|Home | About | Journals | Submit | Contact Us | Français|
An experimental and computational investigation of amido-thiourea promoted imine hydrocyanation has revealed a new and unexpected mechanism of catalysis. Rather than direct activation of the imine by the thiourea, as had been proposed previously in related systems, the data are consistent with a mechanism involving catalyst-promoted proton transfer from hydrogen isocyanide to imine to generate diastereomeric iminium/cyanide ion pairs that are bound to catalyst through multiple non-covalent interactions; these ion pairs collapse to form the enantiomeric α-aminonitrile products. This mechanistic proposal is supported by the observation of a statistically significant correlation between experimental and calculated enantioselectivities induced by eight different catalysts (P 0.01). The computed models reveal a basis for enantioselectivity that involves multiple stabilizing and destabilizing interactions between substrate and catalyst, including thiourea-cyanide and amide-iminium interactions.
Catalysis by neutral, organic, small molecules capable of binding and activating substrates solely via non-covalent interactions—particularly H-bonding—has emerged as an important approach to enantioselective synthesis over the last decade.1 The mechanisms by which such small-molecule catalysts induce high enantioselectivity may be quite different from those employed by catalysts that rely on covalent interactions with substrates. The latter – typified by chiral Lewis acids,2 transition metal complexes,3 and secondary amine derivatives4 – are generally understood to induce enantioselectivity through steric destabilization of all but the dominant pathway. Repulsive steric effects are strongly distance-dependent,5 and expected to be most pronounced in covalently-bound intermediates and transition states. Attractive non-covalent interactions are weaker, less distance-dependent, less directional, and more affected by entropy than covalent interactions.6 This analysis defines the challenge for achieving high enantioselectivity through non-covalent interactions alone: the attractive non-covalent interaction (e.g., H-bonding) that binds the enantioselectivity-determining transition structure to the catalyst may be substantially less distance-dependent than the repulsive steric interactions needed to disfavor competing pathways. If that is the case, then the penalty of avoiding repulsive steric interactions by structural reorganization should be relatively small. However, the conformational constraint required for high stereoinduction may be achieved, in principle, if multiple non-covalent attractive interactions are operating in concert. Enzymes engage such cooperative effects as a general principle,7,8 and we are interested in exploring the extent to which the interplay of weak, non-covalent interactions is also important in highly selective reactions promoted by small molecule catalysts.9
The urea- and thiourea-catalyzed enantioselective hydrocyanation of imines are important examples of small-molecule catalyzed reactions thought to involve only non-covalent substrate-catalyst interactions (Scheme 1).10 In 2002, a preliminary mechanistic analysis of this reaction – based on NMR spectrocopic, kinetic, structure-activity, and theoretical studies – revealed that the urea functionality of 1b is responsible for catalytic activity and that the imine substrate engages in a ground-state interaction with the catalyst via a dual H-bonding interaction to the urea protons (Figure 1).11 This study helped to establish that simple H-bond donors could serve as useful chiral small-molecule catalysts, and laid the foundation for much of the subsequent work in this field.1 However, the mechanism of catalysis, and specifically the question of whether the H-bonding interaction between imine and (thio)urea persisted in the rate- or enantioselectivity-determining transition state, could not be elucidated on the basis of this study.
It was discovered subsequently that urea and thiourea derivatives engage a variety of weakly basic electrophiles, including N-acyl iminium/chloride12 and oxocarbenium/chloride ion pairs,13 in catalytic asymmetric reactions. On the basis of substituent and solvent effects, as well as the well-known anion-binding properties of ureas and thioureas,14 this work led to the proposal that catalysis occurs by abstraction of an anionic leaving group (e.g., Cl−) (Scheme 2).12b,15
The notion that multiple mechanistic channels are available to (thio)urea derivatives in catalysis of nucleophile-electrophile addition reactions—and specifically that H-bonding may occur either directly to the electrophile or to an anionic counterion or nucleophile in the enantioselectivity-determining transition structure—has emerged from theoretical studies of reactions promoted by tertiary amino-thiourea derivatives.16 For example, in the thiourea-catalyzed cyanosilylation of ketones, relatively small differences in activation energy were calculated between mechanisms involving thiourea-ketone and thiourea-cyanide interactions (Scheme 3).16c Analogous considerations may apply to imine hydrocyanation reactions, raising the possibility that the thiourea-imine interactions identified in the preliminary mechanistic work might lie off the productive catalytic cycle.
