Aspartyl β-Turn-Based Dirhodium(II) Metallopeptides for Benzylic C(sp³)–H Amination: Enantioselectivity and X-Ray Structural Analysis

Abstract

Amination of C(sp³)–H bonds is a powerful tool to introduce nitrogen into complex organic frameworks in a direct manner. Despite significant advances in catalyst design, full site- and enantiocontrol in complex molecular regimes remain elusive using established catalyst systems. To address these challenges, we herein describe a new class of peptide-based dirhodium(II) complexes derived from aspartic acid-containing β-turn-forming tetramers. This highly modular system can serve as a platform for the rapid generation of new chiral dirhodium(II) catalyst libraries, as illustrated by the facile synthesis of a series of 38 catalysts. Critically, we present the first crystal structure of a dirhodium(II) tetra-aspartate complex, which unveils retention of the β-turn confirmation of the peptidyl ligand; a well-defined hydrogen bonding network is evident, along with a nearly C₄ symmetry that renders the rhodium centers inequivalent. The utility of this catalyst platform is illustrated by the enantioselective amination of benzylic C(sp³)–H bonds, in which state-of-the-art levels of enantioselectivity up to 95.5:4.5 er are obtained, even for substrates that present challenges with previously reported catalyst systems. Additionally, we found these complexes to be competent catalysts for the intermolecular amination of N-alkylamides via insertion into the C(sp³)–H bond α to the amide nitrogen, yielding differentially protected 1,1-diamines. Of note, this type of insertion was also observed to occur on the amide functionalities of the catalyst itself in absence of substrate, but did not appear to be detrimental to reaction outcomes when substrate was present.

Graphical Abstract

Authors are required to submit a graphic entry for the Table of Contents (TOC) that, in conjunction with the manuscript title, should give the reader a representative idea of one of the following: A key structure, reaction, equation, concept, or theorem, etc., that is discussed in the manuscript. Consult the journal’s Instructions for Authors for TOC graphic specifications.

INTRODUCTION

Given the ubiquity of C(sp³)–H bonds in organic molecules, the manipulation of this structural motif remains one of the most alluring approaches to the modification of complex molecules.¹ Particularly the introduction of nitrogen through nitrene-mediated C(sp³)–H amination is a powerful way to introduce complexity and drastically alter the bioactivity of C(sp³)–H-rich frameworks.² One such approach, amenable to application within complex molecular regimes, is found in intermolecular dirhodium(II)-catalyzed amination of C(sp³)–H bonds, exquisitely exemplified by the work of Du Bois.³ Much effort has been devoted to the development of catalyst platforms that enable high levels of stereocontrol in dirhodium(II)-catalyzed amination, although this remains challenging.⁴ Significant progress has been made through the use of carboxylic acid ligands, including the pioneering efforts of Müller,⁵ Hashimoto,⁶ and Davies⁷ (Figure 1a). More recently, Dauban established the current state-of-the-art of this asymmetric transformation using a similar catalyst (Figure 1b).⁸ For several substrates, enantiomeric ratios up to 94.5:5.5 were achieved. However, a significant drop in enantioselectivity was observed upon introduction of longer alkyl chains or meta-substituents. Other creative approaches to enantioselective C(sp³)–H amination by dirhodium(II) have involved the use of hydrogen-bonding-driven preorganization, as demonstrated by the Bach laboratory,⁹ or a chiral counterion, in the work of Phipps (Figure 1c).¹⁰ The complexity that highly enantioselective approaches frequently require underscores the challenging nature of this asymmetric transformation.

Figure 1.

Selected chiral rhodium complexes and their applications, compared to the current work.

Another inspiring design for chiral dirhodium(II) catalysts may be found in the work of Ball (Figure 1d), in which bis(aspartyl) peptides were identified that can chelate the dirhodium(II) core, concurrently inducing a helical structure in the peptide. These complexes were found to exhibit excellent reactivity and selectivity in asymmetric carbene transfer¹¹ and protein modification.¹² However, one challenge with this catalyst system lies in the formation of both parallel and antiparallel isomers during the complexation reaction, which necessitated purification via preparatory HPLC. In principle, a dirhodium(II) complex with individual aspartic acid ligands would overcome this issue. This led us to wonder whether β-turn-biased aspartic acid tetrameric peptides, which were previously shown by our group to function as highly enantioselective oxidation catalysts,¹³ could function as ligands for dirhodium(II). This could enable a stereochemically well-defined secondary structure, while providing a somewhat globular shape and tightening the pockets around the rhodium catalytic center. Additionally, a significant advantage of short peptide ligands, consisting of readily available chiral building blocks, lies in their ease of synthesis and a high degree of modularity.¹⁴

