Tricyclic Nucleobase Analogs and Their Ribosides as Substrates and Inhibitors of Purine-Nucleoside Phosphorylases III. Aminopurine Derivatives

Etheno-derivatives of 2-aminopurine, 2-aminopurine riboside, and 7-deazaadenosine (tubercidine) were prepared and purified using standard methods. 2-Aminopurine reacted with aqueous chloroacetaldehyde to give two products, both exhibiting substrate activity towards bacterial (E. coli) purine-nucleoside phosphorylase (PNP) in the reverse (synthetic) pathway. The major product of the chemical synthesis, identified as 1,N2-etheno-2-aminopurine, reacted slowly, while the second, minor, but highly fluorescent product, reacted rapidly. NMR analysis allowed identification of the minor product as N2,3-etheno-2-aminopurine, and its ribosylation product as N2,3-etheno-2-aminopurine-N2-β-d-riboside. Ribosylation of 1,N2-etheno-2-aminopurine led to analogous N2-β-d-riboside of this base. Both enzymatically produced ribosides were readily phosphorolysed by bacterial PNP to the respective bases. The reaction of 2-aminopurine-N9-β -d-riboside with chloroacetaldehyde gave one major product, clearly distinct from that obtained from the enzymatic synthesis, which was not a substrate for PNP. A tri-cyclic 7-deazaadenosine (tubercidine) derivative was prepared in an analogous way and shown to be an effective inhibitor of the E. coli, but not of the mammalian enzyme. Fluorescent complexes of amino-purine analogs with E. coli PNP were observed.


Reaction of Chloroacetaldehyde with 2-Aminopurine and Its Riboside
2-Aminopurine riboside reacts rapidly with chloroacetaldehyde (CAA) at room temperature and weakly acidic pH, to give essentially one main product [23], identified as a linear adduct of CAA (1,N 2 -etheno-2-aminopurine-N 9 -β-d-riboside, (3)), readily crystallized from the neutralized reaction mixture. This product is moderately fluorescent in the visible part of the spectrum (see Table 1). The assignment of the 1 H-and 13 C-NMR signals is shown in Table 2. There are also traces of a second, highly fluorescent product, with spectral characteristics similar to the minor product of the reaction of 2-aminopurine with CAA (see next paragraph), but we were unable to isolate this compound in sufficient quantities.
Both abovementioned products revealed moderate to intense fluorescence in neutral aqueous medium, centered at ca. 470 (1,N 2 -etheno-2-aminopurine (1)) and 405 nm (N 2 ,3-etheno-2-aminopurine (2)), with yields approximately 0.18 and 0.73, respectively (see Figure 3 and Table 1). The fluorescence spectra of both isomers were to some extent excitation-dependent (Figure 3), and their fluorescence decays revealed non-exponential behavior (Table 1), the facts possibly related to protomeric equilibria in the ground state (see Discussion). The ionic forms of the two tri-cyclic bases are also fluorescent (see Figure 4, below). Anionic and cationic forms of 1,N 2 -etheno-2-aminopurine (1) emit in the same region as the neutral molecule (460-470 nm), but with different yields (Figure 4a, Table 1). The largest Stokes' shift is observed for the cationic species (9300 cm −1 ). The respective excitation spectra are virtually in line with UV absorption of each form. The anionic form of the non-linear isomer 2 is strongly fluorescent at 390 nm (Figure 4b, blue line), while the cationic form of this compound exhibits two-band fluorescence, with Stokes' shift for the low-energy band exceeding 10,000 cm −1 (see Figure 4b, red curve).    (2) in aqueous medium at various pH: red curves, pH 3; green, pH 6.5 (a) and 7 (b); blue-pH 11.5. Excitation was at 300 nm (a) and 290 nm (b). The spectra of the ionic forms are virtually excitation-independent.  [28], extended for the etheno protons (see Figure 1).