The mechanism of imine hydrocyanation has been subjected to rigorous scrutiny in the context of non-asymmetric methods, whereas current understanding of enantioselective catalytic variants is more qualitative. Hydrocyanation of imines is promoted by polar, protic solvents such as methanol, but proceeds slowly in non-polar or aprotic solvents, and is inhibited under basic conditions.17,18 Insight into the timing and precise mechanism of the σ-bond formation and protonation steps is available from classical studies of nucleophilic additions to carbonyl and imine electrophiles conducted on reactions run in aqueous or alcoholic media.19,20 Under these conditions, solvent molecules can assist in proton transfer steps that may be required for reaction.21 Indeed, experimental and theoretical studies of non-asymmetric imine hydrocyanation have led to the proposal that solvent molecules (e.g., water or methanol) participate in charge separation and promote proton transfer from HCN to the imine, resulting in the formation of a highly reactive iminium/cyanide ion pair that collapses to afford the α-aminonitrile product (e.g., Scheme 4A).22,23 In contrast, asymmetric catalysis of imine hydrocyanation by chiral Lewis or Brønsted acids is generally most effective in non-polar, aprotic solvents, such as toluene.24 Solvent-mediated proton transfer is not possible under these conditions, and most mechanistic proposals for asymmetric imine hydrocyanation invoke direct interactions between imine and Lewis or Brønsted acid (e.g., Scheme 4B),25 rather than formation of an iminium/cyanide ion pair. In these cases, protonation is proposed to occur after nucleophilic addition,25a,b or by the catalyst itself.25c,d,e The mechanisms proposed until now for non-asymmetric and asymmetric imine hydrocyanation are therefore fundamentally different,26 and bear close parallels to the two limiting mechanisms outlined in Scheme 3 for thiourea catalysis of ketone cyanosilylation.
In this paper we describe a detailed experimental and computational study of highly enantioselective imine hydrocyanation reactions catalyzed by recently discovered urea and thiourea derivatives. We have found that the mechanism of thiourea-catalyzed imine hydrocyanation does not involve productive thioureaimine interactions (Figure 1); instead, all data are consistent with a mechanism in which imine protonation by hydrogen cyanide precedes nucleophilic addition (e.g., as in the water-mediated reaction, Scheme 4A). This study has identified a role for each functional group on the chiral catalyst, and has revealed a basis for catalysis and enantioinduction that involves multiple stabilizing electrostatic interactions between catalyst and the bound iminium and cyanide ions.
Our efforts to carry out detailed investigations of imine hydrocyanation reactions catalyzed by thiourea Schiff base catalysts such as 1a–c were limited severely by the structural complexity of the catalysts. However, we discovered recently that N-benzhydryl imines (2) undergo highly enantioselective imine hydrocyanation with the much simpler amido-thiourea catalyst 4a (Table 1).27 This reaction displays broad substrate scope, with aldimines bearing either sp3- or sp2-hybridized substituents undergoing hydrocyanation in high yield and enantioselectivity (entries 1–7). Urea-analogue 4b promotes hydrocyanation to afford α-aminonitriles in similar enantiomeric excess and yield as thiourea 4a (entries 8–10). This observation indicates that these catalysts operate by similar mechanisms as H-bond donors, and that the nucleophilicity at sulfur of the thiourea catalyst 4a does not play a direct role in the catalytic mechanism.
In order to establish the stoichiometry of the rate-limiting transition structure, we carried out a detailed kinetic analysis of the imine hydrocyanation reaction catalyzed by 4a. Rate studies were executed using representative substrate 2a at synthetically relevant concentrations using in situ infrared spectroscopy. The kinetic experiments were executed under homogeneous conditions, in which HCN was generated in situ by addition of controlled amounts of MeOH to solutions of TMSCN in toluene.28 Data obtained from eighteen independent experiments over a range of concentrations of HCN (0.06–0.35 M), 2a (0.01–0.15 M), and catalyst 4a (0.00080–0.0064 M) were used to derive an empirical rate law for the reaction. Kinetic modeling of the experimental data reveals good agreement with the rate law given by eq 1 (Figures 2 and and33):29,30,31
The values of the exponents a and b are both approximately 1, and thus analysis of the data reveals an approximately first-order dependence of rate on both [HCN] and [imine]. In contrast, the value of c is somewhat less than 1, and the deviation from unity in the rate dependence on [cat]tot appears to be statistically significant.
