Lipophilic prodrugs of nucleoside triphosphates as biochemical probes and potential antivirals

The antiviral activity of nucleoside reverse transcriptase inhibitors is often limited by ineffective phosphorylation. We report on a nucleoside triphosphate (NTP) prodrug approach in which the γ-phosphate of NTPs is bioreversibly modified. A series of TriPPPro-compounds bearing two lipophilic masking units at the γ-phosphate and d4T as a nucleoside analogue are synthesized. Successful delivery of d4TTP is demonstrated in human CD4+ T-lymphocyte cell extracts by an enzyme-triggered mechanism with high selectivity. In antiviral assays, the compounds are potent inhibitors of HIV-1 and HIV-2 in CD4+ T-cell (CEM) cultures. Highly lipophilic acyl residues lead to higher membrane permeability that results in intracellular delivery of phosphorylated metabolites in thymidine kinase-deficient CEM/TK− cells with higher antiviral activity than the parent nucleoside.

O ver the last decades, a variety of nucleoside analogues were applied in antitumour and antiviral therapy and still play an important role to combat HIV, herpes virus, hepatitis B and hepatitis C virus infections 1,2 . The target of these nucleoside analogue drugs is the inhibition of the virus-encoded DNA polymerases, such as the HIV reverse transcriptase (RT) 3,4 or the HCV-encoded RNA-dependent RNA-polymerase NS5B (ref. 5), which are the key enzymes in the replication cycle of HIV and HCV, respectively. To date, eight nucleoside analogues have been approved as HIV RT inhibitors (NRTIs) 6 . NRTIs are still used as the backbone of the combined antiretroviral therapy 7 . However, the antiviral efficacy of nucleoside analogues, such as the thymidine analogue 3 0 -deoxy-2 0 ,3 0 -didehydrothymidine 1 (d4T) and 3 0 -deoxy-3 0 -azidothymidine (AZT), is dependent on the activity of host cell kinases metabolizing these nucleoside analogues into their antivirally active triphosphate forms (nucleoside triphosphates, NTPs) [8][9][10][11] .
The stepwise transformation via the nucleoside mono-(NMP) and diphosphates (NDP) into the corresponding NTP often occurs insufficiently because of the high substrate specificity of the involved kinases ( Supplementary Fig. 1). Furthermore, many nucleoside analogues have limitations such as poor biological half-lives, variable bioavailability after oral administration or selection of drug resistance, which reduce their clinical efficacy 12,13 . To overcome these hurdles, the usage of prodrugs of the phosphorylated metabolites have been explored in the past 14,15 .
In the case of d4T 1, the first phosphorylation step to yield its monophosphate form 2 catalysed by the host cell enzyme thymidine kinase (TK) is metabolism-limiting because of the rather modest affinity of d4T 1 to TK as an alternative substrate and because TK activity is S-phase-dependent 11,16,17 . However, to avoid this limitation, it is not possible to apply the charged monophosphorylated metabolite because of the high polarity and thereby extremely poor, if any, membrane permeability. The development of nucleotide prodrugs capable of delivering the monophosphorylated metabolite and thereby bypassing the intracellular activation offered advantages over the use of the corresponding nucleoside analogue 18,19 . Moreover, lipophilicmasked NMPs such as 3 are less vulnerable to degradation by unspecific phosphatases present in the blood. This enhanced not only the plasma half-life but also enables the prodrug to be taken up by cells through passive diffusion 20,21 . A number of successful NMP prodrug strategies were reported in the past that efficiently bypass the nucleoside kinase hurdle. These prodrug forms such as phosphoramidates and cycloSal-phosphate triesters of nucleoside analogues were indeed shown to deliver the NMP either by chemical or enzymatic hydrolysis in the target cells [22][23][24][25][26][27] . In addition to NMPs also the successful delivery of acyclic nucleoside phosphonates such as cidofovir has been reported 28 . However, all these approaches delivered the monophosphor(n)ylated forms of the nucleosides that subsequently needed further phosphorylation into the triphosphate forms by cellular kinases to inhibit their target polymerase.
