Synthesis of Polycyclic Fused Indoline Scaffolds through a Substrate-Guided Reactivity Switch

Zn(II)-catalyzed divergent synthesis of functionalized polycyclic indolines through formal [3 + 2] and [4 + 2] cycloadditions of indoles with 1,2-diaza-1,3-dienes (DDs) is reported. The nature and type of substituents of substrates are found to act as a chemical switch to trigger two distinct reaction pathways and to obtain two different types of products upon the influence of the same catalyst. The mechanism of both [4 + 2] and [3 + 2] cycloadditions was investigated and fully rationalized by density functional theory (DFT) calculations.


■ INTRODUCTION
Functionalized polycyclic fused indoline frameworks are central molecular architectures in nature and pharmaceuticals. 1 As one of the indolines, C2,C3-fused indolines 2 have attracted extensive research effort over the past decades because scaffolds of this type lead to relatively rigid structures that might be expected to show substantial selectivity in their interactions with enzymes or receptors. 3 Representative naturally occurring polycyclic indolines such as vincorine, minfiensine, gliocladin C, kopsnone, pleiomaltinine, and communesin F are shown in Figure 1.
Among the annelated indolines, the pyrroloindoline, pyridazino indoline skeletons and their related structures, can be found in numerous natural bioactive products, marketed drugs, and other functional molecules. 4,5 The desire to build such appealing polycyclic frameworks, particularly those with bridgehead amino acetal C2 carbons, has inspired the development of elegant methodologies over the past several years. Among the reported methods, dearomatization of indoles via cycloaddition reactions 6 has been demonstrated as a reliable approach in converting simple planar aromatic molecules into structurally complex and stereoselective ring systems.
By taking advantage of the unique reactivity of DDs 10 and intrigued by these and our recent findings in the manipulation of indolyl cores, 11 we reasoned that the proper combination of indole and 1,2-diaza-1,3-diene elements might allow us to design a substrate-controlled divergent approach. In this design, DDs would be used as C2N1 or C2N2 units (1,3 or 1,4 dipole synthons) to realize [3 + 2] and [4 + 2] annulation Received: June 23, 2020 Published: August 12, 2020 Article pubs.acs.org/joc reactions of indoles, respectively (Scheme 1). Thus, by tuning the substituents of both substrates upon the influence of the same catalyst, two series of fused indoline-based scaffolds such a s t e t r a h y d r o -1 H -p y r i d a z i n o [ 3 , 4b ] i n d o l e s a n d tetrahydropyrrolo [2,3-b]indoles would be generated with chemodivergence.
The substrate scope with respect to various 2,3-unsubstituted indoles 1a−n and cyclic DDs 2a−h (see the SI for details) was then examined under the optimized reaction conditions, and a variety of tetrahydro-1H-pyridazino [3,4b]indoles (tetracyclic fused ring (6-5-6-6/7/8) systems) 3a−x was synthesized ( Table 1). As shown in Table 1, indoles 1a−n with different electronic characters were suitable for the reaction, with six-membered cyclic DDs giving the relative fused indoline heterocycles 3a−d in moderate to good yields. The Zn-catalyzed [4 + 2] cycloaddition reactions were further extended to seven-and eight-membered cyclic DDs. We were glad to find that the use of seven-membered DDs gave rise to the best results in terms of isolated yields. Also, the wide functional group tolerance was well demonstrated by the fact that both electron-donating  and electronwithdrawing  groups were well tolerated, providing efficient access to the fused indoline heterocycles 3e−s. Interestingly, the use of the 7azaindole substrate also worked well to give the product 3t in 85% isolated yield. The formal [4 + 2] annulation was then extended to DDs bearing cyclooctane, and the reactions furnished the relative products 3u−x with lower yields than those of seven-membered cyclic DDs. Additionally, the generality of the N-terminal protective group on DDs as well as for the N atom of indoles was explored. Remarkably, free N−H indoles were also compatible with this protocol, albeit slightly lower yields were observed, probably owing to the reduced nucleophilicity at C3 and the reduced electrophilicity at C2 of the starting indole (Scheme 1, 3s vs 3p, and 3x vs 3u).
No annulation occurred when five-membered cyclic DD was employed under the optimized reaction conditions (3y, 0%). 12 The relative configurations of cycloadducts 3 were determined by X-ray diffraction analysis of 3e 13 (see the SI for detailed Xray crystallography data), and those of other compounds were assigned by analogy.
