Counterion-Mediated Enantioconvergent Synthesis of Axially Chiral Medium Rings

There are few enantioconvergent reactions in which racemic substrates bearing multiple stereochemical features are converted into products with high levels of diastereo- and enantiocontrol. Here, we disclose a process for the highly enantio- and diastereoselective syntheses of medium ring lactams via an intramolecular counterion-directed C-alkylation reaction. The treatment of racemic biaryl anilides that exist as a complex mixture of enantiomers and diastereoisomeric conformers by virtue of multiple axes of restricted rotation with a quinidine-derived ammonium salt under basic conditions affords medium ring lactams bearing elements of both axial and point chirality via an enolate-driven configurational relaxation process. Thermal equilibration of the syn- and anti-product diasteroisomers has demonstrated that the barriers to bowl inversion are >124 kJ mol-1. We propose that the chiral ammonium salt differentiates between a complex and rapidly equilibrating mixture of enolate and rotational isomers, ultimately leading to highly enantioselective alkylative ring closure. This dynamic and enantioconvergent process offers an operationally simple approach to the synthesis of valuable chiral medium ring lactams for which there are few catalytic and enantioselective approaches.


■ INTRODUCTION
Enantioconvergent catalytic reactions are those in which a racemic starting material is converted directly into an enantioenriched product. There are numerous examples of such transformations that convert racemic substrates bearing a single stereochemical element into the products with high levels of enantiocontrol. 1,2 However, there are relatively few examples in which racemic substrates bearing multiple stereochemical features are converted into complex products with diastereo-and enantiocontrol. Zhou and co-workers have demonstrated that a racemic and diastereoisomeric mixture of α,α′-substituted cyclopentanones can be hydrogenated to a single product with high e.r. and d.r. in the presence of a chiral ruthenium complex (Figure 1a). 3 Zhao and co-workers employed a chiral iridium complex in conjunction with a chiral phosphoric acid catalyst to enable a hydrogen borrowing protocol for the dynamic enantioselective amination of an isomeric mixture of secondary alcohols with almost complete stereocontrol. 4 Kalek and Fu 5 have shown a chiral nickel catalyst can enable enantioconvergent sp 3 −sp 3 cross coupling from two racemic coupling partners (Figure 1b), 6 and Jiang and co-workers have demonstrated enantioselective photoredox radical−radical coupling. 7 We were interested in whether we could apply an enantioconvergent approach to the synthesis of biaryls bearing restricted rotation about the aryl−aryl axis. With the recognition of the increasing importance of the field of axial chirality, there has been an explosion of interest in enantioselective methods for the synthesis of these molecules. 8−12 However, few of these methods are applicable to the synthesis of biaryls that are embedded in rings, despite these motifs being observed in potent bioactive natural products and medicinally relevant compounds. 13−16 This may reflect the inherent challenges of constructing rotationally restricted biaryls, coupled with the difficulties of forming medium ring compounds with their transannular, large angle, and torsional strain. 17−21 In addition, while progress has been made in the enantioselective synthesis of medium ring heterocycles, 22−25 there are few enantioselective catalytic methods for the synthesis of medium ring biaryl compounds. 26−28 We reasoned that a conceptually interesting approach to this class of compounds would be via the base mediated ring-closing C-alkylation of biaryl anilides ( Figure  1c). This poses a number of challenges, as anilides are known to possess significant barriers to rotation about the N-aryl bond when appropriately substituted. 29−31 In addition, the potential for restricted rotation about the biaryl and amide bonds means that these materials exist as a complex mixture of enantiomers and diastereoisomeric conformers. Curran and co-workers 33 have elegantly demonstrated that the barrier to rotation of anilides is very significantly reduced upon enolate formation 32 due to an increased facility for nitrogen pyramidalization in the cross conjugated amide enolate. We postulated that exploitation of this effect could enable rapid conformational and enolate equilibration of our substrates prior to ring closure and that a chiral counterion could differentiate between members of this dynamic ensemble, leading to an enantioconvergent ring closure with control of the axial and point chirality elements. We have previously demonstrated that a chiral counterion is able to differentiate between equilibrating anions in enantioselective O-alkylation reactions, 34 and this tenet extends that observation. 35

■ RESULTS AND DISCUSSION
