Synthesis of Complex Tertiary Glycolates by Enantioconvergent Arylation of Stereochemically Labile α-Keto Esters

Abstract

Enantioconvergent arylation reactions of boronic acids and racemic β-stereogenic α-keto esters have been developed. The reactions are catalyzed by a chiral (diene)Rh(I) complex and provide a wide array of β-stereogenic tertiary aryl glycolate derivatives with high levels of diastereo- and enantioselectivity. Racemization studies employing a series of sterically differentiated tertiary amines suggest that the steric nature of the amine base additive exerts a significant influence on the rate of substrate racemization.

Graphical Abstract

Introduction

The conversion of racemic α-stereogenic ketones to enantiomerically enriched alcohol building blocks through transformation of the carbonyl functionality is an enabling chemical transformation.¹ Transition metal-catalyzed dynamic kinetic hydrogenation reactions, pioneered by Noyori, have been widely employed for the production of enantioenriched secondary alcohols.² The synthesis of tertiary alcohols through the enantioconvergent addition of stabilized carbon nucleophiles to configurationally labile electrophiles is comparatively less common.³ Reported examples employ basic catalysts or additives to promote simultaneous activation of the pro-nucleophile and enantiomerization of the electrophile. Considering this dual role of base in the context of designing other stereoconvergent processes, the transition metal-catalyzed addition of nonstabilized carbon nucleophiles to ketones emerged as a compelling opportunity to generate complex tertiary alcohols not accessible through other methods (Scheme 1, B). The Hayashi-Miyaura type reactions typically rely on the base promoted transmetallation of an organoboron or organosilicon pro-nucleophile to a chiral metal complex.⁴ As an example, the enantioselective addition of arylboronic acids to carbonyl derivatives, including α-keto esters, has been widely developed (Scheme 1, C).⁵ Considering their chemical stability, ease of handling and broad commercial availability,⁶ we envisioned the deployment of arylboronic acids in an enantioconvergent addition to racemic α-keto ester electrophiles electrophiles would facilitate the production of diverse, stereochemically complex glycolate architectures. The purpose of this article is to convey experimental findings related to the dynamic kinetic 1,2-addition of arylboronic acids to racemic α-keto esters (Scheme 1, D).

Scheme 1.

Background and Proposed Enantioconvergent Arylation

Carbonyl electrophiles and their derivatives lacking electron withdrawing functionality (i.e. ketone, ester, or halogen) at the chiral α center are underutilized in dynamic kinetic resolutions (DKR). List and Zhao have reported a dynamic kinetic reductive aminations employing α-alkyl, aryl branched imines that presumably racemize via enamine intermediates.⁷ The cyclohexanecarboxaldehyde derivatives utilized by Ward and co-workers likely racemize via an analogous pathway.^3e Dynamic kinetic hydrogenations of nonactivated aldehydes and ketones have been shown to occur in the presence of tert-butoxide bases.⁸ Nevertheless, considering that facile racemization is essential, the execution of DKRs employing compounds of lower acidity is more challenging.⁹ However, in this context the use of less activated substrates would allow access to heretofore unknown glycolate architectures.

