The aqueous N-phosphorylation and N-thiophosphorylation of aminonucleosides

We demonstrate N-phosphorylation and N-thiophosphorylation of unprotected aminonucleosides in aqueous media. N-Phosphorylations using phosphoric chloride and N-thiophosphorylations using thiophosphoryl chloride were explored as functions of pH using 50-amino-50-deoxyguanosine as substrate. These reagents were compared to phosphodichloridate and thiophosphodichloridate ions, and the methodology was applied to other aminonucleosides. S-Alkylations of the nucleoside N-thiophosphoramidates were investigated as functions of pH and alkylating agent.


Introduction
Nucleoside phosphates are ubiquitous in living systems, with nucleic acids, adenosine triphosphate, and nucleotide sugars being key examples. The synthesis of analogues of these compounds in which the phosphate groups have been modied has been widely explored. [1][2][3][4] Phosphate-modied nucleotides have a number of applications, including investigations of the roles of phosphate in the binding of nucleoside monophosphates to enzymes. 5,6 Modied nucleosides and nucleotides also offer reactive functionalites for bioconjugation. [7][8][9][10] Standard methods for the synthesis of nucleoside phosphates, such as the Poulter 11 and Yoshikawa 12 syntheses, are frequently cumbersome, and require rigorously dry conditions and laborious purication. The use of the phosphoramidite method, while highly efficient for the synthesis of oligonucleotides when the protected nucleoside phosphoramidites are commercially available, becomes more laborious and requires global protection and rigorously dry conditions where the de novo syntheses of non-standard phosphoramidites are required. 13 These problems motivated us to nd a simple method for the synthesis of analogues of nucleoside monophosphates and phosphodiesters, building on our previous work in the area. [14][15][16] Whilst previous syntheses of the 5 0 -N-phosphoramidates of guanosine 17 and other nucleosides 5,6,18 have been reported, these methods suffer from similar drawbacks to the Yoshikawa synthesis. The issues centre on poor nucleoside solubility and the need for dry conditions. In contrast, we nd that the aqueous phosphorylation method is more convenient, where a nucleophilic aminonucleoside allowed phosphorylation to be carried out in water, with high conversion, and good N-selectivity. 14,15 The methodology was extended to the use of thiophosphoryl chloride as a (thio)phosphorylation agent on a series of generic amines, with the advantage that the thiophosphoramidates thus formed can then be S-alkylated to produce analogues of phosphodiester systems. 19 Here we seek to explore, expand and optimise the aqueous phosphorylations and thiophosphorylations of aminonucleosides through pH control, and to compare phosphorylating and thiophosphorylating agents. We have also previously reported on the hydrolysis kinetics of phosphodichloridate (Cl 2 OPO À ) and thiophosphodichloridate (Cl 2 SPO À ) ions, 20 and here we assess their suitability as phosphorylating agents, compared to their counterparts POCl 3 and PSCl 3 .

Results and discussion
pH control and optimisation 5 0 -Amino-5 0 -deoxyguanosine 3 21 was used as a model substrate to determine pH optima for phosphorylation procedures. The 5 0 -amino-5 0 -deoxyguanosine 3 was dissolved in water and adjusted to a predetermined pH value with potassium hydroxide solution. A single equivalent of the phosphorylating agent in acetonitrile was then added slowly to the solution whilst the pH was kept constant using an autotitrator to add potassium hydroxide solution as required.
The 5 0 -amino-5 0 -deoxyguanosine 3 was either phosphorylated or thiophosphorylated using one of phosphoryl chloride 1(O), thiophosphoryl chloride 1(S), potassium phosphodichloridate 2(O) or potassium thiophosphodichloridate 2(S) at pH values of 11, 11.5, 12, and 12.5 (Fig. 1). Aer addition of the phosphorylating/thiophosphorylating agent, the conversion levels were determined by 31 P NMR spectroscopy. The N-phosphoramidate and N-thiophosphoramidate monoesters are unstable, except at high pH, 14,22,23 making them difficult to isolate, however, 5 0 -amino-5 0 -deoxyguanosine 5 0 -N-phosphoramidate has been isolated and characterised by 31 P NMR and 1 H NMR spectroscopies, and these data were used to corroborate the observed signals. 14 The presence of the N-thiophosphoramidate was conrmed by trapping this reactive species through S-alkylation and isolating and characterising the stable adduct, in a manner analogous to our earlier report. 22 Apart from the desired N-phosphoramidate 5 and N-thiophosphoramidate 8, we also observed the formation of bisaminolysis side products 6 and 9, in which the phosphorylating agent was attacked by two equivalents of amine, and inorganic phosphate 7 and thiophosphate 10, which result from the breakdown of the phosphorylating agents. These results mimic those we have seen previously for a range of generic amines 4 (Scheme 1). 19,22 The results show optima at pH 12 for each of the four (thio) phosphorylating agents 1(O), 1(S), 2(O) and 2(S) (Fig. 2). The Fig. 1 The phosphosphorylating agents 1 and 2 used in this study, and the model substrate, 5 0 -amino-5 0 -deoxyguanosine 3.

