DFT-Based Stereochemical Rationales for the Bifunctional Brønsted Acid/Base-Catalyzed Diastereodivergent and Enantioselective aza-Henry Reactions of α-Nitro Esters

Abstract

A pair of chiral bis(amidine) [BAM] proton complexes provide reagent (catalyst)-controlled, highly diastereo- and enantioselective direct aza-Henry reactions leading to α-alkyl-substituted α,β-diamino esters. A C₂-symmetric ligand provides high anti-selectivity, while a non-symmetric congener exhibits syn-selectivity in this example of diastereodivergent, enantioselective catalysis. A detailed computational analysis is reported for the first time, one that supports distinct models for selectivity resulting from the more hindered binding cavity of the C₁-symmetric ligand. Binding in this congested pocket accommodates four hydrogen bond contacts among ligand and substrates, ultimately favoring a pre-syn arrangement highlighted by pyridinium-azomethine activation, and quinolinium-nitronate activation. The complementary transition states reveal a wide range of alternatives. Comparing the C₁ and C₂-symmetric catalysts highlights distinct electrophile binding orientations despite their common hydrogen bond donor-acceptor features. Among the factors driving unusual high syn-diastereoselection are favorable dispersion forces that leverage the anthracenyl substituent of the C₁-symmetric ligand.

Graphical Abstract

Introduction

α-Substitution of amino acids (Figure 1) can modulate their biological activity, including metabolic stability, when compared to their unsubstituted analogues.^1,2,3 α-Alkyl substituents also induce conformational changes when included as monomers in peptides.^1,4 Synthetically, these compounds are interesting due to the difficulty in synthesizing enantioenriched quaternary amine carbons. α,β-Diamino acids are attractive targets for similar reasons,⁵ and the merging of these motifs (Figure 1) can impart the benefits of each to the resulting α-substituted α,β-diamino acids and their derivatives. The vic-diamine can be further constrained to a heterocyclic ring, and as such, is a useful platform for drug discovery where the goal is often to improve small-molecule potency while enhancing resistance toward chemical and enzymatic degradation.^6,7 In addition to their biological impact, α-substituted α,β-diamino acids are interesting synthetic targets due to the difficulty in generating a densely functionalized carbon core containing vicinal stereocenters, with high enantio- and diastereoselectivity.⁸

Figure 1.

Key Features of α-Amino Esters Accessible from a Diastereodivergent aza-Henry Addition Product

While there have been considerable studies on the synthesis of α-amino acids, there are still few intermolecular catalytic asymmetric C-C bond forming methods that are successful in forging the α-quaternary vic-diamine and diamino acid motif with both high diastereoselection and enantioselection, while providing useful isolated yields.^5,9,10 Approaches that address all three of these criteria enhance accessibility, and those that use similar catalysts to achieve reagent control over diastereoselection impart an ease of synthesis. The use of a single catalyst design to provide selective access to both anti- and syn-diastereomers while simultaneously controlling enantioselection—termed diastereodivergent^11,12—is uncommon, but a growing number of transformations have been addressed by combinations of distinct catalysts.^13,14,15 More common are solutions involving the use of chiral nonracemic reagents to activate nucleophile and electrophile independently (vide infra), while single-reagent systems to achieve similar selectivity control are quite rare.^16,17,18

As an expedient solution, we have broadly investigated the organocatalyzed aza-Henry (or nitro-Mannich) reaction using proton complexes of chiral bis(amidine) [BAM] ligands (Scheme 2) with the goal to uncover the design principles necessary to achieve high reactivity and selectivity.^{19,20,21,22,23,24,25,26,27,28,29,30,31} Reactivity using BAM-proton catalysis is a function of nitroalkane acidity and catalyst basicity, with the latter providing external rate control;^25,26 greater overall reaction rate correlates with more Brønsted basic catalyst such as 2 or 3 compared to 7, or with an acidic nitroalkane such as α-nitro ester²¹ or bromonitromethane.²⁷ This trend was observed in the syn-selective aza-Henry summarized in Scheme 1, and in combination with the impact of catalyst symmetry on selectivity, this case provided a launch point for a more detailed consideration of stereochemical rationales.²³ Too much contrast in ligand basicity, however, can lead to catalyst poisoning as it is inhibited by extensive hydrogen bonding by the nitroalkane.²⁵ Most enantioselective aza-Henry reactions are anti-selective, with rare examples of syn-selective additions.³² Shibasaki has developed complementary anti-³³ and syn-selective³⁴ additions of nitroethane to N-Boc imines using Cu(II)/Sm(III) and Yb(III)-based catalysts, respectively, albeit with different ligand designs.³⁵

Scheme 2.