Enzymatic Ribosylation of the Etheno-2-Aminopurine Isomers Using Various Forms of PNP
Both isomers of the etheno-2-aminopurine are substrates for PNP from E. coli in the reverse (synthetic) pathway with R1P as a ribosyl donor. Ribosylation of 1,N 2 -etheno-2-aminopurine (1) as well as N 2 ,3-etheno-2-aminopurine (2) led to substantial changes in the UV absorption ( Figure 5) and fluorescence ( Figure 6) spectra, suggesting that the ribosylation sites may be different from the proton location in the respective base (the latter is possibly N 9 ). We have also noted a striking similarity between UV absorption spectrum of the ribosylation products and the fluorescence excitation spectra of the minor tautomers of the respective bases, measured with observations at the blue edge of the emission spectrum (cf. Figures 2 and 3). The ribosylation rate for the non-linear ε2AP isomer, N 2 ,3-etheno-2-aminopurine (2), is ca. 40-fold higher than that for the linear isomer 1, and comparable to the ribosylation rate of guanine, measured in the same conditions ( Table 3).
Ribosides of 1,N 2 -etheno-2-aminopurin (1) as well as N 2 ,3-etheno-2-aminopurine (2), generated using the E. coli PNP, were subjected to HPLC purification on a milligram scale, and their identification and properties are described in the next paragraph.  Kinetic parameters of the synthetic (ribosylation) reaction, catalyzed by the wild-type and mutated forms of PNP, were determined using standard procedures, and are summarized in Table 3. There are some minor differences between wild-type enzymes (E. coli and calf PNP) and forms mutated in the active site, but without qualitative differences, observed previously for some purine analogs [21,22]. Generally, kinetic parameters for ribosylation of 2 by E. coli PNP and its mutated forms do not differ markedly from those determined earlier for natural purines [12], and the K m values are close to 10 µM, hence comparable to those observed for guanine ribosylation under the same conditions. It may also be of interest that the trimeric calf spleen PNP, much more demanding in respect to substrate structures than the hexameric E. coli enzyme [12], ribosylates N 2 ,3-etheno-2-aminopurine (2) with moderate rate (Table 3), but is apparently inactive towards the second (linear) isomer 1. It may be of interest that N 2 ,3-etheno-2-aminopurine (2) is fairly rapidly ribosylated by the calf PNP, mutated in the active site (N243D). But the ribosylation goes in an essentially similar way as with the E. coli PNP, that is, giving the identical single product (Table 3). Table 3. Kinetic parameters for the enzymatic ribosylation of selected etheno-purine derivatives in 40 mM HEPES buffer, pH 7, by α-d-ribose-1-phosphate, using various forms of PNP (wt = wild type; nr-no reaction detected; nd-not determined). Standard errors are estimated to be~15%.

Properties and Identification of the Enzymatically Produced Ribosides
Reaction of 2-aminopurine riboside with chloroacetaldehyde gives N 9 -riboside (3) of the linear isomer of etheno-2-aminopurine (for the assignment of the 1 H and 13 C-NMR signals see Table 2), revealing spectral properties very similar to those of the respective base (see Section 3.2), but with a single emission band (465 nm) and a single decay time (Table 1). By contrast, the main product of enzymatic ribosylation of 1,N 2 -etheno-2-aminopurine, 4, is characterized by the emission at 400 nm, single decay time, and the UV absorption shifted to the blue by over 20 nm (see Figures 5 and 6, left panels, and Figure 7). This riboside undergoes protonation with pK a~6 .3 (see Supplementary Materials, Figure S1). The compound has been subjected to purification using semi-preparative HPLC, and identified as 1,N 2 -etheno-2-aminopurine-N 2 -riboside ((4), see below).
The riboside produced enzymatically from the non-linear isomer of etheno-2-aminopurine, 5, also differs spectrally from the parent base 2. Its emission spectrum is shifted by~45 nm to 355 nm, and UV absorption reveals fine structure ( Figure 8). Fluorescence decay is mono-exponential, and decay time is lowered to~2 ns, with yield~0.29 ( Table 1). The protonated form of the riboside is also strongly fluorescent, but we did not detect any traces of dual emission (as observed in the emission spectrum of the protonated base, see Figure 4), and the Stokes' shift was moderate ( Figure 8). We conclude that the photo-transformation, observed in the protonated base as two-band emission, is absent in the riboside.  The fluorescence decay times of all the ribosides 3-5, both N 9 -ribo and N 2 -ribo, measured at pH > 7.5, were mono-exponential, in agreement with the view that no protomeric equilibrium is possible in the ground states of the ribosides, at least at their heteroaromatic moiety. Accordingly, fluorescence excitation spectra were in line with the UV absorption. Interestingly, the emission and excitation spectra of N 2 -ribosides 4-5 resemble those of the minor tautomers of both isomeric bases (see Section 3.2). This leads to tentative identification of the minor tautomers of both N 2 ,3-ε2AP (2) and 1,N 2 -ε2AP (1) as N 2 H (see Discussion).
The ribosylation site of N 2 ,3-etheno-2-aminopurine (2) via the enzymatic process was identified as N 2 (for the assignment of the NMR signals in 5 see Table 2), based on the observation of non-vanishing three-bond scalar couplings between H1 and both C2 and C11, as well as between H11 and C1 , in the 1 H-13 C HSQMBC spectrum ( Figure 9). In the absence of other observable scalar couplings between the ribose and the base nuclei the position N 2 is the only ribosylation site congruent with this coupling pattern. The ribosylation product of the linear isomer 1 was identified as 1,N 2 -etheno-2-aminopurine-N 2 -riboside (5), based on the observation of a coupling pattern fully analogous to that of the non-linear isomer.  Table 2).
The ribosides are fairly stable in solution, but we were unable to crystallize them due to small amounts obtained. They were stored as frozen in neutral aqueous solutions.

