Diastereoselective C–H Bond Amination for Disubstituted Pyrrolidines

Abstract

We report herein the improved diastereoselective synthesis of 2,5-disubstituted pyrrolidines from aliphatic azides. Experimental and theoretical studies of the C–H amination reaction mediated by the iron dipyrrinato complex (^AdL)FeCl(OEt₂) provided a model for diastereoinduction and allowed for systematic variation of the catalyst to enhance selectivity. Among the iron alkoxide and aryloxide catalysts evaluated, the iron phenoxide complex exhibited superior performance towards the generation of syn 2,5-disubstituted pyrrolidines with high diastereoselectivity.

Iron dipyrrin complexes catalyze the diastereoselective conversion of aliphatic azides to 2,5-disubstituted pyrrolidines. A combination of experimental and theoretical investigations unveiled the features required for diastereocontrol.

The prevalence of 2,5-disubstituted pyrrolidines in natural products,^[1] pharmaceuticals,^[2] and synthetic applications^[3] has inspired the emergence of numerous stereoselective methodologies^[4] for the synthesis of these N-heterocycles. However, more atom-economical, efficient, and general processes to access complex architectures featured in many of these N-containing compounds are highly desirable. Towards this goal, direct C–H amination methods could provide a streamlined protocol and may serve to enhance the scope of conventional syntheses.

Our group has recently reported a catalytic C–H amination process for the conversion of aliphatic azides to N-heterocycles by using a high-spin iron dipyrrin complex (^AdL)FeCl(OEt₂) (^AdL =1,9-adamantyl-5-Ar-dipyrrin; Ar = 2,6-Cl₂C₆H₃ (1a), mesityl (1b)).^[5] During our studies, we found that exposure of 1-azido-1-phenyl-hex-5-ene to catalytic quantities of 1a in the presence of di-tert-butyldicarbonate (Boc₂O) afforded Boc-2-phenyl-5-vinylpyrrolidine as a 3.9:1 syn:anti mixture of diastereomers (Scheme 1).^[5] The inherent stereocontrol of the iron-catalyzed cyclization, albeit modest, encouraged us to establish whether diastereoselectivity could be enhanced by means of catalyst modification. Herein, we present a dipyrrinato–iron phenoxide complex that furnishes 2,5-disubstituted pyrrolidines with >20:1 syn:anti diastereoselectivity.

Scheme 1.

Diastereoselective iron-catalyzed cyclization.

To develop an understanding of the diastereocontrol of the amination reaction, the diastereoselective cyclization with catalyst 1a was investigated. A series of 1-azido-1-aryl-hex-5-ene substrates were examined to establish the electronic and steric contributions of the aryl ring on the observed selectivity. Both electron-donating and electron-withdrawing substituents (e.g., Me, halides, CF₃) were amenable to cyclization (Table 1). Interestingly, para (3b–3 f) and meta (3g–3i) aryl functionality gave a diastereomeric ratio (d.r.) consistent with our initial findings (3–4:1), whereas ortho substitution (3j–3m) afforded a single diastereomer. Increasing the steric bulk of the R³ group (3p) or exchanging the positions of the phenyl and vinyl functionalities (3q) did not affect diastereoselectivity. However, the cyclization of 1-azido-2-phenyl-hex-5-ene (2r) yielded the corresponding 2,4-disubstituted pyrrolidine (3r) with a low (1.5:1) d.r.^{[5] 1}D NOESY experiments on the deprotected pyrrolidines,^[6] along with X-ray diffraction studies on single crystals of the (^AdL)FeCl[2-(o-(F₃C)C₆H₄)-5-vinylpyrrolidine] adduct (4) (Figure 1a), identified a major syn diastereomer in all instances.

Table 1.

Diastereoselective C–H amination.