However, not in all cases such NMP prodrug strategies were successful. For instance, in the case of AZT the metabolism is limited by the second conversion step, the formation of AZT-diphosphate by thymidylate kinase 10,29 . In this case, a lipophilic prodrug that intracellularly releases the NDP would be desirable. At the same time, this would avoid toxicity caused by the parent nucleoside or the accumulation of the monophosphate form 30 . A further example is 2 0 ,3 0 -dideoxy-2 0 ,3 0 -didehydrouridine (d4U). The parent nucleoside proved to be completely inactive against HIV replication in cell cultures. In contrast, the triphosphate form of d4U is one of the most effective inhibitors of the HIV's RT 31 .
We reported on NDP prodrugs (DiPPro-approach), which selectively released NDPs not only in phosphate buffer (pH 7.3) but for the first time also in CEM cell extracts [32][33][34][35] . These compounds showed very good antiviral activity in TK-deficient CEM/TK À cell cultures infected with HIV, proving the uptake of the compounds and the delivery of at least phosphorylated metabolites, most likely the NDP. The uptake of those compounds was achieved by two acceptor-substituted benzylesters linked to the b-phosphate group of the NDP. The stability of the compounds correlated with the length of the aliphatic residue of the mask 32,33 . However, we have shown that the delivery of the corresponding mono-or diphosphates from prodrug forms (such as the cycloSal-or DiPPro-compounds) did not improve the antiviral activity 34 . Importantly, in the same study we have proven that d4U diphosphate was a very poor substrate for the NDP kinase, the enzyme that is generally accepted to be involved in the conversion of NDPs into their triphosphate forms. Thus, this study showed that the phosphorylation of nucleotide analogues by NDP kinase can be also rate-limiting in the activation process of a nucleoside analogue 34,36 . As a consequence, the development of nucleoside triphosphate prodrugs would be highly interesting and desirable because this would bypass all steps of intracellular phosphorylation and would maximize the intracellular concentration of the ultimately bioactive NTP. Although this has been recognized before 11 , it was also claimed that the development of prodrugs of NTP is chemically not feasible because of the low stability of such compounds 37 . For this reason, NTPs were very rarely used as drug platforms because of their expected poor deliverability and their high sensitivity for enzymatic dephosphorylation. Thus, very few reports on potential triphosphate prodrugs have been reported 38,39 . In addition, in the few reported examples the yields in the chemical synthesis were poor, and additionally the compounds proved hydrolytically very unstable. A difficulty that has to be taken into account in the development of NTP prodrugs is related to the energy-rich phosphate anhydride bonds within the triphosphate unit. Under physiological conditions, these linkages are only kinetically stable because of the charges present at that moiety, which prevent nucleophilic reactions that end up in the cleavage of these anhydride bonds but can be enzymatically cleaved. Interestingly, c-modification of NTPs by esterification or replacement of the c-phosphate group by a phosphonate moiety led to a marked increase in enzymatic stability of the triphosphate unit 40,41 . On the other hand, complete lipophilic modification and thereby neutralization of the charges would significantly increase the reactivity of the triphosphate unit. For completeness, it should be added that the delivery challenge for NTPs has also been addressed by formulating these compounds with cationic nanogels. However, this approach still requires elaboration with respect to their toxicity, immunogenicity and pharmacokinetics 42,43 .
Here we disclose the development of a novel prodrug concept for NTPs that releases directly NTPs with high selectivity by an enzyme-triggered mechanism and thus allows the bypass of all phosphorylation steps normally needed for the activation of a nucleoside analogue ( Supplementary Fig. 2). To achieve this goal, the g-phosphate group of a NTP was modified by esterification with acyloxybenzyl moieties (TriPPPro-compounds). By fine-tuning the lipophilicity through the use of different acyl esters the chemical stability also proved controllable so that the polarity caused by the remaining charges at the aand b-phosphates could be efficiently compensated.