Phosphorolysis of The Ribosides with Various Forms of PNP
The highly fluorescent N 2 ,3-etheno-2-aminopurine-N 2 -β-d-riboside (5), generated enzymatically (Section 3.3) is readily phosphorolyzed in the phosphate buffer by both E. coli and calf PNP. The reaction rates are in the case of E. coli enzyme comparable or even higher than that of guanosine phosphorolysis (Table 4), while those obtained for calf PNP are moderate. The observed spectral changes are reverse in respect to those presented of Figures 4 and 5 for the synthetic process (see Supplementary Materials, Figure S2), and lead to very pronounced fluorogenic effect (not shown). Kinetic analysis revealed relative low K m values for these reactions (Table 4), although they are somewhat higher than those for the synthetic reactions (Table 3). We expect that the human PNP, which is quite similar to the calf enzyme [12], will also react with N 2 ,3-etheno-2-aminopurine-N 2 -β-d-riboside (5), giving the highly fluorescent base 2 as a reaction product, with possible applications to analytical or clinical biochemistry. This point will be addressed in a separate paper. The ribosylated linear isomer, identified as 1,N 2 -etheno-2-aminopurine-N 2 -riboside (4) reacts much slower, and only with the E. coli PNP as a catalyst (Table 4).
We have found that 1,N 6 -etheno-tubercidine (6) competitively inhibits phosphorolysis of purine nucleosides, catalyzed by the E. coli enzyme. The inhibition constant, K i~4 .5 µM, is comparable to that of tubercidine itself [12], some formycin derivatives [31,32], and other good inhibitors of this enzyme [12]. Etheno-tubercidine (6) is therefore a good candidate to observe enzyme-ligand complexes by spectral methods, as shown below. No inhibition of the calf PNP was observed, at least in moderate concentrations (up to 50 µM) that we used.

Fluorescence of Enzyme-Ligand Complexes
Titration of the E. coli PNP with N 2 ,3-etheno-2-aminopurine (2) in the presence of phosphate leads to moderate quenching of the protein fluorescence at 305 nm and formation of fluorescent complexes, visible both in emission and in excitation spectra (see Figure 10). Additionally, the fluorescence excitation spectra reveal fluorescence energy transfer (FRET) from the protein to the complexed ligand, since they show marked enhancement in the region 270-280 nm, where the tyrosine residues of the PNP molecule absorb ( Figure 10). The fluorescence quantum yield of the bound ligand is very high and comparable to that of the free molecule (0.73, see Table 1), as evidenced by difference spectra (no negative contribution at the long-wavelength tail of the spectrum).
The respective difference spectra (Figure 10c,d) confirm moderate quenching of protein fluorescence (right-hand side), evidence of FRET (left), and appearance of the complex emission at 380-400 nm, revealing some fine structure, which is absent in the free ligand spectrum. We did not observe analogous complexes in the absence of phosphate, in spite of the low Michaelis' constants obtained for the synthetic reaction (Table 3).  Etheno-tubercidine (6) forms somewhat similar fluorescent complexes with the E. coli PNP. As shown below, FRET from the protein to the ligand is evident in the excitation spectra (Figure 11), and the ligand fluorescence blue shifted from 415 to ca. 390 nm (Figure 11b,c). In this case more than 60% of protein fluorescence is quenched when protein binding sites are saturated (cf. Figure 11). Very small changes in the excitation spectra suggest that the ligand is bound to the protein as a neutral species; but some irregularities in the difference spectra (Figure 11c) may indicate that the binding sites in the hexameric PNP molecule are not equivalent, but this conclusion needs verification.