During the investigation on the ring size effect of the 1,2diaza-1,3-diene substrate, it was also noted the formation of ring-opened [4 + 2] byproduct 4, highlighting the ease of rearomatization of 3 to give a more stable indole derivative. The sensitivity of 3 to the rearomatization process was confirmed by complete transformation of 3b into 4e in the presence of Amberlyst 15(H) (vide infra, Scheme 4b). This undesirable event appears to be the cause for lowering the [4 + 2] cycloaddition product yields found in some cases. Notably, this pathway remains dominant when the reaction was conducted using N-methyl indole (1a) or 1,2-dimethyl indole (1o) with linear DDs 2j and 2n (Scheme 2) in line with what was previously observed in the reactions of 2,3-(and 3-)unsubstituted indoles with cyclic and noncyclic DDs. 7a,10e More precisely, the reaction of N-methyl indole (1a) with linear DD 2n afforded the more polar ring-opened [4 + 2] product 4a (48% yield). However, thin-layer chromatography (TLC) analysis revealed the presence of a mixture of the diastereoisomers of pyridazine 3z. Consistent with Gilchrist's observation, 7b monitoring the progress of the reaction by 1 H NMR, we detected an initial (preferential) formation of (cis,cis)-3z, which then partially isomerized to its isomer (cis,trans)-3z either during the course of the reaction or during chromatographic separation. Despite the isomerization side reaction, both diastereoisomers were isolated ((cis,cis)/ (cis,trans) ∼ 2:1, 32% combined yield) and characterized (see the SI for details). On the other hand, the reaction of Nmethyl indole (1a) with DD 2j or 1,2-dimethyl indole (1o) with DD 2j or 2a led to the formation of the sole ring-opened [4 + 2] products 4b−d (Scheme 2). Therefore, given the results with the use of both 2,3-and 3-unsubstituted indoles (associated with the [4 + 2] pyridazine-ring-opening reaction) and to further showcase the flexibility of this catalytic annulation strategy, we next moved our attention to exploring the reactivity of C3-blocked indoles (e.g., 3-substituted and 2,3disubstituted indoles) with DDs. To our surprise, the reaction of 3-methyl indole (1p) with linear DD 2n led to a mixture of two cycloadducts, the expected tetrahydro-1H-pyridazino [3,4b]indole compound 3ab and the tetrahydropyrrolo [2,3-b]indole compound 5a 14 in a ratio of approximately 1:1, which could possibly be the result of the above-mentioned two competitive reaction pathways 15 (Scheme 2). Interestingly, when 1,3-dimethyl indole (1q) was used in combination with DD 2j, the exclusive formation of product 5b (46% yield) was detected. As expected, when the reaction was repeated using cyclic DD 2c, the exclusive formation of the corresponding [4 + 2] product 3ad (40% yield) (Scheme 2) was observed.

Scheme 1. Working Hypothesis: Chemodivergent Synthesis of Polycyclic Fused Indoline Scaffolds
Intrigued by the starkly different reaction profile, we next focused our attention on the 2,3-disubstituted indole motif. Unfortunately, the reactions of 2,3-disubstituted indoles such as 2,3-dimethyl indole 1r and 2,3,4,9-tetrahydro-1H-carbazole 1t with cyclic DD such as 2c did not work well, and only a trace amount of the respective formal [4 + 2] cycloaddition product was detected in the complex crude reaction mixture (Scheme 2). Explanations for these findings are not immediately intuited, but the steric effect seems to be playing a major role.
The structures of compounds 5a−s were confirmed by subjecting 5s to N−N bond cleavage using the Magnus method. 16 Treatment of compound 5s with ethyl bromoacetate/Cs 2 CO 3 /MeCN at 50°C followed by heating to 80°C resulted in N−N′ bond cleavage to the corresponding NH-free tetrahydropyrrolo [2,3-b]indole 6a in 64% isolated yield (Scheme 4a).
As a synthetic strategy, this [3 + 2] annulation affords, in a single operation, the structurally rigid 6-5-5 tricyclic subunit Table 1. Scope of the Zn(II)-Catalyzed [4 + 2] Cycloaddition Reaction of 2,3-Unsubstituted Indoles (1) and Cyclic Azoalkenes (2) a,b with a substituent at the 3-position of the indole nucleus, which is the basic structure of pharmaceutically valuable natural products. 4 Besides, this nonclassical approach provides access to functionalized pyrroloindoline systems with substitution patterns that are otherwise inaccessible using tryptamines 17 as precursors.
