Acid-base properties of an antivirally active acyclic nucleoside phosphonate: ( S )-9-[3-hydroxy-2-(phosphonomethoxy)propyl]adenine (HPMPA)

(Abstract) HPMPA is an acyclic nucleoside phosphonate analogue of AMP which displays antiviral properties. Therefore, its acid-base behavior as well as that of related compounds like PMEA, 9-[2-(phosphonomethoxy)ethyl]adenine, are for many reasons (e.g., binding to enzymes, coordination of metal ions) of general interest. HPMPA can accept two protons at the phosphonate and two more at the adenine residue, but not all acidity constants are accessible by potentiometric pH titrations. Therefore, we measured the chemical shifts of the nine non-exchangeable HPMPA protons by 1 H NMR in D 2 O in dependence on pD in the range from 1 to 12. The corresponding results allowed identifying the protonation sites and, transferred to aqueous solution, they gave also the acidity constants. The most basic site is the phosphonate group followed by N1 of adenine. The p K a values increase from ca. –0.27 [–N7(H) + ] via 1.27 [– PO(OH) 2 ] and 4.23 [–N1(H) + ] to 6.86 [–PO(OH) – ]. In the fully protonated species charge repulsion exists between N1(H) + and N7(H) + ; therefore, the affinity of N7 for H + is not correctly reflected by the measured acidity constant ( ca . –0.27). Needed is the intrinsic micro acidity constant which reflects the H + affinity of N7 under conditions where N1 is unprotonated; we abbreviate this species as + H•N7(HPMPA)N1. The corresponding microconstant is estimated to be p k H•N7-N1 N7-N1 ≈ 3.5; the minor species + H•N7(HPMPA)N1 occurs with an estimated formation degree between about 5 to 20%. The basicity of the adenine nitrogens decreases in the order N1>N7>N3.


