N-Heterocycles are essential to drug development. Among the
N-heterocyclic pharmacophores, piperidines and pyrrolidines figure prominently and rank No. 2 and 3 respectively in the FDA approved drugs in the past decade [
1]. While a large percentage of drugs and natural products contain a multiply (>2) substituted piperidine or pyrrolidine core (
Fig. 1A), efficient protocols to assemble these cores are nonetheless nontrivial. For instance, synthesis of vibegron, an anti-overactive bladder agent, requires a long sequence of >10 steps. This multi-substitution issue has significantly limited the exploration of piperidine and pyrrolidine chemical space for drug development campaigns.
The cyclic nitrone scaffold is found in many important functional molecules, including immuno-spin trapping agents [
2,
3], biorthogonal probe [
4], natural products, therapeutic agents [
5–
7], and acts as a highly useful 1,3-dipole in dipolar cycloaddition reactions and a key intermediate to important 5- and 6-membered
N-heterocyclic natural products (
Fig. 1B) [
8–
13]. In contrast to their prominent role, methods for enantioselective synthesis of nitrones are strikingly underdeveloped. During the past few years, Zhang and Zhao [
14], the Kobayashi group [
15] and the Breit group [
16] have made progresses on transition-metal and Lewis acid catalyzed asymmetric intermolecular addition of oxime to olefins. However, catalytic asymmetric synthesis of cyclic nitrones, a key subset of nitrone family [
9], remains highly underdeveloped. The first report on catalytic asymmetric synthesis of cyclic nitrones was reported by Goti and coworkers on the oxidative desymmetrization of 5-membered cyclic hydroxyamine [
17], which obtained only 68/32
er in a single example (
Fig. 1C). Zhang and Zhao
et al. reported in 2019 the first highly enantioselective synthesis of chiral cyclic nitrones
via Cu-catalyzed intramolecular Cope-type hydroamination, but only 5-membered carbocyclic nitrones succumbed to high asymmetric induction [
18]. The Carreira group subsequently reported an elegant Ir-catalyzed highly enantioselective intramolecular allylation of hydroxy-latedalkenyl oximes to deliver a series of 5-, 6- and 7-membered cyclic nitrones with high levels of enantiocontrol (
Fig. 1C) [
19]. To date, however, all enantio-selective syntheses of cyclic nitrones are limited to unimolecular intramolecular versions. Practical, adaptable, and highly enantioselective methods for diversified nitrone synthesis are still in urgent demand.
Pd-catalyzed heterocyclization of
N- and
O-nucleophiles onto pendant alkene has been established as an important tool for the syntheses of various important
N- and
O-heterocycles [
20–
22]. While significant progresses have been made in enantioselective aza-Wacker cyclization [
23,
24] and carboamination [
22,
25,
26] of alkenyl amides in the past decade, the related process of
N-cyclization of alkenyl oxime toward chiral nitrone synthesis remains elusive. Due to the challenging chemoselectivity issue caused by the oxime functionality (N
vs. O), as well as the thermodynamic unfavourability of nitrone due to charge-separation, development of enantioselective synthesis of cyclic nitrone
via alkenyl oxime poses additional challenges as compared to alkenyl amides. For instance, competitive
O-cyclization can intervene, for which the highly enantioselective versions have been realized recently by the Zhao [
27] and Zhang groups [
28]. In contrast, enantiocontrol in
N-cyclization toward nitrone formation remains an unsolved challenge [
29]. Furthermore, an additional challenge comes from competing Heck reaction of the alkene with the electrophile. Herein we disclose the development of highly enantioselective synthesis of 5- and 6-membered nitrones from Pd-catalyzed cyclization coupling of
γ,
δ- and
δ,
ε-alkenyl oximes with (hetero)aryl and alkenyl electrophiles (
Fig. 1D). This protocol enables a facile access to multi-substituted piperidines and pyrrolidines upon facile reductive N‒O bond scission and subsequent nucleophilic addition.
