Quantitative Prediction of Rate Constants for Aqueous Racemization To Avoid Pointless Stereoselective Syntheses

Abstract Racemization has a large impact upon the biological properties of molecules but the chemical scope of compounds with known rate constants for racemization in aqueous conditions was hitherto limited. To address this remarkable blind spot, we have measured the kinetics for racemization of 28 compounds using circular dichroism and 1H NMR spectroscopy. We show that rate constants for racemization (measured by ourselves and others) correlate well with deprotonation energies from quantum mechanical (QM) and group contribution calculations. Such calculations thus provide predictions of the second‐order rate constants for general‐base‐catalyzed racemization that are usefully accurate. When applied to recent publications describing the stereoselective synthesis of compounds of purported biological value, the calculations reveal that racemization would be sufficiently fast to render these expensive syntheses pointless.

Abstract: Racemization has alarge impact upon the biological properties of molecules but the chemical scope of compounds with knownr ate constants for racemization in aqueous conditions was hitherto limited. To address this remarkable blind spot, we have measured the kinetics for racemization of 28 compounds using circular dichroism and 1 HNMR spectroscopy. We show that rate constants for racemization (measured by ourselves and others) correlate well with deprotonation energies from quantum mechanical (QM) and group contribution calculations.Such calculations thus provide predictions of the second-order rate constants for general-basecatalyzedr acemization that are usefully accurate.W hen applied to recent publications describing the stereoselective synthesis of compounds of purported biological value,t he calculations reveal that racemization would be sufficiently fast to render these expensive syntheses pointless.
Thalidomide racemizes in am atter of hours and yet it remains ap oster child for enantioselective synthesis which would not have saved its victims. [1] Thes tatus quo in enantioselective synthesis thus ignores the cruel blind spot that we address in this paper:racemization.
Racemization is ap articular problem because its detection requires chiral analytical methods. [13,14] Hence,f ew reports disclose rate constants for racemization under aqueous conditions. [1,[15][16][17][18][19][20] Chiral centers with certain combinations of substituents have been posited to be prone to general-basecatalyzed racemization although with little supporting data. [21][22][23] We therefore classified stereogenic carbon atoms according to their attached substituents.E ach substituent is identified as one of sixty types, [24] which encompass more than 99.95 %ofall such substituents in the GOSTAR database. [25] Thet en most frequently occurring substituents are listed in Figure 1; the Hr equired for general-base-catalyzed racemization is prominent. [24] Groups labelled *w ere selected for experimental study.

