Nitrogen-containing heterocycles (
N-heterocycles) have been identified as the most commonly employed subunits in the approved molecular entities, as evidenced by the statistical analyses that 88% of all the approved small molecule drugs contain
N-heterocycles during 2015 to 2020 [
1–
6]. Among them, the four membered azacyclobutanes also known as azetidines display unique pharmacokinetic properties as well as enhanced bioavailability that draw attentions from medicinal chemists [
7–
10]. In addition, complex natural products that feature multi-substituted azacyclobutanes as core motifs pose grand challenge in azetidine formation and stereochemical control for synthetic chemists (
Fig. 1a) [
11–
18]. In fact, synthesis of azetidines proved to be more difficult than its 3- to 6-membered counterparts following traditional cyclization mode, due to unfavorable confirmation and the ring strain (25.4 kcal/mol) [
19–
25]. The transition metal catalyzed C—N coupling [
26], aza [1.1.0] ring fragmentation [
27] and azaridine expansion [
28] require directing group and/or complicated substrate prepration [
29,
30]. There is a drastic lack of powerful strategies and methodologies in constructing stereo-controllable multisubstituted azacyclobutanes, not to mention stereochemical editing of the azetidines [
31–
34]. The traditional ultraviolet mediated aza Paterno-Buchi reaction, reported by Tsuge in 1968, offered a straightforward pathway to assemble azacyclobutanes from imines and alkenes (
Fig. 1b) [
35,
36]. Although many efforts had been devoted to improving efficiency and expanding C=N scope, huge room still remains to be improved especially under the context of visible photocatalysis and key problems untackled,
e.g., stereochemical control.
Visible light mediated aza-[2 + 2] cycloaddition reaction was first disclosed by Schindler group [
37,
38], using methyl oximes as C=N motif. This pioneering work was continued by You [
39], Bach [
40], Brown [
41], Luo [
42] and Zhong groups [
43]. However, the C=N motif was largely limited to cyclic "imines" with push (electron-donating groups, EDG)-pull (electron-withdrawing group, EWG) activating handles or using quinoxalinone (
Fig. 1c) [
36,
44]. Simultaneously,
N-heteroarenes represent a large group of compounds that contain C=N bonds, which are completely neglected in visible light mediated aza-[2+2] cycloaddition. The benefits of using
N-heteroarenes are quite obvious because they feature: (i) feedstock availability; (ii) inexpensive; (iii) "C=N" motifs come with various substitution, as well as in versatile ring systems; (iv) easy to derivatize with further transformations. These merits showed that
N-heteroarene are very attractive substrate to be used in aza-[2+2] reaction. Once realized, aza-cyclobutane bearing three contiguous stereocenters can be obtained in one step, which would also offer the opportunity to "edit" current drugs entities with suitable C=N handles for de-planarity [
45–
47].
Keenly attractive as it can be, grand challenges and major limitations must be confronted and addressed. Ultraviolet irradiation is a prerequisition for traditional aza-[2 + 2] cycloaddition implied that most C=N bonds need high transferring energy (
ET) to be excited to their triplet states from photosensitizers or UV light source. For example, the common
N-heteroarene such as benzothiozole, benzoxazole and benzoimidazole requires
ET = 74.2, 81.2 and 94.1 kcal/mol, respectively to be excited to their corresponding triplet state [
48]. Contrastingly, the currently known visible light mediated photosensitizers' triplet energies are all below 73 kcal/mol, which rendered direct triplet excitation of the
N-heteroarenes unfeasible (
Fig. 1d) [
49,
50].