Introduction
Nucleoside phosphates (NPs) and their metal ion complexes play key roles in many metabolic transformations. [1][2][3][4] It is thus not surprising that attempts to exploit nucleoside/nucleotide derivatives as drugs have a long history. For example, the antiviral activity of benzimidazole (= 1,3-dideazapurine) derivatives was first described 5,6 in 1947 and 1--D-ribofuranosylbenzimidazoles have been tested against the influenza B virus, 7,8 as well as against several strains of polio virus and herpes simplex virus. 9,10 Typically, for antiviral activity, these compounds require to be present in phosphorylated form; however, their phosphorylated derivatives suffer from a serious handicap, namely that due to the wide occurrence of non-specific dephosphorylation enzymes 11 phosphorus-oxygen ester bonds are easily cleaved, rendering these anti-metabolites inefficient. 11 This problem is circumvented by using phosphonate derivatives, as the phosphorus-carbon bond cannot be hydrolysed. [11][12][13][14] With the foregoing facts in mind, it is not surprising that acyclic nucleoside phosphonates (ANPs) have been tested for several illnesses and effects such as Schistosomiasis, 15 Malaria 16 and Sleeping Sickness, 17 and have also been shown to activate macrophages. 18 However, more importantly for the present context, the ANPs constitute a class of highly successful antiviral compounds. 11,19,20 (S)-9-[3-hydroxy-2-(phosphonomethoxy)propyl]adenine (HPMPA), 21 which is active against a range of DNA viruses, 21 was the first to be evaluated in 1986, yet it was itself never commercialised as a drug. However, it led to the three structurally related successful drug molecules 19,20  Clearly, understanding the solution chemistry of ANPs, that is, their acid-base and metal ion-binding properties (e.g., [23][24][25][26][27][28][29] ), helps to reveal their antiviral mode of action (e.g. [29][30][31][32][33][34] because it is "Chemistry" that ultimately governs drug action. Evidently, the knowledge of the acid-base properties of liganding sites is important because (a) their metal ion affinities are closely connected to their proton affinities, 35 and (b) basicity also informs about energetics of hydrogen bonding which is important for drug recognition by target enzymes. In the present study we concentrate on HPMPA as a good representative of ANPs. Its dianion, HPMPA 2-, is shown in Fig. 1 together with PMEA 2and AMP 2-, [36][37][38][39][40][41] the natural metabolite that HPMPA mimics. In the course of this study comparisons will be made between these compounds. (phosphonomethoxy)ethyl]adenine (PMEA 2-). It is assumed that the preferred orientation of HPMPA 2corresponds to the one observed for PMEA 2in solution 36 and in the solid state, 37 which resembles the anti conformation of the parent nucleotide, adenosine 5'-monophosphate (AMP 2-), which is also shown in its dominating anti conformation. [38][39][40][41] These compounds are abbreviated as NP 2-= nucleoside phosph(on)ate derivative meaning further that P represents the phosph(on)ate group and N the nitrogen sites of the adenine moiety. HPMPA 2can accept up to four protons, two at the phosphonate group and, as we know from studies with adenosine, 42 two more at the adenine residue, i.e., at N1 and N7. Further protonation at N3 gives rise to an exceedingly acidic species, with pKa ≈ -4.0. 42,43 Therefore, we consider here only the 4-fold protonated species, i.e., H4(HPMPA) 2+ 36 Therefore, we decided to determine the pD dependence of the 1 H NMR chemical shifts of the non-exchangeable aromatic and aliphatic hydrogens of HPMPA in D2O. This procedure will not only provide the acidity constants for the equilibria (1) through (4), but in addition may provide some information about the conformations of the species present in solution.
The buffers used for pH calibration (pH 4.64, 7.00, and 9.00) 36 were from Metrohm AG, Herisau, Switzerland, and they were based on the U.S. National Institute of Science and Technology (NIST) scale (the apparatus was also from Metrohm). 36 In addition, a buffer with pH 1.00 was also used (also NIST scale), which was obtained from Merck AG, Darmstadt, Germany (for further details see 50 ). D2O (>99.8% D), NaOD (>99.9% D) and DNO3 (>99% D) were from Ciba-Geigy AG (Basel, Switzerland). All other reagents were the same as used before in related studies. 27,29,36,50,51 The experiments with HPMPA were carried out exactly as described for PMEA and its derivatives, 36 and this includes the evaluation methods. In brief, HPMPA was dissolved in 100% D2O, lyophilised, and then dissolved again in D2O, at a final concentration of 5 mM, in the presence of 0.1 M NaNO3 to control ionic strength (I). We note that at pD < 1 I is somewhat larger. 36 The pD of samples was measured with a pH meter (Metrohm) that had been calibrated with the buffers in H2O mentioned above, and adjusted with L quantities of NaOD or DNO3.
To obtain pD values, 0.4 pH units were added. 52

1 H NMR spectra of HPMPA at different pD values and site attribution of the resonance signals
The upper part of Fig. 2 shows the full 1 H NMR spectrum of HPMPA at pD 11.5, where the fully deprotonated species HPMPA 2dominates. The underlay in gray between 3.3 and 4.6 ppm highlights the region that encompasses the aliphatic protons, and is enlarged in the middle part of  the single doublet-of-a-doublet observed at pD 11.5). In addition, the splitting pattern (also two doublets-of-a-doublet) for the two CH2(1') protons is also better recognised; at pD 11.5, the two innermost peaks are nearly coinciding.
At pD > 8 the methylene protons neighboring the phosphonate group (P-CH2) are magnetically equivalent and appear as one doublet-of-doublets (see Fig. 2, middle), with the larger splitting due to coupling to 31 P ( 2 J = 9.4 Hz), and the smaller splitting due to 4-bond coupling to H2' ( 4 J = 2.2 Hz). At pD < 8, the two protons become magnetically inequivalent, and give rise to two doublets-of-a-doublet (see Fig. 2, bottom), with the larger splitting due to geminal coupling ( 2 J = 15 Hz), and the smaller due to coupling to 31 P ( 2 J = 9.4 Hz). The 4-bond coupling to H2' is no longer observed. The pD at which the two protons begin to differ coincides with protonation of the phosphonate group; the significance of this observation will be discussed further below.
It may be added that the assignments of all aliphatic protons in Fig. 2(middle) and 2(bottom) agree with those given in the literature for 1-deaza-HPMPA. 55 The assignment of the signals of the two aromatic protons is based on comparisons with PMEA and 5'-AMP ( Fig. 1) and the dependence of the signals on pD. 36,38,56 Note that the signal of H8 is downfield from that of H2.  when needed; at pD < 1 I is somewhat larger). 36 The solid lines represent the computercalculated best fits 36,38 of the experimental data points by using the averaged pKa/av values given in Table 1; the resulting shifts are listed in Table 2 (see also Section 2 and text in Section 3.2)