Product	Yield [%][a]	d.r.[b]	Product	Yield [%][a]	d.r.[b]	Product	Yield [%][a]	d.r.[b]
R3 =C2H3			R3 =C2H3			R3 =C2H3
3a	(72)[c]	3.9:1[c]	3f	36 (36)[c]	4:1[c]	3k	30 (30)[c,e]	>20:1[c]
	77 (72)[d]	>20:1[d]		53 (53)[d]	20:1[d]		52 (45)[d]	>20:1[d]
3b	(63)[c]	3:1[c]	3g	(47)[c]	4:1[c]	3l	37 (36)[c]	>20:1[c]
	75 (67)[d]	>20:1[d]		55 (49)[d]	>20:1[d]		58 (56)[d]	>20:1[d]
3c	44 (41)[c]	4:1[c]	3h	59 (55)[c]	4:1[c]	3m	38 (31)[c]	>20:1[c]
	67 (68)[d]	>20:1[d]		79 (72)[d]	>20:1[d]		58 (51)[d]	>20:1[d]
3d	(49)[c]	3:1[c]	3i	64 (48)[c]	4:1[c]	3n	27 (30)[c]	4:1[c]
	67 (66)[d]	>20:1[d]		74 (67)[d]	>20:1[d]		50 (43)[d]	11:1[d]
3e	38 (41)[c]	3:1[c]	3j	45 (38)[c]	>20:1[c]	3o	22 (16)[c]	3:1[c]
	74 (71)[d]	>20:1[d]		69 (64)[d]	>20:1[d]		20 (23)[d]	10:1[d]
R1 =Ph, R3 =Ph			R1 =C2H3, R3 =Ph			R1 =H, R2 =Ph, R3 = C2H3
3p	34 (37)[c]	5:1[c]	3q		3.5:1[c]	3r	(66)[c]	1.5:1[c]
	43 (48)[d]	16:1[d]		N/A	7:1[d]		71 (72)[d]	2:1[d]

^[a]1H NMR yield (isolated yields in parentheses).

^[b]syn:anti.

^[c]Catalyst 1 a.

^[d]Catalyst 8.

^[e]10 mol% catalyst.

Figure 1.

Truncated solid-state structure^[10] of a) (^AdL)FeCl-[2-(o-(F₃C)C₆H₄)-5-vinylpyrrolidine] (4) and b) (^AdL)Fe(OPh)(THF) (8) at 100 K with 50% probability ellipsoids.

Prompted by these results, DFT studies were undertaken to evaluate the factors leading to the diastereoselectivity manifest by 1a. Starting from crystallographic parameters for 1b and 4, the reaction trajectory for cyclization of 2a was calculated to locate ferric iminyl and amido intermediates (Figure 2a) with metrics matching those for analogous isolated dipyrrin iron species.^[7] The C–H bond activation transition state (TS) identified displays an elongated Fe–N_im bond (1.835 &) and a nearly linear N_im–H–C₄ vector (155°) in line with a hydrogen atom abstraction (HAA) mechanism.^[8] Scanning the potential-energy surface to locate a TS for the subsequent radical rebound step resulted in immediate convergence to the Fe–pyrrolidine adduct. This finding suggests a barrier-less recombination event, which is consistent with our experimental observations using a radical-clock probe to establish a short-lived (<10⁻¹¹ s) carboradical intermediate.^[5] A similar energetic profile was reported for the diruthenium nitrene mediated allylic intramolecular C–H amination.^[9]

Figure 2.

a) Calculated^[12] reaction coordinate for cyclization of 1-azido-1-phenyl-hex-5-ene (2a) mediated by 1b. b) Geometry-optimized TS for the HAA step in the pro-syn (blue) and pro-anti (red) configuration.

The fast recombination rate identified experimentally and theoretically establishes HAA as the selectivity-determining step. As such, the TS for the HAA of each of the two diastereotopic hydrogen atoms were investigated. Calculations revealed that formation of the syn pyrrolidine product is favored by 0.93 kcalmol⁻¹. The energy difference is subtle; however, it predicts a 4.6:1 ratio of diastereomers, corroborating the experimental 3.9:1 d.r.^[5] The optimized pro-anti TS (Figure S11a in the Supporting Information) shows unfavorable interactions of the substrate with both the dipyrrin bridgehead position and the cleft comprised of the adamantyl groups and the Cl ligand. Rotation about the C₄–C₅ bond reduces the steric interference between the vinyl substituent and the mesityl group (Figures 2b and S11b) and allows for a decrease in the C_meso–Fe–N_im angle from 112.1° (anti) to 106.9° (syn), pulling the iminyl fragment away from the cleft. The steric pressure between the α-azido aryl unit and the cleft components is thus minimized, affording a lower energy pro-syn TS, in line with the experimentally favored syn diastereomer. Ortho aryl functionalities induce additional strain to furnish the syn product exclusively, whereas the remote meta and para substituents have no effect on the d.r. assuming no change in TS organization. The modest d.r. found for substrates 2p–2q can also be explained: the planar phenyl ring interacts similarly with the mesityl unit, whereas a vinyl group does not project far enough into the cleft to enhance diastereocontrol.

The insight gained from DFT analysis allowed us to systematically modify our catalyst to improve diastereoselectivity. Among the features proposed to dictate selectivity, the larger Taft steric parameter^[11] of Me versus Cl suggests that the bridgehead mesityl group may be more beneficial than the 2,6-Cl₂-C₆H₃ group in 1a. With respect to the cleft, the adamantyl moieties provide a balance between productive reactivity and diastereoinduction and cannot be altered without compromising the desired chemistry. However, the ancillary anionic ligand can affect the steric profile. To this end, we sought to examine if a series of electronically and sterically modular alkoxide and aryloxide complexes afforded superior catalysis and diastereoselectivity.

