, 11, 539 The rate constants and kinetic isotope effects for the synthesis stages of 13every of 17 isolated compound were determined (Table 1). Three consecutive first- or pseudo first-order reaction models were used.

Table 1. Reaction rate constants and solvent kinetic isotopeisotope effectseffects forfor synthesissynthesis ofof 1212,, 14,14, andand 1010..

pH pH (pD)Solvent Solvent k1, 1/h k , 1/h k2 k, 1/h, 1/h k k,3 1/h, 1/h KIE KIE1 KIEKIE2 KIEKIE3 (pD) 1 2 3 1 2 3 H2O 0.011 ± 0.0010.011 ± 0.170.17 ± 0.02± 0.0460.046 ± ±0.004 7.0 H2O 1.10 ± 0.11 1.31 ± 0.15 0.77 ± 0.07 D2O 0.010 ± 0.0010.001 0.13 ±0.02 0.01 0.0600.004 ± 0.005 1.10 ± 1.31 ± 0.77 ± 7.0 H2O 0.0081 ± 0.00060.010 ± 0.0470.13 ± 0.006± 0.0150.060 ± ±0.002 0.11 0.15 0.07 4.1 D O 1.14 ± 0.08 1.34 ± 0.17 0.68 ± 0.09 D2O 0.00712 ± 0.00040.001 0.035 ±0.01 0.003 0.0220.005 ± 0.002 0.0081 ± 0.047 ± 0.015 ± 4. DiscussionH O 2 0.0006 0.006 0.002 1.14 ± 1.34 ± 0.68 ± 4.1During a series of syntheses of 2-fluorocordycepin from 3′-deoxyinosine cata- 0.0071 ± 0.035 ± 0.022 ± 0.08 0.17 0.09 D O lyzed by E. coli2 PNP (Scheme0.0004 2), we0.003 encountered0.002 the problem of a decrease in the yield of the target product. At the same time, E. coli PNP worked normally: 2-fluoroadenine was produced, while phosphorolysis of 3′-deoxyinosine did not occur (ac- 4. Discussion cording to HPLC data, there was practically no hypoxanthine0 in the reaction mixture).During As a result a series of of asyntheses thorough of study 2-fluorocordycepin of 3′-deoxyinosine from 3 quality,-deoxyinosine an impurity catalyzed of by E. coli PNP (Scheme2), we encountered the problem of a decrease in the yield of the target product. At the same time, E. coli PNP worked normally: 2-fluoroadenine was produced, while phosphorolysis of 30-deoxyinosine did not occur (according to HPLC data, there was practically no hypoxanthine in the reaction mixture). As a result of a thorough study of 30-deoxyinosine quality, an impurity of 20,30-anhydroinosine was detected. Acyclic nucleoside (10) was produced from this impurity during the enzymatic reaction carried out at pH 7.0 and 50 ◦C. This brought up two questions: 1) where the impurity of 20,30-anhydroinosine appears in 3-0deoxyinosine, and 2) how this impurity prevents the phosphorolysis of 3- 0deoxyinosine to 3-deoxyribose-1-α-phopsphate in the active site of E. coli PNP. Acyclic nucleoside 10 was first discovered more than 40 years ago as an impurity in the synthesis of adenine nucleosides. The degradation of 20,30-anhydroadenosine in hot water was investigated in detail in [20]. An intermediate product (10) with opened pyrimidine ring was isolated. After adding NaOH to the reaction mixture, compound 10 was obtained. It was shown that 20,30-anhydroinosine (11) is converted into 10 during the heating of the solution of 11 in 0.1 M NaOH, yet the intermediate products were not isolated [19]. In our synthesis of 30-deoxyinosine, different processes and conditions were used. Nevertheless, we assumed that at the stage of radical dehalogenation of 30-deoxy-30- xylobrom-20,50-diacetylinosine (7), in addition to the target 30-deoxy-30-xylobrom-20,50- diacetylinosine (8), an impurity of 50-acetyl-20,30-anhydroinosine is formed. This impurity is difficult to separate by chromatography on silica gel. After removing acetyl groups by ammonolysis 30-deoxyinosine, samples contained epoxide 11 (up to 17%), detected by high-resolution LC-MS. Thus, we determined the compound that hinders the normal synthesis of 2-fluoroco- rdycepine. We replaced the stage of radical dehalogenation with a catalytic one (Scheme1, route B). Using the “new” 30-deoxyinosine, we performed several enzymatic syntheses of 2-F-Cord in high yields (up to 72%). However, it was difficult to identify the compound preventing the normal functioning of E. coli PNP. It was definitely not 20,30-anhydroinosine itself since enzymatic synthesis began to slow down only on the 3rd day from the beginning of the enzymatic reaction. Biomolecules 2021, 11, 539 14 of 17