A model anilide substrate 1 was constructed in a five-step procedure from N-methyl iodoaniline that included a palladium mediated cross coupling of 2-hydroxybenzene boronic acid hemiester with palladium acetate, N-acylation to introduce the β-keto amide functionality and transformation of a primary alcohol into a bromide leaving group (see Supporting Information p 9). Model substrate 1 is point chiral and racemic and exists as a complex mixture of rotameric forms, as demonstrated by 1 H NMR spectroscopy. With a model substrate in hand, we commenced our investigations of the cyclization reaction by examining the base catalyzed Calkylation process (Table 1). When the substrate was stirred with stoichiometric LiHMDS at RT, a single diastereoisomer of (racemic) product 2 was produced in 66% yield. In contrast, we observed that treatment with tetrabutylammonium ammonium bromide and KOH led to an effective cyclization but favored a different diastereoisomer (d.r. 4:1). This change in diastereoselectivity is likely a consequence of deprotonation and cyclization being faster than conformational equilibration in the presence of stoichiometric LiHMDS, so the diastereoselectivity more closely reflects the geometry attained from the kinetic and irreversible deprotonation. In contrast, with a reversible equilibrium deprotonation in the presence of the ammonium salt, equilibration may be faster than cyclization,  and so, the diastereoselectivity reflects the relative rates of cyclization of the (rotameric) transition states, leading to different diastereoisomers. Subsequently, we turned our attention to the enantioselective process and were pleased to observe that treatment with N-benzylquinidinium chloride 3 in the presence of potassium phosphate affected the desired cyclization, albeit in poor yield and selectivity (Table 1, entry 3). A change to cesium carbonate as base (entry 4) led to an encouraging increase in selectivity (85:15 d.r.; 90:10 e.r.) with a modest 29% yield. Selectivity was further improved with the exploration of a different N-benzylic group in 4 at the expense of yield. The exploration of a range of different solvents with catalyst 4 was mostly ineffective (entries 5−9), but a switch to potassium hydroxide in cyclopentyl methyl ether led to a striking increase in yield (to >99%) without significantly compromising selectivity (95:5 d.r.; 90:10 e.r., entry 10). At this stage, we probed how the leaving group could also affect reaction efficacy, recognizing that the relative rates of the ring closure and the interconversion of the rotameric forms of the substrate could have an impact on selectivity.
The change from a bromide leaving group to a methanesulfonate in 5 lead to a small increase in enantioselectivity (to 93:7 e.r., entry 11), and hence, we subsequently explored how the nature of the N-pendant group on the catalyst could influence enantioselectivity with a focus gave very similar results (entries 11−13). Finally, we observed that catalyst 8 in which the quinoline was functionalized with an O-diphenylmethyl group led to a further increase in selectivity (to 97:3 e.r. and 97:3 d.r., entry 14). In all cases, the minor diastereoisomer was produced in a significantly lower e.r. The absolute configuration of 2 was unambiguously determined by X-ray crystallography (see CCDC 2071682 for the crystal data).We were confident we had discovered the best catalyst for this transformation, and hence, we explored the substrate scope by initially examining the substitution on the aryl ring distal to the lactam nitrogen ( Table 2). Electron-donating groups such as 4-methyl and 4methoxy were well tolerated, and enantioenriched lactams 9 and 10 were obtained in excellent yields and selectivities (9: 95.5:4.5 e.r. and 93:7 d.r.; 10: 95.5:4.5 e.r. and 95:5 d.r). Substrates with halogens in this position, such as fluorine and chlorine, were also transformed smoothly and selectively into the desired lactam products 11 (98:2 e.r.; 98:2 d.r.) and 12 (97.5:2.5 e.r.; 96:4 d.r.). Substituents in the 5-position such as fluorine and bromine have minimal impact on the broad, high selectivity observed; the corresponding medium ring products 13 (97.5:2.5 e.r.; 96:4 d.r.) and 14 (97:3 e.r.; 97:3 d.r.) were isolated in good yields with excellent e.r. and d.r. We also observed good stereocontrol with the cyclization of a substrate bearing fluorine in the 3-position of the distal arene chain to afford product 15 (95.5:4.5 e.r.; 92:8 d.r.). We next examined the impact of the introduction of substituents on the arene proximal to the lactam nitrogen and were able to demonstrate that both electron-rich and electron-deficient groups were well tolerated. In general, the yield and stereoselectivity of the cyclization was insensitive to the identity of the substituent in the 4′-position. We observed that electron-rich substituents including methoxy (16,  ; 99% yield) containing-substrates were all successfully transformed in this