Angewandte Chemie
Communications first-order rate constants for these processes were corrected for hydrolysis side reactions if required. [24] Plotting the firstorder rate constants for racemization or H/D exchange against the concentration of the basic component of the buffer yielded the second-order rate constants for generalbase-catalyzed racemization. These were corrected for reaction temperature and substrate protonation state. [24] Forp redictive modeling,amechanistic understanding is beneficial. Racemization of the stereogenic centers studied here could occur by either an S E 1o ra nS E 2m echanism. For hydantoins (e.g. 4-6)b oth the S E 1a nd S E 2mechanisms have been proposed previously, [18,29] but we have shown that these reactions occur via the S E 1mechanism. [30] Further,Hammett plots for 1a-h show ap ositive slope and better correlation with s À than s suggesting that anegative charge is formed on the stereogenic center in the rate-determining step of the racemization reaction, in line with an S E 1mechanism. [24] Thee xperimental data were correlated with deprotonation energies (DDG(R 1 ,R 2 ,R 3 ), Scheme 1) from B3LYP/6-31 + G** calculations incorporating aqueous solvation using the PCM protocol. [24] Second-order rate constants for generalbase-catalyzed racemization, k gb ,c orrelate well with DDG-(R 1 ,R 2 ,R 3 )f or 1-3 and 4-8. [24] Thes et of compounds was supplemented with literature data for 9-16, [24] leading to the relationship with DDG-(R 1 ,R 2 ,R 3 )s hown at the top of Figure 2. Thel ine of best fit has equation log(k gb ) = À0.20 DDG(R 1 ,R 2 ,R 3 ) À14.28, with an R 2 value of 0.68 and root mean square error of 0.61, that is, reproducing rate constants to within afactor of approximately 4. Clopidogrel is excluded from this analysis due to large experimental uncertainties. [24] Thec omputational procedure was extended to include ag roup-contribution approach that is amenable to rapid analysis of chiral compounds and is described with examples in section S5. Asimple representative (R) of each substituent type was selected and DDG(R 1 ,R 2 ,R 3 )computed with R 1 = R and R 2 = R 3 = H. These DDG(R,H,H) values indicate how much Rs tabilizes an adjacent anion. Thes um DDG-(R 1 ,H,H) + DDG(R 2 ,H,H) + DDG(R 3 ,H,H), for the three non-H substituents around ac hiral carbon atom is referred to here as SDDG. When two or three of the groups provide stabilization through charge delocalization, ac ross-conjugation correction is applied to reflect the reduced ability of the second group to stabilize the anion caused by the presence of the first. [24] Scheme 1. The35compounds studied fall into fourteen chiral carbon atom types (A-N,T able 1) that have one hydrogen attached. [1,[15][16][17][18][19]31] Thes econd-order rate constants for generalbase-catalyzed racemization are plotted against SDDG at the bottom of Figure 2. When ac hiral center type is represented by more than one compound, the mean value of log(k gb )f or all representatives was used and the full range of values is shown as av ertical line.T his prevents any center type from dominating the linear fit.
Fort he phenylglycine esters (A), substituent effects can cause up to alog unit variation from the line of best fit. This is likely to be representative of general substituent effects. [24] Grouping five-membered aromatic rings together (C)m asks variation of 1.6 log units,l ikely reflecting the more direct influence of heteroatoms in aromatic rings.Arelatively diverse set of alkyl substituents in the 5-position on ahydantoin ring (group D)orthiohydantoin (F)causes little variation in rate constants for racemization. In general, variation caused by substitution or structural variation within classes is less than two orders of magnitude and typically less than one order of magnitude.
At the bottom of Figure 2, two subgroups are apparent: those involving ac yclic anion with the potential to be aromatic and those that do not. Forthe non-aromatic set (shown in red) aline of best fit with equation log(k gb ) = À0.11 SDDG À9.81 was found (R 2 = 0.78 and RMSE = 0.40 log units) and for the aromatic anion set (shown in blue) the line of best fit has equation log(k gb ) = À0.26 SDDG À16.95 (R 2 = 0.92 and RMSE = 0.39 log units). TheR MSE for all compounds computed individually is 0.64 for non-aromatic anions (excluding clopidogrel) and 0.37 for aromatic anions,that is,predictions are typically within 5-fold. Although not ap erfect guide,t he group contribution approach provides an easily applied, useful and rapid filtering that can even be used for very large databases.
Forthe particular example of chiral pharmaceuticals,our analysis can be applied to predict half-lives of racemization in physiological conditions.T he rate constants for racemization of thalidomide at different phosphate buffer concentrations compared to that in blood suggest that, in terms of availability of catalytically active general bases,b lood is approximately equivalent to a0.15 m phosphate buffer at pH 7.2. Therefore, with k gb predicted by the QM or group contribution method, the required half-lives can be predicted. [1] Acomprehensive workflow has now arisen:rapid analysis with agroup contribution based method can trigger quantum mechanical calculations,w hich in turn can trigger an experimental protocol (Figure 3). Compounds at high risk of racemization can be avoided and racemization risk can be suppressed by design.
To illustrate the degree to which racemization in aqueous conditions is an overlooked issue,w eh ave surveyed recent editions of leading chemistry journals,using our knowledge of the group contributions,toidentify several articles envisaging biological applications.T his was not an exhaustive search. Compounds described were subject to group contribution calculations.I ti sd isappointing to reveal ( Table 2) that liability to racemize under physiological conditions is more commonplace than would be possible if it were properly understood and controlled, as many chemists seem to believe.
In summary,w ed escribe an approach to quantitatively predict the racemization risk that is generally applicable and allows synthetic chemists to avoid racemization-prone targets or understand erosion of the enantiomeric excess (ee). The approach allows quantitative assessment of the risk of chiral compounds of turning into racemic mixtures when used as pharmaceuticals or for other purposes. [36] Acknowledgements TheUKResearch Council and AstraZeneca are thanked for aPhD studentship (AB) under project number EP/C537572/ 1(EPSRC CASE for New Academics Award), the Kurdistan Regional Government is thanked for aP hD studentship (HOA), Advanced Research Computing at Cardiff (ARCCA) are thanked for computational resources.T he