We considered different possible explanations for the deviation from a strict first-order dependence of rate on [cat]tot. (a) The deviation from first-order kinetics might be explained by irreversible catalyst deactivation. “Same excess” experiments32 were carried out in order to establish whether consistent kinetic behavior (i.e., the same rate law) was maintained throughout the course of the entire reaction. The perfect overlap of kinetic runs executed at different ratios but same excess of [HCN] relative to [imine] (compare red and blue points in Figure 2) reveals that no significant catalyst deactivation occurs, and that product inhibition is insignificant.33 (b) In principle, the deviation from first-order kinetics might also be explained by competitive uncatalyzed imine hydrocyanation under the reaction conditions. However, the initial rate of the uncatalyzed reaction is approximately 5% of the initial rate of the reaction using the lowest catalyst concentration (0.00080 M). In addition, the enantiomeric ratio is independent of [cat]tot under the conditions used in the kinetic analysis (13:1–14:1 er). Both observations indicate that uncatalyzed background reaction is negligible under the conditions of the kinetic analysis. (c) Having ruled out other possible explanations for this effect, we ascribe the small deviation from a first-order rate dependence on [cat]tot reversible formation of an inactive catalyst dimer at high [cat]tot.34,35 However, it is important to note that this effect is small and that the precise origins of this effect do not appear to have any bearing on the mechanism of catalysis or enantioinduction, as the enantiomeric excess of the product is the same at different catalyst concentration (vide supra).
The kinetic data do not distinguish between the two fundamentally different mechanisms in Scheme 5: (a) protonation of the imine to form a transient iminium ion, followed by nucleophilic addition of cyanide anion to form the α-aminonitrile product (Scheme 5A);36,37 or (b) addition of cyanide to an imine to form an α-amidonitrile anion, followed by protonation to form the α-aminonitrile product (Scheme 5B). In order to evaluate the relative timing of the protonation and cyanide addition steps in the thiourea-catalyzed imine hydrocyanation reaction, we sought to assess the sense and degree of charge development at the imine-derived nitrogen in the rate-determining transition state. The first-order dependence on both [imine] and [HCN] observed experimentally indicates that the imine reactant is not associated with HCN or catalyst in the resting state. Thus a mechanism that involves rate-limiting formation of an iminium ion intermediate (e.g., step 1 in Scheme 5A) will be characterized by an increase in positive charge at nitrogen and faster rates are expected with substrates that better stabilize that positive charge. In contrast, a mechanism that involves an α-amidonitrile intermediate is expected to be characterized by faster rates with substrates that better stabilize negative charge. To distinguish between these possibilities, we measured the rates of hydrocyanation of substituted benzaldehyde-derived imines promoted by thiourea catalyst 4a and urea analogue 4b (Figure 4).38 Hammett plots reveal negative ρ-values that are similar in magnitude (ρthiourea = −2.7 ± 0.2; ρurea = −2.5 ± 0.2),39 consistent with substantial positive charge buildup in the transition state compared with the imine reactant.40,41,42 These data provide compelling evidence against the involvement of an α-amidonitrile anion (Scheme 5B) in the (thio)urea-catalyzed imine hydrocyanation, and point instead to an iminium ion mechanism (Scheme 5A).43,44
Schreiner’s achiral thiourea (5)45 is also an efficient catalyst for hydrocyanation of imine 2a. Kinetic analysis under conditions similar to those used with chiral catalyst 4a affords an adequate fit to eq 2:46
A Hammett study of the hydrocyanation catalyzed by 5 reveals that electron-rich substrates react more rapidly than electron-poor ones (Figure 5), as is the case for reactions catalyzed by chiral catalysts 4a and 4b. The derived ρ value of −3.3 ± 0.1 is slightly larger than in the enantioselective reactions, indicating that imine hydrocyanation catalyzed by 5 involves a somewhat greater degree of positive charge buildup than imine hydrocyanation catalyzed by 4a or 4b. Nonetheless, because the kinetic behavior and ρ values in imine hydrocyanation reactions catalyzed by 4a and 5 are similar, it appears reasonable to conclude that these catalysts promote hydrocyanation by analogous mechanisms.47
The rate of imine hydrocyanation is only slightly lower with Schreiner’s catalyst (5) than with the chiral amidothiourea 4a (compare Figures 4 and and5,5, or data in the Supporting Information), indicating that an H-bond donor alone is sufficient for efficient catalysis of this reaction.48 In order to glean insight into the mechanism of stereoinduction in the asymmetric reaction, the impact of catalyst structural properties on the rate and enantioselectivity of hydrocyanation of imine 2a was assessed by a combination of kinetic analysis using in situ IR spectroscopy and ee determinations using chiral HPLC analysis.49 Analysis of the data indicates that replacing the 3,5-bis-trifluoromethylanilino group with less electron-deficient anilines leads to a substantial decrease in activity without substantial change in enantioselectivity (Table 2.2, entries 1–3). Urea-derived catalyst 4b is slightly less active and enantioselective than 4a (entries 1 and 4). In contrast, altering the properties of the amino acid-derived portion of the catalyst leads to substantial decreases in both activity and enantioselectivity (entries 1, 5–10). The (R)-enantiomer is the major product in reactions promoted by (S)-amino acid-derived catalysts in all but one case (entry 8).50 The observation that catalyst 4a is both the most enantioselective and the most reactive catalyst identified in this series raises the possibility that the stereochemical outcome of the reaction is a result of not only selective destabilization of the transition state leading to the minor (S)-enantiomer, but also of selective stabilization of the transition state leading to the major (R)-enantiomer.
