Probing the Binding Requirements of Modified Nucleosides with the DNA Nuclease SNM1A

SNM1A is a nuclease that is implicated in DNA interstrand crosslink repair and, as such, its inhibition is of interest for overcoming resistance to chemotherapeutic crosslinking agents. However, the number and identity of the metal ion(s) in the active site of SNM1A are still unconfirmed, and only a limited number of inhibitors have been reported to date. Herein, we report the synthesis and evaluation of a family of malonate-based modified nucleosides to investigate the optimal positioning of metal-binding groups in nucleoside-derived inhibitors for SNM1A. These compounds include ester, carboxylate and hydroxamic acid malonate derivatives which were installed in the 5′-position or 3′-position of thymidine or as a linkage between two nucleosides. Evaluation as inhibitors of recombinant SNM1A showed that nine of the twelve compounds tested had an inhibitory effect at 1 mM concentration. The most potent compound contains a hydroxamic acid malonate group at the 5′-position. Overall, our studies advance the understanding of requirements for nucleoside-derived inhibitors for SNM1A and indicate that groups containing a negatively charged group in close proximity to a metal chelator, such as hydroxamic acid malonates, are promising structures in the design of inhibitors.


Introduction
The nuclease SNM1A (also known as DCLRE1A) is part of the interstrand crosslink repair machinery in mammalian cells [1]. Depletion of SNM1A increases sensitivity to crosslinking agents [2][3][4][5], although the precise mechanisms behind this are still under investigation. Current research suggests that SNM1A digests past crosslinks from an incision 5 to the site of the lesion; this role and other proposed pathways are summarised in a recent review [1]. SNM1A has been reported to have 5 -exonuclease as well as endonuclease activity, although endonuclease activity requires a blocked 5 -end and stoichiometric amounts of enzyme [6]. Exonuclease activity requires a 5 -phosphate group and is processive for high molecular weight substrates [7]. The active site of SNM1A is postulated to contain a di-metal centre, based on similarity with other members of the β-CASP family [8]. However, only one metal ion has been observed in crystal structures of SNM1A to date [9]. The active site is located within a wide binding groove, which permits the processing of bulky substrates such as DNA containing interstrand crosslinks [8].
Like other DNA damage repair proteins [10], SNM1A is a potential target for overcoming resistance to crosslinking agents in cancers [11]. Inhibitors reported to date are limited to o-phenanthroline [12], cephalosporins [13], and a modified nucleoside inhibitor [14]. In our previous work designing nucleoside inhibitors, the hydroxamic acid modification emerged as the most promising group [14]. Nucleosides 1 and 2 ( Figure 1a) were found to inhibit SNM1A to varying degrees, and the more potent compound 1 was found to have an IC 50 value of 139 µM. However, both compounds likely only target one of the two postulated metal ions in the active site of SNM1A. Therefore, in this work, we expand the

Synthesis
The synthesis of the 5′-alkylated targets was adapted from Zlatev et al. [28] and is shown in Scheme 1. Starting from 5′-deoxy-5′-iodothymidine (3), protection of the 3′-hydroxy group as the silyl ether and of the nucleobase with the benzyloxymethyl group was carried out in 87% and 95% yield, respectively. This protected iodide 5 then underwent nucleophilic substitution with benzyl methyl malonate in 66% yield. Removal of the silyl ether of malonate ester 6 afforded alcohol 7 in 71% yield, and catalytic hydrogenation removed both the benzyloxymethyl group and the benzyl group to give carboxylic acid 8 In recent crystal structures of SNM1A, malonate was observed bound to the metal ion in the active site [9]. This highlights the suitability of the group as a metal chelator. The malonate group has found application both as a scaffold for a hydroxamic acid group [15][16][17][18][19][20][21][22] and as a phosphate bioisostere in many different contexts, including analogues of phosphosugars [23][24][25][26] and nucleotides [27,28]. By combining the malonate group with the hydroxamic acid group, binding of both metal ions is envisaged (Figure 1b). The substituents of the malonate group were varied to include esters and carboxylate salts to obtain further information about the binding mode.
