Remarkable Levels of Enantioswitching in Catalytic Asymmetric Hydroboration

Abstract

TADDOL-derived phosphites and phosphoramidites are effective ligands for rhodium-catalyzed asymmetric hydroborations of β,γ-unsaturated amides, achieving up to 99% ee. However, the sense of stereoinduction, R or S, is surprisingly dependent on rather subtle features of the ligand. For example, catalysts employing a TADDOL phenylphosphite and those using the closely related N-methylaniline-derived phosphoramidite of the same configuration give opposite enantiomers of the product. Those derived from optical antipodes give the same product with virtually the same enantioselectivity as illustrated above. The different stereochemical outcomes may reflect fundamental differences in catalyst structure, reactivity or reaction mechanism.

Accessing either enantiomer of a chiral product via asymmetric catalysis is typically achieved by preparing both enantiomers of the catalyst. Occasionally, efficient enantioswitching, that is, producing each enantiomer using similar non-enantiomeric chiral catalysts, has been achieved by changing substituents on a ligand while preserving its absolute configuration.^1,2 Although the structural changes needed to effect enantioswitching are difficult to predict a priori, or even rationalize after the fact, such processes are of special interest and hold the potential to yield fundamental insight into the origin of enantioselectivity in asymmetric catalysis. We now report striking examples of directed rhodium-catalyzed asymmetric hydroboration for which rather subtle changes in the structure of a TADDOL-derived ligand, for example, comparing a phenylphosphite to an N-methylaniline-derived phosphoramidite, lead to nearly complete reversal of stereochemistry. Several factors influencing the efficiency of enantioswitching and some surprising characteristics of the catalysts are found.

The catalyzed hydroboration reaction, first reported in 1985,³ is of renewed interest.⁴ Building upon the work of Evans⁵ and Gevorgyan,⁶ we reported that TADDOL-derived monophosphites are excellent ligands for carbonyl-directed rhodium-catalyzed asymmetric hydroborations of (E)- and (Z)-trisubstituted-β,γ-unsaturated amides.⁷ BINOL-derived phosphoramidites afford very selective catalysts for the related disubstituted alkenes and certain vinyl arene substrates.^8,9,10

Several surprising observations were made while studying a series of TADDOL derivatives, (TADDOL)PX (L1–L4), in the reaction of amide 1.¹¹ Small variations in the aryl substituents give rise to subtle changes in the shape of the phosphorus ligand.^7,8a,12,13 A series of phosphites, including phenylphosphites (i.e., L1A–L4A), and phosphoramidites (i.e., L1–L4 with B–G) was used in conjunction with Rh(nbd)₂BF₄ to effect asymmetric hydroboration of 1 with pinacolborane (PinBH). The (3R)-2:(3S)-2 enantiomer ratios (er) determined after oxidative workup are summarized in Figure 1.

Figure 1.

The enantiomer ratio (er) for the formation of R/S-2 varies as a function of the TADDOL subunit (i.e., L1–L4) while the sense of asymmetric induction (3R or 3S) is a function of (TADDOL)PX substituent, X (X = A–G).

Phenylphosphites L1A–L4A afford predominantly the (3R)-β-hydroxyamide 2 (86–93% ee, 72–80% yield); L3A is the most selective among the ligands examined. In contrast, N-methylaniline-derived phosphoramidites L1B–L4B give the enantiomeric product, (3S)-2 (77–92% ee); L1B and L2B perform nearly equivalently. It is not simply a consequence of phosphoramidite versus phosphite that determines the R/S stereochemical course of the reaction. The N,N-dibenzylamine-derived L3F affords (3R)-2 in 91% ee. Other phosphoramidites give only modest levels of asymmetric induction. The N-benzylaniline-derived phosphoramidites L1C–L4C favor the (3S)-2 (4–40% ee). The indoline (i.e., L1D–L4D) and isoindoline (i.e., L1E–L4E) derivatives give near racemic product in the L1 series and a modest excess of the (3R)-2 for L2–L4 derivatives (i.e., 13–40% ee). The N,N-dimethylamine derivative L1G affords (3S)-2 in 34% ee.