Addition of KOR reagents to 1b did not cleanly produce monosubstituted complexes, as the anionic species (^AdL)Fe-(OR)₂(μ²-K) were readily formed. Alternatively, metathesis from the triflate complex (^AdL)Fe(OTf)(THF)₂ (5 ; see the Supporting Information) provided the alkoxide or aryloxide solvento-adducts (^AdL)Fe(OR)(THF) (Scheme 2, 6–13). All paramagnetic complexes 5–13 were analyzed by ¹H NMR and combustion analysis to assess purity and single-crystal X-ray diffraction to establish a coordination sphere typical of high-spin ferrous dipyrrin complexes. Although the individual bond metrics are unremarkable, the size of the OR group changes the geometry at the Fe center from trigonal monopyramidal for iron phenoxide complexes 8–13 (Figure 1b) to distorted tetrahedral for the alkoxide variants 6–7 (Figures S3, S4), in which the large alkoxide ligand resides outside the cleft created by the two adamantyl units.

Scheme 2.

Synthesis of catalyst variants.

With catalysts 6–13 in hand, we began canvassing the respective performance for the cyclization of test substrates 1-azido-1-(4-chlorophenyl)-hex-5-ene (2b) (catalysts 6–8) or 1-azido-1-phenyl-hex-5-ene (2a) (catalysts 9–13) (Table 2). Although 1a yielded modest d.r. (3:1), employing the OC-(CF₃)(Me)Ph alkoxide variant (6) improved the selectivity to 13:1; however, introduction of the larger alkoxide OC-(CF₃)Ph₂ (7) resulted in both diminished conversion and d.r. In fact, lowering the steric demands of 6 by using phenoxide 8 provided the highest diastereoselectivity (>20:1) for the syn diastereomer. Ortho substitution on the aryl unit does not alter the observed d.r., but the yield diminishes with increased steric bulk of the ortho position. Finally, para-substituted phenoxide complexes preserve d.r. relative to 8, consistently offering the best yields (66–82%).

Table 2.

Catalyst screening as a function of ancillary ligand.

R	Complex	Yield [%]	d.r.[c]
C(CF3)(Me)Ph	6	54[a]	13:1
C(CF3)Ph2	7	26[a]	6:1
Ph	8	67[a]	>20:1
o-C6H3Cl2	9	42[b]	16:1
o-C6H3Br2	10	42[b]	>20:1
o-C6H3iPr2	11	49[b]	>20:1
p-C6H4F	12	66[a]	>20:1
p-C6H4OMe	13	82[a]	>20:1

Catalysts 6–8: reaction with 1-azido-1-(4-chlorophenyl)-hex-5-ene (2 b); 9–13: reaction with 1-azido-1-phenyl-hex-5-ene (2 a);

^[a]Isolated yields.

^[b]1H NMR yields.

^[c]syn:anti.

Both catalysts 8 and 13 afforded similar yields with problematic substrates (Table S1), therefore, the more readily accessible complex 8 was used to further explore the scope of stereoinduction. Compared to 1a, catalyst 8 displays uniformly elevated isolated yields and remarkable selectivity for the syn diastereomer (>20:1) regardless of the aryl substitution (Table 1). Unlike the alkoxide variants, the phenoxide unit can engage in π interactions with the iron center; these interactions may lock the phenoxide ligand within the Fe–dipyrrin plane and induce more steric clashes with the substrate. Furthermore, given our recent study on the role of ancillary ligands in ferric dipyrrin complexes,^[13] we propose that the enhanced Fe–OPh covalency stabilizes the Fe–iminyl intermediate, improving cyclization yields. The only exceptions for this substrate class are entries 2n (d.r. 11:1) and 2o (d.r. 10:1). The benzylic C–H bond proximal to the iminyl moiety (2n) may promote other reactions, decreasing both the d.r. and pyrrolidine yield. The reason for the diminished selectivity of the naphthyl-containing substrate remains unclear. In the absence of sterically demanding α-azido groups (2q–2r), d.r. are low, corroborating the proposed steric role of the anionic ligand.

The results described demonstrate the viability of catalytic diastereoselective N-heterocycle synthesis via direct C–H amination. DFT studies guided a systematic catalyst design leading to an optimized diastereoselective cyclization by using a phenoxide ancillary ligand. This ligand is likely involved in π-bonding with the iron center and resides in the Fe–dipyrrin plane, thus creating a more directional steric profile than the chloride or alkoxide variants examined. The Fe–OPh interaction may also contribute to the improved catalytic performance and highlights the interplay of electronic and steric factors required for robust catalysts.