The computed [4 + 2] energy reaction paths starting from the cisoid-1,2-diaza-1,3-diene·ZnCl 2 ·catalytic complex (cisoid-DD·ZnCl 2 ) leading to the complex endo-cycle·ZnCl 2 and to exo-cycle·ZnCl 2 are reported in Figure 3a; since the reaction is highly exoergonic, both reaction trajectories go through a typical reactant-like transition state [TS] ‡ having pericyclic ) are shown in Figure 3b. The computations show clearly that the observed high diastereoselectivity toward the formation of the slightly less stable (cis,cis)-3ab pyridazino indoline ((cis,cis) → (cis,trans), ΔG°= −2.66 kcal mol −1 ) is obtained under kinetic control. Indeed, since its endo cyclic precursor is substantially more stable than the exo adduct (ΔΔG ‡ = −1.70 kcal mol −1 , mainly for the lack of the steric clashes of the two methyl groups; see Figure 2a), the two associated activation energy barriers are very different (ΔG ‡ = 9.02 vs 10.46 kcal mol −1 ); thus, the endo path is kinetically more favorable. Interestingly, in both [TS] ‡ , the ratio between the two forming C−C and C−N single bonds is about 1.3 (Figure 3b), which is symptomatic of an asynchronous concerted transition state. 21 The comparison of the [3 + 2] cycloaddition energy diagram of the two stepwise mechanisms with that of the concerted cycloaddition suggested by Gilchrist et al. with very similar substrates 7b,8 shows clearly that the latter mechanism is not active in our case ( Figure 4).
Finally, as a corollary of the above-reported computations, we used them to evaluate the order of magnitude of the product ratio [(cis,cis)-pyridazinio indoline (3ab) Since the two-reactant catalytic complexes (the cisoid-DD· ZnCl 2 and the transoid-DD·ZnCl 2 ) are in equilibrium, and their interconversion is much faster than the cycloaddition reaction rates, it is possible to apply the Curtin−Hammet equation, 22 which, in our case with a ΔΔG ‡ = [TS] endo ‡ − [TS1] ‡ = 0.50 kcal mol −1 , gave a ratio of 7:3, (cis,cis)-3ab and pyrazole indoline 5b, respectively. We reckon that this result is Table 2. Scope of the Zn(II)-Catalyzed [3 + 2] Cycloaddition Reaction of 2,3-Substituted Indoles (1) and Linear Azoalkenes (2) a,b fair enough, considering the chemical accuracy attainable via the used model chemistry.
Combining the above experimental results, DFT studies, and available literature, 7,10e a reasonable mechanism for these annulation processes is summarized in Scheme 3. Two competing (and independent) reaction pathways for both the tetrahydro-1H-pyridazino [3,4-b]indole and tetrahydropyrrolo- [2,3-b]indole derivatives appeared to take place upon initial ZnCl 2 activation of the 1,2-diaza-1,3-diene substrate. The [4 + 2] cycloaddition (path a) can be simply rationalized as a concerted inverse hetero-Diels−Alder reaction. The preference for an endo cycloaddition transition state, which requires the cisoid conformation for DD 2 (II), supports the high observed diastereoselectivity for product 3. 23 Alternatively, [3 + 2] annulation (path b) can be viewed as proceeding via a stepwise process. Regioselective 1,6-addition of the indole nucleophile 1 on activated DD 2 (I) that is in a transoid conformation affords the zwitterionic intermediate IV, which undergoes intramolecular 5-exo-trig cyclization collapsing to the fivemembered azomethine imide V. The subsequent 1,3-H shift furnishes via intermediate VI the tetrahydropyrrolo [2,3b]indole product 5 and restores the ZnCl 2 −diene catalytic complex. 24 The fact that the indole 1q gave both [4 + 2] and [3 + 2] cycloadducts using cyclic (R 4 ≠ H) and linear (R 4 = H) DDs (3ad vs 5b) supported this mechanism scenario. Likewise, the borderline example of Scheme 2 in which both cycloadducts 3ab and 5a concurrently formed 15 from 1p and 2n illustrates the delicate balance and subtle nuances between the two annulation processes. It is evident that, in the presence of additional substituents on the indole ring (R 3 ≠ H), the [3 + 2] mode of addition becomes competitive since the concerted [4 + 2] pathway is more susceptible to steric inhibition. Moreover, it was quite interesting to note that when sixmembered cyclic 1,2-diaza-1,3-diene 2i was reacted with 1s, the exclusive formation of the [4 + 2] cycloaddition product 3ae was observed (Scheme 4c). Similarly, the use of linear 1,2diaza-1,3-diene 2t yielded the product 3af (Scheme 4d). Our control experiments illustrate that the absence of EWG groups like esters, amides, or phosphonates in the C4 position of the starting DD (R 4 = H; R 5 ≠ CO 2 R, CONR 2 , and PO(OR) 2 ), which likely disfavors the proton transfer process (V → VI), also privileged the [4 + 2] mode of addition.