While the extent of enantioreversal is very high for (Z)-1 with monophosphoramidites L1B–L4B, the isolated yields of 2 are significantly lower (35–45%) than those obtained using monophosphites L1A–L4A (72–80%). A side product, tentatively identified as an isomeric δ-hydroxyamide, is formed in substantial amounts (25–32%) but low enantiomeric excess from 1. It presumably arises via rhodium-catalyzed alkene isomerization followed by hydroboration. Although rhodium-catalyzed alkene isomerization is well known,¹⁴ it has not generally been problematic with β,γ-unsaturated amides. For example, (Z)-3 affords β-hydroxyamide 4 in 80% yield using either phosphite L3A or phosphoramidite L4B; alkene isomerization is apparently not a significant competing side reaction. Phenylphosphite L3A gives (3R)-4 (96% ee). Phosphoramidite L4B exhibits the expected enantioreversal, however, the level of enantioselectivity favoring (3S)-4 is somewhat lower, 80% ee (Figure 2). Other β,γ-unsaturated amides possessing trisubstituted alkenes bearing all alkyl substituents give similar results.

Figure 2.

Other trisubstituted substrates, for example (Z)-3, exhibit enantioreversal but to a somewhat lesser degree.

The disubstituted alkene, (E)-5 (R = (CH₂)₂Ph), affords (3S)-6 (90–99% ee) using monophosphites L1–4A; L3A and L4A give the highest enantioselectivity (Table 1). In contrast to the trisubstituted alkene substrates discussed above, the N-methylaniline-derived phosphoramidite ligands L1–4B afford poor to moderate levels of enantioselectivity varying from 33% ee favoring (3S)-6 to 55% ee favoring (3R)-6. The indolinyl (D) and isoindolinyl (E) derivatives are superior, giving (3R)-6 in up to 97% ee with L1D or L2D. Furthermore, enantioswitching now strongly depends upon the TADDOL scaffold. In contrast to L1D and L2D, L3D and L4D give predominantly the enantiomeric product (3S)-6. Using L4D the enantioselectivity reaches 90% ee.

Table 1.

The er and Sense of Enantioselectivity (i.e., R or S) Vary Widely as a Function of the TADDOL Scaffold (i.e., L1–L4) and the Nature of X (i.e., A–G) with Unsaturated Amide 5.

X	L1		L2		L3		L4
	(3S)	(3R)	(3S)	(3R)	(3S)	(3R)	(3S)	(3R)
A	99	1	95	5	99.5	0.5	99.5	0.5
B	22.5	77.5	30	70	60	40	66.5	33.5
C	15	85	50	50	50	50	50	50
D	1.5	98.5	1.5	98.5	75	25	95	5
E	5.5	94.5	6.5	93.5	28	72	50	50
F	91.5	8.5	91.5	8.5	98	2	70	30
G	15	85	25	75	20	80	38	62

^a)The β-hydroxyamide is favored over the γ-regioisomer in all cases (2–15:1). The yield of 6 (R = CH₂CH₂Ph) is ligand dependent and varies from 27–78%. The lowest yields generally reflect incomplete reaction and are not corrected for recovered starting material; the reaction conditions were not optimized. See supporting information for a complete summary of conversions and yields.

While computational studies have addressed the mechanism of rhodium-catalyzed hydroborations using rhodium chloride catalysts, the conclusions are not directly applicable to variants employing dissociable counterions (e.g., BF₄⁻) or two point binding substrates.¹⁵ Furthermore, prior studies suggest that several reaction pathways are close in energy and mechanistic details may vary depending on the exact conditions employed.^15b,16 It is, therefore, not feasible at this time to develop a complete mechanistic rationale accounting for the observed enantioreversal. Nonetheless, the (Z)- and (E)-isomers of unsaturated amide 7 prove useful in identifying several characteristics relevant to enantioswitching.