Our strategy is taking advantage of the relatively easy to excite diene moiety and speculate that the resulting diradicals would trigger a dearomative aza-[2 + 2] cycloaddition to achieve the goal (
Fig. 1e). It is not without challenge to realize the design,
e.g., (a) how to control chemoselectivity of diene excitation; (b) the competitive 4
π cyclization of diene and/or dearomative [4 + 2] cycloaddition could be interfering side reactions; (c) whether the triplet energy of diene sufficient to dearomatize the
N-heteroarene. Last and the most challenging task is how to control the stereoselectivity of the multi-substituted azetidine, provided that dienes are notorious for their difficulty to control diastereoselectivity in aza-[2 + 2] reaction [
51]. Herein, we disclose the discovery of a protonic solvent accelerated reversible dearomative aza-[2 + 2] cycloaddition between dienes and
N-heteroarenes under white light, allowing stereochemical-selectively synthesis of multisubstituted aza-cyclubtanes.
We started with
sub-A (
Scheme 1a), which can be prepared through amide coupling between the
N-(thiazol-2-ylmethyl)aniline and cinnamic acid (see Supporting information for details). When irradiating
sub-A with blue LED in the presence of 2 mol% of photosensitizer (PC), no desired product can be detected except for starting material recycling (
sub-A: 78% and 10% yield, respectively) and olefin isomerization (
prod-Ⅰ: 22% and 80% yield, respectively in entries 1 and 2). The predicted oxidative 6
π cyclization
prod-Ⅱ could be detected in 76% yield, along with 16% of
prod-Ⅰ (entry 3). The olefin isomerization implied that the energy transfer took place, but the triplet diradicals' life time is too short. And thioazole dearomatization was difficult compared to the 6
π cyclization as the high
ET photosensitizer,
PC-3 only triggered the latter reaction.
With the above information in hand, we designed dienamide
1a (
Scheme 1b) [
52–
56]. Delightedly,
1a proved to be a suitable substrate with MeCN under blue LED irradiation with 2 mol% of
PC-2 as photosensitizer (entry 4). The desired aza-[2 + 2] products were obtained in 93% overall yield, albeit with less
dr control (
2a/3a = 1/1.7). Several photosensitizers (
PC-2~5) and solvents were tested (entries 5~10) and the Ru(phen)
3(PF
6)
2 (
PC-4) in acetonitrile proved to be the optimal choice, yielding combined aza-[2 + 2] product
2a (31%) and
3a (69%) with very high level of mass balance (entry 6). Surprisingly, we observed that the diastereoselectivity of aza-[2 + 2] reaction could be altered and reached to 8.8:1, when performed in CH
2Cl
2 (entry 9). The reason is not very clear and was largely attributed to solvent effect on the triplet state life time. Following this logic, it was found that when adding trifluoroethanol (TFE) as co-solvent, the diastereoselectivity of the dearomative aza-[2 + 2] could be increased to 11:1 (entry 11). Further elongating reaction time to 60 h, the dr ratio can be reached to 15:1 without affect reaction efficiency (
2a in 90% and
3a in 6% yield, entry 12). Delightedly, the major C1
β-configured isomer
2a could be isolated in 94% yield even under irradiation of white light (entry 13).
Towards this end, the optimal conditions were successfully established for the dearomative aza-[2 + 2] cycloaddition. The reaction scope was explored. A broad series of dienamide coupled
N-heteroarene substrates with different steric and electronic properties have been examined (
Scheme 2). Delightedly, good to high yields of diazabicyclo[3.2.0]heptanones were obtained, with good diastereomeric control. It was found that changing the steric of N substituents of dienamides (
1a-1d) does not affect the efficacy, and 85%~94% yields and single diastereomers (
2a-2d) were obtained. Gratifyingly, X-ray crystallographic analysis was conducted with
2b and the stereochemistry was verified. When
N-substituents were switched to aryl groups (
2e-2i), the aza-[2 + 2] reaction's efficiency were decreased (62%−83% yield), while maintaining good dr control (
dr > 10:1). Considering that the
N-benzyl substituted product
2j could be isolated in 87% yield as a single diastereomer. We speculated that
π-conjugation with
N-aryl moieties might be blamed for decreased efficiency of this reaction. Two substrates were designed
1k and
1l to test our hypothesis, since the sterically hindered 2,6-dimehtylphenyl and pyrrolyl groups will not form
π-conjugation with the dienamide part thanks to the atropisomeric strain. As expected, the desired products
2k and
2l were isolated in 81% yields for both with decent
dr control (10:1 and 7:1
dr, respectively).