Here, we report our findings that canonical β-turn-biased tetrameric peptides indeed provide a suitable chiral ligand platform to rapidly generate a library of chiral dirhodium(II) complexes, as illustrated by the facile preparation of 38 new catalysts. We obtained the—to our knowledge—first crystal structure of a dirhodium(II) aspartyl peptide complex, which unveiled a retention of the ligand β-turn confirmation, as well as a near-C₄ symmetry. The synthetic utility of this system is illustrated through benzylic C(sp³)–H amination with levels of enantioselectivity at or above the state of the art. We further explored the potential for self-amination of the catalyst via intermolecular C(sp³)–H amination of amidic N(CH₃) groups present in the peptide ligands.

RESULTS AND DISCUSSION

To assess the ability of β-turn aspartyl tetramers to coor-dinate to a dirhodium(II) center, our initial experiments involved combining the peptide Boc-Asp(H)-D-Pro-Acpc-Leu-OMe with dirhodium(II) trifluoroacetate under reaction conditions analogous to those reported by Ball (see Figure 2).^11b Pleasingly, at 50 °C in the presence of DIPEA, the labile trifluoroacetate ligands were readily displaced by each aspartyl peptide, giving the fully exchanged tetrakis complex Rh₂(Boc-Asp-D-Pro-Acpc-Leu-OMe)₄ in 51% yield. The UPLC chromatogram below highlights the straight-forward nature of this complexation reaction. The resulting compound was well-behaved, showed excellent solubility in many organic solvents, and was easily purified via column chromatography.

Figure 2.

Synthesis of the initial dirhodium(II) peptide complex, with UPLC chromatograms of the crude reaction mixture and the purified complex.

Initial results for the asymmetric amination of benzylic C(sp³)–H bonds using this complex, with 2,2,2-trichloroethoxysulfonamide (TcesNH2) as the nitrene precursor, were very encouraging (see Table 1). However, several parameters were critical for optimization. For example, at room temperature, using 1 mol % of Rh₂(Boc-Asp-D-Pro-Acpc-Leu-OMe)4, significant quantities of product could be obtained when the reactions were run in number of solvents, but initially only minimal enantioselectivity was observed (entries 1–4). When ethyl acetate was used as the solvent, hints of enantioselectivity appeared, and the amination product could be obtained in 12% yield and 60:40 er (entry 5). By switching to i-PrOAc, enantioselectivity increased to 66:34 er, while yield increased to 41% (entry 6). Varying the solvent further to t-BuOAc resulted in a drop in both yield and enantioselectivity (entry 7). The reaction temperature was also critical, as lowering reaction temperature to 4 °C led to an increase in both yield and enantioselectivity to respectively 56% and 71:29 er (entry 8). Yet, lowering the temperature further to −10 °C led to ablation of both yield and enantioselectivity. Additives such as MgO had a minor impact on reaction outcomes (entry 10), and reducing the catalyst loading below 1 mol % also had a small detrimental effect (entry 11). Accordingly, we designated the conditions of entry 8 as the optimized reaction conditions for a broader screen of catalyst structures.

Table 1. Optimization of reaction conditions for the benzylic amination of 4-ethylanisole.

.

entry	solvent	T (°C)	variation	yielda (%)	erb
1	C6H6	rt	-	43%	49:51
2	PhCF3	rt	-	22%	48:52
3	MeCN	rt	-	9%	55:45
4	TFE	rt	-	29%	53:47
5	EtOAc	rt	-	12%	60:40
6	i-PrOAc	rt	-	41%	66:34
7	t-BuOAc	rt	-	13%	61:39
8	i-PrOAc	4	-	56%	71:29
9	i-PrOAc	−10	-	14%	65:35
10	i-PrOAc	rt	no base	34%	65:35
11	i-PrOAc	rt	0.5 mol % [Rh]	32%	64:36

^aDetermined by ¹H NMR in CDCl₃ using trimethyl 1,3,5-benzenetricarboxylate as internal standard.