catalytic transformation with high levels of enantio-and diastereoselectivity. The introduction of strong conjugating 5′-electron withdrawing groups such as nitro and cyano was also investigated; substrates containing these groups cyclized in moderate yields and high levels of enantioselectivity to afford 24 (37:63 d.r.; 55% yield; 97.5:2.5 e.r. for minor diastereoisomer; 44:56 e.r. for major diastereoisomer) and 25 (70:30 d.r.; 62% yield; 96.5:3.5 e.r. for major diastereoisomer). These substrates also cyclized with significantly diminished d.r., which may reflect modulation of the starting substrate rotational barriers. It has been demonstrated that conjugating electron withdrawing groups para to the nitrogen increase the barriers to N−C rotation in axially chiral anilines. 36,37 In anilides, para-electron withdrawing groups have been shown to reduce the barrier to amide N−CO rotation very significantly. 38,39 In 24, the barrier to N−Ar rotation is likely to be much higher than in a compound such as 2. This is a consequence of ground state stabilization through delocalization of the (anilide) nitrogen with the conjugating nitro group. This will have an impact on the N−Ar rotational barrier in the reactive enolate intermediate that will likely slow down conformational relaxation relative to the rate of ring closure, resulting in the lower observed diastereoselectivity. Additionally, the installation of substituents on both aryl rings is also possible: 26 (99:1 e.r., 95:5 d.r., and 77% yield) and 27 (98.5:1.5 e.r., 95:5 d.r., and 86% yield) were obtained successfully with excellent stereoselectivities and good yields.
The configurational stability of the related dibenzolactam compounds 35 has been demonstrated by Natsugari and coworkers, 41 and hence, we were confident that the products of our cyclization would be unlikely to change relative or absolute configuration under the reaction conditions. This is consistent with the observation that there was no change in e.r. or d.r. for 2 in toluene solution at 298 K for several weeks. The X-ray crystal structure (Figure 2A) of anti-2 demonstrates that the biaryl lactam adopts a deep bowl-like arrangement in which the two aryl rings are twisted out of conjugation (torsion angle of 59°) and the N-methyl amide populates an E-configuration. The smaller C7 methyl group occupies a pseudoaxial position with the methyl ester substituent in an equatorial arrangement. The X-ray crystal structure of syn-2 is relatively similar with the torsion between the aryl rings being slightly larger (62.8°). The amide is also E-configured by virtue of the overall bowl geometry, which is slightly twisted from an optimal boat conformation, likely to minimize eclipsing strain between the two sp 3 carbons in the ring. Although 2 is, in principle, a twoaxis system by virtue of the N-aryl and biaryl−biaryl bonds, we anticipated that inversion of the bowl shape of the eight membered ring would occur via a concerted process; this is consistent with the mechanism proposed for a related system by a detailed computational study. 40 We assumed that the allcarbon quaternary stereocenter adjacent to the lactam carbonyl would be invariant under thermal conditions, which precludes the enantiomerization processes pictured ( Figure 2B). Thus, we anticipated that we would observe the (independent) interconversion of two pairs of diastereoisomers under thermal equilibration.
To probe the magnitude of the barriers to rotation, we heated an enantio-and diastereoenriched m-xylene solution of 2 (starting composition: 98:2 d.r. (anti/syn); 99:1 e.r. (anti); 56:44 e.r. (syn)) to 90°C and followed the thermal interconversion of these compounds over time with chiral stationary phase HPLC. 42−44 The 98:2 (anti/syn) ratio of diastereoisomers was converted to an equilibrium mixture of 15:85 (anti/syn) in 270 min, and from this, we can calculate the barriers to bowl inversion to be ΔG 363 K ‡ = 124.1 kJ mol −1 (for the anti-to syn-conversion) and ΔG 363 K ‡ = 129.8 kJ mol −1 (for the syn-to anti-conversion). 45 This appears to be broadly consistent with our original observation; barriers of this magnitude mean that interconversion at room temperature is practically nonexistent. The syn-isomer in which the methyl group populates a pseudoequatorial position is the most stable, which is also consistent with other observations of the relative size of a methyl ester vs a methyl group. 46 A consequence of this equilibration is an increase in the population of syn-2 (and a decrease in that of anti-2); while this is reflected in the diastereoisomeric ratio above, syn-2 is also enantiomeric with ent-syn-2, which means that the e.r. of the syn-isomer increases from 54:46 to 96.5:3.5 ( Figure 2C). As the total enantioselectivity in the system remains constant throughout the equilibration process, we observe a compensatory decrease in the e.r. of anti-2 from the 99:1 starting point to the same enantiomeric ratio of 96.5:3.5.