Firstly, high levels of enantioselectivity and remarkably efficient enantioswitching are observed for either alkene geometry. Using phosphite L4A, (Z)- and (E)-7 each afford (3S)-8 in excellent enantiopurity (96% and 94% ee, respectively, Table 2, entries 1 and 2). Phosphoramidite L1D affords (3R)-8 in 97% and 96% ee from (Z)- and (E)-7, respectively (entries 3 and 4). Using deuterated borane, PinBD, the isomeric (Z)- and (E)-7 substrates afford diastereomeric products using either ligand indicating that the catalyzed addition of boron and deuterium (hydrogen) across the double bond is overall stereospecific and syn for either ligand.¹⁷

Table 2.

(E)- and (Z)-7 Exhibit High Enantioswitching.^a)

^a)Reactions were run as follows: 1% Rh(nbd)₂BF₄, 2.1% ligand, 2.0 equiv PinBH, rt, 24 h.

Matched and mismatched combinations of L4A and L1D were examined (Table 2, entries 5–8). Recognizing of course that these heterocombinations could give rise to a mixture of three distinct 2:1 L:Rh complexes,¹⁸ the “matched” pair is a pseudo-racemate consisting of one equivalent of the indolinyl phosphoramidite derived from the L or (4R,5R)-isomer of tartaric acid (i.e., (4R,5R)-L1D) with one equivalent of the phenylphosphite derived from (4S,5S)-tartrate (i.e., (4S,5S)-L4A). Surprisingly, the matched pseudo-racemate combination is only moderately less selective than either enantiopure ligand used separately. (3R)-8 is produced in 76–80% ee (Table 2, entries 5 and 6), more selective than the homochiral (i.e., mismatched) combination, (4R,5R)-L4A plus (4R,5R)-L1D. The latter affords the opposite enantiomer, (3S)-8, albeit in only 46–50% ee (entries 7 and 8).

The phenylphosphite L4A, indolinyl phosphoramidite L1D and the matched/mismatched combination catalysts were evaluated for non-linear effects¹⁹ in the catalyzed reaction of (E)-7 with PinBH. The surprising results are shown in Figure 3. Graph A shows the change in enantiomeric purity of 8 as a function of the enantiomeric purity of phenyl phosphite L4A. The dashed line serves as a reference for a purely linear relationship. Viewing left-to-right, pure (4R,5R)-L4A to pure (4S,5S)-L4A, the data show a negative non-linear effect for L4A. In contrast, indolinyl phosphoramidite L1D (Graph B) exhibits a strong positive non-linear effect. Graph C shows two sets of data obtained by holding one component of the matched/mismatched combination catalysts constant while varying the enantiomeric purity of the second component. The red data points are obtained varying L4A; the blue data vary L1D. The data for each show little deviation from linearity, suggesting that the mixed “(L4A)(L1D)Rh(I)” complex, rather than “(L4A)₂Rh(I)” or “(L1D)₂Rh(I)”, dominates the reaction.

Figure 3.

Non-linear effects in the catalyzed hydroboration of (E)-7 with PinBH. Negative and positive non-linear effects for the catalysts using phosphite L4A (graph A) and phosphoramidite L1D (graph B), respectively, contrast with the largely linear effects found for equimolar combinations of L4A and L1D (graph C). The dashed lines are for reference only.

In summary, amide-directed rhodium catalyzed asymmetric hydroboration exhibits remarkable levels of enantioswitching. Relatively small changes in ligand substituents result in complete enantioreversal without changing the absolute stereochemistry of the ligand. Comparing non-linear effects for the phenylphosphite L4A, indolinyl phosphoramidite L1D and the apparent matched/mismatched combination catalysts suggests that, while the ligands employed are structurally quite similar, the reactivity of each catalyst differs substantially from the others. The different stereochemical outcomes therefore may reflect significant and fundamental differences in catalyst structure, reactivity and/or perhaps reaction mechanism leading to enantioreversal. Further work is in progress to gain a better understanding of the mechanistic basis for enantioswitching.