Our study commenced by optimizing the Pd-catalyzed asymmetric cyclization of
γ,
δ-alkenyl oxime
1a with 4-tolyl iodide. Initial scouting experiments revealed that a suitable combination of base and media is critical for the enantiocontrol, with Cs
2CO
3 and methyl
tert-butyl ether (MTBE) being optimal (Tables S1 and S2 in Supporting information). Reactions with a series of commercial monodentate and bidentate ligands were then examined, but were found unsatisfactory with respect to enantiocontrol (entries 2–5,
Table 1). In contrast, a type of Sadphos ligand developed by the Zhang group [
31] turned out to be highly competent, retrieving cyclic nitrone
3aa in 96.7/3.3
er. Notably, using Pd(dba)
2 as precatalyst appreciably facilitated the conversion, allowing the reaction to be completed at 50 ℃ within the same time span with concomitant improvement of enantioselection (entries 6–9,
Table 1). This suggests that reductive elimination should be turnover limiting. We substituted PhBr or PhOTf for PhI, and the results were not as satisfactory (entries 10 and 11,
Table 1).
Our attention then shifted to the coupling of
δ,
ε-alkenyl oxime to deliver 6-membered cyclic nitrones. The above standard conditions for 5-membered nitrone synthesis with
L1 was found to give poor enantioselectivity for
4na (entry 1,
Table 2). After extensive trials (Table S3 in Supporting information), we found that SadPhos ligands bearing a bulkier PAd
2 (Ad = 1-admantyl) analogues [
31] could enable the reaction of oxime
1n to deliver the 6-membered nitrone
4na in higher enantioselectivities compared to its PCy
2 counterpart, but in a low yield (entry 2,
Table 2). Moreover, we found that changing the Ar moiety posited
α to the nitrogen from 3,5-di-
t-Bu-4-MeOC
6H
2- (
L6) to 3,5-di-
t-Bu-4-
n-BuOC
6H
2- (
L7) had a positive effect on the enantioselectivity (entry 2
vs. 3). When ArI was replaced with ArBr and the temperature was lowered to 90 ℃, the yield was not improved (entry 4). To our surprise, the yield can be significantly improved by increasing the catalyst loading (entries 5 and 6,
Table 2), which we attributed to appreciable product sequestration (see DFT studies,
vide infra). Eventually, 4 mol% catalyst loading enabled to produce the product
4na in a good enantiomeric ratio (95/5
er) and a high yield of 90% at 80 ℃ with aryl bromide as coupling partner (entry 7,
Table 2).
The scope of the enantioselective synthesis of 5-membered nitrones summarized in
Fig. 2A showcased the robustness of the conditions to accommodate substituents of both electronic and steric nature. Substitution at various positions of the aryl groups of the alkenyl oxime have little effect on the efficiency or enantioselectivity (
3ba-
3ia), except for 2-thienyl, which led to a compromised 73.1/26.9
er likely due to competitive coordination with the catalyst which changed the ratio of
anti/
syn aminopalladation route (
vide infra). Single crystals of
3fa were obtained which enabled determination of its absolute configuration to be (
R) by X-ray diffraction studies. Spirocyclic nitrone
3ja was obtained in a good yield and 96.9/3.1
er. The reaction of
α,
α-disubstituted oximes led to
3ka and
3la in high
ers. To our delight, the reaction of the prochiral bis-homoallylated oxime retrieved the corresponding cyclization products
3la/3la' in 4.8:1
dr as readily separated diastereomers, both with high enantioselectivity. Alkyl substitution at the internal carbon of the alkene did not affect the reactivity, and remarkably, a higher level of enantiocontrol was realized (
3mb).
Next, aryl halides bearing electronically distinct groups were examined and found compatible (
3ab-
3al). Remarkably, the sterically congested
ortho-substitution is also tolerated (
3am and
3an). Fused aromatic hydrocarbons such as naphthalene and pyrene (
3ao-
3ap), and heterocycle such as quinoline can also be tolerated (
3aq). Alkenyl halides were also found competent, as showcased in the comparable reactivity and selectivity in nitrone
3ar (59%, 95/5
er). The use of 1-alkynyl bromide as the electrophile retrieved nitrone
3as in a low yield with only nominal enantioselectivity, indicating the importance of an electron-rich π-system for the enantiocontrol. In this case, switching to a XantPhos-hybrid ligand, PC-Phos
L8 [
30] enabled high enantiocontrol, albeit the yield still needs further optimization (Table S4 in Supporting information).