The cleavage of the prodrug moieties is initiated by an enzymatic hydrolysis of the phenolic acyl-ester. This reaction leads to a spontaneous cleavage of the benzyl C-O bond forming ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9716 first a monomasked intermediate of the NTPs. This process is repeated for the second mask so that finally the NTP is formed 32 . Lipophilic aliphatic esters have proven to be suitable for prodrugs to allow entering the cell independently of nucleoside transporters but to enable removal of the lipophilic prodrug moieties by cellular esterases/lipases 44 . Moreover, fatty-acid esters are known to be taken up by the mononuclear phagocyte system. These cells are important in the pathogenesis of AIDS and are considered to be reservoirs for HIV particles 45,46 .
We report on the synthesis of NTP prodrugs 3 bearing two identical 4-alkanoyloxybenzyl-(a-k), 4-alkoxycarbonyloxybenzyl-(l-n) and 4-aminocarbonyloxybenzyl groups (o-q), their hydrolysis properties in different media, the hydrolysis mechanism, primer extension assays and their anti-HIV activity. In addition, the synthesis of the monomasked (4-alkyloxybenzyl)-d4TTPs 4a,e,j is reported. As a model nucleoside analogue, d4T 1 was used to allow a comparison of the TriPPPro-compounds 3 with the DiPPro-compounds.

Results
Synthesis of TriPPPro-d4T triphosphate prodrugs 4. For the synthesis of TriPPPros-d4TTP prodrugs 3 a convergent strategy using a dicyanoimidazol (DCI)-mediated coupling of an appropriate phosphoramidite 5 and d4TDP 6 to form the energetically rich pyrophosphate moiety in the last step was performed ( Fig. 1). First, d4T 1 was prepared in good overall yields according to a three-step protocol reported by Horwitz 47 . From that compound, d4TDP 6 was prepared by applying the cycloSal technique (55% yield) because it has been reported that acceptor-substituted cycloSal nucleotides gave access to diphosphorylated compounds by using tetra-n-butylammonium phosphate as a nucleophile 48 . Therefore, the 5-chloro-substituted cycloSal-phosphate triester 7 was synthesized starting from d4T 1 with 5-chlorosaligenylchlorophosphite 8 followed by oxidation with tert-butylhydroperoxide to give the product 7 as a mixture of two diastereomers in high yields.
Despite its very difficult chromatographic properties, the resulting crude (n-Bu) 4 N þ -salt 6 was purified by automatic RP-18 flash chromatography without additional ion exchange.
The hygroscopic tetra-n-butylammonium salt form of d4TDP 6 was co-evaporated in dimethylformamide (DMF) and dried in high vacuum before the coupling reaction to ensure dry reaction conditions. The use of tetra-n-butylammonium counterions afforded a higher reactivity and better solubility in organic solvents. D4TDP 6 was then reacted with a series of phosphoramidites 5 in a very fast DCI-mediated coupling reaction and was oxidized 32,33 .
Inspired by recently published biodegradable linear polymers that were degraded by an acid-induced cascade reaction, we also synthesized carbamate derivatives 3o-q (ref. 49). In this case, 4-hydroxybenzyl alcohol 10 was first protected with tert-butyldimethylsilyl chloride to give compound 11 followed by an esterification using 4-nitrophenyl chloroformiate. To trap the excess of the chloroformiate, triethylene glycol monomethyl ether was added 49 . The obtained carbonate 12 was then converted with t-Boc-protected dimethylethylenediamine 13 and after acid-catalysed cleavage of both protecting groups the methylcarbamate 14 was isolated in 98% yield. Finally, phenylcarbamates 15 were synthesized starting from 14 in an one-pot reaction including tetramethylsilane (TMS) protection, coupling with the corresponding acyl chloroformiates 16 and desilylation in a yield of up to 59% (Fig. 2).