Chemical shifts in dependence on pD
Like with PMEA, 36 and as expected, also for HPMPA four protonation reactions are discernible from the combined plots (Fig. 3). Inflection points are at pD ca. 0, 2, 5, and 7.5, with different protons responding to a different extent to the four (de-)protonation reactions. With the exception of H8, for all protons curve fits for all four pKa values defined in eqns (1) to (4) could be carried out (Table 1).
Chemical Shift (ppm)  from the eight curves that allowed a fit (i.e., without that for H8). These three values (Table 1, last line) were now kept constant and a value for pK D 4 (HPMPA) D for the H8 proton chemical shift dependence was estimated by systematically varying the constant. The "best" value (-0.05 ± 0.4) was selected based on two criteria, namely, the error square sum of the fit and a reasonable shift difference (i.e. < 1.5 ppm). The final weighted means from the process described are given as pKa/av in the terminating row of Table 1. The four pKa values are similar to those found for PMEA, 36 which is unsurprising as the hydroxymethyl group is not expected to have a significant influence on any of the four protonation reactions.
In order to determine the chemical shifts () for all nine protons in the five differently protonated HPMPA species, the weighted means pKa/av (Table 1) were set as constants in a second round of curve fitting. Results of this analysis are listed in Table 2; the solid lines in Fig.   3 are those resulting from these fits. For the aromatic protons, the chemical shifts are similar to those of PMEA, 36 again in line with expectations, as the hydroxymethyl group is not expected to have a significant influence on the chemical shifts of these protons.  Table 1 and also text in Based on the chemical shift values () listed in Table 2 one may calculate shift differences () between different protonated species. These values are helpful in ascertaining the sites to which the various equilibria (1) to (4) pertain. In a first approximation, we expect that the hydrogens that are closest to the site of protonation will experience the largest changes in chemical shift.
Thus, both Fig. 3 and Table 3 furnish information about the site at which a particular acid-base reaction occurs: In the fully deprotonated HPMPA 2species (Fig. 1) the most basic site is the -PO 3 2residue which readily accepts a proton. 23,36 This is confirmed by the P-CH' hydrogen experiencing the largest shift difference upon the first protonation (    The next protonation step occurs at N1 and therefore significant downfield shifts are observed for H2 and H8 upon protonation of D(HPMPA) - (Table 3, column 4), which leads to the zwitterionic form D2(HPMPA) ± [eqn (3)]. That H8 is also sensitive to acid-base reactions occurring at N1 is not surprising as electron density changes are easily dispersed through the  system of the purine base, as is also observed with other adenine derivatives. 38,57 Further protonation leads to D3(HPMPA) + , and this reaction occurs again at the phosphonate residue, as the marked downfield shifts for both hydrogens of P-CH2 demonstrate (Table 3,  as is demonstrated by the pronounced downfield shift for H8 (Table 3, column 2), together with a smaller downfield shift for H2. This conclusion agrees with results obtained for related systems. 36,43,57 N3 is never protonated in systems discussed up to now, which agrees with previous experience. 42,43 Hence, one may conclude that the basicity of the various ring nitrogens of the adenine residue decreases in the order N1 > N7 > N3. 36 Generally, protonation of a site in the vicinity is expected to lead to downfield shifts, as a reduction in electron density will normally reduce electronic shielding of the proton of interest.
The data collected in Table 3 show some interesting deviations from this general principle.
Firstly, the first protonation of the phosphonate group gives rise to "wrong-way" (i.e. upfield upon protonation) shifts for the aromatic protons, most clearly observed for H8 (-0.069 ± 0.029; Table 3, column 5). Interestingly, H3' also experiences a wrong-way shift upon this protonation reaction (-0.033 ± 0.004 ppm; Table 3, column 5). In all cases, such wrong-way shifts are thought to be a consequence of changes in the dominant conformation in dependence on protonation state, combined with anisotropy of the magnetic properties of the phosph(on)ate group. 58 Secondly, the chemical shift trends for the P-CH2 protons (Table 3 and Fig. 3) also show some peculiar behaviour: whereas P-CH' is -as expected -strongly affected by the (de)protonation of the neighboring phosphonate group at pD 7.4 (= 0.195 ± 0.015 ppm), P-CH" barely senses this protonation step (= 0.022 ± 0.005 ppm). Together with the observation that the first protonation of the phosphonate group also renders the two protons magnetically inequivalent, this suggests that this protonation reaction leads to the predominance of a conformation in which one of the P-CH2 protons either senses little from the change in electron density due to the protonation reaction, or, more likely, two opposing effects cancel each other out. It is also noteworthy that the pKa values for the two first protonation steps (8.439 ± 0.198 and 5.358 ± 0.209) determined from the curve fitting for P-CH" deviate significantly from the overall averages; this is due to both transitions being ill defined for this proton.
Either way, the seemingly paradoxical chemical shift trends, i.e. wrong-way shifts of H8, H2 and H3' and the behaviour of the P-CH" proton, are in each case manifestations of the same concept: the change in charge and bound H + due to (de-)protonation of the phosphonate group leads to a change in conformational preferences and dynamics, similar to the situation in the parent nucleotide. 59 Fully deprotonated purine nucleotides are known to favour anti conformations, in which H8 faces the phosphate group, whilst protonation diminishes the prevalence of anti conformations. 38,59 These conformational preferences, together with anisotropic effects of the phosph(on)ate group, 58 are thought to be responsible for the chemical shift trends for H8 in 5'-AMP. 38,60 The magnitude of the wrong-way shifts for either PMEA 36 or HPMPA is smaller than that for 5'-AMP, which is probably owed to the larger conformational flexibility of the acyclic analogues. Nonetheless, these similarities in conformational preferences with their parent compounds may contribute to their effectiveness as substrate analogues for both kinases and nucleic acid polymerases.
We note that the data in Table 3 also suggest that H8 appears to experience a wrong-way shift upon the second protonation of the phosphonate group (-0.104 ± 0.071 ppm; Table 3, column 3), but would caution that the datapoints for H8 deviate considerably from the fitted curve in this pH range (Fig. 3) -which of course coincides with the pK D 3 (HPMPA) D value for which no fitting was possible for this proton. The plot in Fig. 3 therefore does not support the existence of a wrong-way shift for this step.