Diastereodivergency as a Function of Catalyst Symmetry: Diastereo- and Enantioselective Additions of α-Substituted α-Nitro Esters to Imines^23,36

Scheme 1.

Correlation of Design Principles and Outcomes for BAM-Catalyzed aza-Henry Reactions: Independent Manipulation of Reactivity and Selectivity with an Unsymmetrical Catalyst

The aza-Henry reaction³² between α-substituted nitroacetates and imine electrophiles (Figure 1) is the only case for which highly diastereoselective reactions have been realized in favor of both diastereomers using the same catalyst design (Scheme 2).^23,36 The individual diastereo- and enantioselective synthesis of anti-adducts 1 has several solutions. Jorgensen developed a copper(II) bis(oxazoline) catalyst effective with PMP-imines (anti-1, P=^p(MeO)C₆H₄).³⁷ High diastereoselection is achieved when using a co-catalytic amount of quinine, since the use of triethylamine instead provided 1:1 dr. Shibasaki reported an elegant approach involving a bimetallic (Ni-Ni) catalyst and N-Boc imines to afford the anti-adducts (anti-1, P=Boc) with high selectivity.^38,39 Chen introduced a thiourea/disubstituted amine catalyst for N-Boc imine electrophiles,⁴⁰ and the Huang-Dong thiourea/guanidine catalyst was similarly effective.³⁸

Only two catalysts have provided entry to the less-accessible syn-adducts using the aza-Henry approach. In the same year, we²³ and Ooi⁴¹ disclosed distinct metal-free catalysts effective in achieving syn-diastereoselective aza-Henry additions with N-Boc imine electrophiles. The Ooi catalyst is an innovative betaine system based on the close proximity of alkoxy and tetraalkyl ammonium functionality, leading to syn-adducts in up to 6:1 dr.

Despite these discoveries, the underlying determinants of diastereoselection in the aza-Henry reaction remain unclear. The majority of catalyst-controlled reactions favor anti-selectivity,⁴² regardless of the substitution level of the nitroalkane.³² Using BAM catalysis, for example, anti-selectivity is observed in all cases except one (vide supra). Given the unusual selectivity favoring syn-adducts 1 when using BAM catalyst 2 (Scheme 2), we established a goal to achieve high anti-selectivity for 1 using the same basic design elements. This would provide a straightforward organocatalytic synthesis pathway to each of the four stereoisomers of α-quaternary α,β-diamino acids by simply choosing the proper BAM catalyst. We reported recently that symmetrical catalyst 3 complements unsymmetrical catalyst 2 by providing adducts with high anti-selectivity while preserving high enantioselectivity.³⁶ In this report we detail the first computationally-based study of catalyst 2 within these reaction pathways and advance new hypotheses for selectivity through evaluation of catalyst-substrate interactions.^43,44,45 The summary findings reveal unanticipated details related to catalyst design that underpin this example of diastereodivergent synthesis with generally conserved catalyst features.

Results and Discussion

General Trends for Stereoselection

The aza-Henry reaction of α-nitro esters provides a complementary alternative to enantioselective α-amino acid synthesis based on glycine Schiff base reagents (e.g., O’Donnell’s enantioselective phase transfer catalyzed alkylations).⁴⁶ α-Nitro esters require, in principle, far less basic conditions for activation by Brønsted base. The nature of the contacts between substrate(s) and catalyst, however, are less understood. The development of models for reagent-mediated diastereoselection and enantioselection might provide a basis for generalizing the use of α-nitro esters.

Intriguingly, catalysts 2 and 3 behaved similarly from the standpoint of reactivity and enantioselectivity, but favored opposite diastereomers. Based on product homochirality at the benzylic aminomethyl carbon (R), we inferred that the imine electrophile-catalyst binding with 2 and 3 is generally conserved. This case then provides a unique opportunity for stereochemical analysis by using a modest change in catalyst structure as a single variable for consideration. The key differences between these catalysts are the increased crowding in the quinolinium binding pocket of unsymmetrical catalyst 2, and the creation of a cavity defined on one boundary by the anthracenyl group. Ester size of the nucleophile is a second steric influence within the binding pocket of each catalyst.