^bDetermined by chiral HPLC.

Over the course of this study, we examined a total of 38 readily prepared aspartyl peptide complexes, of which 12 illustrative examples are presented in Table 2 (see Table S1 for the complete set of catalysts and screening results). Examination of all four diastereomers of the original ligand sequence (entries 1–4) revealed a strict preference for alternating L-D-L stereochemistry at the Asp, Pro and Leu residues, respectively. That is, this stereochemical array uniquely resulted in appreciable enantioselectivity, while the three diastereomers delivered near-racemates. While a complete understanding of this phenomenon remains elusive, we and others have observed similarly strict stereochemical requirements for the formation of robust β-turn and β-hairpin structures,¹⁵ and it seems likely that these structural features are promoting enantioselectivity-inducing conformations. Further modification of the peptidyl ligand revealed additional requirements for selectivity. Changing the 1-aminocyclopropane-1-carboxylate (Acpc) residue to 2-aminoisobutyrate (Aib, entry 5) revealed an ablation of selectivity to 56:44 er, highlighting a strict preference for a cyclopropyl ring. We have observed in a number of cases that the τ-angle associated with the i+2 residue can have a large influence on enantioselectivity.^15b We presume that these effects stem from precise peptide geometries associated with multivalent catalyst-substrate interactions. Introducing alanine in place of leucine at the C-terminus showed near complete erosion of selectivity (entry 6), congruent with its lower steric demands. Isoleucine at this position, on the other hand, unveiled a promising increase in yield and enantioselectivity to 73% and 79:21 respectively (entry 7). This led us to explore additional β-branched amino acid residues, including cyclohexylglycine (Chg), which afforded a jump in enantioselectivity to 82:18 (entry 8). Attention was then turned to the C-terminal protecting group, leading to an additional enhancement of enantioselectivity. An N,N-dimethylamide in place of the O-methyl ester gave 90:10 er and 81% yield (entry 9). A pyrrolidinyl amide (entry 10) gave equal enantioselectivity, albeit at slightly lower yield (73%). Interestingly, the presence of an N-terminal carbamate proved essential, and modification to an amide completely ablated any enantioselectivity, even when the steric bulk was retained (entry 11). The tert-butyl group of the Boc carbamate was important, as substitution with a methyl group lead to ablation of stereoselectivity (entry 12). Increasing the length of the coordinating acid side chain by switching from aspartic acid to glutamic acid (entry 13) did not prove effective, giving near-racemic products. Of all of the catalysts explored, Rh₂(Boc-Asp-D-Pro-Acpc-Chg-NMe2)₄ (C1, entry 9) and Rh₂(Boc-Asp-D-Pro-Acpc-Chg-Pyrr)₄ (C2, entry 10) proved most selective under the conditions explored, and were thus selected for further study.

Table 2. Screening of selected peptidyl ligand sequences.

entry	[Rh]	yielda (%)	erb
1	Rh2(Boc-Asp-D-Pro-Acpc-Leu-OMe)4	56	71:29
2	Rh2(Boc-D-Asp-D-Pro-Acpc-Leu-OMe)4	30	53:47
3	Rh2(Boc-Asp-Pro-Acpc-Leu-OMe)4	27	53:47
4	Rh2(Boc-D-Asp-Pro-Acpc-Leu-OMe)4	35	48:52
5	Rh2(Boc-Asp-D-Pro-Aib-Leu-OMe)4	50	56:44
6	Rh2(Boc-Asp-D-Pro-Acpc-Ala-OMe)4	42	45:55
7	Rh2(Boc-Asp-D-Pro-Acpc-Ile-OMe)4	73	79:21
8	Rh2(Boc-Asp-D-Pro-Acpc-Chg-OMe)4	80	82:18
9	Rh2(Boc-Asp-D-Pro-Acpc-Chg-NMe2)4 (C1)	81	90:10
10	Rh2(Boc-Asp-D-Pro-Acpc-Chg-Pyrr)4 (C2)	73	90:10
11	Rh2(Piv-Asp-D-Pro-Acpc-Chg-NMe2)4	31	50:50
12	Rh2(MeOCO-Asp-D-Pro-Acpc-Chg-NMe2)4	28	54:64
13	Rh2(Boc-Glu-D-Pro-Acpc-Leu-OMe)4	40	42:58

^aDetermined by ¹H NMR in CDCl₃ using trimethyl 1,3,5-benzenetricarboxylate as internal standard.