The 4-hydroxybenzyl alcohols bearing acyl-ester groups at the phenol and the carbamates were converted into phosphoramidites 5 in high yields as published before 32,33 . Finally, d4TDP 6 was mixed with 1.7 equivalents (eq.) of a corresponding phosphoramidite 5 and co-evaporated with acetonitrile. Then, the mixture was dissolved in a minimum of acetonitrile because achieving a high concentration was crucial for the success of the coupling reaction. In case of compounds with long acyl residues (RZC 11 H 23 ), tetrahydrofurane (THF) was added to accomplish complete solubility of the reagents. In some cases, the conversion of d4TDP 6 was not complete. In these cases, all volatile components were removed in vacuum, the residue was redissolved and further 1.0 eq. of the phosphoramidite 5 and 0.8 eq. of DCI were added. After another minute of stirring, the reaction mixture was oxidized. After oxidation the quantitative consumption of d4TDP 6 was confirmed using high-performance    Supplementary Fig. 3). Next, the solvent was removed in vacuum and the crude product was purified with RP-18 chromatography using gradients of water/ acetonitrile or water/THF as eluents. For compounds 3a-f, l-n, o-q the tetra-n-butylammonium ions were exchanged by ammonium ions using Dowex 50WX8 and the chromatography was repeated.
Synthesis of monomasked triphosphate prodrugs 4. In addition to the TriPPPro-compounds 3, the monomasked acyloxybenzyl-NTP derivatives 4 were synthesized as well (Fig. 3). Such monoesterified substances were described by others as potential triphosphate prodrugs 38,39,50 . Several synthesis routes mainly based on DCC-activated coupling were published. In our recent studies with DiPPro-prodrugs, we isolated such monoesterified compounds by simple hydrolysis 33 .
An efficient access to such compounds was developed in this study. The monoesterified NTPs were prepared starting from 4-acyloxybenzyl alcohol 9 and its conversion into the 5-nitro-cycloSal-triester 17. Despite the high reactivity of this compound, the purification by preparative thin-layer chromatography (TLC) was successful. The benzyl-(5-nitro-cycloSal)phosphate triesters 17 were obtained in high yields (up to  89%). Next, triesters 17 gave the monomasked acyloxybenzyl-d4TTPs 4 in yields of 26-30% by the addition of d4TDP 6.
Stability studies. The TriPPPros-d4TTP prodrugs 3 and the intermediates 4 were incubated in PBS (25 mM, pH 7.3), or were exposed to pig liver esterase (PLE) in PBS and to human CD 4 þ T-lymphocyte cell extracts to study their stability and the product distribution. The hydrolysis mixtures were analysed by means of analytical RP-18-HPLC. The calculated half-lives (Table 1) were determined for the first removal of one masking unit (t 1/2 (1)) to yield the intermediate 4 and the second hydrolysis step (t 1/2 (2)) to give the triphosphate 19.
Chemical stability in PBS at pH 7.3. In PBS, the stability of TriPPPro-d4TTP prodrugs 3a-h,l-n increased with increasing alkyl chain lengths (Table 1). However, the half-lives of more lipophilic compounds 3i-k decreased because of altered solubility behaviour or micelle formation. The half-lives of the intermediates 4 were always considerably higher than those of their precursors 3 because of the increase in charges leading to repulsive interaction with an incoming nucleophile. Moreover, formation of the three nucleotide forms 2, 6 and 19 were observed ( Supplementary Fig. 4).
Hydrolysis study using esterase. Next, we examined the enzymatic stability of prodrugs 3a-n by incubation with PLE in PBS, pH 7.3. The cleavage of the masking units for 3b-g occurred much faster than that in PBS, demonstrating a significant contribution of the enzymatic cleavage ( Table 1). As observed in the chemical hydrolysis studies, the cleavage of the second masking group proceeded much slower. According to the substrate specificity of PLE we determined the lowest half-lives for 3c-f and 3m. Shorter as well as longer alkyl residues in the ester moiety of the masking group led to increased half-lives. In addition, d4TTP 19 was delivered by enzymatic activation of 3a-n but was also found to be the sole metabolite from 4a,e,j as long as the enzymatic cleavage occurred rapidly ( Supplementary Fig. 5). The carbamate-functionalized prodrugs 3o-q were not substrates for PLE, as expected.