Results and Discussion

In light of the considerations described above, the β-alkyl, aryl substituted α-keto ester derivative 1a was chosen as a model substrate for this transformation (Table 1). Our group has previously developed dynamic kinetic resolutions of α-keto esters that occur in the presence of tertiary amines;^{2l-n, 3a} therefore, we reasoned that an amine base would promote substrate racemization. Sterically hindered Hünig’s base (ⁱPr₂NEt) was initially selected in an effort to minimize interference with the Rh(I)-catalyst through nonproductive binding. A substoichiometric quantity of potassium hydroxide was employed because analogous conditions promote the Hayashi–Miyaura arylation of isatins and 1a is sensitive to stoichiometric hydroxide base.¹⁰ An initial evaluation of ligands revealed that the Ph-substituted norbornadiene derived ligand A developed by Hayashi and co-workers¹¹ provided promising levels of enantioselectivity, although low conversion was observed under these conditions (entry 1). Further screening showed the 4-CF₃C₆H₄-and 3,5-(CF₃)₂C₆H₃-substituted analogues B and C provided higher levels of enantioselection; however, conversion remained low (entries 2 and 3). The supposed low acidity of these substrates caused us to wonder if a simple kinetic resolution was occurring under these conditions, but this possibility was ruled out by isolation of racemic unreacted 1a from entry 3. Interestingly, the benzyl substituted ligand D provided low enantioselectivity slightly in favor of opposite enantiomer, while also exhibiting drastically lower levels of diastereocontrol over the formation of 3a (entry 4). Switching the inorganic base promoter from potassium hydroxide to CsF while increasing the loading to 3.0 equiv allowed for full conversion to the desired aryl glycolate, albeit with a striking drop in enantioselectivity (entry 5). Simply replacing Hünig’s base with triethylamine restored the previously observed levels of enantioselectivity (entry 6). Further increasing the amount of triethylamine to 6.0 equiv provided higher levels of enantioselectivity (entry 7), although a longer reaction time was necessary to achieve full conversion under these conditions. Satisfactory levels of enantioselectivity were achieved when chloroform was used as solvent in place of methylene chloride (entry 8). At this stage of optimization it was noted that both the 4-CF₃C₆H₄- and 3,5-(CF₃)₂C₆H₃-substituted norbornadiene ligands B and C provided identical levels of enantioselectivity (entries 8 and 9). Running the reaction at 60 °C does not influence the enantio- or diastereoselectivity of the process (entry 10). Substituting the ethyl ester of 1a with bulkier ^tBu or Bn groups (entries 11 and 12, respectively) did not result in improved enantioselectivity. Finally, although phosphine^5b and phosphite^5a ligands have been utilized in Hayashi–Miyaura-type arylation reactions of α-keto esters, a complex of triphenylphosphite (E, entry 13) as well as hydroxy[(S)-BINAP]rhodium(I) dimer (F, entry 14) failed to catalyze this transformation.

Table 1.

Reaction Optimization

entrya	cat.	org. base	inorg. base	conv (%)b	drb	erc
1	A	DIPEA	KOH	59	>20:1	80:20
2	B	DIPEA	KOH	56	>20:1	90:10
3	C	DIPEA	KOH	40	>20:1	90:10
4d	D	DIPEA	KOH	61	2.7:1	43:57
5e	C	DIPEA	CsF	>95	>20:1	70:30
6e	C	Et3N	CsF	>95	20:1	89:11
7e,f	C	Et3N	CsF	85	>20:1	92:8
8e,f,g,h	C	Et3N	CsF	>95	>20:1	94:6
9e,f,g,h	B	Et3N	CsF	>95	>20:1	94:6
10e,f,g,h,i	B	Et3N	CsF	>95	>20:1	94:6
11j,k,e,f,g,h	C	Et3N	CsF	>95	>20:1	93:7
12j,l,e,f,g	C	Et3N	CsF	>95	14:1	91:9
13e,g,h,j	E	Et3N	CsF	trace	-	-
14e,g,h,j	F	Et3N	CsF	trace	-	-

^aAll reactions were conducted on a 0.10 mmol scale.

^bDetermined by ¹H NMR analysis of the crude reaction mixture.

^cDetermined by HPLC using a chiral stationary phase.

^dReaction time = 36 h.

^e3.0 equiv CsF.

^f6.0 equiv of Et₃N.

^gCHCl₃ as solvent.

^hReaction time = 48 h.

ⁱReaction was run at 60 °C.

^j3.0 equiv PhB(OH)₂,

^kSubstrate ester = ^tBu.

^lSubstrate ester = CH₂Ph

At this juncture we sought to understand the large contribution to product enantioselectivity associated with the superficially similar structure of the amine base additive. We hypothesize that this difference might arise from a faster rate of starting material racemization under the action of triethylamine. Using Hünig’s base in conjunction with low inorganic base concentration resulted in high levels of product enantioselectivity and the unreacted starting material recovered from the reaction was not enantioenriched (entry 3, Table 1), suggesting that an efficient dynamic kinetic resolution is occurring under these conditions. We postulate that under conditions of low inorganic base concentration the arylation reaction is slow relative to Hünig’s base promoted racemization (k_rac > k_fast)⁹ resulting in a dynamic kinetic resolution. However, in entry 5 the higher loading of CsF results in a faster arylation reaction for both substrate enantiomers, presumably due to higher rates of transmetallation, while the rate of racemization by Hünig’s base occurs too slowly for efficient dynamic kinetic resolution.¹² To gain further insight into this phenomenon and to provide support for our hypothesis we studied the rate of racemization of 1a using an array of tertiary amine bases (Figure 1). At room temperature racemization with triethylamine was rapid; within eight minutes the extent of racemization had reached 87% and complete racemization occurred after 20 min. Tri-n-butylamine exhibited a noticeably slower racemization profile, but was still nearly complete within 20 min. In contrast to triethylamine and tri-n-butylamine, the alkyl branched Hünig’s base displayed a slow racemization profile, and 1a was still measurably enriched after approximately 1.7 h at room temperature. When studied at 40 °C in the presence of 6 equiv of Hünig’s base, racemization of 1a was enhanced but complete racemization only occurred after 1 h. Thus, although Hünig’s base exhibits greater thermodynamic basicity than triethylamine it is less effective at promoting the racemization of 1a.¹³ Finally, N-methylpyrroldine, which possesses lower thermodynamic basicity than triethylamine,¹⁴ displayed the fastest racemization profile, promoting complete racemization of 1a in under two minutes. The observed trend suggests the kinetic basicity of the tertiary amine exerts a larger influence on the racemization of 1a than its thermodynamic basicity. This observation may prove to be generally important in the de novo design novel dynamic kinetic resolutions involving enolizable carbonyl substrates.