^bDetermined by chiral HPLC.

We then turned our attention to evaluating other reaction parameters and substrate scope. While our screening studies had employed an excess of substrate (3.0 equiv.), we wished to employ 1a and its congeners as the limiting reagent. Accordingly, we examined the sulfamate nitrene precursor as a possible variable for optimization (see Table 3). In studies employing catalyst C1, we found that, when using 1.0 equivalent of the less reactive substrate 2-ethylnaphthalene (1b), the sulfamate TcesNH2 gave the amination product in 93:7 er, but in only 42% yield (entry 1). The electron-poor character of the nitrogen protecting group was deemed essential, as ethoxysulfonamide as the nitrene precursor gave a mere 14% yield (entry 2, er not determined). Increasing the electron-withdrawing properties of the protecting group by using 2,2,2-trifluoroethoxylsulfonamide (TfesNH₂, entry 3) resulted in an increase in yield to 54% and 95:5 er. By increasing the number of sulfamate equivalents from 1.2 to 1.5, yield was further increased to 64% while enantioselectivity remained the same. Introducing an additional CF₂ group resulted in a similar yield but with lower enantioselectivity (93:7, entry 4). Therefore, TfesNH₂ was selected as the nitrene precursor to explore the scope of this reaction.

Table 3. Sulfamate optimization.

entry	RNH2	yield (%)	er
1		42%	93:7
2		14%	nd
3		54% 64% c	95:5 95:5
4		65%c	93:7

^aDetermined by ¹H NMR in CDCl₃ using trimethyl 1,3,5-benzenetricarboxylate as internal standard.

^bDetermined by chiral HPLC.

^c1.5 equivalents of sulfamate instead of 1.2 equivalents.

The capacity of catalyst C1 for enantioselective benzylic C(sp³)–H amination is shown in Figure 3a. In some instances, C2 was used because of higher yield or enantioselectivity. Using the optimized sulfamate TfesNH₂, 4-ethylanisole amination product 2a was obtained in 86% yield and 95.5:4.5 er. Importantly, enantioselectivity was retained upon extension of the alkyl chain, giving amination products 2c and 2d in 95.5:4.5 and 95:5 er, respectively. This is of particular interest, since reported dirhodium(II) catalysts for enantioselective benzylic amination often exhibit appreciable erosion of er with elongated alkyl substrates.⁸ To further highlight this complementarity to existing catalysts, 3-phenylpropyl acetate and its butyl analog were investigated, to give products 2e and 2f in respectively 54% yield, 94.5:5.5 er and 63% yield, 93.5:6.5 er. Also 3-phenylpropyl bromide was efficiently aminated to give 2g in 53% yield, 93:7 er. Next, the 4-position of the substrate was explored. Enantioselectivity remained high with a 4-phenyl substituent, giving product 2h in 77% yield and 95:5 er. Electron-withdrawing substituents at the 4-position were also tolerated, although this required an excess of substrate due to their lower reactivity, as was observed with 4-bromo derivative 2i (68% yield, 93.5:6.5 er). Even the 4-trifluoromethyl derivative 2j proceeded with considerable enantioselectivity (95:5 er), although yield for this substrate was lower at 24%, reflecting its slower rate of reaction, as has also been noted by others with this substrate.^8b The reaction was tolerant towards amide substituents, and 4-acetamido product 2k was obtained in 66% yield and 90.5:9.5 er. Interestingly, moving the acetamido group to the meta position increased enantioselectivity, giving product 2l in 95.5:4.5 er. A phenyl group was not strictly required, as 1,2,3,4-tetrahydrocarbazole 2m was afforded in 89.5:10.5 er. The regiochemistry of 2m was confirmed via X-ray crystallography. Significantly lower enantioselectivity (71:29 er) was observed for ring-constrained substrate 2n. The catalyst system proved sensitive to steric factors, and was not tolerant towards an ortho-substituent, as only trace product was obtained in the case of 2o. Also, a free alcohol was shown to inhibit reactivity (2p). Despite these limitations, catalysts like C1 and C2 represent a clear advance in the state of the art of enantioselective C(sp³)–H amination.