Scheme 1
The observed products and side-products of the phosphorylation and thiophosphorylation reactions. R ¼ 5 0 -deoxyguanosyl. optima were especially pronounced for the thiophosphorylation reactions; the greatest levels of conversion were achieved using the thiophosphodichloridate ion 2(S), with a maximum of 97% conversion to thiophosphoramidate 8 at pH 12, as opposed to 87% at pH 11. The thiophosphorylation reactions with thiophosphoryl chloride 1(S) reect the same trend, with a maximum of 93% at pH 12, and minimum of 79% at pH 11.
The effect of pH on the oxyphosphorylation appears to be less pronounced. The greatest level of conversion was again achieved using the phosphodichloridate ion 2(O) at pH 12, with 83% to 76% being observed over the tested pH range. The reactions with phosphoryl chloride 1(O) were marginally poorer with conversions in the range 80% to 72%. Analysis of the byproducts revealed a clear advantage to the use of the phosphodichloridate 2(O), namely, the lack of formation of bis aminolysis product 6.
Considering the structure of the substrate, 5 0 -amino-5 0deoxyguanosine 3 and the (thio)phosphorylating agents, several factors could account for these trends. High pHs are required to stabilise the (thio)phosphoramidate products, however, they also serve to solubilise the 5 0 -amino-5 0 -deoxyguanosine 3 substrate through ionisation of the N-H of the guanine. At higher pHs, the ionisation of the cis-diol functionality could contribute to a change in reaction outcome (pK a $ 12.5). 24 In the case of the dichloridate ions 2(O) and 2(S), hydroxide-promoted hydrolysis of the (thio)phosphorylating agent is unlikely in light of our previous studies upon these species, which reveal at pH-k obs proles. 20 With trichlorides 1(O) and 1(S), their enhanced reactivity could render them more susceptible to hydrolysis, however, their hydrolysis products are the dichloridates 2(O) and 2(S), which permit additional opportunity for nucleophilic attack (loss of the "third" chloride is likely to be unselective towards the nature of the nucleophile, and thus would reect nucleophile concentration, where water is clearly the most abundant entity). Solubility phenomena and the partitioning of reaction partners between different phases could also inuence reaction outcome signicantly, however, the effects of these properties are difficult to predict.
5 0 -Amino-5 0 -deoxyuridine was formed by adapting a single step procedure developed by Hata et al. to convert uridine to 5 0azido-5 0 -deoxyuridine 14, 27 which was reduced to the desired amine using triphenylphosphine, 26 and isolated as its hydrochloride salt 15$HCl (Scheme 2).
To our knowledge 5 0 -amino-5 0 -deoxycytidine 20 has not been reported. We protected the exocyclic amino group of cytosine 16 via benzoylation. 28 The protection of the 2 0 -and 3 0 -hydroxyl groups and the tosylation of the 5 0 -hydroxyl group were adapted from procedures developed by Winans and Bertozzi. 29 The crude tosylated material was reacted with sodium azide in DMSO in an adapted literature procedure 30 to afford azide 18 which was isolated by precipitation in a large volume of water. The protecting groups were removed, 31 and 5 0 -azide 19 was isolated by cation exchange chromatography. The azide was then reduced with triphenylphosphine 26 and the amine was isolated as the dihydrochloride salt 20$2HCl (Scheme 3).
The conversion levels observed in the experiments are shown in Fig. 3. For the thiophosphorylations, conversions were generally high, especially for the reactions of 5 0 -amino-5 0 -deoxyguanosine 3 with thiophosphodichloridate 2(S) and 5 0 -amino-5 0deoxyadenosine 11 with thiophosphoryl chloride 1(S), with conversions of 97% and 99% respectively. A notable exception to the high thiophosphorylation conversions is the reaction with 3 0 -amino-3 0 -deoxythymidine 12; this is likely due to the more sterically hindered environment of the 3 0 -amine nucleophile in this example.
The use of thiophosphoryl chloride 1(S) gave greater levels of conversion than the thiophosphodichloridate ion 2(S), with the exception of the reaction with 5 0 -amino-5 0 -deoxyguanosine 3. Unlike the thiophosphodichloridate ion, thiophosphoryl chloride shows very limited solubility in water. This may give rise to the improved selectivity, despite thiophosphoryl chloride's greater reactivity, possibly through reaction at solvent-water interfaces rather than in homogeneous solution.
In other cases, the use of potassium thiophosphodichloridate 2(S) afforded fewer side-products such as the bis thiophosphoramidate 9. For the oxyphosphorylating agents 1(O) and 2(O), the trend appears to be reversed; the phosphodichloridate 2(O) reactions tended to give the greatest levels of conversion, possibly due to the reduced reactivity of 2(O) allowing better mixing and thus greater selectivity than with phosphoryl chloride 1(O). In general, the levels of conversion for the oxyphosphorylations are more modest than those of the thiophosphorylations with a maximum of 83% for the reaction of 5 0 -amino-5 0 -deoxyguanosine 3 with phosphodichloridate 2(O). 3 0 -Amino-3 0 -deoxythymidine again showed the poorest conversions, at 44% and 51% for the reactions with phosphorylating agents 1(O) and 2(O) respectively.

