General Principles for Yield Optimization of Nucleoside Phosphorylase‐Catalyzed Transglycosylations

Abstract The biocatalytic synthesis of natural and modified nucleosides with nucleoside phosphorylases offers the protecting‐group‐free direct glycosylation of free nucleobases in transglycosylation reactions. This contribution presents guiding principles for nucleoside phosphorylase‐mediated transglycosylations alongside mathematical tools for straightforward yield optimization. We illustrate how product yields in these reactions can easily be estimated and optimized using the equilibrium constants of phosphorolysis of the nucleosides involved. Furthermore, the varying negative effects of phosphate on transglycosylation yields are demonstrated theoretically and experimentally with several examples. Practical considerations for these reactions from a synthetic perspective are presented, as well as freely available tools that serve to facilitate a reliable choice of reaction conditions to achieve maximum product yields in nucleoside transglycosylation reactions.

The biocatalytic synthesis of natural and modified nucleosides with nucleoside phosphorylases offerst he protecting-groupfree direct glycosylationo ff ree nucleobasesi nt ransglycosylation reactions.T his contribution presents guidingp rinciples for nucleoside phosphorylase-mediated transglycosylations alongside mathematical tools for straightforwardy ield optimization. We illustrate how product yields in these reactions can easily be estimated and optimized using the equilibriumc onstants of phosphorolysis of the nucleosides involved. Furthermore, the varying negative effects of phosphate on transglycosylation yields are demonstrated theoretically and experimentally with severale xamples. Practical considerations for these reactions from as ynthetic perspective are presented, as well as freely availablet ools that serve to facilitate ar eliable choice of reaction conditions to achieve maximum product yields in nucleoside transglycosylation reactions.
Nucleosides are highly functionalized biomolecules essential for the storage of information as DNA and RNA,c ellular energy transfer and as enzyme cofactors. Modified nucleosides are widely employed as pharmaceuticals for the treatment of cancers and viral infections. [1] Consequently,their synthetic accessibility is crucial.H owever,t he preparation of nucleosides and nucleoside analogues by conventional synthetic methods heavily relies on protectingg roups and, thus, suffers from poor atomic efficiency and low yields. [2][3][4][5] Biocatalytic methods offer the efficient and protecting group-free synthesis of pyrimidine and purine nucleosides. The use of nucleoside phosphorylases (NPases) for the preparation of nucleosides andt heir analogues in transglycosylation reac-tions is firmly established [6] and numerouse xamples of enzymatic or chemoenzymatic syntheses can be found in the literature. [7][8][9][10][11][12][13][14] NPases catalyze the reversible phosphorolysis of nucleosides to pentose-1-phosphates (Scheme 1, I). In transglycosylation reactions, af orward and ar eversen ucleoside phosphorolysis are coupledi nsitu to glycosylate af ree nucleobase with the pentose-1-phosphate generated by the first reaction (Scheme 1, Ia nd II). Formally,t his equals ad irect glycosylation of the nucleobase to yield anucleoside of interest. Conveniently,n ature has provided an arsenal of robustb iocatalysts that offer ab road substrate spectrum, excellent tolerance to harsh reactionc onditions as well as perfect regio-and diastereoselectivity at the C1' position. [11,12] Despite their great versatility,e nzymatically catalyzed nucleoside transglycosylation reactions have previously suffered from an unclear interrelation between yields and the employed enzymes and starting materials. Particularly,t he impact of differents ugar donors and/or nucleobases as wella sv arying phosphate concentrations on the product yield had remained unclear until recently.T he pioneering work of Alexeev et al. [15] demonstrated that yields of nucleoside transglycosylation reactions involving uridine and adenosine can be accurately predicted based on the equilibrium constants of phosphorolysis of the sugar donor and the product nucleoside. They concluded that the ratio of the equilibrium constants of the sugar donora nd the product nucleoside (K 1 /K 2 )d etermines maximum product yields and that an excesso fs ugar donor is further beneficial. On the other hand, increasing phosphate concentrationsw ere shown to have an egative impact on product yields. However,A lexeev and colleagues [15] based their calculations on the assumption that the concentration of phosphate is constant and furthermoreo nly investigated one example of ah igh-yielding NPase-catalyzed transglycosylation. As ac ontinuation of the considerations of Alexeev et al. [15] we explored this reaction system from ap ractical synthetic perspective and developed au niversally applicable equation for yield prediction. We show that the yield-diminishing effect of phosphate strongly depends on the equilibrium constant of phosphorolysis of the nucleoside of interest (K 2 ). This important feature, which proved critical in the biocatalytic preparation of pharmaceutically relevant pyrimidine nucleosides has thus far not been described theoretically or experimentally.A longside our freely available Pythonc ode for precise yield predictions (see below) we also provide as implified equation for the estimation of product yield that allows for straightforwarda nalytical solutionsi nsteado ft he numerical solutions previouslyr equired.