The kinetic and mechanistic analyses described above indicate that thiourea-catalyzed imine hydrocyanation is accompanied by positive-charge buildup on the imine, and likely involves the intermediacy of an iminium ion. The structure-activity-enantioselectivity relationship data provide qualitative information about the critical features of the catalyst, revealing for example that the structures of both the amide and the amino-acid derived portion of the catalyst are critical for both enantioselectivity and activity. In principle, further information about substrate-catalyst interactions might be obtained from spectroscopic or crystallographic analyses of substrate-catalyst complexes. However, such complexes are expected to be weakly bound, and the rates of formation and dissociation of such complexes will almost certainly be much greater than the rate of the catalytic reaction.51 Under these conditions, spectroscopic analyses are unlikely to provide insight into interactions that occur in transition states.52,53
In contrast, computational chemistry is well-suited to the characterization of complex transition states in organic chemical reactions,54 including in the context of asymmetric organocatalysis.55 We thus turned to computational methods to obtain deeper insight into the reaction mechanism and to identify catalyst-substrate interactions responsible for catalysis and asymmetric induction.56 Catalyst 8a and model substrate 9a have limited conformational flexibility, were therefore selected as starting points for comparing energies of a variety of transition structures. Initial calculations were executed at the B3LYP/6-31G(d) level, and all energies refer to uncorrected electronic energies at this level of theory, unless noted otherwise. To test the validity of using this computational method, key transition structures were recalculated at the MP2/6-31G(d) and M05-2X/6-31+G(d,p) levels.57
To establish a reference point for analysis of transition structures, ternary complexes of imine, HCN, and catalyst were identified. Complexes 10a–c involving either thiourea-cyanide or thiourea-imine interactions are each stabilized by three hydrogen bonds and were found to be nearly isoenergetic.
One possible mechanism for imine hydrocyanation involves direct imine activation by catalyst via an imine–thiourea complex.11 Within such a mechanism, a possible role of the amide might be to activate HCN or HNC to generate a more active nucleophile.25b Transition state calculations indicate that a mechanism involving a combination of amide–HNC and thiourea-imine interactions has a high activation barrier but is viable (Scheme 6). Because this mechanistic proposal suggests a direct role for the amide, it might also help explain the strong effect of amide structure on reaction rate and enantioselectivity. The computational analysis reveals that proton transfer from one of the N-protons of the thiourea occurs to the imine nitrogen simultaneously with cyanide addition.58 This proton transfer would be expected to lead to positive charge development on the imine in the rate-limiting transition structure and therefore appears to be consistent with the results of the Hammett analysis (Figure 4).
However, both rate data comparing the urea and thiourea catalysts and isotopic labeling data allow us to rule out the mechanism in Scheme 6. Thioureas are more acidic than ureas by approximately 6 pKa units, and this difference in acidity holds across a broad range of substitution patterns.59 If the thiourea or urea catalyst undergoes deprotonation prior to or during the rate-limiting step, then thiourea 4a is expected to be substantially more active than urea 4b. Indeed, the computational analysis of the thiourea-bound-to-imine mechanism (Scheme 6) predicts that the activation energy for the urea-catalyzed reaction is 5.3 kcal/mol greater than for the thiourea-catalyzed reaction.60 However, this effect is not borne out experimentally: 4a catalyzes imine hydrocyanation of a range of substrates only slightly more rapidly than 4b (2–3-fold, see Figure 4 and Table 2).