Derivatives utilised either one carboxylic acid group of the malonate group to connect the group to the nucleoside via an amide linkage (N-linked), or alkylation at the α-carbon (C-linked) was employed to install the malonate group onto the nucleoside. The nucleobase used for this family of modified nucleosides is thymine, allowing for direct comparison with the previously evaluated compounds [14]. The 5 -C-linked series in particular is directly related to the previously reported compound 2, as its malonate group is at the same distance from the deoxyribose core as the hydroxamic acid group in compound 2. However, the malonate modifications were introduced not only at the 5 -position, but also at the 3 -position and as a linker between two nucleosides (Figure 1c) to probe the interactions between the nucleoside and the active site of recombinant SNM1A in gel-based assays.

Synthesis
The synthesis of the 5 -alkylated targets was adapted from Zlatev et al. [28] and is shown in Scheme 1. Starting from 5 -deoxy-5 -iodothymidine (3), protection of the 3hydroxy group as the silyl ether and of the nucleobase with the benzyloxymethyl group was carried out in 87% and 95% yield, respectively. This protected iodide 5 then underwent nucleophilic substitution with benzyl methyl malonate in 66% yield. Removal of the silyl ether of malonate ester 6 afforded alcohol 7 in 71% yield, and catalytic hydrogenation removed both the benzyloxymethyl group and the benzyl group to give carboxylic acid 8 in 86% yield. For use in assays, the sodium carboxylate salt 9 was required, so ion exchange was carried out following hydrogenation. Due to contamination with ion-exchange resin, purification by reversed-phase preparative TLC was employed and gave target 9 in a low yield of 22%. Carboxylic acid 8 was used as an intermediate for targets 10 and 11. Following ester hydrolysis and ion exchange to furnish carboxylate salt 10, reversed-phase preparative TLC was again used to purify compound 10, which was obtained in 55% yield. Carboxylic acid 8 was also employed in the synthesis of hydroxamic acid 11, which was obtained in 15% yield via aminolysis with hydroxylamine [29]. Targets 9 and 11 were isolated and used as mixtures of interconverting diastereomers. Unfortunately, synthesis of a 1,3-dihydroxamic acid is not possible using aminolysis with hydroxylamine, as the treatment of a 1,3-diester with hydroxylamine under basic conditions results in cyclisation to form isoxazolidine-3,5-diones [30,31]. in 86% yield. For use in assays, the sodium carboxylate salt 9 was required, so ion exchange was carried out following hydrogenation. Due to contamination with ion-exchange resin, purification by reversed-phase preparative TLC was employed and gave target 9 in a low yield of 22%. Carboxylic acid 8 was used as an intermediate for targets 10 and 11. Following ester hydrolysis and ion exchange to furnish carboxylate salt 10, reversed-phase preparative TLC was again used to purify compound 10, which was obtained in 55% yield. Carboxylic acid 8 was also employed in the synthesis of hydroxamic acid 11, which was obtained in 15% yield via aminolysis with hydroxylamine [29]. Targets 9 and 11 were isolated and used as mixtures of interconverting diastereomers. Unfortunately, synthesis of a 1,3-dihydroxamic acid is not possible using aminolysis with hydroxylamine, as the treatment of a 1,3-diester with hydroxylamine under basic conditions results in cyclisation to form isoxazolidine-3,5-diones [30,31]. Scheme 1. Synthesis of 5′-modified C-linked targets 9, 10 and 11.
For the N-linked targets, amide coupling between 5′-amino-5′-deoxythymidine (12) and monomethyl potassium malonate was employed (Scheme 2) to furnish ester 13 in 54% yield. This target served as an intermediate in the synthesis of targets 14 and 15. Ester hydrolysis followed by ion exchange afforded sodium carboxylate 14 in 98% yield. Hydroxamic acid 15 was obtained in 83% yield by treating ester 13 with hydroxylamine under basic conditions [29].