To confirm the prodrug concept and thus the direct successful release of the biologically active triphosphate metabolite, the prodrug 3e was exposed to PLE and the hydrolysis monitored with RP-18-HPLC. After complete consumption of 3e as well as its intermediate 4e the solvents were removed. For the template/ primer extension assay (Fig. 4a), HIV RT was incubated with the PLE hydrolysate as such (T*) or with the PLE hydrolysate in the additional presence of dCTP (T*, C), or dCTP þ dGTP (T*, C, G) or all natural 2 0 -deoxynucleotides (N*). Interestingly, an immediate DNA chain termination was observed after incorporation of d4TMP 2 (derived from the incoming d4TTP 19 that was released from the prodrug by PLE), while the control reaction containing all four natural NTPs in the absence of the PLE hydrolysate (N) showed full extension of the primer (Fig. 4a). In addition, TriPPPro-TTP 3r was synthesized (Fig. 5) and also investigated in the same way. As expected, the template/ primer extension assay showed efficient DNA elongation (Fig. 4b). The T* lysate resulted in termination of the polymerization because of the lack of the next complementary nucleotide (dCTP; position 26 nt), whereas the reaction proceeded till position 28 nt in the presence of both T* and dCTP (T*, C). The primer could be fully extended till 30 nt when T* was added in the presence of dCTP and dGTP, as also the N* and N samples could.
Hydrolysis in cell extracts. The hydrolysis of the TriPPProcompounds 3 was further investigated in human CD 4 þ T-lymphocyte CEM cell extracts. Again, the half-lives of the prodrugs 3a-n correlated well with chain length and were significantly lower than the half-lives in PBS buffer (Table 1). Thus, an enzymatic cleavage reaction took place as described above for the PLE studies. Furthermore, we observed the formation of the corresponding intermediates 4a-n that had lower half-lives than their parent prodrugs. This assumption was proven by hydrolysis of the synthesized intermediates 4a,e,j. In contrast to hydrolysis studies with PLE, in addition to d4TTP 19 d4TDP 6 was also detected as a major component in the CEM cell extracts most probably because of the presence of hydrolytic enzymes such as phosphatases and esterases ( Supplementary Fig. 6).
The formation of d4TDP 6 was clearly not a result of unselective cleavage of the prodrug 3. Investigations in cell Table 1 | Hydrolysis half-lives of TriPPPro-d4TTPs 3 and monoesterified d4TTPs 4 in different media.  extracts starting from d4TTP 19 led to a rapid degradation (t 1/2 ¼ 0.63 h). For this reason d4TDP 6 (t 1/2 ¼ 59 h) accumulated under these conditions. On the other hand, only very small amounts of d4TMP 2 were formed. Nevertheless, it was proven that the triphosphate of d4T 19 was successfully released in biological media such as CD4 þ T-lymphocyte extracts.
Antiviral evaluation. TriPPPro-compounds 3a-n were evaluated for their ability to inhibit the replication of HIV. For this purpose, HIV-1-or HIV-2-infected wild-type CEM/0 as well as mutant TK-deficient CEM cell cultures (CEM/TK À ) were treated with the prodrugs 3. As can be seen in Table 1, all compounds showed virtually similar activities against HIV-1 and HIV-2 as the parent nucleoside d4T 1. A somewhat increased antiviral activity with increasing lipophilicity resulting from their advantageous permeability was observed. In addition, all prodrugs with RZC 8 H 17 were also highly potent in CEM/TK À cells, whereas d4T 1 lacked relevant anti-HIV activity as expected in this TK-deficient cell model (EC 50 :150 mM). It should also be noticed that none of the prodrugs 3 were endowed with a significantly higher cytotoxicity than the parent d4T 1 compound (Table 2).