A second aspect of our working hypothesis for diastereocontrol is the role that hydrogen bonding between catalyst and nitronate-NO₂ might play. Since anti-selectivity is observed in the addition of nitroethane and nitroacetate with 7 and 9, respectively,^19,21 the loss of diastereoselection was originally ascribed to the possibility of a hydrogen bond between the imine-bound catalyst and nitroalkane in the pre-transition state assembly (Scheme 3). In other words, the change from nitroalkane to α-substituted nitro ester offered a second hydrogen bond acceptor—the ester functionality—after considering the azomethine as the first and dominant hydrogen bond acceptor. The possibility that this hydrogen bond could be transitional, evolving from ligand-bound to product-bound, appeased our concern for the details of its coordination number and the accordant geometric restrictions.

Scheme 3.

Determinants of Diastereoselection: Hypothesized Role of Nitro Group as a Hydrogen Bond-Acceptor, and Ester Competition for the Catalyst Hydrogen Bond

Our development of a computationally-driven model for enantioselection in the aza-Henry reaction of nitromethane to N-Boc imines is based on a hydrogen bond donor-acceptor interaction between catalyst and Boc-carbonyl oxygen, leading (R,R)-cyclohexane diamine-derived BAM complexes to provide R-configured benzylic amines. This catalyst-benzylic amine (of both diastereomers) correlation of configuration is conserved across all BAM-catalyzed aza-Henry reactions to-date. That is, it is the rule, not the exception, for diastereomeric aza-Henry products of BAM-catalysis to be homochiral at the benzylic amine carbon. This regularity with which benzylic amine carbon and catalyst antipode correlate suggests a robust catalyst-electrophile complexation.

Computational Analysis of Substrate and Catalyst Interactions

Recent studies produced a detailed depiction of C₂-symmetric BAM catalyst-substrate interactions using computational analysis.⁴³ This groundwork advanced a hydrogen bond donor-acceptor interaction between the quinolinium and Boc-imine, respectively, that leads to Si-face selectivity when using the (R,R)-cyclohexane diamine catalyst backbone. This analysis became a point of departure for our hypothesis that a coordination complex is conserved in the transition states leading to each diastereomer, owing to the homochirality of the aminomethyl carbon (Scheme 2). A second hydrogen bond donor-acceptor interaction might be a key determinant of aza-Henry reactions exhibiting low diastereoselection, one that must be managed (i.e., minimized) in order to return to high diastereoselection. In the case of nitro esters, the ester is presented as a hydrogen bond acceptor that might compete with the nitro group for a hydrogen bond donor, or provide a nitro-group surrogate for a dipole-dipole interaction in the transition state (Scheme 3).

Following this reasoning, complexes exhibiting a crowded imine-binding pocket might be more selective toward the preferred hydrogen bond acceptor (i.e., nitro). Moreover, the ester carbonyl in our studies was sterically manipulated to deter its hydrogen bond acceptor ability by increasing the size of the ester substituent (-CO₂R), a change that would also increase the steric penalty of its role as an opposing dipole. In the following section, we provide computationally-founded transition state models based on these hypotheses.

Mechanistic Analysis

Our previous computational investigation of the BAM-catalyzed aza-Henry reaction focused on the C₂-symmetric H,Quin-BAM•HOTf (7) catalyzed addition of nitromethane and nitroethane to N-Boc imines.⁴³ From this analysis it was shown that 1) a synclinal arrangement between the nitronate NO₂ and Boc-imine nitrogen was favorable, in part, owing to secondary orbital overlap, 2) two-point hydrogen-bond contacts between the amidinium and N-Boc imine, favoring polar ionic hydrogen bond-nitrogen and polar covalent hydrogen bond-oxygen was vital for regioselection (nitronate C vs. O), and 3) cooperativity among noncovalent interactions was important for determining diastereoselectivity. This study also supported rate-limiting nitroalkane activation by the catalyst, consistent with our observation that higher Brønsted basicity of BAM-catalysis translates to a faster overall aza-Henry reaction in most cases.²⁵ Furthermore, the selective orientation of the N-Boc imine electrophile and coordinated nitronate combined to result in the observed enantioselection, determined by the chiral cyclohexane diamine backbone.