Nucleoside transglycosylation reactions are generallyc onsidered as formal glycosylations of an ucleobase B2,w hich yields the corresponding nucleoside of interest, N2.H ere, as tarting nucleoside, N1,i su sed as ag lycosylation agent with the purpose of donating the sugar moiety. In an enzymec ascade, the sugar donor N1 is subjected to phosphorolysis yieldingapentose-1-phosphate (P1P), which is consumed in the sequential reactionw ith nucleobase B2 to produce N2 (Scheme 1). The yield of this reactioni sg enerally defined as the formation of N2 in respectt oB2,n eglecting the other reagents P1P, B1, N1 and phosphate. Indeed, inorganic phosphate only plays ac atalytic role as it is used in the first step butl iberated again in the following reaction.
Generally,y ields in NPase-catalyzed transglycosylations are dictatedb yt he equilibrium constraints of the two half reactions Iand II [Eqs. (1), (2)]: where K 1 and K 2 are the apparent equilibrium constants of phosphorolysis of the sugar donor and product nucleoside,r espectively, [  are the equilibrium concentrations of the nucleosides andb ases. Alexeev et al. [15] previously solved this system of equations by assuming aconstant concentration of phosphate and numerically solving the resulting cubic equation.
When we attempted to apply their equations to the synthesis of the pharmaceutically relevant nucleoside 5-ethynyluridine we were unablet oo btain results that were in agreement with experimental HPLC data, as their formula yielded negative values for this case (Table S1 in the Supporting Information). Therefore, we sought to establish am athematical tool that allows general applicabilitya nd reaction optimization of all nu-cleoside transglycosylations. Bypassing the simplification made by Alexeev and co-workers,w ei mplemented the systemo f equilibrium constraints (1) and (2) including all reagents as variables in aP ython code to obtain more precise predictions (see externally hostedP ython code). [16] Numerical solutions of this system allowed theoretical examination of the effect of phosphate and sugar donor excess on the product yield, considering ar easonable range of equilibriumc onstants. [17] Approaching zero phosphate concentration, the maximum (ideal) product yield can be obtained, but at higher phosphate concentrationsa na pparent loss of yield can be observed due to phosphorolysis (or non-synthesis) of the product nucleoside (Figure 1). Whereas the K 1 /K 2 ratio (equal to K N )d ictates the maximumy ield with minimal phosphate, K 2 determines the extent of yield loss in the presence of phosphate. Ah igh K N in the order of 5-15 promises good to excellent yields (i.e., > 90 %) with only moderate excesses (i.e., twofold) of the sugar donor.O nt he other hand, reactions with al ow K N requireagreat excesso fs ugar donor to facilitate yields upward of 50 %. Interestingly,t he effect of phosphate varies between systems with the same K N ,w hich results from the fact that high K 2 values dictate ag reater degree of phosphorolysis of N2 at non-negligible phosphate concentrationseven at great excesso ft he sugar donor N1 (Figure 1). Notably, whereas potentialf ormation of intermediate pentose-1-phosphate needs to be considered for ar ealistic assessment and prediction of synthetic yield, we only observed less than four percentage points of deviation from the ideal yield for any nucleoside transglycosylation with < 0.3 equiv of phosphate in the reaction conditions we covered with our considerations. [16] To validate thesep redictionse xperimentally and demonstrate the varying impact of phosphate on the product yield, we prepared as eries of natural and base-modified ribosyl nucleosides from their respective nucleobases, using uridine as a sugar donor. Fitting of the experimental data to the equilibrium constraints [15,16] yielded equilibrium constants K 1 and K 2 very similar to those reported previously [17] and revealed a great range of apparent equilibrium constants K 2 (0.01 to 0.35 at 60 8C, pH 9) and K N (0.4 to 16.0). In all cases,t he experimental yields determined by HPLC agreedw ell with the predictions obtained for different phosphate concentrations ( Figure 2). Our data emphasize that, as illustrated in Figure 1, particularly the yields of transglycosylation reactions with high K 2 values suffer enormously from phosphate concentrations higher than strictly necessary.F or instance, adenosine formation (K 2 = 0.01) was impacted only minorly by the addition of 10 equiv of phosphate (92 %i deal yield vs. 88 %e xperimental yield with 10 equiv of phosphate), but 5-ethynyluridine yield (K 2 = 0.35) dropped by more than 30 percentage points under the same conditions (53 %i deal yield vs. 22 %e xperimental yield with 10 equiv of phosphate;F igure 2). Thus, steep losses in yield should be expected for products with ah igh K value (K 2 ), whereas the synthesiso fn ucleosides with low K values tolerates significant amountso fp hosphate (Figure 2). Consistent with our predictions, we only observed small deviations from the maximum yield in the experimentsw ith 0.2 equivalents of phosphate. Thus, the concentration of phosphate should be  Figure 1. Impact of different K 1 and K 2 valuesont ransglycosylation yieldand phosphate gap. Realistic K 1 and K 2 valuesw ere assumed basedo nr ecently reportede quilibrium constants. [17] The graphsf or maximum yield (max.;b lack), 0.1 equiv (green), 1equiv (blue)a nd 10 equiv (red) of phosphate were plotted using numericalsolutionso ft he system of equilibrium constraints (1) and(2) calculatedw ith the Pythonc ode describedint he external SupportingI nformation. [16] Figure 2. Biocatalytic synthesis of nucleosides by transglycosylation.R eactions were performedw ith 1mm uridine as sugardonor (K 1 = 0.16),0 .5 mm nucleobase, 32 mgmL À1 pyrimidineN Pase (2.5 UmL À1 )a nd 66 mgmL À1 purine NPase (5.0 UmL À1 )in5 0mm glycine buffer at pH 9a nd 60 8Cw ith either 0.1 mm (0.2 equiv in respect to the starting base), 0.5 mm (1 equiv)or5m m (10 equiv)K 2 HPO 4 in at otal volume of 1mL. Experimental yield (^)w as determined by HPLC considering conversionoft he free nucleobase to its corresponding ribosyln ucleoside. Predictions( blue, lightb lue, turquoise and greenc olumns)w ere carried out with the Python code describedi nt he external SupportingInformation. [16] The values for the maximum yield (max.;b lue) can also be obtained from Equation (4). kept lowi ns ynthetic nucleoside transglycosylations to obtain maximum yield. For these cases, calculation of maximum (ideal) conversion provides ac lose approximation of the yield and allows for the use of as implified formula.