Discussion
We reported on the first successful direct intracellular delivery of NTPs using prodrug technology. D4T triphosphate prodrugs 3a-q were prepared via a convergent route using phosphoramidite chemistry (Fig. 1). Despite complete and selective conversion, TriPPPro-NTP prodrugs 3 were obtained in yields between 27 and 66%. We assumed that the loss in yield may be the result of a cleavage of the band g-phosphate anhydride bond of the prodrugs during work-up. This assumption was supported by the detection of d4TDP 6 and the bis(benzyl)phosphate diester after chromatography. Alternative purification methods such as extraction and precipitation were investigated but proved to be  inefficient. Nevertheless, by this method a large number of TriPPPro-d4TTPs 3 bearing various acyloxybenzyl-masking units were obtained. Moreover, this synthesis strategy also showed to be applicable to the synthesis of TriPPPro-compounds bearing other pyrimidine or purine nucleoside analogues. In addition, the intermediates 4a,e,j were synthesized using the cycloSal method and were obtained in moderate chemical yields. This method, which is based on the cycloSal strategy, is a reliable method for the synthesis of polyphosphate diesters comprising esters at both ends of the polyphosphate group (Fig. 3) 48,51,52 .
In general, three reactions should be considered in the hydrolysis pathways of TriPPPro-nucleotide prodrugs 3. First, the designed pathway yielding the NTP; second, a concurrent reaction that involved a nucleophilic reaction at the g-phosphate leading to the formation of d4TDP 6; and third, a nucleophilic reaction at the b-phosphate that would lead to d4TMP 2. Figure 6 summarizes all three possible hydrolysis pathways leading to the different phosphorylated nucleotide species. As shown in Fig. 6, to release d4TTP 19 two successive cleavage processes were necessary (path A). Thus, in addition to hydrolysis pathway A, also a reaction at the gand b-phosphate groups took place as side-reactions. Owing to the presence of a second energetically rich pyrophosphate bond, but despite its additional negative charge, the half-lives were found to be lower than those published recently for the DiPPro-d4TDPs 33 . However, the triphosphate 19 was the predominant product formed ( Supplementary Fig. 4). Moreover, after the starting material 3e was completely consumed, there was no further increase in the amounts of d4TMP 2 and d4TDP 6, which again points to the fact that the intermediate selectively delivers the triphosphate while the mono-and the diphosphate are formed from the starting TriPPPro-compounds only. A comparable behaviour has also been observed for the DiPPro-compounds 33 .
In contrast, for the carbamate derivatives 3o-q the cleavage of the masking groups occurred only once leading to the intermediates 4o-q. Therefore, it was concluded that the delivery mechanism should be different as compared with the ester-bearing masking groups. Owing to the highly stable carbamate functions present in the masking group, the first masking group cannot be cleaved by the original mechanism that involves a cleavage within the ester/carbamate residue. To gain more insights into this, a hydrolysis experiment of derivative 3o was conducted in 18 O-labelled water to yield 4o. Surprisingly, the 18 O-label was found in the cleaved benzyl-alcohol and not at the phosphate, which was convincingly confirmed using mass spectrometry ( Supplementary Fig. 7). Two different interpretations are possible for this result: (i) a S N 1-type reaction took place forming a benzyl cation and the monomasked NTP intermediate 4. The cation is then trapped by addition of water or (ii) a S N 2-type reaction took place instead in which the labelled water displaces the monomasked NTP intermediate 4 (Fig. 6, hydrolysis pathway A 2 ). In addition to hydrolysis in PBS, pH 7.3, 4p was hydrolysed under slightly acidic conditions (pH 6.0), although in comparison with the physiological pH conditions no difference in its hydrolysis behaviour was observed. Because of the very long hydrolysis time periods, a cleavage of the glycosidic bond in d4T 1 resulted in the appearance of the nucleobase thymine. The amount doubles every 63 h; however, this aspect has not been further considered in these investigations because it was irrelevant in the enzyme or cell extract incubations and in the case of other nucleoside analogues.