This picture is consistent with our hypothesis that both diastereomers evolve from very similar catalyst-imine binding orientations, an assumption driven in part by the conservation of azomethine facial selectivity determined by diamine absolute configuration. As a result, we hypothesized that high diastereoselectivity is achieved by differences in catalyst-nitronate binding.⁴⁷ Given this premise and our previously reported model for stereoselection in C₂-symmetric H,Quin-BAM Brønsted acid catalysis, we sought a deeper understanding of C₁-symmetric H,Quin-BAM Brønsted acid catalysis. Specifically, the manner in which nitronate facial selectivity is achieved concomitant with preservation of azomethine facial selectivity.

With this intent, as a point of departure to better understand catalyst-nitroalkane interactions, we initially investigated the structure of the salt formed from nitronate (formed in situ by α-nitro ester deprotonation of 11) binding to 2-methylamino pyridinium (10) and quinolinium 15, the latter serving as a truncated form of a typical BAM catalyst (Figure 2). Since it has been shown that the 4-methoxy or 4-pyrrolidinyl substituents affect reactivity substantially, with little or no impact on selectivity,²⁵ the use of a simple quinolinium (10) unsubstituted at the 4-position for these studies is reasonable. Meanwhile, the model ester 11 was chosen to represent hindered α-nitro esters (Figure 2, left-hand side). The salt presents a delocalized nitroalkane conjugate base, wherein one or both anion stabilizing groups can engage in a hydrogen bonding interaction with the catalyst when in the Z-configuration. The lowest energy orientations between α-nitro ester and amidinium are those involving nitronate binding (13 and 14). Compared to amidinium binding to the enolate oxygen, for which a bifurcated acceptor was located (12), nitronate binding through both oxygens of the nitro group (13 or 14) was favored by nearly 6 and 12 kcal mol⁻¹, respectively in the model for the anthracenylpyridine-containing catalyst. The difference in energy between nitronate 13 and 14 also equated to nearly 6 kcal mol⁻¹, with the large ester favored in an anti relationship to the ring. The same preference was observed upon analysis of the quinoline binding element (16 – 18), for which the anti relationship of the quinoline and large ester is 1.5 kcal mol⁻¹ lower in energy (Figure 2, right-hand side). These calculations clearly illustrate the measurable, conformation-driven energy differences in α-nitro ester activation through amidine binding. Nonetheless, it remained unclear which isomer would be lower in energy when considering the chiral catalyst (vide infra), where the cyclohexane diamine backbone was present.

Figure 2.

Computational analysis of α-nitro ester conjugate base orientations relative to amidiniums modeled with a truncated version (10, 15) of catalyst 2 calculated at the TOLUENE(SMD)/M06-2X/6-311++G(2d,2p)//M06-2X/6-31G(d) level. (Ar = 2,6-ⁱPr₂-C₆H₃ (11), energies reported as Gibbs free energies).

During exploration of electrophile activation, we assessed amidinium complexation using catalyst 2 complexed with 4b. In a prior model for H,Quin-BAM•HOTf (7) catalysis it was found that both the nitrogen and carbamate oxygen of the N-Boc imine serve as hydrogen bond acceptors⁴³ and by analogy a similar two-point binding orientation for N-Boc imine 4b to favorably interact with catalyst 2 was anticipated. With this in mind, a conformation search of the analogous complex using the truncated anthracenyl pyridine of 2 was performed revealing, as anticipated, two-point binding of the amidinium to both the carbonyl-oxygen⁴⁸ and imine nitrogen was favored (N-H-O, 1.89 Å vs. N-H-N 1.91 Å), (19 and 20 in Figure 3). In this respect, the large planar surface of the imine was situated coplanar with the anthracene ring for both cases, namely the less favored s-cis imine bound complex and the favored s-trans imine containing complex. Two key distinct differences between these complexes, however, are centered around relief of steric strain, owing to the distal-spatial relationship of the large tert-butyl group and the amidinium moiety, and more favorable hydrogen bonding interactions to the carbonyl oxygen atom, as observed in structure 19. Since the anthracene ring is perpendicular to the pyridine ring, the overall orientation of the imine is unique, however, this does not result in a reversal of azomethine π-bond exposure (that would lead to the heterochiral benzyl amine products).