Considering ideal (intermediate-free) coupling of the two half reactions, Ia nd II, the terms for phosphate and P1P would cancel in the mathematical consideration of this system, as the productiono ft hese in one half reactioni sc ompensated by the consumption in the other.C onsidering the net reaction, one may therefore define [Eq. (3)]: with definitions from above. Solving this equation for the concentration of the product nucleoside, N2,y ields only one physically possible solution that can be used to calculate ideal (phosphate-and P1P-free) yields of nucleoside transglycosylations with variable initial concentrationso ft he sugar donor N1 and the nucleobase B2,[N1] 0 and [B2] 0 ,respectively [Eq. (4)]: Thus, the maximum yield (at zero phosphate) can be calculated easily from Equation (4) to reflect ar ealistic estimate of the experimentaly ield if < 0.3 equiv of phosphate are used. Ideal yields for conversions employing ar ange of pyrimidine and purine ribosyl and 2'-deoxyribosyl nucleosides [17] with different reaction conditions including sugar donor excess and temperature, can be calculated with an Excel sheet freelya vailable from the externally hosted Supporting Information. [18] These considerations andp revious findings [15,16] bear several practical implications for NPase-catalyzed nucleoside transglycosylations. First, ah igh K 1 /K 2 ratio (high K N )l eads to excellent yields which can be achieved with moderate excess of the sugar donor,a sm entioned by Alexeev and colleagues, [15] and estimated easily with Equation (4). Second,p yrimidine nucleosides serve better as sugar donors than purine nucleosides. [17] From ap ractical point of view,u ridine and thymidine recommend themselves as ribosyl and 2'-deoxyribosyl donor,r espectively,d ue to their simple commerciala vailability and high K value. Third, phosphate concentration in the transglycosylation reactionshould generally be kept as low as possible to prevent loss of product yield. This becomes especially important in the synthesis of nucleosidesw ith high K values, such as pyrimidine nucleosides like 5-ethynyluridine. Thus, 0.1-0.3 equivalents of phosphate in respect to the startingb ase may presenta n appropriate trade-off between reactions peed and maximum yield. Ap otential workflow for the fruitful application of the methodology presented in this work is suggested in the Supporting Information.
Given the easy accessibility of apparent equilibrium constants of phosphorolysis of any nucleoside of interest, the tools for yield prediction presented in this work aid the straightforward design and optimization of nucleoside transglycosylations to facilitate high yields in NPase-catalyzed reactions. Exact yield prediction of transglycosylations may be performed with our Phyton code considering phosphate [16] and practical estimations forideal yield can easily be obtainedf rom Equation (4). [18] Experimental Section Enzymatic nucleoside transglycosylations were performed with 0.5 mm nucleobase, 1mm uridine as sugar donor,3 2mgmL À1 pyrimidine NPase (2.5 UmL À1 ;E -PyNP-0002, BioNukleo GmbH, Berlin, Germany) and 66 mgmL À1 purine NPase (5.0 UmL À1 ;E-PNP-0002, Bi-oNukleo GmbH) in 50 mm glycine buffer at pH 9a nd 60 8Cw ith either 0.1 mm (0.2 equivalents in respect to the starting base), 0.5 mm (1 equiv) or 5mm (10 equiv) K 2 HPO 4 in at otal volume of 1mL. Reaction mixtures were prepared from stock solutions and started by the addition of the enzyme(s). Time to equilibrium was approximated via UV/Vis spectroscopy. [19] Allowing for additional time after apparent reaction completion, the reactions were stopped after 1h by quenching samples of 100 mLi na ne qual volume of MeOH and analyzed by HPLC. All experimental and calculated data are available online. [16,18]