We then explored the scope of the 6-membered nitrones with a panel of substrates (
Fig. 2B). Firstly,
δ,
ε-alkenyl oximes with different aryl moieties substituted at the para and/or meta position were well-tolerated, affording the desired products in good yields and excellent enantioselectivities (
4oa-4sa). The absolute configuration of the products was determined to be (
S) by the crystal structure of
4na, which is the same sense of enantiocontrol with the 5-membered nitrones. The alkyl substituted alkenyl oxime
1t can also deliver the corresponding nitrone
4ta in a moderate yield and comparable level of enantiocontrol. Likewise, spirocyclic nitrone
4ua was obtained in a good yield and higher er of 93.9/6.1. Racemic oxime
1v gives the diastereoisomers
4va/4va' with 85% yield in 1:1
dr, both were formed in a high enantio-selectivity. The
O-tethered substrate was also competent, affording tetrahydro-1,4-oxazine
N-oxide
4wa in a high yield and 92/8
er. Dihydroisoquinoline
N-oxide
4xa, featuring an aldehyde-derived nitrone moiety, was also successfully synthesized in 33% yield and a good enantiomeric ratio of 86/14. Indole-derived nitrone
4ya can also be successfully prepared in a good yield yet with inferior enantiocontrol as compared to carbon-tethered ones. The relatively poor performance of these heterocycle-tethered and benzo-derivatives can be attributed to the rigidity of the tethers that prevent a better alignment of the two approaching moieties (oxime and alkene) in the cyclization transition structures, which may make the inner-sphere migratory insertion mechanism more favorable. Next, aryl halides bearing substituents of distinct electronic properties were tested under the optimized conditions, and the corresponding products (
4nb-4ni) were obtained in 33%~85% yields with excellent enantioselectivities (89.5/10.5 ~ 96.5/3.5
er). Electron-rich heteroaromatic halides are also tolerated (
4nj and
4nk). The aryl triflate derived from estrone proceeded smoothly, and the desired product
4nl was furnished in 93.5/6.5
er, albeit with a low yield due to competing Heck reaction. The aryl bromide with 3,4-disubstitution and a chiral tetrahydrofuran moiety also showed excellent reactivity, and the corresponding nitrone
4nm was afforded in 92.4/7.6
er. Unfortunately, the substrate bearing an alkene moiety with a substituent at the internal carbon did not result in formation of the expected 6-membered nitrone
4nn, probably due to unfavored cyclization step.
Revelation of the stereochemical course of the cyclization is key to development of highly enantioselective protocols, as
syn- and
anti-type cyclization processes usually exert distinct requirements for chiral discrimination. Mechanistic studies on Pd-catalyzed cyclization of alkenyl amides revealed that both
cis-amino-palladation and aza-Wacker type
anti-aminopalladation are possible [
28]. To discriminate these two pathways, stereo-defined deuterium labeled alkenyl oxime
1a-d was prepared and subjected to the optimized reaction conditions (
Fig. 3A). The product, deuterated nitrone
3am-d, was subsequently reduced by LiAlH
4 to the corresponding
trans-configured hydroxyamine, which underwent reductive N‒O cleavage to secondary amine
6a-d. Partial configurational scrambling was observed at this step, likely due to elimination to the imine followed by reduction. Finally, a Pd
0-catalyzed C‒N coupling delivered the cyclization product, benzopyrrolizidine derivative
7a-d. An inner-sphere process is expected to result in deuterated benzopyrrolizidine
7a-d with the deuterium atom posited
trans to H
α, while the outer-sphere mechanism would lead to a
cis-relationship, and
vice versa. Comparison of the
1H NMR of
7a and
7a-d, as well as the 2D NOESY spectrum of
7a indicated that
7a-d has a
trans-arrangement of D and H
α, suggesting that the reaction underwent exclusive
trans-aminopalldation
via outer-sphere mechanism (Figs. S1 and S2 in Supporting information). The deuterated substrate
1a-d was then used as a probe to examine the stereochemical outcome of the cyclization under different conditions. As shown in
Fig. 3B, the nature of base and ligands play an important role in the cyclization pathway.
