Peroxide-Mediated Release of Organophosphates from Boron-Containing Phosphotriesters: A New Class of Organophosphate Prodrugs

Phosphate mono- and diesters can be liberated efficiently from boryl allyloxy (BAO) and related phosphotriesters by H2O2. This protocol was applied to the release of a phosphorylated serine derivative and the nucleotide analogue AZT monophosphate. Nucleotide release in the presence of ATP and a kinase provides a diphosphate, demonstrating that this method can be applied to biological processes.

T he phosphate group is an essential component of a wide range of biologically active molecules with disparate structures including nucleotide analogues (1) 1 and secondary metabolites such as fostriecin (2) (Figure 1). 2 Nucleoside/ nucleotide anticancer and antiviral agents illustrate the challenges of employing phosphorylated molecules as therapeutic agents. 1 Changes to nucleobases and/or the sugar unit can lead to inefficient nucleoside phosphorylation, which is necessary for disrupting nucleic acid chain elongation, thereby diminishing therapeutic efficacy. Nucleotide analogues subvert the initial phosphorylation problem since the phosphate group is already present at the 5′-site of the structure. However, the charged group substantially diminishes cell permeability and mitigates biological activity. 3 Phosphate prodrugs address this problem (Figure 2), as seen in the clinical agents sofosbuvir (3) and remdesivir (4). 4 These neutral agents enter cells and release nucleotide analogues through esterase-initiated cascades. Other phosphate prodrugs 5 employ the esterase labile pivaloylmethyl and S-acyl-2thioethyl groups, while cyclic phosphotriester groups are activated by acidic hydrolysis and cytochrome P450 oxidation.
Applications of organophosphate prodrugs have been impressive, though site-selective release is challenging due to the ubiquitous presence of the releasing enzymes. Enhancing the targeted release capacity for phosphate-containing drugs offers the potential to increase their efficacy while minimizing off-target effects. Hydrogen peroxide is generated in numerous disease states, including cancer, viral infection, neurodegeneration, reperfusion injury, and arthritis. 6 Elevated levels of H 2 O 2 and its ability to initiate chemical reactions create the potential for its use in drug release, illustrated by the release of fluorophores 7 and cytotoxins 8 from boronates in cells and animals. Our prior studies in which boronate oxidation initiates fragmentation reactions that release polar groups through the generation of boron enolates from boryl allyloxy (BAO) groups 9 or hemiacetal intermediates from α-boryl ethers 10 led us to explore the potential of peroxide-mediated phosphate release. Herein we describe the development of a new phosphate group-release strategy based on the peroxidemediated oxidation of borylated phosphoesters (Scheme 1).
The process is rapid and efficient and can be applied to the release of a range of phosphate monoesters and diesters. Structural variants are explored to demonstrate the capacity for changing the byproduct without compromising the release efficiency. Biologically and medicinally relevant structures are liberated, and nucleotide release provides a viable substrate for subsequent kinase-mediated phosphorylation.
Proof-of-concept substrates for demonstrating phosphate release were prepared through classical phosphoramidite chemistry (Scheme 2). The substrate for phosphodiester release was prepared from 5 through sequential additions of nPrOH and allylic alcohol 6 9 followed by oxidation with tBuOOH to yield 7. The phosphomonoester-releasing substrate 8 was prepared by swapping the order of the nucleophile additions. These reactions are viable without intermediate phosphoramidite purification, though significantly higher throughput is achieved with purified phosphoramidites. The different yields for these reactions show that 6 is less nucleophilic than nPrOH and that the first step is more challenging than the second. 1 H NMR served as a convenient tool for monitoring phosphate release. Substrates were prepared as a solution in CD 3 CN and D 2 O in the presence of buffer (pH 7.4) followed by the addition of excess H 2 O 2 ·urea (30 equiv) in D 2 O at 37°C . The final concentration of the substrate was 2 mM with a CD 3 CN:D 2 O ratio of 7:3 (v/v). The resulting H 2 O 2 concentration is higher than cellular levels, 11 though prior related studies 10 show that this concentration predicts cellular responses. Product concentrations were quantified against an internal standard. The products of these reactions were validated by comparison to independently prepared and characterized phosphates. 12 Time courses for the oxidative cleavage reactions of 7 and 8 are shown in Figure 3. The cleavage of 7 proceeded rapidly and efficiently via the intermediate boron enolate, with 94% of the starting material being consumed within 18 min and phosphate 9 being produced in an equal amount. This experiment showed a pseudo-first-order rate constant of 2.3 × 10 −3 L mol −1 sec −1 and a half-life of 301 s. 12 The reaction with 8 was more complex since two cleavage events are required. Spectral overlap prohibited monitoring the formation of the phosphomonoester, though acrolein production is easily detected. The  reaction showed a rapid consumption of 8, with the concentration of the monocleavage product 10 increasing and then ultimately becoming negligible at 24 min. Subsequent NMR analysis confirmed that the product was phosphate 11. Acrolein was formed in an 88% yield based on the two equiv that are expected from the cleavage of two BAO groups. The second BAO cleavage is significant, since it showed that phosphate dianion release is faster than boron enolate protonation. Additionally, the absence of a significant buildup of the monocleavage product indicates that the second cleavage is not substantially slower than the first. The pinacol esters partially hydrolyze to boronic acid intermediates prior to oxidation, although this does not impact cleavage efficiency. Phosphate release was not observed in the absence of peroxide or solely in the presence of urea. 12 Serveral substrates were synthesized and served to highlight the scope of the process ( Table 1). The results are reported as a function of starting material consumption and product formation after 18 min for consistency in cleavage rate comparisons, unless otherwise noted.