In the thiourea-bound-to-imine mechanism (Scheme 6), a thiourea proton is incorporated into the α-aminonitrile product.25c,d,e Identification the origin of the N-proton in the α-aminonitrile product would therefore provide a straightforward test of the plausibility of this mechanistic proposal. Amido-urea derivative 7b was identified as an efficient and enantioselective imine hydrocyanation catalyst in which the urea N–H resonances could be monitored by 1H-NMR spectroscopy under the catalytic reaction conditions. When deuterium cyanide rather than hydrogen cyanide is used as a cyanide source, exchange of the N-protons of 7b occurs slowly, both in the absence and presence of imine. Hydrocyanation of imine 2c proceeds more rapidly than hydrogen/deuterium exchange, and nearly quantitative deuterium incorporation into the α-aminonitrile product is observed by 1H-NMR spectroscopy well before one catalytic turnover is reached (eq 3 and Figure 6). Under these conditions, only small amounts of exchange between the urea N-protons and DCN are observed (compare the integration of the N-protons in Figures 6A, 6B, and 6C). 61 These observations lead to the conclusion that the source of the N-proton on the α-aminonitrile is HCN rather than catalyst, and this is inconsistent with the mechanism depicted in Scheme 6. It is thus necessary to consider alternative mechanistic possibilities for imine hydrocyanation.62
As discussed above, the proposed mechanism for water-mediated imine hydrocyanation involves proton transfer from HCN to imine to form an iminium/cyanide ion pair, followed by collapse to form the α-aminonitrile product (Figure 2A).22 We considered the possibility that chiral amido-thiourea catalysts might promote hydrocyanation via an analogous mechanism. The computational analysis depicted in Scheme 7 indicates that HCN-mediated imine hydrocyanation can proceed by thiourea-assisted proton transfer from HCN to imine to form a transient catalyst-bound ion pair that collapses to form the α-aminonitrile product (A → C). The iminium ion in the transition structure (B) is stabilized by a tight H-bond to the amide moiety of the catalyst, whereas the cyanide anion is stabilized by a dual H-bond to the thiourea moiety of the catalyst. The computational analysis with this simplified catalyst suggests that this process is formally concerted but highly asynchronous: the C–H bond between cyanide and iminium ion is partially broken in the transition structure, but almost no C–C bond formation has occurred at this point on the reaction coordinate. The reaction coordinate with this simplified model involves rate-limiting proton transfer followed by barrierless cyanide addition.
An analogous mechanism that uses hydrogen isocyanide (HNC) as the active nucleophile was also identified by computational methods (Scheme 8).63 This mechanism for imine hydrocyanation involves two discrete steps. The first step (A → C) involves proton transfer from HNC to imine to generate a catalyst-bound iminium/cyanide ion pair (C). In order for the cyanide anion to be able to approach the iminium ion in a Bürgi-Dunitz trajectory,64 further charge and spatial separation is necessary. The rate-limiting step in the overall transformation involves structural reorganization of this ion pair to form a transition structure with a long forming C–C bond that collapses to the α-aminonitrile (C → E). The overall activation barrier for HNC addition is slightly lower than for HCN addition. However the key stabilizing substrate-catalyst interactions in the transition structures (Schemes 7B and and8D)8D) are nearly identical: both transition structures involve a network of medium-length H-bonds between cyanide and thiourea (1.9–2.2 Å), between iminium and cyanide (2.1–2.2 Å), and between amide and iminium ion (2.2 Å).65 Because of the overall similarities between these mechanistic hypotheses, and the limited relevance of the simplified model system used in this study, we have not made further efforts to distinguish between these pathways.66
The mechanistic hypothesis that either HCN or HNC protonates the imine to generate a catalyst-bound ion pair (Schemes 7 and and8)8) is consistent with the experimental data: (a) each of the proposed mechanisms involves positive-charge buildup on the imine, and is thus consistent with the Hammett analysis; (b) each proposed mechanism involves imine protonation by the nucleophile rather than the catalyst, and is thus consistent with the experimental observation that α-aminonitrile N-proton originates from the nucleophile rather than from the catalyst. In addition, the overall calculated activation barrier for hydrocyanation via an iminium/cyanide ion pair is 20 kcal/mol lower than hydrocyanation by direct addition to a catalyst-bound imine.67
The HCN- and HNC-addition transition structures leading to the (R)- and (S)-enantiomers of α-aminonitrile are depicted in Figures 7 and and8.8. Each of the two sets of structures differs with respect to the orientation of the substrate tert-butyl group relative to the catalyst, but both possess the same cyanide-thiourea and iminium-amide interactions. The HCN- and HNC-addition transition structures leading to each enantiomer of product are remarkably similar, and differ only in the orientation of the cyanide anion relative to the catalyst and iminium ion. The similarities in the transition structure geometries thus provide an explanation for why transition structures formally originating from tautomeric nucleophiles have such similar energies. Nevertheless, the HNC-addition is more favorable than HCN-addition transition structures in every case examined, and our analysis is therefore focused on HNC-addition mode for the remainder of this paper.68
Calculations of the HNC-addition transition structures—including full geometry optimizations—at the M05-2X/6-31+G(d,p) and MP2/6-31G(d) levels yield nearly identical transition structures and relative energies (Supporting Information). This mechanistic proposal is thus consistent with the experimental data, and the geometries of the transition structures are consistent across different theoretical methods; we thus conclude that an imine hydrocyanation mechanism involving a catalyst-bound iminium/cyanide ion pair is viable. However the simplified model system examined in this study does not serve to address the high enantioselectivity observed experimentally in the catalytic reaction.69 In order to account for the experimental catalyst structure-activity data (Table 2) and attempt to develop an understanding of the basis of stereoinduction, it is necessary to evaluate catalyst and substrate structures that are more relevant to the catalytic, enantioselective reaction.