Figure 1. Influence of Base Structure on Racemization Rate.

a) Trial conducted at 40 °C. 6.0 equiv Hünig’s base.

With optimal reaction conditions in hand we began to study the scope of the process with respect to the arylboronic acid component (Table 2). It should be noted that while catalysts B and C provide identical levels of selectivity for product 3a, in certain cases it was found that one catalyst was more selective for a particular substrate. Ultimately, electron-rich arylboronic acids were found to be suitable reaction partners as the p-tolyl adduct 3b was formed in high yield with high levels of diastereo- and enantiocontrol. Electron-poor arylboronic acids could also be used; however, in the case of p-fluoro- and p-chlorophenylboronic acid a larger excess was required to achieve good yields. Nevertheless, high levels of diastereo- and enantioselectivity were still observed for addition products 3c and 3d. Substitution of the arylboronic acid at the m-position was also tolerated. For instance, the m-methoxy and m-tolyl adducts 3e and 3f were obtained in good yield, with high levels of stereocontrol. Electron-withdrawing substituents were also tolerated at this position and the use of m-chlorophenylboronic acid afforded the desired arylation product 3g in good yield with high levels of stereocontrol. Polyaromatic boronic acids were also suitable substrates for this transformation, as the 2-naphthyl adduct 3h could be obtained in good yield with similarly high levels of diastereo- and enantiocontrol. The sterically demanding o-methoxy adduct 3i was formed in good yield with high levels of enantiocontrol, although in this instance a relatively large excess of the boronic acid substrate was required to achieve full conversion. Finally, we found that even unprotected 6-indoylboronic acid could be employed, furnishing adduct 3j, while maintaining reaction efficiency. It should be noted that at this stage of optimization certain electron poor arylboronic acid substrates cannot be used, as the 4-pyridyl and 5-indazole adducts 3k and 3l were not formed. In addition, the reaction with 2-thienylboronic acid only reached 11% conversion after 36 h under the optimized reaction conditions (not shown). Efforts to address these limitations are currently underway in our laboratory.

Table 2.

Scope of Dynamic Kinetic Arylation Reaction.

^aReactions run on 0.1 mmol scale for 48 h or 60 h (see SI for individual reaction times and boronic acid equivalents), reported yields and er values are averages of two runs. Values in parentheses represent recrystallized yields and enantiomeric ratios.

^bCatalyst B employed.

^cCatalyst C employed.