Fluorescent Isomers of the Etheno-2-Aminopurine
Vinyl chloride, a known chemical mutagen and carcinogen, acts as a modifier of nucleobases, in particular, of adenine and guanine moieties [14,15], which upon this chemical modification change the respective coding properties, leading to the mutagenic effect [33,34]. In addition some of the bases modified in this way exhibit additionally marked fluorescence [1][2][3][4], which make them good candidates for fluorescent probes in enzymological research.
We found two isomeric products of the reaction of CAA with 2-aminopurine, both revealing intense fluorescence in aqueous medium. In both cases, fluorescence was excitation-dependent and decays were non-exponential. The most likely interpretation of this fact is the N 9 H-N 7 H tautomerism, confirmed for some purines and analogs [35][36][37], and suggested for 1,N 6 -ethenoadenine [21], but other tautomeric forms, like N 2 H or N 3 H, cannot be excluded (see Figure 12). The similarity between properties of the minor tautomers and N 2 -ribosides strongly suggests participation of the N 2 H protomers 1d, 2c. Large Stokes' shifts observed for the cationic species of both isomers suggests photo-transformations. Such photo-transformations (e.g., excited-state proton transfer and a resulting photo-tautomerism) are not infrequent among the fluorescent purine analogs and derivatives [38].

Enzymatic Syntheses of the Tri-Cyclic Ribosides and Their Properties
In this and preceding papers [21,22], we have described substrate and inhibitor properties of several tri-cyclic nucleobase analogs towards the enzyme purine-nucleoside phosphorylase. Many of the investigated compounds, in particular, etheno-adenosine and one isomer of etheno-guanosine were found to be excellent substrates for the bacterial (E. coli) type of PNP, and the respective bases were easily ribosylated in the reverse process [21,22]. Enzymatic ribosylation of the tri-cyclic nucleobase analogs and similar compounds leads to non-typical ribosides, which are nevertheless good substrates for PNP, as shown previously for the N 6 -ribosylated 1,N 6 -ethenoadenine [21], and analogous isoguanine derivatives [22].
Both isomers of etheno-2-aminopurine (1 and 2) are ribosylated using R1P as a ribose donor and the E. coli PNP as a catalyst, but the ribosylation site is N 2 , rather than N 9 . Somewhat similar ribosylation pattern was previously reported for etheno-adenine [21] and etheno-isoguanine [22], but in both cases different enzyme forms led to different ribosylation products. By contrast, the linear isomer of etheno-guanine was rapidly ribosylated on the "canonical" N 9 nitrogen [22].
This ambiguity in the ribosylation sites in enzymatic reactions, catalyzed by PNP, was also previously observed in 8-azapurines [39,40], and probably results from the plasticity of the active site of this class of enzymes. For example, in the X-ray crystal structure of hexameric PNP from H. pylori, bound to an inhibitor formycin A (8-aza-N 9 -deazaadenosine), both the standard (anti) and non-standard (syn) conformations of the inhibitor were revealed [41]. In another study, acyclo-guanosine, an inhibitor of mammalian PNPs, with the acyclic chain bound to N 7 position of the purine base, was found in the inverted ('up-side-down') position in the active site of calf PNP, with the chain located in the place normally occupied by the ribose [42]. Finally, N 3 -β-d-ribofuranosyladenine and N 3 -β-d-ribofuranosyl-hypoxanthine, with sugar moiety attached to the N 3 position of the base, were found to be non-conventional substrates of purine nucleoside phosphorylase from E. coli and calf [43].
Some of the fluorescent tri-cyclic ribosides are also moderate to good substrates of mammalian PNPs, known to be homologous to human enzymes [12]. This makes them potential indicators of PNP activity in biological or clinical samples. As an illustration, we have recently shown that human erythrocytic PNP activity can be measured fluorimetrically in 1000-fold diluted hemolysates using the N 6 -ribosylated 1,N 6 -ethenoadenine as an artificial substrate [44]. We expect that the new compounds presented in this work, particularly the ribosides of N 2 ,3-etheno-2-aminopurine (2), will exceed in sensitivity other fluorescent indicators of PNP activity (to be published elsewhere).