For 5-membered cyclic nitrone, when the strong base
tBuONa was used, the reaction proceeds through both inner-sphere migratory insertion pathway and outer-sphere
anti-aminopalldation, with the former being slightly more favored, regardless of the ligand being used (slots a~c,
Fig. 3B). Under the catalytic conditions of Pd/
L1, the reactions under different bases show that as the basicity of the base decreases, consistently higher percentage of outer-sphere
anti-aminopalldation took place, which is also in line with the increase of enantiocontrol (slots c~e,
Fig. 3B). In contrast, for the synthesis of 6-membered cyclic nitrone, the pathway selectivity favors outer-sphere
anti-aminopalldation even with NaO
t-Bu as the base, and ligand also plays a role (slots f~h,
Fig. 3B). Yet, under the catalytic conditions of Pd/
L6, use of a weaker base also increased the
anti/
syn cyclization selectivity, albeit to a smaller extent (slots h~j,
Fig. 3B).
To gain insights into the mechanism regarding the origin of enantiocontrol and rate-determining step(s), DFT calculation was performed first to explore the potential energy surface of the 5-membered nitrone formation reaction (
Fig. 4A). Our calculation reveals that oxidation addition of Pd
0L* (L* =
L1) with PhI has a low barrier (4.1 kcal/mol) and a net free energy downhill of 25.0 kcal/mol. For the subsequent ligand exchange of Ph-PdIL
* with alkene moiety of the Cs-countered oxime anion, we sampled the trajectories of the alkene
via both
Re- and
Si-attack patterns, and for both attacking patterns, two representative approaching poses of the oxime were evaluated (denoted by a and b; wherein the poses having the oxime Cs part being proximal to the Pd‒O moiety are named with a, and those having oxime Cs part being proximal to the Pd-Ph moiety are named with b). The four cyclization transition states TS2 feature long C‒N bonds (2.2–2.8 Å, Fig. S6 in Supporting information), among which the
Si-b pathway is most energetically favored with an activation energy of 19.3 kcal/mol (
Si-TS2-b), leading to (
S)-configured cyclization product
Si-Ⅳ-b. In comparison, the lowest energy
Re-face attacking transition state (
Re-TS2-b) is disfavored by 4.2 kcal/mol. The irreversible nature of the cyclization step, as inferred from the 27.2 kcal/mol activation energy of the reverse event (
Si-TS2-b
vs Si-Ⅳ-b), suggests that the cyclization step is enantio-determining. As it is experimentally observed that electron deficient dba (dibenzylidene acetone) facilitates the reaction (
vide supra), the reductive elimination step, with a slightly higher barrier of 21.2 kcal/mol (
vs. 20.3 for the cyclization), should be involved in the rate-limiting step(s). In addition, the reductive elimination step of the
Re-
b pathway also has a higher barrier compared to the
Si-b pathway (
Re-TS3-b
vs. Si-TS3-b). Furthermore, the transition structures of the two
cis-aminopalladation pathways were also located and found to be energetically less favored than the above two lowest
trans-aminopalladation TSs, and are very close in energy (Fig. S5 in Supporting information), consistent with experimental observation that the more
cis-aminopalladation is involved in the cyclization, the lower the enantiocontrol was observed.
Interaction region indicator (IRI) analysis [
31] revealed key differences in the weak interaction regions within the two cyclization TSs arising from the two interacting fragments,
i.e., the oxime Cs and the Ph-PdL* moiety (
Fig. 4B). The two notable regions are the cation···π interaction between the Cs
+ and the Pd-Ph, which is present in both TSs but looks stronger in
Si-TS2-b (regions Ⅰ/Ⅰ'), as well as a dispersion interaction region between the alkyl tether and the ligand's P-Cy moiety which is much more extensive in
Si-TS2-b but only sporadic in
Re-TS2-b (regions Ⅱ/Ⅱ').