Synthesis of the 3′-modified targets was attempted analogous to the 5′-modified compounds. For the C-linked targets, initially nucleophilic substitution of a mesylate was envisaged (Scheme 3a). Inversion of the 3′-hydroxy group of protected thymidine 16 was carried out by sequential mesylation and hydrolysis [32] in 90% yield. Mesylation [33] of the alcohol 17 proceeded in 90% yield. However, substitution of this mesylate 18 with benzyl methyl malonate was unsuccessful, with the N-methylated elimination product 20 obtained in 46% yield instead of the desired product 19. An alternative strategy to functionalise the 3′-position using protected 2,3′-anhydrothymidine [34] 21 and methyl triflate as an activating agent as reported by Saha et al. [35] was unsuccessful in our hands. Only Scheme 1. Synthesis of 5 -modified C-linked targets 9, 10 and 11.
For the N-linked targets, amide coupling between 5 -amino-5 -deoxythymidine (12) and monomethyl potassium malonate was employed (Scheme 2) to furnish ester 13 in 54% yield. This target served as an intermediate in the synthesis of targets 14 and 15. Ester hydrolysis followed by ion exchange afforded sodium carboxylate 14 in 98% yield. Hydroxamic acid 15 was obtained in 83% yield by treating ester 13 with hydroxylamine under basic conditions [29].
Molecules 2021, 26, x FOR PEER REVIEW 4 o the hydrolysed product 23 was obtained after quenching the reaction, with none of desired product 22 observed (Scheme 3b).
We then turned our attention to the N-linked targets. Commercially available AZT (24) was reduced to aminothymidine (25) by catalytic hydrogenation in 99% yield (Scheme 4). Amine 25 was then used in an EDC-mediated amide coupling with monomethyl potassium malonate, which afforded target 26 in 34% yield. Treatment of ester 26 with KOH followed by ion exchange produced carboxylate salt 27 in 82% yield, while aminolysis of this ester using the same conditions as for the 5′-modified nucleosides afforded hydroxamic acid 28 in 70% yield. We then turned our attention to the N-linked targets. Commercially available AZT (24) was reduced to aminothymidine (25) by catalytic hydrogenation in 99% yield (Scheme 4). Amine 25 was then used in an EDC-mediated amide coupling with monomethyl potassium malonate, which afforded target 26 in 34% yield. Treatment of ester 26 with KOH followed by ion exchange produced carboxylate salt 27 in 82% yield, while aminolysis of this ester using the same conditions as for the 5 -modified nucleosides afforded hydroxamic acid 28 in 70% yield.
For the final group of compounds, the malonate group forms a phosphodiester replacement and links two nucleosides. To this end, a 5 -modified carboxylic acid and a 3 -amino nucleoside were combined (Scheme 5). For the synthesis of the dinucleosides, protecting groups for the 5 -and 3 -hydroxy groups were used to aid solubility. The carboxylic acid nucleoside 29 was obtained from protected malonate ester 6 by catalytic hydrogenation. The benzyloxymethyl group of 6 was only partially removed in this step, but this hemiaminal did not interfere with further reactions. The other nucleoside, 5 -protected 3 -aminothymidine 31, was synthesised from DMTr-protected AZT 30 [36]. Reduction of azide 30 furnished amine 31 in 95% yield. EDC-mediated coupling between nucleosides 29 and 31 afforded the fully protected dinucleoside 32 in 76% yield. Deprotection of the 3 -and 5 -hydroxy groups was achieved using TBAF and TFA, respectively, in 92% yield for each step. This sequence afforded target 34, which was further derivatised. Sodium carboxylate 35 was furnished from ester 34 in 89% yield by ester hydrolysis and subsequent ion exchange, and hydroxamic acid 36 was obtained in 74% yield from ester 34 by aminol-ysis. As observed with the 5 -modified C-linked compounds 9 and 11, the dinucleosides 34-36 were also isolated and used as mixtures of interconverting diastereomers. For the final group of compounds, the malonate group forms a phosphodiester replacement and links two nucleosides. To this end, a 5′-modified carboxylic acid and a 3′amino nucleoside were combined (Scheme 5). For the synthesis of the dinucleosides, protecting groups for the 5′-and 3′-hydroxy groups were used to aid solubility. The carboxylic acid nucleoside 29 was obtained from protected malonate ester 6 by catalytic hydrogenation. The benzyloxymethyl group of 6 was only partially removed in this step, but this hemiaminal did not interfere with further reactions. The other nucleoside, 5′-protected 3′-aminothymidine 31, was synthesised from DMTr-protected AZT 30 [36]. Reduction of azide 30 furnished amine 31 in 95% yield. EDC-mediated coupling between nucleosides 29 and 31 afforded the fully protected dinucleoside 32 in 76% yield. Deprotection of the 3′-and 5′-hydroxy groups was achieved using TBAF and TFA, respectively, in 92% yield for each step. This sequence afforded target 34, which was further derivatised. Sodium carboxylate 35 was furnished from ester 34 in 89% yield by ester hydrolysis and subsequent ion exchange, and hydroxamic acid 36 was obtained in 74% yield from ester 34 by aminolysis. As observed with the 5′-modified C-linked compounds 9 and 11, the dinucleosides 34-36 were also isolated and used as mixtures of interconverting diastereomers.