Finally, a very important result from the studies conducted with the monomasked intermediates 4a,e,j was the finding that exclusively d4TTP 19 was formed from these compounds. In addition, the hydrolysis studies conducted in the presence of PLE clearly led to the selective formation of two different NTPs (d4TTP and TTP) from the corresponding prodrug forms.
In conclusion, because of a successful cell membrane passage of the TriPPPro-compounds 3 and subsequent intracellular enzymatic hydrolysis, which led to the direct intracellular formation of phosphorylated d4T metabolites such as d4TTP 19 or at least d4TDP 6, marked anti-HIV activity in CEM/TK À cell cultures was observed while the parent nucleoside d4T 1 lacked significant activity in this cell assay. Thus, although the TriPPPro-compounds are still charged at the phosphate groups, obviously the modification at the g-phosphate group by lipophilic, bioreversible moieties gives the molecule sufficient lipophilicity to penetrate the cell membrane. To the best of our knowledge, we provided the first direct proof of the successful application of masked triphosphates that obviously are able to efficiently enter the cells and to directly deliver a higher phosphate derivative, most likely d4TTP 19. Because the concept should be generally applicable to natural nucleosides and a broad variety of nucleoside analogues, a novel way to deliver the corresponding bioactive triphosphate form of these nucleosides without any need for further enzymatically catalysed phosphorylation has been discovered. This concept seems to be very interesting for application with nucleoside analogues that show severe limitations in their activation to give the corresponding NTPs. Moreover, we are convinced that this approach is not limited to HIV treatment but can also be used for other viral targets and cancer and can also be used as a delivery for non-natural NTPs as biochemical tools in Chemical Biology approaches.

Methods
General. All reactions were carried out under dry conditions and at room temperature. Solvents and reagents: Acetonitrile, THF and DMF were purchased from Acros Organics (Extra Dry over molecular sieves) and dried with activated   (d) are quoted in parts per million (p.p.m.) downfield from TMS and calibrated on solvent signal. The 31 P-NMR chemical shifts (proton decoupled) are also quoted in p.p.m. using phosphoric acid as the external standard. Mass spectrometry: high resolution mass spectrometry (HRMS and electrospray ionization (ESI) mass spectra were acquired with a VG Analytical Finnigan ThermoQuest MAT 95 XL spectrometer. MALDI measurements (matrix: 9aminoacridine (9AA)) were performed with a Bruker UltrafleXtreme spectrometer. Infrared spectroscopy: IR spectra were recorded on a Bruker Alpha P FT-IR at room temperature in the range of 400-4,000 cm À 1 .
General procedure A: preparation of 4-acyloxybenzyl alcohols 9. 4-Hydroxybenzyl alcohol 10 (1.1 eq.) and triethylamine (1.0 eq.) in THF were cooled down to 0°C. The corresponding acyl chloride (1.0 eq.) in THF was added dropwise and the mixture stirred for 1-2 h. The precipitate was removed by filtration and the solvent evaporated in vacuum. The residue was diluted with CH 2 Cl 2 and washed once with saturated sodium bicarbonate solution and once with water. The organic layer was dried with Na 2 SO 4 and the solvent was removed in vaccum. The crude material was purified using column chromatography to give compound 9.

4-(Hydroxymethyl)phenyl-methyl(2-(methylamino)ethyl)-carbamate 14.