At this time, the original 20, 30-anhydroinosine was already transformed into at least three nucleosides of a different structure (according to LC-MS data). How could these nucleosides be isolated from reaction mixtures, and how do they affect the functioning of PNP? It would be valuable to find inhibitors of E. coli PNP. Obviously, it is necessary study the degradation mechanism of 20,30-anhydroinosine in a buffer solution at 50 ◦C thoroughly and attempt to isolate the degradation products quantitatively. However, in H2O, the degradation proceeded rather quickly, and the content of intermediates did not exceed 3–5%. For this reason, we decided to replace H2O with D2O. This should slow down all chemical processes and allow a larger amount of intermediates to be accumulated in the solution. Our assumption was fully confirmed: the amount of intermediate nucleosides in the reaction mixture increased from 3–5% to 15%. However, at pH 7.0, the formation of the acyclic compound 10 proceeded rather quickly. We reduced the pH of the solution from 7.0 to 4.1; as a result, the rate of formation of 10 decreased by five times. In Supplementary Ma- terials, we provide the HPLC profile (Figure S7) of the inosine epoxide degradation reaction mixture at the optimum time (day 4) when the amount of intermediate degradation products is maximal. The conditions that allowed us to accumulate the degradation products in sufficient quantities were: 10—10 mg (17%), 12—5 mg (9%), 14—8 mg (13%). The structure of these nucleosides was deduced from NMR spectra (see Supplementary Information). It is clear that inosine epoxide ring opening in phosphate buffer is a multi-stage process. Biomolecules 2021, 11, x FOR PEER REVIEW 15 of 17 Our goal was to identify the chain of transformation - which one of the intermediates forms from 11 primarily and which one is an immediate precursor of 10. UV-spectroscopy helped us to understand this process. Each of the isolated nucle- determineosides has the own transformation absorption maximum: chain of the nucleoside inosine epoxide10—276, 11:11 chain—247, of 12transformation—254, and 14 —is 11310–12 nm.–14 We–10 investigated. the change of UV spectra of reaction mixtures upon incubation of aqueousThe reaction solutions proceeds for 0 tothrough 312 h. Analysisthe formation of the of dynamics intermediates of the 12 spectra and 14 allowed. A possible us to mechanismdetermine theof compound transformation 10 formation chain of thein D inosine2O is shown epoxide in Scheme11: chain 3. of transformation is 11–12Performing–14–10. the reaction in D2O allowed us to localize the proton positions and deuteriumThe reaction addition. proceeds This is through useful thefor formationtransient ofstate intermediates and intramolecular12 and 14 .rearrange- A possible mentsmechanism visualization. of compound 10 formation in D2O is shown in Scheme3.

SchemeScheme 3. 3. AA possible possible mechanism of of hydrolysis hydrolysis of of epoxide epoxide11 11with with transient transient states states and intramolecularand intramolecular rearrangements. rearrangements.

The rate constants and kinetic isotope effects for the synthesis stages of each isolated compound were determined (Table 1). It can be seen that the fastest stage is the synthesis of intermediate 14, while the slowest one is the synthesis of cyclonucleoside 12. At pH 4.1, the first stage is 1.4 times slower, and the next two stages are 3-4 times slower than at pH 7.0. The kinetic isotope effect of the solvent in the second stage is greater than in the first. In the synthesis of compound 10, the reverse isotopic effect is observed. Neither compound 11 nor compounds formed from 11 affect E. coli PNP activity (Fig- ure 12). Therefore, the reason for the slowdown of transglycosylation reaction in the presence of 11 is still unclear. Probably some short-living intermediates interact with the active site, with their lifetime being too short for us to isolate them. We believe that nucleosides with a structure similar to the compound shown in Scheme 3 could be effective inhibitors of E. coli PNP.