Intrinsic basicity and development of a microconstant scheme
For H2(HPMPA) ± the constants pK H 2 (HPMPA) H [= 4.23;  (3) and (4). Moreover, and this is of equal significance, the (de)protonation reactions occur at different parts of HPMPA, that is, at the phosphonate group and the adenine moiety ( Fig. 1) and the protonation sites will sense only little from each other due to the distance between them. This is different for the acid-base reactions that involve N1 and N7; both sites are at the adenine residue and therefore a proton at N1 will affect the properties of N7, as there is repulsion between these two positively charged sites -even more so since they are part of the same  system. In other words, a positive charge "at N1" will facilitate deprotonation at N7(H) + .
To learn the true proton affinity of N7 we therefore must consider the situation in a species which has a free, unprotonated N1 site; we write this species as + H•N7(HPMPA)N1 (or even shorter as + H•N7-N1). The consequence of this is that there must also be species in which N1 is protonated, namely + H•N7(HPMPA)N1•H + (or shorter + H•N7-N1•H + ). This in turn means that there are two ways from N7(HPMPA)N1 to form the species with a diprotonated adenine residue, beginning protonation either with N1 and followed by that of N7, or vice versa. If we ignore for now the phosphonate residue, we can summarise the situation regarding the two N sites as given in the equilibrium scheme shown in the upper part of Fig. 4.
This scheme defines the reactions for the micro acidity constants, and in the lower part of Fig. 4 the interrelations are given between these microconstants and the measured macro acidity constants. The definitions provided in the lower part of Fig. 4 follow known routes. 63 Fig. 4 shows that there are four unknown micro acidity constants, but only three independent equations interrelating them with the macroconstants; thus, one of the microconstants must be obtained independently.  Table 4; footnote a). follows eqn (7) which is valid for adenine derivatives; as error limits we now use one standard deviation (see Table 4):