Figure 3.

Computational analysis: bidentate binding of the aldimine to anthracenyl amidine. Comparison of s-trans and s-cis conformations, structures were computed at the SMD_(TOLUENE)/M06-2X/6-311++G(2d,2p)//M06-2X/6-31G(d) level of theory.

Next, in order to determine the contribution of relative product energies to selectivity we modeled the individual diastereomers in the absence of catalyst to reveal the syn-product was lower in energy than the anti-product by 0.6 kcal mol⁻¹ (Figure 4). Notably, this trend is consistent with all of our past work, in which thermodynamic conditions that establish either reversible aza-Henry at ambient temperature, or reversible deprotonation at the α-position of the nitroalkane (when possible), led to maximal syn:anti ratios of 2-3:1, but more often ≈ 1:1. It is clear that anti-selectivity is both kinetically favored in our work, and torsional strain alone is insufficient to provide a high degree of diastereoselectivity.

Figure 4.

Computational analysis: syn-product is 0.5 kcal mol⁻¹ lower in energy than anti-product, R = 2,6-ⁱPr₂-C₆H₃. Structures were calculated at the SMD_(TOLUENE)M06-2X/6-311++G(2d,2p)//M06-2X/6-31G(d) level of theory.

Extrapolating from these truncated models to a more representative picture of the full catalyst system, we constructed four transition state models corresponding to Re- and Si-stereofacial imine addition in the absence of catalyst, with the relative orientation of the nitronate flipped by 180°. In these transition state models, the imine and nitronate were positioned parallel to one another at a carbon-carbon bond-forming distance of 2.00 Å. These structures were then imported into the program MacroModel, the internal coordinates of the imine C=N bond, as well as nitronate C=N bond and directly connected oxygen atoms (each assigned a formal charge of ‒1) were frozen. A model of catalyst 2 was then positioned near (~ 4.00 Å) each of these rigid template transition state models and Monte Carlo conformational searches (MCCS) with the OPLS3 force field were performed.⁶⁷ From these searches, an array of conformers for Re- and Si-stereofacial addition leading to anti- and syn-products of addition were generated. These conformers were then optimized at the M06-2X/6-31G(d) level of theory. From these optimizations, transition states were located, possessing unique (homo-/heteronuclear) hydrogen bonding manifolds leading to anti- or syn-products with (2S,3R), (2R,3S), (2R,3R), or (2S,3S)-configurations, labeled as syn-(2S,3R)-TS, anti-(2R,3R)-TS, etc. The terms syn-(2S,3R) and anti-(2R,3R) indicate the relative and absolute configurations of the product.

Emerging from these calculations was a preferred arrangement of the catalyst in which the nitrogens of both amidinium rings were oriented downwards, as opposed to an up-down alignment. The preference for a diequatorial (vs. diaxial) conformation of the cyclohexyl ring backbone of the catalyst was another general feature of the lowest energy transition states. Collectively, these structural aspects are captured in the four lowest anti- or syn-selective transition state models leading to products with (2S,3R), (2S,3S), (2R,3S) or (2R,3R)-configurations (Figure 5) (see SI for additional transition state structures and energies). Among these transition states, syn-(2S,3R)-TS was favored with a relative Gibbs free energy 0.5 kcal mol⁻¹ below diastereomeric anti-(2S,3S)-TS, 1.1 kcal mol⁻¹ below enantiomeric syn-(2R,3S)-TS and 2.0 kcal mol⁻¹ below anti-(2R,3R)-TS. Accompanying these trends were activation barriers corresponding to 10.7 kcal mol⁻¹, 16.6 kcal mol⁻¹, 14.6 kcal mol⁻¹ and 14.0 kcal mol⁻¹, respectively. The greater stability of syn-(2S,3R)-TS, displaying a carbon-carbon bond forming distance of 2.08 Å, was traced to subtle differences in hydrogen bonding, minimization of repulsive steric interactions and an optimal alignment of the anthracenyl- and phenyl rings, clearly depicting the importance of dispersion interactions.⁴⁹ For instance, this favored transition state exhibited a pair of two-point hydrogen bonds offering both imine activation and nitronate orientation. The two-point heteronuclear hydrogen bond distances to the nitronate were 1.67 Å and 1.98 Å, while the two-point homo/heteronuclear hydrogen bonds contributing to imine activation measured 1.64 Å and 1.75 Å. Salient to this structure, both amidine arms of the catalyst were pointed in the same direction, somewhat like a crown, while the large tert-butyl of the N-Boc group and large diisopropyl ester of the nitronate were oriented away from the catalyst backbone, far removed from unfavorable steric interactions. Meanwhile the anthracene ring of the catalyst effectively contributes to the stabilization of this transition state by maximizing dispersion interactions upon contact with the neighboring phenyl ring.