The free energy profile of the reaction of the 6-membered nitrone formation with Ligand
L7 revealed a significantly different scenario (
Fig. 4C). The large P-substituents makes the cyclization step not only basically non-selective, but also even slightly favor
Re-face attack (which leads to the un-favored
R-product antipode). In contrast, the reductive elimination step shows a strong preference for the
Si-b pathway (
Re-TS3-a_6
vs. Si-TS3-b_6), leading to the experimentally favored product antipode. At the experimental temperature (373 K), the cyclization step should be more reversible, resulting in both cyclization and the subsequent reductive elimination step to be engaged in the enantio-control, while cyclization is likely rate-determining. In contrast to the 5-membered nitrone formation, both reductive elimination TSs feature a broken Pd‒O bond, which may expedite the reductive elimination. In
Si-TS3-b_6, the distance between Pd and O atoms is significantly longer (3.86 Å
vs. 3.12 Å in
Re-TS3-a_6), which may play a role in facilitating the reductive elimination. It is also noted that the (
R)-product, albeit being the unfavored stereoisomer, forms a much more stable complex with the catalyst, (
R)-V, which may rationalize the requirement of higher catalyst loading (entries 5~7,
Table 2).
IRI analysis of the two selectivity-determining reductive elimination TSs (
Fig. 4D) shows that the oxygen is enclosed in a dispersion region between the aliphatic linker of the oxime and the
t-Bu of the ligand's bulky aryl chiral side arm in
Si-TS3-b_6 which does not have a counterpart in
Re-TS3-a_6 (regions Ⅲ/Ⅲ'). In addition, the C—H···π interaction between one of the P-adamantly moieties and the Pd-pH adopts head-on pose in
Si-TS3-b_6, which could be more efficient compared to the side-on pose in
Re-TS3-a_6 (regions Ⅳ/Ⅳ
'). These two interaction regions may contribute to the large energy difference for the enantio-differentiation. It is thus likely that, as inferred in
Fig. 4B, ii, the lower overall enantiocontrol in formation of the 6-membered nitrone may be caused by poor pathway selectivity, or a more pronounced sabotage by the minor non-selective inner-sphere pathway. Combined with the information in 5-membered nitrone formation, the calculation agrees well with experimental observation that the aryl moiety of the aryl halide electrophile is important for the enantiocontrol.
We then performed preliminary downstream elaboration of the chiral cyclic nitrone products. First, 5-membered nitrone
3aa undergoes dipolar cycloaddition with methyl propionate to afford bicyclic dihydroisooxazole
8aa in 94% yield and high diastereoselectivity (
Scheme 1A). In addition, the 6-membered nitrone
4na can be facilely deoxygenated with TiCl
3 and reduced to the chiral pipyridine
10na in an overall yield of 70% with no encroachment of stereoselectivity (
Scheme 1B).
In summary, we have developed a highly enantioselective synthesis of various 5- and 6-membered nitrones via Pd-catalyzed coupling cyclization of γ,δ- and δ,ε-alkenyl oximes with (hetero)aryl, alkenyl and alkynyl oximes with tailored sulfinamidephosphine ligands. The products could be facilely transformed to important multi-substituted chiral N-heterocycles, such as bicyclic dihydroisooxazole and cyclic imines. Reactions with isotope-labeled substrate revealed a highly base and solvent dependent cyclization pathway distribution which significantly impacts enantiocontrol. Under the optimized conditions, the anti-aminopalladation cyclization prevails and the bidentate ligand exerts high differentiation for the enantiocontrol while the non-selective cis-aminopalladation is minimized. DFT calculations revealed a more demanding cyclization step as moved from 5-membered annulation to the 6-membered one, and provided insights into the factors for the enantiocontrol in both senarios. The addressing of the multiple selectivity issues for enantiocontrol has significant implication for related heterocyclization processes and asymmetric cyclization reactions at large.