The phosphotriester 12 releases phenyl phosphate 13 somewhat more slowly than the breakdown of 8 (entry 1). While 90% of the starting material was consumed after 18 min, dicleavage product was formed in 63% yield, and 27% of the corresponding monocleavage product remained. This proceeded to 70% and 23% of the respective products after 25 min. 12 We speculate that the rate arises from solubility issues, as turbidity was observed during the course of the reaction. The phosphorothioate 14 releases thiophosphate 15 in 79% yield after 18 min and 94% yield after 49 min, with turbidity again being observed as a potential source of the slowed release. Delivering thiophosphates, which are substantially more stable toward enzymatic cleavage than phosphates, 13 is significant, regardless of the release rate. Phosphodiesters that contain an aryl and an alkyl group cleave selectively as shown in the conversion of 16 to 17. Aryl phosphates can be cleaved enzymatically to release monoalkyl phosphates. 14 Compound 18, in which the BPin group is replaced by the recently reported and more easily handled EPin group, 15 releases 9 extremely efficiently, albeit somewhat more slowly. Consumption with quantitative product release did not occur until 32 min. Boronate 18 is significantly more hydrolytically stable than 7, which could be useful for applications in which the boronic acid analogue does not readily traverse a cell membrane. The diminished rate of release will be valuable when a slower phosphate production is therapeutically beneficial.
Acrolein is an inhalation toxin that can sequester glutathione, 16 leading to concerns about its formation during prodrug cleavage, though the release of a glutathione scavenger could augment the potency of these agents in numerous applications. 17 Although our studies with deliberate acrolein release through the cleavage of BAO-containing molecules showed no toxicity, 12 the potential for a competitive biological response inspired us to prepare oxidatively cleavable boronate analogues that mitigate byproduct electrophilicity. The α-boryl phosphate 19 reacts with peroxide to form 9 within 12 min, indicating that this group reacts even faster than the BAO group. Boronates 20 and 22 (secBAO) are homologues of the original vinyl boronate and release methyl vinyl ketone and crotonaldehyde, respectively, in addition to 21. These byproducts are significantly less electrophilic than acrolein. 18 The phosphates were prepared by the addition of the alcohols to dimethylphosphoryl chloride, since the phosphoramidite protocol was unsuccessful. Both of these substrates release dimethyl phosphate in the presence of H 2 O 2 , though 20 showed an alternate, slower mechanism for release in the  12 This study shows that a range of options exist should byproduct release lead to a competitive biological response. Serine derivative 23 served as a model for the conditional release of phosphoserine. The process was conducted with the initial concentration of 23 being 0.2 mM in biologically relevant PBS buffer, which is an order of magnitude more dilute than the previous studies to observe the capacity for oxidative cleavage at low substrate and oxidant concentrations. Serine phosphate 24 was formed in 89% yield after 60 min and quantitatively after 140 min as determined by HPLC, showing that phosphorylated amino acid units can be released under conditional control in peptides and proteins. 19 Insterestingly, replacing the BAO groups of 8 with p-borylbenzyl groups 7,8 resulted in oxidation but only monocleavage. 12 Thus, the BAO group is substantially more effective for releasing dianionic phosphates. 20 We explored the potential for nucleotide release through the preparation of AZT-derived phosphate 25 in which the boronic acid (HO-BAO group) proved desirable for isolation purposes (Scheme 3). 21 1 H NMR studies showed that 25 reacts with H 2 O 2 as expected to release AZT monophosphate (AZTMP) 26 in 91% yield after 25 min. Moreover, the negative control phosphate 27 proved to be completely inert under these conditions. 12 Successful nucleotide analogue release led us to explore the viability of using peroxide-release to initiate the enzymatic phosphorylation of 26 in the presence of ATP to form AZT diphosphate 28 and ADP ( Figure 4A). The F105Y mutant of thymidine monophosphate kinase (TMPK) was expressed and served to promote phosphorylation in this study. 22 The phosphorylation assay ( Figure 4B) employs the conversion of luciferin to oxyluciferin to provide an ATP-dependent luminescent signal in which ATP consumption through the phosphorylation of 26 can be determined by a luciferase assay.
Prior to conducting phosphorylation studies, we showed that ATP is stable to H 2 O 2 , 25, and a mixture of 25 and H 2 O 2 for 1 h. The assays were conducted by combining the substrate (2 μM), ATP (10 μM), and TMPK. Hydrogen peroxide (30 equiv) was added for the oxidative release studies of 25. After 1 h CellTiter-Glo reagent was added, and luminescence was measured after 5 min. The results are shown in Figure 5. AZTMP showed a 20% reduction in luminescence, indicating complete conversion to 28. The negative controls 27 and H 2 O 2 -free 25 failed to initiate ATP consumption. The combination of 25 and H 2 O 2 showed a level of ATP consumption equivalent to that of the positive control 26, confirming that oxidative cleavage provides a nucleotide analogue that can engage in enzymatic reactions.
We have shown that a range of organoboronates can be incorporated into phosphotriesters to release phosphoesters upon exposure to H 2 O 2 . The reactions are rapid and efficient, even at low substrate and peroxide concentrations. Applying the protocol to AZT monophosphate release demonstrated that the released nucleotide is a suitable substrate for enzymatic phosphorylation, validating the compatibility of the process with biomolecular transformations. The elevated levels of H 2 O 2 in numerous disease states indicate that this approach will be applicable to the development of site-selective phosphate-containing prodrugs.   The data underlying this study are available in the published article and the Supporting Information.