Fluorescent Complexes
Various forms of PNP form fluorescent complexes with purines and their analogs [17]. Long ago Porter et al. [45] reported a fluorescent complex of calf PNP with guanine, ascribing its fluorescence to the anionic form of the ligand. Later investigations with fluorescent 8-azaguanine derivatives and calf PNP [46,47] instead suggested neutral ligand as the emission source. The E. coli PNP is an interesting object of this kind of experiments since its molecule does not contain tryptophan and its native fluorescence is located near 305 nm [12], allowing easy observations of emission of the complexes, as illustrated by Kierdaszuk et al. [31,32]. In the latter case, it was possible to identify individual protomeric forms of the ligand formycin A (8-aza-N 9 -deazaadenosine), when bound to the PNP molecule.
Somewhat similar fluorescent complexes and FRET were previously observed with E. coli PNP complexed with formycin A and its N-methyl derivatives as ligands [31]. Now, we present evidence for the highly fluorescent base-enzyme complexes, observed in the presence of phosphate. We did not observe analogous complexes in the absence of phosphate, in spite of the low Michaelis' constants obtained for the synthetic reaction (Table 3). This is in line with the fact that for both mammalian and bacterial PNPs, due to complex mechanisms of catalysis exhibited by these two enzyme families, K m does not describe the affinity of most substrates, e.g., [12,16,48] adequately. This also suggests that, in agreement with previous mechanistic considerations and stabilization pattern exhibited by substrates [49], the purine base is the only E. coli PNP substrate that cannot bind to the enzyme molecule in the absence of either a second substrate (in this case R1P) in the synthetic reaction, or phosphate, a substrate in the reverse phosphorolytic path, in the latter case forming a so-called dead-end complex.
Previous papers have shown that the E. coli PNP molecule can selectively bind individual tautomeric forms of some ligands [31,32]. At present, we cannot identify which of possible tautomeric structures is bound, but the minor tautomer, visible in solution spectra as an inflection, and postulated to be the N 2 H protomer 2c, is evidently not responsible for the above changes since the emission spectra of the complex differ markedly from those of the N 2 -riboside 5 (cf. Figures 10 and 5). Our interpretation of the presented data is that N 7 or N 9 are likely locations of the proton in the complex.
The etheno-derivative of 2-aminopurine-N 9 -β-d-riboside ( Figure 13) was prepared according to Virta et al. [23], and crystallized from the neutralized reaction mixture. Its structure was confirmed by NMR data, which were in agreement with the published results [23], and by MS: comp.  The reaction of chloroacetaldehyde with tubercidine (7-deazaadenosine) was carried out as previously described [24], with the modification that an aqueous CAA was applied instead of the distilled reactant. The product purified by semi-preparative HPLC in milligram quantities (approximate yield > 60%). The spectral parameters were in agreement with those previously published [24]. See Supplementary Materials for more details.
Product separation and purification were performed by HPLC on a UFLC system from Shimadzu (Kyoto, Japan) equipped with UV (diode-array) detection at 260, 280 and 315 nm, and a fluorescence detector. The column used was a Kromasil reversed-phase, semi-preparative C-18 column (250 × 10 mm, 5-µm particle size). Elution was initially (10-15 min) isocratic, followed by a water-methanol gradient (usually 10-30% methanol for 40 min, see Supplementary Materials for details). All buffers were of analytical grade and showed no fluorescence background.