The successful synthesis of the dinucleosides 34-36 completes a series of twelve nucleosides. These can be divided into four groups: 5′-modified C-linked, 5′-modified Nlinked, 3′-modified N-linked, and dinucleosides. Each group consists of three compounds each, with an ester, sodium carboxylate and a hydroxamic acid analogue.

Biological Evaluation
The twelve modified nucleosides ( Figure 2a) were evaluated as inhibitors of recombinant SNM1A to evaluate binding to the active site, using a previously reported gelbased assay with the same 21-mer oligonucleotide substrate [14]. The nucleosides were incubated with SNM1A for 5 min prior to the addition of the fluorescently tagged 21-mer oligonucleotide substrate and further incubation for 1 h. In the event of the nucleoside binding to the active site, digestion of the substrate proceeds less efficiently (Figure 2b). SNM1A removes nucleotide monomers from the 5′-end, generating a shorter oligonucleotide that in turn is a substrate for SNM1A and can be digested further. The size of the oligonucleotide products, analysed by gel electrophoresis, therefore indicates how strongly a compound binds to the nuclease.  The successful synthesis of the dinucleosides 34-36 completes a series of twelve nucleosides. These can be divided into four groups: 5 -modified C-linked, 5 -modified Nlinked, 3 -modified N-linked, and dinucleosides. Each group consists of three compounds each, with an ester, sodium carboxylate and a hydroxamic acid analogue.

Biological Evaluation
The twelve modified nucleosides ( Figure 2a) were evaluated as inhibitors of recombinant SNM1A to evaluate binding to the active site, using a previously reported gel-based assay with the same 21-mer oligonucleotide substrate [14]. The nucleosides were incubated with SNM1A for 5 min prior to the addition of the fluorescently tagged 21-mer oligonucleotide substrate and further incubation for 1 h. In the event of the nucleoside binding to the active site, digestion of the substrate proceeds less efficiently (Figure 2b). SNM1A removes nucleotide monomers from the 5 -end, generating a shorter oligonucleotide that in turn is a substrate for SNM1A and can be digested further. The size of the oligonucleotide products, analysed by gel electrophoresis, therefore indicates how strongly a compound binds to the nuclease.  11,14). The methyl ester analogues were only active for the 5′-C-linked series, which contains an additional carboxylate group, and the dinucleoside series ( Figure 2c, lanes 4, 13). To further investigate the potency of the different nucleosides, concentration-dependence experiments were carried out for the nine nucleosides that showed inhibition (Figures 2 and S1-S4), with concentrations ranging from 1 mM to 1 µM. Although partial digestion of the substrate by SNM1A was observed at concentrations below 1 mM, the strongest inhibitor 11 was found to impede hydrolysis at concentrations as low as 33 µM The results of this assay for the twelve modified nucleosides are shown in Figure 2c. Lane 1 shows the oligonucleotide substrate in the absence of SNM1A, while lane 2 shows maximal digestion in 60 min, in the absence of any modified nucleoside. Thymidine (T) was used as a negative control to ensure that any inhibitory effects in lanes 4-15 can be attributed to the nucleoside modification, as thymidine has no effect on the hydrolysis of the substrate (Figure 2c, lane 3). Encouragingly, nine out of twelve malonate-based nucleosides showed some inhibition at 1 mM concentration (Figure 2c). Only three nucleosides (Figure 2c, lanes 7, 8, 10) had no impact on the activity of SNM1A.