Figure 5.

Transition state geometries computed at the M06-2X/6-31G(d) level of theory, providing reported free energies relative to syn-(2S,3R)-TS. Bonds in green represent the various homo- and heteronuclear hydrogen bonding interactions. Bond distances are reported in Å.

By comparison, anti-selective transition state anti-(2S,3S)-TS displayed a shorter carbon-carbon bond-forming distance of 2.00 Å. A unique aspect of this structure was stabilization of the carbonyl oxygen atom of the aryl ester by the quinolinium moiety, in which a heteronuclear hydrogen bond contact N(4)-H•••O(9), measuring 1.74 Å persists. Nearby, the reactive nitronate was held in place by the flanking, antiparallel anthracenyl- and phenyl rings, as well as by a single hydrogen bond contact, with a defining nitronate oxygen O(8) distance to the pyridinyl (N(1)) hydrogen atom of 1.89 Å. Meanwhile, the two anilino groups adjacent to the cyclohexyl ring stabilized the imine component, with respective homo- and heteronuclear bond distances of 2.11 Å and 1.66 Å. Furthermore, the large nitronate ester group was forced underneath the backbone of the catalyst (i.e., cyclohexyl ring), thus minimizing steric interactions.

A distinct feature of syn-(2R,3S)-TS (the disfavored enantiomer of the major diastereomer), having a carbon-carbon bond-forming distance of 2.03 Å, was reversal of the hydrogen bond network necessary for substrate activation. Moreover, imine activation took place via hydrogen bonding between the N-Boc carbonyl oxygen of the imine and the anilino nitrogen of the quinolinium amidine arm of the catalyst, while the quinolinium moiety engaged in a homonuclear hydrogen bond contact with the nitrogen atom of the imine. Defining these homo- and heteronuclear hydrogen bond contacts N(4)-H•••N(6) and N(3)-H•••O(5) were bonds with distances of 1.63 Å and 1.87 Å. Additionally, the nitronate was bound to the catalyst through two-point heteronuclear hydrogen bonds with distances of 1.73 Å and 1.83 Å. Associated with this hydrogen bonding was arrangement of the large nitronate ester- and tert-butyl groups away from the catalyst.

Lastly, anti-(2R,3R)-TS, with a carbon-carbon bond forming distance of 2.15 Å, was found to be a rather loose transition state, in which the imine and nitronate components were more remote from the catalyst and less influenced by its stereochemical environment. With this was a reduced level of favorable non-covalent interactions, such as dispersion interactions, as well as the lack of steric directing effects enabling stereocontrol, respectively. This transition state featured two, single hydrogen bond contacts; one to the N-Boc carbonyl oxygen atom and the other to the nitrogen atom of the imine, and two-point heteronuclear hydrogen bonding to the nitronate. The former hydrogen bonding mode involved homo/heteronuclear N(2)-H•••N(6) and N(1)-H•••O(5) bonding with distances of 2.23 Å and 1.59 Å, while the latter involved heteronuclear N(3)-H•••O(7) and N(4)-H•••O(8) hydrogen bonding with distances of 1.69 Å and 1.80 Å. Coupled to this hydrogen bonding, the large ester of the nitronate was orientated away from the catalyst cyclohexyl ring backbone, residing near the quinolinium and imine aryl groups, while the tert-butyl group of the imine was essentially unencumbered.