A broad range of enantioselectivities are observed in imine hydrocyanation with different amido-(thio)urea catalysts (Table 2), and we sought to determine whether computational methods can be used to account for the observed trends in a quantitative manner.70,71 For such an analysis to be meaningful, the computational models must incorporate all the catalyst and substrate structural features necessary for high enantioselectivity. As discussed above, both the identity of the amide and the amino acid components play crucial roles in defining the enantioselectivity of the catalyst. In contrast, the precise structure of the anilino group has little effect on enantioselectivity (Table 2, entries 1–3), so we selected the simple phenyl-substituted catalyst 6a, together with an imine that undergoes hydrocyanation with high ee, N-benzhydryl imine 2a, for our model. In the following discussion, energies of all calculated intermediates and transition structures are provided relative to the stable complex between catalyst, substrate, and HCN shown below.
Scheme 9 provides a representation of the calculated intermediates and transition states in the lowest energy pathway for the addition of HNC to imine 2a catalyzed by thiourea 6a. The first two steps of this mechanism (A → C → E) involve formation of a catalyst-bound iminium-cyanide ion pair. This ion pair then undergoes structural rearrangement to form new ion pair in which the iminium ion is associated to the amide component of the catalyst (E → G). In this model system, there is a small barrier calculated for the collapse of this ion pair to form the (R)-α-aminonitrile product (G → I). As with the simpler model system described above, the highest energy transition state, and therefore the rate-limiting step, is associated with the ion pair reorganization step (compare Scheme 8, C → E with Scheme 9, E → G). In short, the basic calculated HNC-mediated imine hydrocyanation mechanisms are predicted to correspond closely in the small and large model systems; the mechanisms differ only in the number of intermediates and transition structures that are observed prior to and after the rate-limiting step.
An analogous computational analysis of the formation of the (S)-α-aminonitrile yields a similar reaction coordinate (Supporting Information), in which a transition structure akin to F is predicted to represent the rate-limiting step (E = +19.4 kcal/mol). The calculated relative energies of the ion pair rearrangement transition structures (ΔΔE‡ = 2.4 kcal/mol) thus correspond well with the experimentally observed preference for formation of the (R)-α-aminonitrile. This analysis assumes that the calculated rate-limiting step, E → G, is enantioselectivity-determining. It is important to note that the calculations also predict the same sense and similar magnitude of enantioselectivity if the cyanide-addition step, G → I, is enantioselectivity-determining. Thus comparison of the energies of transition structure H (Scheme 9) and the corresponding structure leading to the (S)-α-aminonitrile (Supporting Information) also reveals a significant preference for formation of (R)-α-aminonitrile (ΔΔE‡ = 2.1 kcal/mol). The computational analysis therefore provides support for either iminium-ion rearrangement or cyanide addition as enantioselectivity-determining steps in the catalytic reaction, although it predicts a higher barrier to the former.
We explored alternative approaches to establish the identity of the enantioselectivity-determining step with greater certainty. In principle, the question of which of transition structures F and H in Scheme 9 is rate-determining might be addressed by examining secondary isotope effects in the hydrocyanation reaction. A key difference between these transition structures is that H undergoes partial sp2 → sp3 rehybridization of the electrophilic carbon atom, whereas F does not. Secondary α-hydrogen/deuterium kinetic isotope effects (KIEs, kH/kD) are sensitive probes of changes in carbon hybridization state,72 and inverse α-KIEs are observed in reactions thought to involve significant sp2 → sp3 rehybridization (typically, kH/kD ≈ 0.75 – 0.90).73
However, in reactions of carbonyl derivatives that involve reversible protonation or Lewis acid complexation of the carbonyl prior to nucleophilic addition, it is also necessary to consider secondary β-hydrogen/deuterium equilibrium isotope effects (EIEs) on lone pair basicity.74 Accurate and systematic measurements of β-EIEs on nitrogen basicity have been reported only recently, and these studies have revealed that β-deuterium substituted amines are more basic than their β-protio analogues in aqueous solution.75 By extrapolation of the experimental data, a maximum β-EIE on basicity of ≈ 46 cal/mol is predicted when the C–H or C–D bond is anti-periplanar to the basic nitrogen (KH/KD = 0.93 at 298 K).76 This effect is reproduced using harmonic frequency calculations at the MP2/6-31G(d,p) level (KH/KD = 0.91).77 C–H stretching frequencies involving sp2-hybridized carbons are affected by the presence of an anti-periplanar lone pair;74,78 because secondary isotope effects can arise any time a vibrational mode is perturbed in the course of a reaction, an EIE might be expected in imine protonation as well.