Thiophosphoramidate S-alkylation
The S-Alkylation of thiophosphoramidates 21a-d(S) could provide access to mimics of naturally occurring phosphodiesters such as CMP-Neu5Ac or analogues for mechanistic studies. 4 Additionally, there are many commercially available alkylating agents, and this reaction could allow the rapid development of a diverse range of derivatives. We conducted preliminary studies in this vein recently, where the products of the N-thiophosphorylation-S-alkylation of 5 0 -amino-5 0 -deoxyguanosine 3 and 5 0 -amino-5 0 -deoxyadenosine 11 were prepared and isolated from non-pH-controlled reactions, and these results served as our starting point. 22 Here we investigate the pH optimisation of this procedure. Using our model substrate, 5 0amino-5 0 -deoxyguanosine 3, we performed N-thiophosphorylation, and the resulting nucleoside-N-thiophosphoramidate anion was then S-alkylated with one of three alkylating agents, each at xed pHs using an autotitrator (Scheme 5).
We chose benzyl chloride, methyl iodide, and 2-bromoethanol as alkylating agents, using guanosine-N-thiophosphoramidate 4(S) as the substrate. Guanosine-N-thiophosphoramidate 4(S) was prepared under optimised conditions, where >95% of the product represented the desired material, and this material was used directly in S-alkylation procedures. The conversion levels determined by 31 P NMR spectroscopy for each of the alkylating agents at each pH are illustrated in Fig. 4.
To conrm the intermediacy of the unstable, unalkylated precursor N-thiophosphoramidate 4(S), three S-alkylation experiments were performed on 4(S) at pH 12, and the products were isolated by ion exchange chromatography and characterised. Isolated yields of 95%, 63%, and 74% were recorded for the S-methyl 22a, benzyl 22b, and ethan-2-ol 22c systems. S-Alkylation efficiency appeared to be largely unaffected by pH, at least within the range that we studied, with conversions levels being $80%. Surprisingly, S-alkylation occurs at lower conversion levels than N-thiophosphorylation, which may be due, in part, to the hydrolysis of the desired products under the high pH conditions. In particular, the formation of a nascent "good" thiolate leaving group through S-alkylation could represent a likely pathway for hydrolysis, and we are currently studying the hydrolysis kinetics of the S-alkylated thiophosphoramidates to give greater insight.