Analysis of the size of oligonucleotide products remaining after the incubation for each compound family revealed the following: the 5 -C-linked series produced three compounds that led to almost full inhibition at 1 mM concentration, with malonate hydroxamic acid 11 emerging as the most potent overall (Figure 2c, lane 6). The dinucleoside series also afforded three active compounds that impeded SNM1A activity (Figure 2c, lanes [13][14][15] and was only slightly less effective in this assay than the 5 -C-linked series (Figure 2c To further investigate the potency of the different nucleosides, concentration-dependence experiments were carried out for the nine nucleosides that showed inhibition ( Figure 2 and Figures S1-S4), with concentrations ranging from 1 mM to 1 µM. Although partial digestion of the substrate by SNM1A was observed at concentrations below 1 mM, the strongest inhibitor 11 was found to impede hydrolysis at concentrations as low as 33 µM ( Figure 3, lanes 4-7) and its activity was found to be concentration dependent. This implies that the malonate group is not coordinating exclusively through the carbonyl oxygen atoms but requires substituents with the potential to hold a negative charge to achieve effective chelation. The 5′-C-linked malonate series (Table 1, entries 1-3) resulted in the strongest binding. This finding is not surprising given the additional carboxylate group that is present compared to all other derivatives. It is therefore likely that this carboxylate participates in coordination to the metal centre. As observed in Figure 2c and Table 1, the most potent compound is the hydroxamic acid malonate 11 (Table 1, entry 3).   The results of identical assays for the other modified nucleosides are summarised in Table 1 (for gel images see Figures S1-S4). All compounds that were active at 1 mM concentration also have a partial inhibitory effect at 333 µM, except dinucleoside 34 (Table 1, entry 10). Only the malonate hydroxamic acid 11 (Table 1, entry 3) was effective at lower concentrations. In three of the four series, the ester analogue is the weakest inhibitor (Table 1, entries 4, 7, 10; Figure 4). This result is consistent with the initial screen and can be explained by the lack of an acidic proton compared to hydroxamic acids or the lack of negatively charged atoms compared to carboxylate salts, making chelation unfavourable.
This implies that the malonate group is not coordinating exclusively through the carbonyl oxygen atoms but requires substituents with the potential to hold a negative charge to achieve effective chelation. The 5 -C-linked malonate series (Table 1, entries 1-3) resulted in the strongest binding. This finding is not surprising given the additional carboxylate group that is present compared to all other derivatives. It is therefore likely that this carboxylate participates in coordination to the metal centre. As observed in Figure 2c and Table 1 the strongest dinucleosides was significantly lower than that of the strongest 5′-modified nucleoside 11 (Table 1, entry 3) which contained an additional carboxylate group. The benefits of a second nucleoside therefore do not counteract the negative effect of removing a carboxylate group. The trends observed in these gel-based assays are summarised in Figure 4. . Trends observed from the analysis of modified (di)nucleosides in this work. Potency of the modifications increases in the following order: methyl ester malonamide < sodium carboxylate malonamide < hydroxamic acid malonamide ≈ monomethyl sodium malonate < disodium malonate < hydroxamic acid sodium malonate. For the site of modification, the observed trend is 5′-Nlinked < 3′-N-linked < dinucleoside < 5′-C-linked.
Comparison with the previously reported compounds 1 and 2 (Figure 5a,b) shows that compound 11 has a slightly lower potency than compound 1 (Figure 5b, lanes 4-7 vs. [8][9][10][11], which emerged as the best inhibitor in our previous work [14]. Structurally, compound 11 is an analogue of hydroxamic acid 2, which differs only in the presence of an additional carboxylate group (Figure 5a). This additional group enhances binding significantly, as observed in Figure 5b (lane 4 vs. 12), which points towards interactions between the carboxylate group and the active site and supports the presence of a second metal ion. The binding mode of compound 11 to two metal ions shown in Figure 5c is supported by the different activities of nucleosides 2 and 11 and the observed binding mode of the malonate anion in crystal structures of SNM1A [9]. It is also consistent with the requirement for groups with the potential to hold a negative charge and the enhanced activity of the 5′-C-linked series, which contains an additional carboxylate group. It is, however, important to note that these trends could also be the result of other interactions such as hydrogen bonding with amino acid residues in the active site. The weaker binding of the carboxylate 10 compared to the hydroxamic acid 11 (Table 1, entries 2 vs. 3) can be attributed to the strong chelating ability of the hydroxamic acid group resulting from the formation of a five-membered ring [37], as shown in Figure 5c. Trends observed from the analysis of modified (di)nucleosides in this work. Potency of the modifications increases in the following order: methyl ester malonamide < sodium carboxylate malonamide < hydroxamic acid malonamide ≈ monomethyl sodium malonate < disodium malonate < hydroxamic acid sodium malonate. For the site of modification, the observed trend is 5 -N-linked < 3 -N-linked < dinucleoside < 5 -C-linked.