To test this notion, we calculated the EIE for protonation of a model imine at the MP2/6-31G(d,p) level using harmonic frequency calculations (eq 4).79,80 This analysis predicts a significant inverse EIE on imine basicity comparable in magnitude to the inverse EIE observed experimentally on tertiary amine basicity. Thus the direction of the secondary β-EIE on imine basicity is expected to be the same as the direction of the secondary α-KIE in a reaction involving sp2 → sp3 rehybridization in the transition state. In the asymmetric imine hydrocyanation, an inverse KIE of kH/kD = 0.88 ± 0.03 is observed (eq 5).81 An isotope effect of this magnitude allows us to rule out a late transition state involving a largely sp3-hybridized electrophilic carbon,73 but it is consistent with a transition state that involves either early C–C bond formation or iminium ion rearrangement.82 Therefore, the experimental secondary isotope effect data are consistent with either F or H (Scheme 9) as the rate- and enantioselectivity-determining transition structures. Given that the calculations predict a very low barrier for the cyanide addition step (G → I), we proceeded in our analysis with the assumption that the iminium ion rearrangement (E → G) is enantioselectivity-determining.
The experimental enantioselectivity data listed in Table 2 reveal a strong impact of relatively small changes in catalyst structure on enantioselectivity in the hydrocyanation of 2a.83 Diastereomeric transition structures akin to F depicted in Scheme 9 for addition of HNC to imine 2a were calculated at the B3LYP/6-31G(d) level for a series of catalysts. Experimental data listed in Table 2 from reactions run with catalysts 4a–4h were compared with calculations executed using the N-phenyl catalysts 6a–6h (Figure 9). A plot of the relative energies of the (S)- and (R)-transition structures versus the energy difference between (S)- and (R)-transition states calculated from experimental ee measurements yields a strong correlation and a statistically significant positive slope (Figure 10).84 Single-point energy calculations at the M05-2X/6-31+G(d,p) and MP2/6-31G(d) levels result in similar plots (Figures 11 and and1212).85,86 Computations at each level of theory reproduce the experimentally observed preference for the (R)-enantiomer of product, the higher observed enantioselectivity with 4d compared with 4e, the reduced enantioselectivity with secondary amide catalyst 4g compared with tertiary amide catalyst 4a, and the preference for the (S)-enantiomer of product with catalyst 4f. 87
The computational data are strongly consistent with the experimental observation that the enantioselectivity of imine hydrocyanation is strongly dependent on the structure of the catalyst. However, comparison of imine hydrocyanation transition structures using catalysts that induce high enantioselectivity in hydrocyanation (e.g., catalyst 6a, Figure 13) with those that induce low enantioselectivity (e.g., catalyst 6f, Figure 14) reveals few obvious differences. In particular, there are no apparent “steric clashes” that might explain why some transition structures are significantly higher in energy than the one leading to (R)-3a with catalyst 6a. We therefore undertook a quantitative analysis of structural differences between transition structures using different catalysts. As described below, this analysis has led to the conclusion that enantioselectivity can be attributed primarily to differential stabilization of the iminium cation.
Because imine hydrocyanation is catalyzed by simple H-bond donors (i.e., Schreiner’s catalyst, 5), we considered whether the basis for enantioselectivity might be tied to the dual H-bond interaction between cyanide and (thio)urea. The strength of H-bonds is distance-dependent;6 thus, if the cyanide–thiourea H-bond strength changes substantially among the four transition structures, the H-bond lengths in the four transition structures would be expected to be substantially different (d1 and d2 in Figure 15). However, the length of the dual H-bond between thiourea and cyanide does not differ substantially amongst the (R)- and (S)-transition structures with either catalyst: the eight cyanide nitrogen–thiourea proton distances among the four transition structures vary from 1.97 to 2.11 Å (Figures 13 and and14).14). The small differences in H-bond length among the different structures can be attributed primarily to changes in binding symmetry rather than differences in overall binding strength: the average of the two H-bond lengths of the four structures in Figures 13 and and1414 ranges from 2.01–2.05 Å. We reach the same conclusion from a complete analysis of all sixteen HNC-addition transition structures calculated with imine 2a and catalysts 6a–6g (Supporting Information). Plots of the sum of the thiourea–cyanide bond lengths (d1 + d2) versus calculated enantioselectivity reveal that there is no statistically significant trend in H-bond length within either the (R)- or (S)-transition structures between highly and poorly enantioselective catalysts (Figure 16). Thus, there is no correlation between thiourea–cyanide H-bond length and enantioselectivity, and other explanations must be evaluated for the basis for enantioselectivity.