Conclusions
Effective N-oxyphosphorylation and N-thiophosphorylation of aminonucleosides 3, 11, 12, 15 and 20 was achieved through pH control, with high levels of conversion. Phosphodichloridates 2(O) and 2(S) were competent alternative phosphorylation agents, that eliminated the formation of undesired bis-amino(thio)phosphoramidates, while retaining similar levels of conversion. The S-alkylation of guanosine-N-thiophosphoramidate 4(S) showed no signicant pH-sensitivity across the pH range 9-12, although conversion levels were not quantitative. The combination of efficient N-(thio)phosphorylation and S-alkylation gives straightforward access to phosphodiester mimics from a range of unprotected aminonucleosides derived from the common nucleobases.

N4-Benzoylcytidine
In an adapted literature procedure, 28 cytidine (2.784 g, 11.4 mmol) and benzoic anhydride (2.818 g, 12.5 mmol) were placed in a round bottomed ask with dry methanol (300 ml) and heated at reux with stirring. Aer 1 h, additional benzoic anhydride (2.774 g, 12.3 mmol) was added, followed by further additions (2.785 g, 12.3 mmol aer 2 h, 2.794 g, 12.4 mmol aer 3 h). Heating was maintained for a total of 5 h. Aer allowing the reaction vessel to cool for 18 h, the precipitate was collected by vacuum ltration. Drying over P 2 O 5 under vacuum yielded the desired product (2.786 g, 70%), mp 236-238 C (from methanol) (lit., 32 238-240 C); n max /cm À1 3421, 3307, 3161, 1644 (CO); d H (400 MHz, (CD 3 ) 2 SO) 3.61 (1H, ddd, J 12.3, 5.2, 3.1, Fig. 4 The conversion levels by 31 P NMR spectroscopy of S-alkylations of guanosine-N-thiophosphoramidate 4(S) to S-alkylthiophosphoramidates 22a-c using methyl iodide, benzyl chloride, and 2-bromoethanol, respectively.  N4-Benzoyl-2 0 ,3 0 -O-isopropylidenecytidine (17) In an adapted literature procedure, 29 N4-benzoylcytidine (2.79 g, 8.03 mmol), 2,2-dimethoxypropane (11 ml), 4Å molecular sieves, and tosic acid monohydrate (0.45 g, 2.37 mmol) were placed in a round-bottomed ask with DMF (44 ml). The mixture was heated at 40 C for 2.5 h, before Amberlyst® A-21 anion exchanger (1.19 g) was added and heating was continued for a further 30 min. The mixture was ltered through Celite®, the solvent was then removed under vacuum from the ltrate, and the residue was recrystallised from water to yield the product  (20 ml) and DCM (40 ml) were placed in a round-bottomed ask and heated at reux for 2.5 h. The solution was then diluted with chloroform (120 ml) and washed with hydrochloric acid (0.5 M, 5 Â 50 ml) and saturated sodium hydrogencarbonate solution (2 Â 50 ml). The organic layer was dried over MgSO 4 and the solvent was removed under vacuum to yield the crude product, which was used in the next reaction without further purication. (1.53 g).  Based on a literature procedure, 31 5 0 -Azido-5 0 -deoxy-N4-benzoyl-2 0 ,3 0 -O-isopropylidene cytidine (732 mg, 1.78 mmol) was placed in a round-bottomed ask with a 1 : 1 mixture of methanol and concentrated ammonium hydroxide (90 ml) and was stirred at room temperature for 18 h. The solvents were removed under vacuum, and the residue was stirred in a 9 : 1 mixture of tri-uoroacetic acid and H 2 O for 3 h at room temperature before the solvents were removed under vacuum. The residue was dissolved in H 2 O and introduced to a protonated SP Sepharose® Fast Flow cation exchange column (1.6 cm i.d. Â 22 cm, 5 ml min À1 ow rate). The nucleoside was eluted with a 3% ammonium hydroxide solution (diluted from 35% (w/w) ammonium hydroxide solution) and the eluted fractions were freeze-dried. The residue was then dissolved in methanol and ltered, and the solvent was removed from the ltrate under vacuum to give the solid product.  5 0 -Azido-5 0 -deoxycytidine (400 mg, 1.49 mmol) was placed in a ask with triphenylphosphine (790 mg, 3.01 mmol) and pyridine (6 ml). The solution was stirred at room temperature for 3 h before further triphenylphosphine (791 mg, 3.02) and pyridine (6 ml) were added. The solution was stirred for an additional 3 h at room temperature before ammonia solution (50 ml, 35% w/w) was added, whereupon immediate precipitation was observed. The mixture was stirred overnight before being diluted with water (100 ml) and extracted with chloroform (3 Â 100 ml). Residual chloroform was removed from the aqueous layer under vacuum, and the aqueous extracts were lyophilised. The resulting solid was dissolved in ethanol (60 ml) and hydrogen chloride gas was bubbled through the solution. The addition of diethyl ether (300 ml) precipitated the product, which was isolated by ltration,  Adapting a literature procedure, 26 3 0 -amino-3 0 -deoxythymidine (1.00 g, 3.74 mmol) and triphenylphosphine (1.54 g, 5.87 mmol) were dissolved in pyridine (8 ml) and stirred at room temperature for 1 h. Ammonia solution (30 ml, 35%) was then added, and the mixture le to stir overnight. The suspension was diluted with water (30 ml) and extracted with chloroform (3 Â 30 ml) before being lyophilised. The solid residue was dissolved in ethanol (100 ml) and hydrogen chloride gas was bubbled through the solution until precipitation was observed. The precipitate was isolated by ltration, and washed with a small quantity of diethyl ether. Additional product was obtained by adding diethyl ether (500 ml) to the ltrate, and again ltering and washing the precipitate.