5. Conclusions It is undesirable to use modified nucleosides synthesized with the help of radical dehalogenation (in particular, using Bu3SnH) as substrates of nucleoside phosphorylases. 2′,3′-Anhydronucleosides form during this reaction and the main product is difficult to separate from them. An admixture of 2′,3′-anhydroinosine was detected in some species of 3′-deoxyinosine synthesized using radical dehalogenation. It was found that derivatives of 2′,3′-anhydroinosine inhibit the forming of 1- α-phospho-3-deoxyribose during the synthesis of 2-fluorocordycepin analogs. A mechanism of hydrolysis of 2′,3′-anhydroinosine in D2O is fully determined for the Biomolecules 2021, 11, 539 15 of 17

Performing the reaction in D2O allowed us to localize the proton positions and deuterium addition. This is useful for transient state and intramolecular rearrangements visualization. The rate constants and kinetic isotope effects for the synthesis stages of each isolated compound were determined (Table1). It can be seen that the fastest stage is the synthesis of intermediate 14, while the slowest one is the synthesis of cyclonucleoside 12. At pH 4.1, the ﬁrst stage is 1.4 times slower, and the next two stages are 3-4 times slower than at pH 7.0. The kinetic isotope effect of the solvent in the second stage is greater than in the ﬁrst. In the synthesis of compound 10, the reverse isotopic effect is observed. Neither compound 11 nor compounds formed from 11 affect E. coli PNP activity (Figure 12). Therefore, the reason for the slowdown of transglycosylation reaction in the presence of 11 is still unclear. Probably some short-living intermediates interact with the active site, with their lifetime being too short for us to isolate them. We believe that nucleosides with a structure similar to the compound shown in Scheme3 could be effective inhibitors of E. coli PNP.

5. Conclusions It is undesirable to use modified nucleosides synthesized with the help of radical dehalogenation (in particular, using Bu3SnH) as substrates of nucleoside phosphorylases. 20,30-Anhydronucleosides form during this reaction and the main product is difficult to separate from them. An admixture of 20,30-anhydroinosine was detected in some species of 30-deoxyinosine synthesized using radical dehalogenation. It was found that derivatives of 20,30-anhydroinosine inhibit the forming of 1-α-phospho-3-deoxyribose during the synthesis of 2-fluorocordycepin analogs. A mechanism of hydrolysis of 20,30-anhydroinosine in D2O is fully determined for the first time. Two intermediates containing deuterium are isolated. The structure of those compounds was confirmed by UV-, mass-, and NMR- spectroscopy.

Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/biom11040539/s1, Figure S1: The HPLC profile of the reaction mixture of 2-fluorocordycepin synthesis, Figure S2: A mass-spectrum of unknown nucleoside (10), Figure S3: The data of LC-MS 30-dIno (9) with 16% of compound (11), Figure S4: The HPLC profile of the reaction mixture obtained during 2-fluorocordycepin synthesis. Synthesis of 2-F-Cord, Figure S5: HPLC profile of the reaction mixture obtained during hydrolysis of epoxide 11 in 5 mM potassium-phosphate buffer (pH 7.0), Figure S6: HPLC profile of the reaction mixture obtained during hydrolysis of 11 in 5 mM potassium- phosphate buffer (appeared pH 7.0), D2O, Figure S7: HPLC profile of the reaction mixture obtained during hydrolysis of 11 in 5 mM potassium-phosphate buffer (appeared pH 4.1), D2O, Figure S8: A fragment of 1H NMR spectrum of 12, Figure S9: A fragment of 2H NMR spectrum of nucleoside 12, Figure S10: A fragment of 13C NMR spectrum of compound 12. NMR data for nucleosides 10, 11, 12, 13, 14— pages S-8 – S-32. Author Contributions: Conceptualization, A.L.K. and I.D.K.; methodology, A.L.K. and I.V.F.; software, A.L.K. and I.V.F.; validation, A.L.K. and I.V.F.; formal analysis, A.L.K. and I.D.K.; investigation, A.L.K., J.A.T., I.V.F., A.O.A., M.Y.B., O.I.L., E.V.D., K.V.A., A.S.P. and R.S.E.; data curation, I.D.K. and A.L.K.; writing—original draft preparation, A.L.K., I.V.F. and I.D.K.; writing—review and editing, A.L.K. and I.D.K.; visualization, A.L.K. and I.V.F.; supervision, I.D.K., I.A.M. and A.I.M. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data presented in this study are available on request from the corresponding author. Conflicts of Interest: There is no conflicts to declare. Biomolecules 2021, 11, 539 16 of 17