Spectral Measurements
Fluorescence spectra were measured on a Varian Eclipse instrument (Varian Corp., Palo Alto, CA, USA), and UV absorption kinetic experiments were performed on a Cary 5000 (Varian Corp., Palo Alto, CA, USA) thermostated spectrophotometer. Fluorescence yields were determined relative to tryptophan (0.15) or 1,N 6 -ethenoadenosine in water (0.56, see ref. [2]). Emission spectra were measured in semi-micro cuvettes, pathlength 4 mm, to diminish the inner-filter effect. Typical spectral resolution was 2.5 nm. The ionization constants (pK a values) were determined spectrophotometrically using 20-50 mM phosphate and/or acetate buffers.
NMR measurements were run in DMSO-d 6 at 25 • C on an Avance III HD 800 MHz spectrometer Bruker (Bruker BioSpin AG, Fällanden, Switzerland) equipped with a cryogenically-cooled triple resonance (HCN) probe (the sample identified as the N 2 -riboside of 1,N 2 -etheno-2-aminopurine (3)) and on a Bruker Avance III HD 500 MHz spectrometer equipped with a room-temperature triple resonance (HCN) probe. For all samples, the following spectra were acquired: a standard 1D proton spectrum, a 1 H, 1 H CLIP-COSY [51] and a gradient-selected 1 H, 13 C HSQC [52]. Phase-sensitive gradient-selected HMBC spectra tuned for 2 and 8 Hz J couplings were acquired for the N 2 riboside of 1,N 2 -etheno-2-aminopurine (3), while gradient-selected 1 H, 13 C HSQMBC ( [53], modified by using hard pulses) were run for all other samples-tuned for optimal detection of 12 Hz J couplings in the case of 1, N 2 -etheno-2-aminopurine or two separate experiments tuned for 2 Hz and 8 Hz J couplings otherwise. 1 H chemical shifts were referenced by the field-locked substitution method using a (less than) 1% sample of TMS in DMSO-d 6 , and 13 C chemical shifts were referenced using the unified chemical shift scale [54]. The 2D spectra were processed using the TopSpin 3.6.1 software package (Bruker) and inspected by the Sparky program [55] with manual peak-picking.

Mass Spectrometry
The structure and purity of the new compounds 1-2 and 4-5 were confirmed using the high-resolution mass spectrometry with positive electrospray ionization HRMS (+) ESI. Mass spectra were recorded on a Thermo Scientific QExactive spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Spray voltage was 3800 V, capillary temperature 320 • C.

Enzymes and Enzymatic Reactions
Recombinant E. coli PNP, calf spleen PNP, and their mutated forms were expressed in E. coli and purified according to the procedures described earlier [56,57]. Enzyme concentrations were calculated per monomer.
Enzymatic ribosylation reactions were carried out in 1 mL cuvettes (pathlength 4 mm) in~50 mM HEPES buffer, pH 7.3, using 0.5 mM R1P as a ribose source. Reactions were followed fluorimetrically (see Section 2.4). On a larger scale (2-4 mg), the reactions were run in Eppendorf tubes, volume 2-3 mL, using either R1P (3-10 mM) or 7-methylguanosine as ribose sources. Products were purified by semi-preparative HPLC, concentrated and the final solutions kept frozen. Their structure has been confirmed by NMR (see Section 2. Phosphorolysis reactions were run in 40-50 mM phosphate buffer, pH 7.0 or 6.5, and followed spectrally or fluorimetrically. Kinetic parameters were calculated by standard methods.

Conclusions
We have shown that 2-aminopurine easily reacts with aqueous chloroacetaldehyde to give two fluorescent, isomeric products 1 and 2. Both products are substrates for the bacterial (E. coli) purine nucleoside phosphorylase, but in both cases the ribosylation site has been found to be N 2 rather than N 9 . The new ribosides are fluorescent and potentially useful as fluorescent probes. The antibiotic tubercidine (7-deazaadenosine) also reacts with aqueous chloroacetaldehyde to give the fluorescent product 6. This product is an inhibitor of the E. coli PNP and forms fluorescent complexes with the enzyme.