In contrast to the C-linked malonates, of the N-linked 5 -modified nucleosides, only one compound (15) binds SNM1A. Its lower activity (Table 1, entry 6 vs. 3) compared to the corresponding C-linked hydroxamic acid 11 is possibly due to the increased distance of the hydroxamic acid group from the deoxyribose core in 15, and/or the absence of the additional carboxylate group. However, the 3 -modified nucleosides 26-28 (Table 1, entries 7-9) also contain a malonamide group, and two compounds in this series inhibited SNM1A at 333 µM concentration (Table 1, entries, 8,9). This implies that the relative positioning of the group compared to the deoxyribose core is a causative factor in the lack of potency of 5 -N-linked nucleosides ( Figure 4). For dinucleosides, the lowest concentrations at which inhibition was observed for dinucleosides 35 and 36 are the same as for the 3 -modified analogues ( Table 1, entries 8 vs. 11, 9 vs. 12). However, the ester derivative 34 showed some inhibition at 1 mM concentration (Table 1, entry 10), while the corresponding 3derivative 26 was inactive (Table 1, entry 7). These results point towards the importance of the second nucleoside as a recognition element. Nonetheless, the potency of the strongest dinucleosides was significantly lower than that of the strongest 5 -modified nucleoside 11 (Table 1, entry 3) which contained an additional carboxylate group. The benefits of a second nucleoside therefore do not counteract the negative effect of removing a carboxylate group. The trends observed in these gel-based assays are summarised in Figure 4.
Comparison with the previously reported compounds 1 and 2 (Figure 5a,b) shows that compound 11 has a slightly lower potency than compound 1 (Figure 5b, lanes 4-7 vs. [8][9][10][11], which emerged as the best inhibitor in our previous work [14]. Structurally, compound 11 is an analogue of hydroxamic acid 2, which differs only in the presence of an additional carboxylate group (Figure 5a). This additional group enhances binding significantly, as observed in Figure 5b (lane 4 vs. 12), which points towards interactions between the carboxylate group and the active site and supports the presence of a second metal ion. The binding mode of compound 11 to two metal ions shown in Figure 5c is supported by the different activities of nucleosides 2 and 11 and the observed binding mode of the malonate anion in crystal structures of SNM1A [9]. It is also consistent with the requirement for groups with the potential to hold a negative charge and the enhanced activity of the 5 -C-linked series, which contains an additional carboxylate group. It is, however, important to note that these trends could also be the result of other interactions such as hydrogen bonding with amino acid residues in the active site. The weaker binding of the carboxylate 10 compared to the hydroxamic acid 11 (Table 1, entries 2 vs. 3) can be attributed to the strong chelating ability of the hydroxamic acid group resulting from the formation of a five-membered ring [37], as shown in Figure 5c.