An investigation on the substituents effects on diene motif was further conducted. When changing diene terminal substitutions, no obvious para-substitution effect could be observed. Only single diastereoisomers were found and the isolated yields ranged from 67% to 91% yield (2m-2p). It was noted that furyl group was well tolerated under such redox-neutral conditions and desired product 2q was obtained in 75% yield as a single isomer. It is worthy to point out that even trienamide substrate 1r could be successfully converted to the desired aza-[2 + 2] cycloadduct 2r in very high chemo- and diastereo-selectivity (73% yield, single isomer). The alkyl substituent at α-position of dienamide was also investigated (1s). It was gratifyingly to find the ethyl-substituted products 2s in good yield (94%), whose relative configurations were confirmed by X-ray crystallography. Moreover, vicinal quaternary carbon centers could be established (2t in 45% yield) with tetra-substituted olefin 1t and 2u was obtained in 30% yield due to unexpected decomposition. Interestingly, the diastereoisomers (3v and 3w) could be isolated as the only products (92% and 85% yield, respectively), if TFE was not used as cosolvent. The relative configuration of 3w was confirmed by X-ray crystallography.
Furthermore, the scope of
N-heteroarenes was investigated. It was found that benzoxazole (
1x), benzoimidazole (
1y) even quinoline (
1z) could be successfully dearomatization and afforded the desired aza-[2 + 2] cycloadducts
2x-2z in 82%−95% yield. Notably, the isolation of
3z and its epimer
2z (95% yield combined in 3.7:1
dr) represented the first example of photocatalytic dearomtization of quinoline in a [2 + 2] mode instead the previously reported [4 + 2] cycloaddition type [
57–
64]. Certainly, a few substrates are reluctant to undergo the intended aza-[2 + 2] reaction like
1ab and
1ac (only starting materials were recycled). The thioazole substrate (
1ad) was found to undergo a hetero Diels-Alder cycloaddition affording
4c as the only product.
Our next mission is to elucidate the reason why adding TFE as co-solvent is able to control the diastereoselectivity. To probe this, we first conducted two kinetic monitoring experiments using
1a. It was found that within the presence of TFE, the kinetic product
3a was formed as the major isomer (72% yield, see Supporting information for details) in 2 h. And then the amount of
3a was decreased with time until < 7% yield after 36 h of irradiation). Meanwhile, the yield of thermodynamic product
2a was correspondingly increased to 92%. In addition, their yields of
2a and
3a remained unchanged even after 60 h irradiation under blue LED light (
Scheme 3a). On the other hand,
1a was also subjected to TFE-free standard conditions and the ratio between product
3a and
2a was slowly eroded from 2.4:1 to 1.2:1, featuring kinetic isomer
3a as the major product. These indicated that an inter-conversion from
3a to
2a could be existing or a reverse aza-[2 + 2] cycloaddition could be operating within the presence of TFE. We observed that only trace amount of kinetic product
3a could be detected by subjecting
2a under standard conditions. Contrastingly, 90% yield of
2a was isolated if exposing
3a under standard conditions. This revealed a thermodynamic control and implied why
2a could be observed as a single diastereomer (
Scheme 3b). Then we treated
2a and
3a with standard conditions but without light irradiation.
1a could be isolated in 12% and 98% yield, respectively (
Scheme 3c). These data revealed that a reverse aza-[2 + 2] cycloaddition was triggered under acidic additive (TFE forming H-bonding with the tertiary amine in
2a/3a). Contrastingly, a cyclobutane analogue (
2ag/3ag) would not undergo such a process, even within the presence of TFE (
Scheme 3d). Toward this end, we have elucidated the mechanism for diastereoselective control of our aza-[2 + 2] reaction. To validate that the dienamide was key for photosensitizer triplet energy transfer, we conducted a Stern-Volmer quenching experiment with substrate
1a, its subunits dienamide
1ae and 2-methyl benzothiazole
1af. The slope values indicated that dienamide moiety (
1a and
1ae) is key for energy transfer from the excited triplet state of photosensitizer, while the benzothiazole moiety (
1af) should not be responsible (
Scheme 3e).