Further attempts to estimate the intrinsic proton affinity of N7
As the preceding result is not very satisfying, we aimed at a further independent evaluation, based on the reasoning that a proton at N1 is expected, due to charge repulsion, to facilitate the deprotonation of N7(H) + . This charge effect, ∆ pKa, of N1(H) + on N7(H) + can be calculated for 9-methyladenine (9MeA) by using the results from Fig. 5  These derived values, namely -0.13 ± 1.38 and also 3.49 ± 1.38 (from Section 3.5) can be inserted into the upper and lower circles, respectively, of the microconstant scheme given in Fig.   5. The other microconstants in the scheme can now be calculated, making it complete. Unfortunately, the large error of 3.49 ± 1.38 (= pk H•N7-N1 N7-N1 ) propagates to pk H•N7-N1•H N7-N1•H = -0.13 ± 1.38. Despite this, things are overall conclusive because the previous result (Section 3.5) for the same microacidity constant was -0.18 ± 0.19 (Fig. 4); in other words, the two rather different evaluation methods lead to the same result. This also holds for R * [eqn (8)] which with 10 0.60 (= 4.0) gives a formation degree of about 20% for the H•N7(HPMPA)N1 ± tautomer, in excellent agreement with the previous 18% (Section 3.5).

Conclusions and a further revealing observation
There is a further most interesting observation: For 9-methyladenine (9MeA) the following two micro acidity constants have been determined, 64  (± 1.39) (see Fig. 5). In other words, 9MeA is a good mimic for the adenine part of HPMPA and therefore it is of relevance to note that the + H•N7(9MeA)N1 tautomer occurs with a formation degree of 7%. 64 It may therefore be concluded that the H•N7(HPMPA)N1 ± tautomer is certainly a minority species, but that it is definitely formed, the formation degree being between about 5 and 20%.

(Figs 4 and 5)
. This result is of relevance for the formation of macrochelates of phosph(on)atecoordinated metal ions with N7. [65][66][67][68][69] However, one has to be aware that these N7 interactions are weak and therefore easily affected by other weak metal ion interactions, e.g., with the ether oxygen, the ether oxygen in combination with N3, or the hydroxyl group of the -CH2-OH residue in M(HPMPA) (cf. ref. 29) and related complexes. 45 Such very weak interactions provide pitfalls: To mention just one, which follows from the above observation 29 that metal ions may interact with N3, but only in combination with the supporting ether oxygen-metal ion interaction, forming thus fused 5-and 7-membered chelate rings (cf. Fig. 1). This means, the phosphonate group of HPMPA 2is the primary binding site to which the metal ion is coordinated, followed by the ether oxygen and N3. In this context it is interesting to note that the acidic properties of N3(H) + in the + H•N3(9MeA)N1,N7 tautomer have been estimated 42 as pk H•N3(9MeA)N1,N7 N3-N1,N7 ≈ 2.45. Using 9MeA as a mimic for HPMPA (see above) one may assume in a first approximation that its (N3)H + acidity is approximately described by the mentioned micro acidity constant. Overall, the basicity of the adenine nitrogens decreases clearly in the order N1>N7>N3.

Conflicts of interest
There are no conflicts to declare.