Overall, our studies show that an extended metal-chelating group, such as malonate hydroxamic acids, leads to enhanced binding to the nuclease SNM1A, consistent with the presence of two metal ions in the active site. The synthesis and evaluation of methyl ester, carboxylate and hydroxamic acid analogues of each point of installation revealed that a group with the potential to hold a negative charge is required for binding. Variation in the positioning of the malonate groups relative to the deoxyribose group revealed that of the N-linked compounds, modification of the 3 -position led to increased potency compared to installation at the 5 -position. Interestingly, malonamide-linked dinucleosides were found to be slightly superior to their 3 -modified analogues, which highlights the inclusion of two nucleosides as recognition elements as a promising avenue for future nucleoside-based inhibitors. Overall, our studies show that an extended metal-chelating group, such as malonate hydroxamic acids, leads to enhanced binding to the nuclease SNM1A, consistent with the presence of two metal ions in the active site. The synthesis and evaluation of methyl ester, carboxylate and hydroxamic acid analogues of each point of installation revealed that a group with the potential to hold a negative charge is required for binding. Variation in

Materials and Methods
3.1. General 1 H and 13 C NMR spectra were recorded on Bruker 400 MHz or 600 MHz system spectrometers in DMSO-d 6 , CDCl 3 , CD 3 OD, acetone-d 6 or D 2 O relative to residual DMSO (δ H = 2.50 ppm, δ C = 39.52 ppm), CDCl 3 , (δ H = 7.26 ppm, δ C = 77.16 ppm), CD 3 OD (δ H = 3.31 ppm, δ C = 49.00 ppm), acetone-d 6 (δ H = 2.05 ppm, δ C = 29.84 ppm) or D 2 O (δ H = 4.79 ppm) [38]. Chemical shifts are reported in ppm and coupling constants are reported in Hertz (Hz) and accurate to 0.2 Hz. 13 C NMR spectra are proton decoupled. NMR spectra were assigned using HSQC, HMBC, DEPT and EXSY experiments. Modified nucleosides are numbered according to standard nucleoside conventions. Mass spectrometry measurements were carried out on a Bruker ESI or APCI HRMS. Melting points were measured using a Griffin melting point apparatus and are uncorrected. Infrared (IR) spectra were obtained on a Perkin Elmer spectrophotometer. Flash column chromatography was carried out using silica gel, particle size 0.04-0.063 mm, purchased from Sigma Aldrich or VWR. TLC analysis was performed on TLC Silica gel 60 F 254 plates purchased from Merck and visualised by UV irradiation (254 nm), ninhydrin stain (1.5 g ninhydrin, 5 mL AcOH, 500 mL 95% EtOH), anisaldehyde stain (9.2 mL p-methoxybenzaldehyde, 3.75 mL AcOH, 338 mL 95% EtOH, 12.5 mL conc. H 2 SO 4 ) and iodine. Ion-exchange resin refers to Diaion WT01S(H) resin, which was purchased from Alfa Aesar and activated by consecutive washes with acetone, MeOH, 1 M NaOH (Na form only), H 2 O and MeOH. Preparative reversed-phase TLC was carried out on TLC silica gel 60 RP-18 F 254 S plates purchased from Merck. THF and CH 2 Cl 2 were dried using a PureSolv MD solvent purification system.

-O-(tert-Butyldimethylsilyl)-5 -deoxy-5 -iodothymidine (4):
Nucleoside 4 was prepared according to a modified procedure [39]. Iodide 3 (10.0 g, 28.4 mmol) was dissolved in dry DMF (20 mL) under argon and cooled to 0 • C. Imidazole (2.51 g, 36.9 mmol) and TBDMS-Cl (5.56 g, 36.9 mmol) were added and the reaction mixture was stirred at 0 • C for 6 h. After this time, TLC analysis (EtOAc) showed the consumption of starting material (R f = 0.5) and the formation of the product (R f = 0.8). The reaction was quenched by the addition of MeOH (5 mL), diluted with EtOAc (250 mL) and H 2 O (100 mL). The layers were separated and the organic layer was washed with brine (100 mL), dried over MgSO 4 , filtered and concentrated. The residue was recrystallised from EtOH to afford the desired product 4 as a white crystalline solid (11.59

-Amino--deoxythymidine (25):
Amine 25 was prepared according to a published procedure [41]. AZT (24) (1.00 g, 3.74 mmol) was dissolved in MeOH (35 mL) and added to a dried flask containing Pd/C (10%, 150 mg) and H 2 was bubbled through the suspension while stirring at r.t. for 2 h. After this time, TLC analysis (EtOAc-MeOH, 4:1) showed the complete consumption of starting material, (R f = 0.8) and the formation of the product (R f = 0.1). The reaction mixture was filtered through celite and the filtrate was concentrated to afford the desired product 25 as a white foam (893 mg, 99%).