Abbreviations

2-Cl-Cord 2-chlorocordycepin 2-F-Ade 2-fluoroadenine 2-F-Ado 2-fluoroadenosine 2-F-Cord 2-fluorocordycepin 30-dIno 30-deoxyinosine ADA adenosine deaminase Cord cordycepin (30-deoxyadenosine) DMSO dimethyl sulfoxide HMBC heteronuclear multiple bond correlation spectroscopy HSQC heteronuclear single quantum correlation spectroscopy Hyp hypoxanthine Ino inosine KIE kinetic isotope effects PNP purine nucleoside phosphorylase

References 1. Cunningham, K.G.; Manson, W.; Spring, F.S.; Hutchinson, S.A. Cordycepin, a metabolic product isolated from cultures of Cordyceps militaris (Linn). Nature 1950, 166, 949. [CrossRef] 2. Cunningham, K.G.; Hutchinson, S.A.; Manson, W.; Spring, F.S. 508. Cordycepin, a metabolic product from cultures of Cordyceps militaris (Linn.) link. Part I. Isolation and characterization. J. Chem. Soc. 1951, 2, 2299–2300. [CrossRef] 3. Kaczka, E.A.; Trenner, N.R.; Arison, B.; Walker, R.W.; Folkers, K. Identiﬁcation of cordycepin, a metabolite of Cordyceps militaris, as 30-deoxyadenosine. Biochem. Biophys. Res. Commun. 1964, 14, 456–457. [CrossRef] 4. Rich, M.A.; Meyers, P.; Weinbaum, G.; Cory, J.G.; Suhadolnik, R.J. Inhibition of human tumor cells by cordycepin. BBA 1965, 95, 194–204. [CrossRef] 5. Ng, T.B.; Wang, H.X. Pharmacological actions of Cordyceps, a prized folk medicine. Pharm. Pharmacol. 2005, 57, 1509–1519. [CrossRef][PubMed] 6. Jin, Y.; Meng, X.; Qiu, Z.; Su, Y.; Yu, P.; Qu, P. Anti-tumor and anti-metastatic roles of cordycepin, one bioactive compound of Cordyceps militaris. Saudi J. Biol. Sci. 2018, 25, 991–995. [CrossRef][PubMed] 7. Nakamura, K.; Shinozuka, K.; Yoshikawa, N. Anticancer and antimetastatic effects of cordycepin, an active component of Cordyceps sinensis. J. Pharmacol. Sci. 2015, 127, 53–56. [CrossRef][PubMed] 8. Tang, J.P.; Qian, Z.Q.; Zhu, L. Two-step shake-static fermentation to enhance cordycepin production by cordyceps militaris. Chem. Eng. Trans. 2015, 46, 19–24. 9. Suman, K.V.; Lundback, T.; Yeheskieli, E.; Sjoberg, B.; Gustavsson, A.-L.; Svensson, R.; Olivera, C.G.; Eze, A.A.; Koning, H.P.; Hammarstrom, G.J.; et al. Structure−Activity Relationships of Synthetic Cordycepin Analogues as Experimental Therapeutics for African Trypanosomiasis. J. Med. Chem. 2013, 13, 45–58. 10. Vodnala, S.K.; Lundbäck, L.; Yeheskieli, E.; Sjöberg, B.; Gustavsson, A.-L.; Svensson, R.; Olivera, G.C.; Eze, A.A. Structure-activity relationships of synthetic Cordicepin analogues as experimental therapeutics for African Thrypanosomiasis. J. Med. Chem. 2013, 56, 9861–9873. [CrossRef][PubMed] 11. Filler, R. Fluorine in medical chemistry: A century of progress and a 60-year retrospective of selected highlights. Future Med. Chem. 2009, 1, 777–791. [CrossRef] 12. Strunecka, A.; Patocka, J.; Connett, P. Fluorine in medicine. J. Appl. Biomed. 2004, 2, 141–150. [CrossRef] 13. Liu, P.; Sharon, A.; Chu, C.K. Fluorinated nucleosides: Synthesis and biological implication. J. Fluor. Chem. 2008, 129, 743–766. [CrossRef] 14. Begue, J.-P.; Bonnet-Delpon, D. Recent advances (1995–2005) in ﬂuorinated pharmaceuticals based on natural products. J. Fluor. Chem. 2006, 127, 992–1012. [CrossRef] 15. Esipov, R.S.; Gurevich, A.I.; Chuvikovsky, D.V.; Chupova, L.A.; Muravyova, T.I.; Miroshnikov, A.I. Overexpression of Escherichia coli genes encoding nucleoside phosphorylases in the pET/Bl21(DE3) system yields active recombinant enzymes. Protein Expr. Purif. 2002, 24, 56–60. [CrossRef][PubMed] 16. Barai, V.N.; Zinchenko, A.I.; Eroshevskaya, L.A.; Zhernosek, E.V.; De Clercq, E.; Mikhailopulo, I.A. Chemo-enzymatic synthesis of 3-deoxy-β-d-ribofuranosyl purines. Helv. Chim. Acta 2002, 85, 1893–1900. [CrossRef] 17. Barai, V.N.; Zinchenko, A.I.; Eroshevskaya, L.A.; Zhernosek, E.V.; Balzarini, J.; De Clercq, E.; Mikhailopulo, I.A. Chemo-enzymatic synthesis of 3-deoxy-β-d-ribofuranosyl purines and study of their biological properties. Nucleos. Nucleot. Nucl. Acids 2003, 22, 751–753. [CrossRef][PubMed] 18. Denisova, A.O.; Tokunova, Y.A.; Fateev, I.V.; Breslav, A.A.; Leonov, V.N.; Dorofeeva, E.V.; Lutonina, O.I.; Muzyka, I.S.; Esipov, R.S.; Kayushin, A.L.; et al. The Chemoenzymatic Synthesis of 2-Chloro- and 2-Fluorocordycepins. Synthesis 2017, 49, 4853–4860. 19. Mengel, R.; Muhs, W. Urnwandlung von Inosin in 20- und 30-Desoxy- sowie 20,30-Anhydroinosin. Liebigs Ann. Chem. 1977, 10, 1585–1596. [CrossRef] Biomolecules 2021, 11, 539 17 of 17