With this work, we have demonstrated that the nature and type of substituents of both 1,2-diaza-1,3-diene and indole substrates are critical factors dictating chemoselectivity in the annulation process. Notably, the presence of a H atom in the C3 position of the indole ring is responsible for the observed ring-opened [4 + 2] product 4. As already evidenced, this event becomes prevailing when N-methyl indole (1a) or 1,2dimethyl indole (1o) is used as the nucleophile. To our surprise, when R 3 = H, neither the formation of the [3 + 2] annulation product nor the ring-opened [3 + 2] product of type 7 described by Tan and co-workers was observed. 25 This result shows that when R 3 = H, the indole rearomatization process from 3 (and/or eventually from intermediate IV) to 4 is the preferred one.

■ CONCLUSIONS
In conclusion, we have developed substrate-dependent divergent annulation reactions 26 of indoles with 1,2-diaza-1,3-dienes. By virtue of the versatility of these latter in switching reactivities, efficient synthesis of two types of polycyclic fused indoline scaffolds tetrahydro-1H-pyridazino- [3,4-b]indoles and tetrahydropyrrolo [2,3-b]indoles was achieved. The DFT study revealed that [4 + 2] cycloadditions are concerted but quite asynchronous, while [3 + 2] reactions go undoubtedly through a stepwise mechanism. Our approach expands the scope of polycyclic fused indoline synthesis and increases the flexibility of synthetic strategies toward heterocycle-based scaffolds. Remarkably, the reactions feature a high step-and atom-economy, high chemo-and diastereoselectivity, broad substrate scope, good functional group tolerance, and readily accessible starting materials. The successful construction of unique rigid polycyclic skeletons, particularly those with challenging bridgehead N,N-aminal quaternary centers, enriches the chemistry of both indoles and 1,2-diaza-1,3dienes.

■ EXPERIMENTAL SECTION
General Experimental Details. Indoles 1a, 1l, 1m, 1o, 1p, 1r, and 1s are commercially available reagents and used without further purification. N-Alkylindole derivatives 1b−k, 1n, and 1q were prepared from corresponding commercially available NH-indoles following literature procedures. 27 3,4-Disubstituted indoles 1t−z were synthesized from corresponding phenylhydrazine hydrochlorides as starting materials via Fisher indole synthesis according to the literature. 28 [3 + 2] cyclization: stepwise mechanism (blue path) vs the concerted mechanism (red path) in CH 2 Cl 2 at 298 K. The energies (kcal mol −1 ) are reported with respect to the transoid-DD·ZnCl 2 and In species. For clarity, the H atoms of transition-state structures have been omitted. the corresponding hydrazones following literature procedures. 29 Chromatographic purification of compounds was carried out on silica gel (60−200 μm). TLC analysis was performed on preloaded (0.25 mm) glass-supported silica gel plates (Kieselgel 60); compounds were visualized by exposure to UV light and by dipping the plates in 1% Ce(SO 4 )·4H 2 O and 2.5% (NH 4 ) 6 Mo 7 O 24 ·4H 2 O in 10% sulfuric acid, followed by heating on a hot plate. All 1 H NMR and 13 C NMR spectra were recorded at 400 and 100 MHz, respectively, using dimethyl sulfoxide (DMSO)-d 6 or CDCl 3 on K 2 CO 3 as the solvent. Chemical shifts (δ scale) are reported in parts per million (ppm) relative to the central peak of the solvent and are sorted in a descending order within each group. The following abbreviations are used to describe peak patterns where appropriate: s, singlet; d, doublet; t, triplet; q, quartet; sex, sextet; m, multiplet; and br, broad signal. All coupling constants (J value) are given in hertz (Hz).
Structural assignments were made with additional information from gradient correlation spectroscopy (gCOSY), gradient heteronuclear multiple quantum correlation (gHMQC), gradient heteronuclear multiple bond correlation (gHMBC), and nuclear Overhauser enhancement spectroscopy (NOESY) experiments. Fourier transform infrared (FT-IR) spectra were obtained as Nujol mulls or neat. Highand low-resolution mass spectroscopies were performed on a Micromass Q-ToF Micro mass spectrometer (Micromass, Manchester, U.K.) using an electrospray ionization (ESI) source. Melting points were determined in open capillary tubes and are uncorrected. Elemental analyses were within ±0.4 of the theoretical values (C, H, N).
The relative configurations of diastereomers 3z were assigned by means of two-dimensional (2D) NOESY experiments.