To gain some insight into the mechanism of this enantioselective transformation, we probed the rotameric preferences of a model system 28 and the cyclization precursor mesylate 30 (Figure 3). Model compound 28 incorporates a diagnostic para-fluoro substituent on the biaryl ring but does not possess the activated ortho-hydroxymethyl substituent needed for cyclization. 28 has 4 signals in the proton decoupled 19 F NMR spectrum, consistent with the anticipated slow rotation of the N-aryl and the amide N−CO bonds. Of the four rotameric species, two existed with very low population (>20:1) and were barely visible in the 19 F spectrum, suggestive of high energy barriers to reach these conformational states. The NOE studies performed at 253 K where interconversion between the rotamers could not be detected indicated the two major species correlated with conformers in which the amide C�O was anti-to the aryl ring, 47,48 and hence, the minor forms are anticipated to correspond to the carbonyl being syn to the aryl group. We employed variable temperature 19 F 1D selective exchange spectroscopy (EXSY) NMR experiments to estimate the barrier to rotation of the N-aryl bond as 79.6 kJ mol −1 (see Supporting Information p 73 for details); this corresponds to a half-life of approximately 11 s at 298 K.
Upon formation of the enolate of 29 with stoichiometric LiHMDS, the spectrum simplifies to a single signal at −118 ppm. Low temperature 19 F NMR experiments did not lead to The complexity of this system made accurate determination of all of three rotational energy barriers intractable. A relatively low-energy interconversion could be monitored by 2D 19 F EXSY experiments carried out at 253−303 K that indicated barriers ΔG 298 ‡ to be in the range of 63−69 kJ mol −1 . This is likely to be the aryl−aryl rotation with the increase in this specific barrier in 30 vs 28, a consequence of the introduction of the ortho-substituent on the biaryl system. This orthosubstitution is likely to also increase the barrier of the N-aryl rotation, but higher temperature studies via 19 F NMR to probe this presented a complex pattern of peak broadening without complete signal coalescence at temperatures up to 358 K. This suggests that the remaining two interconversion processes are of much higher energy, although precise barriers could not be determined (see Supporting Information p 77 for details).
From these data, we are able to suggest a plausible mechanism for the cyclization reaction ( Figure 3B). The mesylate 5 exists as a racemic mixture of enantiomers and enantiomeric conformers. In the presence of base and the ammonium salt, reversible deprotonation of the 1,3-dicarbonyl ablates the stereogenic center; formation of this enolate disrupts the conjugation in the amide, enabling pyramidalization of the aniline nitrogen and lowering the barrier to rotation about the N-aryl bond very significantly, leading to rapid and reversible configurational and conformational equilibration under the reaction conditions. We postulate that the chiral ammonium counterion is able to select from this rapidly equilibrating ensemble to enable highly diastereo-and enantioselective cyclization, leading to the observed medium ring products.

■ CONCLUSION
We have demonstrated that the highly enantio-and diastereoselective synthesis of axially chiral medium rings can be accomplished through an enantioconvergent counterion mediated cyclization. In this stereodynamic process, the formation of an enolate enables rapid interconversion of multiple isomeric forms of the starting substrate, between which the chiral counterion differentiates in the cyclization step. This process offers a general approach to ring-constrained biaryls and will likely find an application in materials and medicinal chemistry. ■ ASSOCIATED CONTENT
Experimental procedures and characterization data (PDF)

Funding
The EPSRC has provided financial support for a studentship (to T.B.) via the Centre for Doctoral Training in Synthesis for Biology and Medicine (EP/L015838/1). The Royal Society has kindly provided a K. C. Wong fellowship (to J.-Y.D.; NIF \R1\182030). We are grateful to Owen Smith for the X-ray crystallographic analyses.

Notes
The authors declare no competing financial interest.