Next, we explored the scope of the reaction with respect to the α-keto ester reaction partner (Table 2). Substrates bearing electron donating substituents at the para-position of the aryl ring were suitable reaction partners. For example, the p-tolyl substituted product 3m was obtained in good yield with high levels of stereocontrol. Higher levels of enantioselectivity were observed with this substrate when 2-naphthylboronic acid was employed as a nucleophile furnishing the addition product 3n. Apparently, the electron-rich p-methoxy substituted substrate was subject to facile racemization under the reaction conditions, as product 3o could also be obtained in good yield with high levels of stereocontrol. An ortho-F substituted α-keto ester was subject to phenylboronic acid addition, producing 3p in acceptable yield and high diastereoselectivity and decent levels of enantiocontrol. The o-tolyl product 3q was afforded in 57% yield, and 94:6 er, while the m-tolyl product 3r was formed in 88% yield with 96:4 er, suggesting that the steric nature of the α-keto ester aryl component has a slight impact on reaction efficiency and enantioselectivity. A 2-naphthyl substituted α-keto ester could also be used, affording addition product 3s with high levels of enantio- and diastereoselectivity. Product 3s could be enriched to 97.5:2.5 er following a single crystallization. Notably, arylation of an unprotected 3-indole substituted α-keto ester with 6-indoleboronic acid afforded bis(indole) adduct 3t in good yield with high levels of selectivity. The use of 2-naphthyl and m-tolylboronic acid was also successful with this α-keto ester (see supporting information). Larger alkyl substituents at the β-position were tolerated and the β-ethyl substituted product 3u was obtained in good yield with acceptable levels of enantiocontrol. Product 3u could be obtained as a single enantiomer in acceptable yield following a single recrystallization. The arylation reaction exhibited functional group chemoselectivity in the presence of competing functionality as the bromoaryl and β-allyl substituted product 3v was obtained in acceptable yield with high levels of stereocontrol. Notably, less than 10% of Heck-type co-products¹⁵ were observed during formation of 3v. Additionally, no Suzuki-type products are observed in this process. The branched diester product was formed in excellent yield. Although lower levels of enantiocontrol were observed in this reaction, product 3w can be accessed as a single enantiomer in acceptable yield following a single crystallization. Finally, although pyridine containing boronic acids are not successful reaction partners at this stage of optimization, a 3-pyridyl substituted α-keto ester was tolerated under the reaction conditions and afforded arylation product 3x in good yields with high diastereo- and enantiocontrol. Product 3x could also be recovered as a single enantiomer, albeit in lower yield, after a single crystallization. Substrates bearing only aliphactic substitution at the β-position have note been tested at this juncture; presumably these substrates are less acidic and would be challenging to implement under the present reaction conditions.

Having learned the scope of the DKR arylation process, we sought to examine the effect of increasing the scale of the reaction while simultaneously decreasing the catalyst loading (Scheme 2). The 4-bromo substituted α-keto ester 1y underwent arylation with 2-naphthylboronic acid on 1 mmol scale using 0.5 mol % of the catalyst (1 mol % Rh) to afford 3y in good yield with high levels of diastereo- and enantiocontrol. Product 3y could be recrystallized to 99:1 er allowing the absolute stereochemistry of 3y to be determined via X-ray crystallography. The configuration of the other arylation products 3a–3x were assigned by analogy.¹⁶ The observed stereochemistry can be attributed to the C2-symmetric nature of (R,R)-catalyst C which enforces high levels of enantiocontrol in this reaction by effectively blocking the shaded quadrants in the stereochemical model shown in Scheme 2; the bulky sp³ center is guided to the top left quadrant. The diastereoselectivity of this transformation is in accord with the Felkin-Ahn model.¹⁷

Scheme 2.

Mmol scale arylation and stereochemical model.

Aromatic interactions appear to be important for achieving high levels of enantio- and diastereoselectivity as evidenced by the inferior results using the benzyl substituted catalyst D.

Finally, considering the sterically encumbered nature of the tertiary alcohol installed in the arylation reaction we wondered if this functionality could be leveraged in downstream transformations. Preliminary findings have been promising. For instance, unsaturated alcohol 3v undergoes iodoetherification to tetrahydrofuran 4v in 66% yield albeit without diastereoselectivity. The diastereomers of 4v were easily separated by silica gel column chromatography.

Conclusion

In summary, we have developed an enantioconvergent arylation of racemic β-alkyl substituted α-keto esters catalyzed by a chiral rhodium-diene complex. A wide range of complex aryl glycolate derivatives could be obtained in good yields with high levels of stereocontrol. Notably, despite the longstanding use of transition metal catalysts in dynamic kinetic hydrogenations, this is the first use of analogous catalysts for the installation of C-C bonds in a dynamic kinetic addition to carbonyl electrophiles. Considering the substantial number of commercially available arylboronic acid derivatives and the well-recognized biological activity of the glycolic acid substructure,¹⁸ this chemistry opens the door to a diverse array of interesting building blocks. Although racemization rate is central to efficient dynamic kinetic resolutions⁹ it is rarely discussed or studied in detail, here we have shown that the racemization of less acidic β-alkyl/aryl substituted α-keto esters is strongly linked to the steric size of a tertiary amine additive. Preliminary results show that the products of this reaction can be utilized in additional downstream transformations including the synthesis of valuable tetrahydrofuran derivatives. Extension of this work to other classes of nonstabilized carbon centered nucleophiles is currently underway in our laboratory and will be reported in due course.

Scheme 3.

Iodoetherification of 3v.