Taken together, these trends suggest that homo/heteronuclear hydrogen bonding to the imine chiefly governed enantioselectivity, while diastereoselectivity is governed by steric factors, catalyst-substrate hydrogen bonding, and dispersion. The anthracenyl pyridine substructure provides a unique opportunity for the large nitronate ester to be reoriented to favor its Si face for bonding to imine carbon. This hypothesis is further supported by a parallel analysis of low energy transition states when using C₂-symmetric catalyst 3 which favors the anti-diastereomer. In these cases, the nitronate has only quinolinium rings with which to bind (c.f. anti-(2S,3S)-TS).⁴⁷ That the catalyst-imine binding determines enantioselection is a unifying principle supported by all BAM-catalyzed aza-Henry reactions,^19–31 for which the correlation between product amine configuration and catalyst antipode is currently without exception.

As a last tool for probing these transition state structures, we employed non-covalent interaction (NCI) plots (Figure 6). Directly visible from these NCI plots is the presence of several stabilizing non-covalent interactions (green isosurfaces), as seen in the favored lowest energy transition state (syn-(2S,3R)-TS), displaying a π-π stacking interaction between the anthracene ring of the catalyst and phenyl ring of the N-Boc imine substrate. In comparison, transition state syn-(2R,3S)-TS exhibits a C-H/π interaction between the anthracene ring of the catalyst and phenyl ring of the N-Boc imine substrate, while in transition states anti-(2S,3S)-TS and anti-(2R,3R)-TS, analogous interactions were not present. Incidentally, less prominent, yet favorable, van der Waals interactions between the tert-butyl group of the imine and the anthracene ring of the catalyst were found to persist as stabilizing elements for these transition states.

Figure 6.

NCI plots of transition state structures syn-(2S,3R)-TS, syn-(2R,3S)-TS, anti-(2S,3S)-TS, and anti-(2R,3R)-TS. Green arrows point to substrate interactions with anthracene corresponding to CH/π or π- π interactions of syn-(2R,3S)-TS, anti-(2S,3S)-TS, and anti-(2R,3R)-TS.

Comparative Analysis of Alternatives

This catalyst system is the first to deliver nitro ester addition with a high degree of diastereocontrol, using catalyst symmetry properties alone to selectively manipulate the configuration at the α-nitro ester carbon. In other select cases of diastereodivergence, however, co-catalysts have been identified to influence diastereocontrol. For example, Jorgensen noted that a cinchona alkaloid (quinine) could increase diastereoselection from 2:1 (triethyl amine) to 14:1 anti:syn in the addition of a 2-nitro propanoate to an N-PMP glyoxaldimine.⁵⁰ Shibasaki’s homodinuclear nickel complex is effective in the addition of α-substituted α-nitro esters when aryl and aliphatic N-Boc aldimine electrophiles are used, delivering the β-amino-α-nitro esters in high diastereo- and enantioselection that favors the anti-addition products.³⁸ These same products are favored when a stilbene diamine-derived thiourea catalyst is applied, one that contains a Brønsted basic disubstituted amine in close proximity to the hydrogen bond-donor.⁵¹ A C₂-symmetric thiourea/guanidinium catalyst has also been reported for the anti-selective synthesis.⁵² α-Substituted nitrto alkanes with a phosphonate ester also deliver their addition products with high anti-diastereoselection, and similarly require large phosphonate ester groups (ⁱPr₂CH-).²⁴ Additionally, Miao has reported anti-selectivity for the addition of α-nitro methyl propanoate to phosphoraldimine electrophiles, using a thiourea derived from a cinchona alkaloid.⁵³ syn-Selectivity is favored with Ooi’s phase-transfer catalyzed addition of substituted α-nitro esters and N-Boc imines.⁴¹ This contrasts the He-Liu benzotriazole-containing quaternary ammonium salts used in an anti-selective phase-transfer catalyzed addition of α-nitro ethyl propionate in toluene/brine.⁵⁴