20. Robins, M.J.; Fouron, Y.; Mengel, R. Nucleic Acid Related Compounds. 11. Adenosine 20,30-ribo-Epoxide. Synthesis, Intramolecu- lar Degradation, and Transformation into 30-Substituted Xylofuranosyl Nucleosides and the lyxo-Epoxide. J. Org. Chem. 1974, 39, 1564–1570. [CrossRef] 21. Robins, M.J.; Mengel, R.; Jones, R.A.; Fouron, Y. Nucleic Acid Related Compounds. Transformation of Ribonucleoside 20,30-O- Ortho Esters into Halo, Deoxy, and Epoxy Sugar Nucleosides Using Acyl Halides. Mechanism and Structure of products. J. Am. Chem. Soc. 1976, 98, 8204–8213. [CrossRef][PubMed] 22. Robins, M.J.; Wilson, J.S.; Madej, D.; Low, N.H.; Hansske, F.; Wnuk, S.F. Nucleic Acid-Related Compounds. 88. Efﬁcient Conver- sions of Ribonucleosides into Their 20, 30-Anhydro, 20(and 30)-Deoxy, 20,30-Didehydro-20, 30-dideoxy, and 20,30-Dideoxynucleoside Analogs. J. Org. Chem. 1995, 60, 7902–7908. [CrossRef]