The calculated rate-limiting step (E → G, Scheme 9) is rearrangement of an ion pair: in intermediate E, the iminium ion forms an H-bond to cyanide anion; in intermediate G, the iminium ion forms an H-bond to the oxygen atom of the catalyst carbonyl group. The reaction coordinate thus formally involves transfer of the iminium ion from cyanide to carbonyl, and the transition structure represents a point on the reaction coordinate where the H-bonds from iminium ion to both cyanide (d3) and amide (d4) are each partially formed (Figure 17).88 A straightforward way to estimate the overall H-bond strength to the iminium ion as the rearrangement proceeds is to measure the sum of the O–H and N–H bond lengths in the HNC-addition transition structures (d3 + d4).89 Plots of the sum of the H-bond lengths versus calculated enantioselectivity reveal that d3 + d4 varies only slightly across the transition structures leading to the (R)-enantiomer (Figure 18). In contrast, there is a positive correlation between d3 + d4 and enantioselectivity within the (S)-transition structures. Comparison of the most and least selective catalysts is striking: whereas d3 + d4 is significantly different between the (R)- and (S) addition transition structures with optimal catalyst 6a (4.58 versus 5.00 Å, Figure 13), this difference is negligible with catalyst 6f (4.51 Å for both (R)- and (S)-transition structures, Figure 14). This analysis leads to a simple conclusion: the basis for enantioselectivity may be traced to different degrees of iminium ion stabilization amongst different catalysts and diastereomeric transition structures, and stabilization of the iminium ion is achieved by H-bonding interactions between the iminium ion N–H and both the amide C=O and the thiourea-bound cyanide ion.
This study provides experimental and theoretical support for a mechanism for amido-thiourea catalyzed imine hydrocyanation involving formation of an iminium/cyanide ion pair that is bound to catalyst through multiple non-covalent interactions. This mechanism is analogous to the mechanism proposed for non-asymmetric imine hydrocyanation in polar, protic solvents, in which hydrogen cyanide activates the imine toward nucleophilic attack. In contrast, the data are inconsistent with mechanisms that involve direct imine activation by thiourea. The proposal that a H-bond donor catalyst promotes nucleophilic addition to a basic substrate without a direct interaction between the catalyst’s acidic and substrate’s basic functional groups may be relevant to catalytic asymmetric nucleophilic additions of other protic nucleophiles.90 While stabilization of the cyanide anion by the thiourea is important for catalysis, differences in enantioselectivity cannot be traced to the degree of anion-stabilization. Instead, we conclude that the degree of stabilization of the iminium ion is principally responsible for controlling enantioselectivity in asymmetric imine hydrocyanation.
Small-molecule organocatalysts promote asymmetric transformations via a variety of fundamentally different activation mechanisms—including covalent catalysis,4 general acid-base catalysis,16 and non-covalent catalysis91,92,93—and these activation modes closely parallel those of enzymatic reactions.94 However, small-molecule organocatalysts appear to differ from their macromolecular counterparts in their propensity to bind substrates. The measurement of substrate binding constants is a hallmark of enzymatic kinetic analyses,95 and it is a guiding principle in structural biology that the binding orientation of the substrate in an enzyme active site at least partially relevant to the transition structure.96 In contrast, kinetically observable substrate binding does not occur in the reaction studied herein, and computational studies suggest that multiple, fundamentally different binding modes are in principle isoenergetic (i.e., complexes 10a–10c). Imine–urea binding was observed both kinetically and spectroscopically in the context of imine hydrocyanation catalyzed by 1b (Figure 1),11 but the results described in this paper suggest that this mode of imine binding lies off the productive catalytic cycle.97,98,99
This difference in the relevance of the ground state catalyst-substrate structure to the mechanism of catalysis may be ascribed simply to the structural simplicity of these small-molecule catalysts compared with enzymes. There is no well-defined and structurally rigid “binding pocket” in catalysts such as 4a; instead, the functional groups necessary for catalysis are exposed to solvent and accessible to a broad range of small molecules, whether or not they are substrates for catalysis. Indeed, the ability of these catalysts to accept many different small-molecules is a prerequisite both for broad substrate scope within any single reaction and for the ability of a single catalyst to promote multiple different reactions. A corollary of this analysis is that enantioselectivity is under Curtin-Hammett control and elucidation of its basis therefore requires detailed understanding of transition structure geometries. The present study demonstrates that weak interactions operating in transition states of polar asymmetric reactions are amenable to quantitative characterization using modern but increasingly inexpensive computational methods.100 We suggest that systematic application of hybrid experimental-computational approaches such as those described in this paper will prove valuable in the characterization of other enantioselective catalytic processes that rely on non-covalent interactions.
This work was supported by the NIH (GM-43214) and by fellowship support to SJZ from the American Chemical Society and Roche. We thank Dr. Lars P. C. Nielsen for helpful discussions.
Supporting Information Available: Complete experimental procedures, characterization data, 1H NMR spectra used in isotopic labeling and isotope effect experiments, kinetic data in tabular format, geometries and energies of calculated stationary points, computed harmonic frequencies used in isotope effect calculations, and complete ref 56. This material is available free of charge via the Internet at http://pubs.acs.org.