At 0°C to a stirred solution of 1.6 g compound 13 (8.6 mmol, 1.2 eq.), 1.9 ml diisopropylethylamine (1.4 g, 11 mmol, 1.6 eq.) and catalytic amounts of 4-dimethylaminophenol in 30 ml toluene 2.8 g activated carbonate 12 (7.0 mmol, 1.0 eq.) was added. The reaction mixture was allowed to warm to rt and stirred for 16 h. The solution was diluted with ethyl acetate and washed with 1 M HCl, followed by saturated aqueous NaHCO 3 . The organic phase was dried over Na 2 SO 4 , filtered and the solvent was removed by evaporation. For deprotection the residue was redissolved in a mixture of CH 2 Cl 2 /trifluoroacetic acid (1:1 v/v). After 1 h the volatile components were removed by evaporation and the residue was co-evaporated twice with toluene. The crude material was purified by automatic RP-18 chromatography (water/acetontrile gradient). Yield: 1.6 g (6.9 mmol, 98%) yellowish oil. TLC (CH 2 Cl 2 / General procedure B: preparation of bis-methylcarbamates 15. 4-(Hydroxymethyl)phenylcarbamate 14 (1.0 eq.) was dissolved in THF and cooled down to 0°C. After addition of imidazole (1.3 eq.) and TMSCl (1.2 eq.) the mixture was stirred for 2 h. Triethylamine (3.1 eq.) and the corresponding alkyl chloroformate (3.0 eq.) were added. The reaction was kept for 1.5 h at rt and then diluted with 30 ml of 1 vol% concentrated (12 M) HCl in EtOH for desilylation. The solution was stirred for 1 h and the solvent evaported in vacuum. The residue was dissolved in CH 2 Cl 2 and washed with saturated aqueous NaHCO 3 . The organic phase was dried over Na 2 SO 4 , filtered and the solvent was removed in vacuum. The crude material was purified using column chromatography. General procedure C: preparation of bis-(4-acyloxybenzyl)-N,N-diisopropylphosphoramidites 5. Dichloro-N,N-diisopropylphosphoramidite (1.0 eq.) was dissolved in THF and cooled to 0°C. Triethylamine (2.3 eq.) and the corresponding 4-acyloxybenzylalcohol 9 (2.1-2.2 eq.) in THF were added dropwise. The reaction mixture was kept at 0°C for 18-24 h. After filtration, the solvent was removed by evaporation. The crude products were purified either using column chromatography or using preparative TLC (chromatotron).
Anti-HIV activity assay. Inhibition of HIV-1(III B )-and HIV-2(ROD)-induced cytopathicity in wild-type CEM/0 and TK-deficient CEM/TK À cell cultures was measured in microtitre 96-well plates containing B3 Â 10 5 CEM cells ml À 1 infected with 100 CCID 50 of HIV per millilitre and containing appropriate dilutions of the test compounds. After 4 À 5 days of incubation at 37°C in a CO 2 -controlled humidified atmosphere, CEM giant (syncytium) cell formation was examined microscopically. The EC 50 (50% effective concentration) was defined as the compound concentration required to inhibit HIV-induced giant cell formation by 50%.
Primer extension reactions. The used polymerase HIV RT was obtained from Roboklon. The primer and template were purchased from Life Technologies.
Primer sequence: 5 0 -TTGGATAGGAGGAAGTCCTGGTTGC-3 0 Template sequence: 5 0 -AGACAAACCTATCCTCCTTCAGGACCAACG-3 0 The primer extension assays were performed under the following conditions: The primer was labelled using [g 32 P]-ATP according to standard techniques. After 5-min incubation at 95°C in 20 mM Tris-HCl (pH 7.6) and 50 mM NaCl, the hybridization/annealing of the primer to the template strand was achieved by a cooling phase from 95 to 20°C over 3 h. The final assay solution (20 ml) consists of 2.5 mM dNTPs or hydrolysate, 1 Â reaction buffer (50 mM Tris-HCl (pH 8.6), 10 mM MgCl 2 and 40 mM KCl), 0.02 mM of hybridization and 0.2 units of the enzyme, which was incubated at 37°C for 10 min. The reaction was stopped by heating up to 80°C for 3 min. The assays were separated using a denaturating PAGE (15%). The results were visualized by phosphorimaging.