Lithium azide
Following a literature procedure, 33 sodium azide (6.52 g, 100 mmol) and lithium sulfate (6.90 g, 62.8 mmol) were co-dissolved in water (35 ml) and stirred for 10 min. Ethanol (175 ml) was slowly added, and the mixture was stirred for a further 10 min. The precipitate was removed by ltration, and the solvent was removed from the ltrate under vacuum. The solid residue was dried on the high vacuum line to yield the product (4.818 g, 98%). 5 0 -Azido-5 0 -deoxyuridine (14) With some modications, a literature procedure was followed. 27 Uridine (5.00 g, 20.5 mmol), triphenylphosphine (8.05 g, 30.5 mmol), tetrabromomethane (10.2 g, 30.5 mmol), and lithium azide (3.695 g, 75.5 mmol) were stirred together in DMF (100 ml) for 4 h at room temperature. The greater part of the solvent was removed under vacuum, and the residue was extracted with chloroform (100 ml) and water (100 ml), with the aqueous layer being retained. The aqueous layer was washed again with chloroform (2 Â 100 ml) before being lyophilised. The viscous residue was puried by column chromatography, using an isocratic 5 : 1 mixture of chloroform and methanol. The resulting oil was dissolved in 5 : 1 chloroform-methanol and adsorbed on to a plug of silica, which was washed with 100% chloroform to remove residual DMF, then released using the 5 : 1 chloroform-methanol solution to yield the puried product.  Adapting a literature procedure, 26 5 0 -azido-5 0 -deoxyuridine (404 mg, 1.50 mmol) and triphenylphosphine were (1.66 g, 6.33 mmol) were dissolved in pyridine (1.6 ml) and stirred for 2 h. Ammonia solution (35%, 12 ml) was then added, and the mixture le to stir overnight. Water (40 ml) was added, and the mixture was extracted with chloroform (3 Â 40 ml). Residual chloroform was then removed from the aqueous layer under vacuum, and the remaining solution was lyophilised. The resulting powder was then dissolved in ethanol (20 ml) and heated at reux until dissolved. Hydrogen chloride gas was bubbled through the solution until precipitation occurred, and the pure hydrochloride salt was isolated by ltration and washing with a small quantity of ethanol.

Phosphorylating agents
The preparation of potassium phosphodichloridate and potassium thiophosphodichloridate has been reported previously. 20,34 Phosphorylation procedure The aminonucleoside (0.500 mmol) was placed in a thermostated reaction ask maintained at 25 C and water and potassium hydroxide were added as required to make a 5 ml solution at the desired pH. The phosphorylating agent (1.50 ml, 0.333 M in MeCN) was added using a Hamilton® microlitre syringe. The addition took place over 10 min with vigorous stirring, and with the tip of the syringe below the surface of the reaction mixture. Throughout the experiment, the pH was kept constant using a 1 M solution of potassium hydroxide, added by the autotitrator system. The experiment was considered to be complete when the autotitrator needed to add negligible quantities of potassium hydroxide solution to the reaction mixture. Aer the completion of the reaction, the organic solvent was removed under vacuum and aqueous remainder was lyophilised.