Figure 3.

(a) Scope and selected limitations of the amination reaction. ^aobtained with C2 instead of C1. ^bperformed with 2.0 equiv. of substrate and 1.0 equiv. of TfesNH₂. (b) 1.0 mmol scale example of the enantioselective amination reaction. (c) Deprotection of the Tfes group.

Reactions readily translated from screening scales to more moderate scales. When 2b was examined on a 1.0 mmol scale, the product was delivered in even higher yield at 78%, with enantioselectivity retained (Figure 3b, 94.5:5.5 er). To further illustrate the synthetic utility of this transformation, we showed that the relatively uncommon 2,2,2-trifluoroethoxysulfamoyl protecting group was easily removed by heating in a H₂O/1,4-dioxane mixture to afford free amine 2b’ (Figure 3c). After preparing the N-benzoyl derivative, yield and er were determined at 38% and 95:5, respectively. Comparison of the optical rotation of 2b’ to literature values¹⁶ allowed us to assign the (R)-configuration to the product. While the stereochemistry of the other compounds might be assigned by analogy—and this convention is used to draw the product structures—a rigorous determination of the absolute configuration is presently only known for compound 2b.

During substrate scope exploration, our catalyst system also proved competent in catalyzing a different type of reactivity. In an attempt to aminate arylacetamide 1q (Figure 4a), bearing electron-deficient benzylic C(sp³)–H bonds, the benzylic amination product was not formed, but instead aminal 2q was formed in 59% yield (entry 1), resulting from amination of one of the methyl groups of the N,N-dimethylamide. Similar reactivity has been well-characterized in the case of metal-free radical pathways,¹⁷ Fe catalysis,¹⁸ and Cu catalysis.¹⁹ A dirhodium(II)-catalyzed nitrene insertion into ⍺-amino C(sp³)–H bonds has been observed in an intramolecular context,²⁰ and a limited number of intermolecular examples have been described involving N-alkylamides^3g and N-alkylamines.²¹ Notably, we observed that Rh₂(esp)2 delivered the product in lower yield at 38% (entry 2). We explored a potential asymmetric variant of this reaction, but initial studies have thus far only afforded racemic 1,1-diamines (see Figure 4b).

Figure 4.

(a) Rhodium-mediated formation of a 1,1-diamine product in the presence of an N,N-dimethylamide. (b) Attempted asymmetric version of the N-adjacent amination.

This finding also raised the question whether amination of the N-alkyl moieties present in the catalyst itself might be taking place under the reaction conditions, potentially leading to consumption of nitrene precursor, deactivation of the catalyst, and/or lowered enantioselectivity. We therefore conducted a self-amination study wherein the catalyst was subjected to 10 equivalents of TfesNH₂ in the absence of any substrate (Figure 5). After 20 h, the catalyst was fully consumed and a complex mixture of products appeared, as shown in the LCMS chromatogram. The mass spectrum below shows a statistical mixture of catalyst amination products of varying degrees, ranging from 5 to 11 aminations in total. Since we anticipated that this self-amination would play an important role in reducing yield or enantioselectivity, we prepared a series of catalysts with deactivated N-alkyl positions (see Table S1, entries 35–38). However, none of these performed better than their regular N-alkyl counterparts, suggesting that the self-amination may not be relevant, or at least not deleterious to reaction outcomes under the catalytic reaction conditions. Notably, as the catalyst functions at the 1 mol % loading level, its effective concentration is quite low throughout the course of the reaction, and a significant amount of self-amination might be suppressed.

Figure 5.

Self-amination of C1 as analyzed by LCMS.