A rate-differential protonic solvent-mediated plausible mechanism were proposed (
Scheme 4a).
1a was excited to the triplet state
1aT from its ground singlet state
1aS via EnT process within the presence of photosensitizer under irradiation.
1aT underwent a dearomative addition leading to diradical
INT-1T, which upon intersystem crossing (forming its open shell singlet state
INT-1OSS), led to major kinetically favored aza-[2 + 2] product
3a and thermodynamical
2a as minor products. However,
3a could be quickly (
fast in
Scheme 4a) converted back to
1a under acidic solvents such as TFE, probably due to hydrogen bond activation of azetidine that weakens C—N bond. On the contrary,
2a displayed relative stability (very slow in
Scheme 4a) under acidic solvent conditions (
Scheme 4b). Such a rate-differential retro [2 + 2] reaction combined with the fast forward dearomative aza-[2 + 2] cycloaddition led to net accumulation of thermodynamic
2a, and realized diastereoselectivity.
The robustness of the current aza-[2 + 2] reaction were tested by subjecting complex ingredients in the reaction solvents. These "liquids" were used directly after commercial purchase (
Fig. 2). We found adding commercially available American bourbon or rum did not affect the efficiency of this reaction yielding
2a/3a in 98% and 97% combined yield, respectively. The different dr ratio might imply that there are certain acidic ingredients in American bourbon that play a similar role as TFE. Surprisingly, the reaction yielded quantitative yield of
2a/3a combined, even if the coffee concentrate were used as co-solvent, which are known to contain complex alkaloids, organic acids and lipids.
With the confidence gained from the above test, we set off to explore the aza-[2 + 2]'s synthetic applicability. Since we are able to access the densely decorated azetidines bearing contiguously stereocenters, it would be ideal to do ring expansions, which allow us to access the more commonly encountered aza five- and six-membered rings (
Scheme 5). We attempted a Meisenheimer rearrangement starting from azetidine
2a, the desired isoxazol product
7 could be obtained in 91% yield. Then a scaffold hopping from vinyl azetidine
5 to piperidine
8 (70% yield) was realized through photoredox mediated reductive quenching of
PC-4 with compound
5. The net [4 + 2] cycloadduct proved practicality in forging piperidines from dienamide. Additionally, the C—N bond could be formally disconnected and inserted by "C=N" bond or "C=O" bond of TsC=N=O, leading to the formation of
9 (33% yield) and
9′ (56% yield). The structure of
9 was unambiguously verified by crystallography. The sulfur atom could be removed by treating compound
2a with catalytic amount Co
2(CO)
8 under CO atmosphere. A dihydro carbazole skeleton can be formed in 55% yield. Last but not least, a selective oxidation was performed with
2a using hydroperoxide, obtaining benzothiazole dioxide
6. It allowed sequential oxidation of azetidine
2a possible.
In summary, we have disclosed the first example of a proton coupled reversible aza-[2 + 2] reaction, leading to highly diastereoselective azetidines. Feedstock N-heteroarenes (containing C=N bonds) were dearomatized through the aza-[2 + 2] under white light via energy transfer, which provide an important solution to the limited scope of aza Paterno-Buchi reaction. The rate differential retro [2 + 2] process was observed and harnessed to address the long standing stereochemical problems (controlling all three contiguous stereocenters dr > 19:1) of aza-[2 + 2] cycloadditions with diene substrates. Over 26 examples were demonstrated with good efficiency (up to 94% yield). The mechanism investigation and synthetic applications were carried out, providing new perspective in elevating the power of aza-[2 + 2] cycloaddition.
PDF
Full
121
65
关闭全屏
No. 86 Xueyuan South Road, Haidian District, Beijing
010-62662699(Operation and Technical Support)
qkjq@cast.org.cn