The design hypothesis that led to the discovery of this rare case of diastereodivergence in the aza-Henry reaction involved a frustration of hydrogen bonding by the ester in α-nitro esters 11, as depicted in Scheme 3. Evidence for an interaction of this type was observed in anti-(2S,3S)-TS where the catalyst acted as hydrogen bond donor to the ester sp²-oxygen. The four transition states in Figure 5 reveal a more dramatic repositioning of the substrates in the catalyst binding pocket, leading to the change in nitronate facial selectivity while conserving facial selectivity at the azomethine. These details underscore the complexity that underpins the phenomenon of diastereodivergence, despite the relative simplicity of catalysts 2 and 3. Methods developed by Carretero⁵⁵ and Ooi⁴¹ document substrate-controlled diastereoselection, and diastereodivergence related to a reversal of elelctrophile π-face selectivity and catalyst electronic character. In contrast to their system, the reagents and reactions studied here provide a complementary example wherein a change in nucleophile facial selectivity drives diastereodivergence, ultimately controlled by catalyst structure.

Conclusions

In summary, the catalytic asymmetric synthesis of α-substituted anti-α,β-diamino acid precursors catalyzed by a chiral bis(amidine) complex is compared to the analogous syn-selective variant reported earlier.²³ Key to both developments was a synergistic interaction between the chiral catalyst and sterically large ester to generate high diastereoselection. Interestingly, the anti diastereomer is favored utilizing a C₂ symmetric catalyst, while syn-selectivity is achieved using a C₁ symmetric catalyst. The work described here looks carefully at reagent-controlled diastereoselection, and the means by which diastereoselectivity is reversed by the transition from a C₂- to C₁-symmetric catalyst without major effect on enantioselection. The empirical trends and computational analysis implicate a combination of steric factors, catalyst-substrate hydrogen bonding and dispersion as the reason for the reversal of diastereoselection. Specifically, directing the imine and nitronate into its most favorable diastereomeric conformation were two-point homonuclear- and heteronuclear hydrogen bonding interactions. Accompanying this favored hydrogen bonding network of the syn-(2S,3R)-TS transition state was the indispensable role of the anthracene ring that helped lock the imine in a rigid conformation by maximizing dispersion interactions, ultimately helping govern diastereoselection.

Experimental Section

Computational Methods

Density functional theory (DFT) calculations were performed using Gaussian 09⁵⁶ and all geometry optimizations employed the M06-2X⁵⁷ functional with a 6-31G(d) basis set. This approach was ultimately arrived at upon screening multiple methods, including the use of the B3LYP and B3LYP-D3 functionals, which includes Becke’s three-parameter gradient-corrected exchange functional⁵⁸ and the LYP correlation functional⁵⁹ of Lee, Yang and Parr, as well as Grimme’s dispersion correction method⁶⁰ for the latter utilizing a 6-31G(d) basis set. Moreover, the M06-2X functional was found to be optimal and has been proven to be superior to B3LYP in terms of providing accurate model systems taking dispersion and ionic hydrogen-bonding interactions into account.⁴⁷ Given the large size of these structures, however, it was not feasible to use a more complex basis set, as the computational demand was too large. The optimized geometries were verified as either transition state structures (one imaginary frequency) or minima (zero imaginary frequencies) by frequency calculations. Intrinsic reaction coordinate (IRC) calculations were performed to confirm that all transition state structures were linked to relevant minima. Notably, numerous transition state geometries and addition modes were considered, from which the lowest energy models were included. The energies obtained from the lowest-energy optimized structures at the M06-2X/6-31G(d) level of theory were further refined by single-point calculations performed at the M06-2X/6-311++G(2d,2p)⁶¹ level of theory using the solvation model density (SMD) continuum solvation model with the default parameters of toluene (ε = 2.38) to account for solvent.^62,63 The thermal corrections to the Gibbs free energies (temperature = 298.15 K, pressure = 1 atm) computed at the lower level of theory (M06-2X 6-31G(d)) were added to the electronic energies obtained from the single-point calculations to provide the reported Gibbs free energies. The key word integral=grid=ultrafine was used for all calculations. The reported non-covalent interaction (NCI) visualizations (isovalue = 0.3, min = −0.03 and max = 0.07) were generated with Schrödinger Jaguar.⁶⁴ The 3D images of all optimized geometries were generated with CYLview.⁶⁵ GaussView5⁶⁶ was used to construct all structures prior to optimization and to visualize the output from the Gaussian 09 calculations. Monte Carlo conformational searches (MCCS) with an OPLS3⁶⁷ force field were performed on the full catalyst systems using the Schrödinger Macromodel program.⁶⁸