Dirhodium(II) aspartyl peptide complexes are notoriously difficult to crystallize,²² and an X-ray crystal structure of an elaborate complex such as that presented here has not been reported. We also were unable to identify any conditions that led to crystallization of enantiopure catalyst C1. However, crystallization of racemates is a technique well-known in peptide and protein crystallography to obtain crystals from recalcitrant samples.²³ Thus, we prepared the enantiomer of catalyst C1, combined the two compounds in a 1:1 mixture, and observed facile crystallization from a range of solvents, including methyl acetate. The X-ray reflections produced by these crystals were weak and subsequently integrated to a resolution of 1.1 Å, but we are confident in the model determined from the resulting electron density maps (Figure 6a–d, also see SI Section 7.1). In the solid state, complex C1 exhibits a relatively globular shape with a nonpolar surface (Figure 6a). Both of the rhodium axial sites are occupied by a methyl acetate solvent molecule. Although bond lengths could not be accurately determined from the available data, it is clear from the general shape of the structure that the β-turns remain intact (see Figure 6b). Figure 6c highlights another distinctive feature; the different peptide chains exhibit close proximity to one another. In other words, there likely is a hydrogen bond contact between the C=O of the N-terminal Boc-protecting group and the N–H of the 1-aminocyclopropane-1-carboxylate residue of an adjacent tetrapeptide, resulting in a well-defined, rigid framework of tetrapeptide units surrounding the dirhodium core. Perhaps the most striking characteristic of the solid-state structure lies in the pseudo-C₄ symmetry of the complex, as depicted in Figure 6d. Here, one rhodium atom is flanked by four cyclohexyl rings, while the other is flanked by four prolyl rings. Both rhodium axial sites appear to be accessible, as evidenced by the presence of a coordinating methyl acetate molecule at both sites. This proximity suggests that decoration of the ligand proline or cyclohexyl rings could enhance enantioselectivity. However, two 4-aminoproline derivatives assessed in our ligand screen did not prove effective (Table S1, entries 9–10).

Figure 6.

Crystal structure of rac-C1·(MeOAc)₂ and solution state characterization of C1 by ¹H NMR. The axial methyl acetate ligands are removed for clarity. (a) Space-filling model viewed along the Rh-Rh bond axis. (b) Isolated view of the dirhodium core and ligand β-turn. (c) Axial view, showing the proximity of the different tetramer ligands to one another. (d) Equatorial view of the complex showing the inequivalent rhodium sites, one flanked by four cyclohexyl rings, the other by four prolyl rings. (e) Comparison of the ¹H spectrum of the free peptide ligand and the dirhodium complex.

Given the extant literature data, the prevailing wisdom in dirhodium(II)-catalyzed carbene transfer postulated that only D₂-or C2-symmetric complexes were capable of efficient enantioinduction.²⁴ This line of thinking was challenged by later results that complexes such as Rh₂(S-PTTL)₄ showed high enantioselectivity in carbene transfer and aziridination reactions while displaying inequivalent axial sites in the solid state, analogous to our findings.²⁵ However, it must be noted that there can be discrepancies between the solid-state structure of the complex and its structure in solution.²⁶ The ¹H NMR spectrum of C1 is shown in Figure 6e, and is compared to the spectrum of the free ligand. Only a single set of ligand peaks was observed, suggesting a structure with a high degree of symmetry in solution, or perhaps rapid exchange dynamics. In either limiting case, a well-defined structure exists in both the solid state and in solution, providing a reasonable model for future catalyst design of next-generation dirhodium(II) peptide complexes.

CONCLUSION

In conclusion, we have described a novel class of chiral dirhodium(II) carboxylate complexes with near-C₄ symmetry, based on aspartyl β-turn-biased tetrapeptides, that showed utility in asymmetric benzylic C(sp³)–H amination. The catalysts are modular and easy to prepare and purify, as illustrated by the facile generation of a library of 38 catalysts. After optimization of the peptidyl ligand sequence, this system proved complementary to existing methods, retaining high enantioselectivities for longer-alkyl ben-zenes and meta-substitution. We succeeded in obtaining a crystal structure of the optimized catalyst via racemic crystallization, and showed that the β-turn of the tetramer ligands was retained. The solid-state structure displayed a near-C₄ symmetry, resulting in inequivalent rhodium sites, where one was surrounded by four cyclohexyl rings, and the other by four prolyl rings. Additionally, it was demonstrated that the dirhodium(II) aspartate catalysts were competent catalysts for the amination of NC–H bonds of N-alkylamides, yielding differentially protected 1,1-diamines. Control studies supported the idea that self-amination of the catalyst was plausible under the reaction conditions, but catalysts with deactivated N-alkyl positions did not show improved reactivity or enantioselectivity. By virtue of their modularity and ease of synthesis, aspartyl β-turn-biased ligands provide a unique platform for dirhodium(II) catalyst development, that could allow us to address complex selectivity questions in C(sp³)–H bond amination and beyond.