One-pot synthesis of polycyclic isoindolines using isoindole umpolung

Abstract

Using an isoindole umpolung strategy, a one-pot synthesis of polycyclic isoindolines was accomplished. In this reaction, the in situ-generated nucleophilic isoindoles were converted to electrophilic isoindoliums via protonation, which underwent a Pictet-Spengler-type cyclization to afford a variety of polycyclic isoindolines in good yields.

Introduction

Isoindolines and isoindolinones are prevalent in natural products (e.g., Berberis alkaloids nuevamine and chilenamine) and synthetic bioactives, including staurosporine, which possesses cytotoxic, antiproliferative, and antimicrobial activity, and the aristolactams, which possess immunosuppressant, antiplatelet, antimycobacterial, and neuroprotective activities [1]. However, while isoindole derivatives are ubiquitous, the parent structures, isoindoles (1, Scheme 1), are not very common. Isoindoles are highly reactive moieties that participate in a variety of reactions [2,3]. Electrophilic aromatic substitutions (Scheme 1A), in which the isoindole serves as the nucleophile, are among the most commonly reported reactions of isoindoles [2,3a,b]. Resonance donation from the lone pair on the isoindole nitrogen renders the α-position nucleophilic–a feature of isoindole reactivity that has been exploited in these reactions, as well as in Michael reactions [3c] and intramolecular cyclizations [3b]. Isoindoles also often serve as highly reactive electron-sufficient dienes in Diels-Alder reactions [2,3c]. However, to our knowledge, little work has been done to investigate the role of isoindoles as electrophiles. This is because polymerization occurs when the isoindole reacts incompletely with an electrophile to form an electrophilic species, which then reacts with unreacted isoindole [2]. Protonation of an isoindole is known to occur at the 1-position, giving an isolable isoindolium salt [3a], which polymerizes as described.

Scheme 1.

(A) Examples of isoindole reactivity. (B) Our approach.

We envisioned an approach in which the isoindole could be converted to an electrophile by protonating C1 to form the corresponding isoindolium ion, rendering the C3 position electrophilic and opening up the possibility for reactions with various nucleophiles (Scheme 1B). To avoid decomposition by polymerization and oxidation, we planned to use excess acid to completely convert the isoindole to the isoindolium. In order to prove our hypothesis, we first synthesized N-butylisoindole (1a, Scheme 2) from 2-bromomethylbenzaldehyde (2a) and butylamine (3a, 1.2 equiv.) in the presence of triethylamine (1.2 equiv.) in dichloromethane by adapting the procedure reported by Voitenko, et al. [3c]. Following an aqueous workup, the isoindole was isolated and taken up in deuterated chloroform. The ¹H NMR spectrum (Fig. S1) showed quantitative conversion of aldehyde 2a to the corresponding isoindole 1a. The signal at 7.10 ppm, which corresponds to the protons at the 1- and 3-positions, served as the diagnostic peak for the isoindole [4]. The isoindole was then treated with a large excess of TFA, and quantitative conversion of isoindole 1a to the isoindolium species 4 (Scheme 2) was observed by ¹H NMR (Fig. S1). The isoindole diagnostic peak at 7.10 ppm disappeared and two new singlets at 9.30 ppm and 5.24 ppm, integrating to one and two protons, respectively, were observed [4]. The former was downfield of the isoindole diagnostic peak, as expected for an iminium C—H, and the latter was 1.86 ppm upfield of the isoindole diagnostic peak, consistent with protonation of the isoindole at C1-position. Taken together, these indicate quantitative conversion of the isoindole to the isoindolium. These studies were then repeated in a one-pot procedure in which the isoindole was formed in situ in deuterated chloroform, and then converted to the isoindolium by the addition of excess TFA directly to the ¹H NMR sample. Similar results were observed, demonstrating that a one-pot procedure can also be used to quantitatively form the isoindole and isoindolium.

Scheme 2.

Preliminary studies [6].

In the Pictet-Spengler reaction, an iminium ion serves as the electrophile, which is subsequently attacked by an aromatic nucleophile. Hence, we decided to develop a Pictet-Spengler-type reaction to synthesize polycyclic isoindolines, reminiscent of natural product nuevamine [1]. Tryptamines were chosen in our model system, as they have been utilized extensively in Pictet-Spengler cyclizations [5], and the β-carboline structures formed are found in bioactive molecules and hormones [7]. Herein, we report a facile one-pot procedure that provides an efficient route to polycyclic isoindolines.

The reaction conditions were optimized systematically (Table 1). Initially, the isoindole was isolated prior to acidification. Various solvents, protic acids, and equivalents of acid were investigated. Based on literature precedent, TFA was used as the default protic acid catalyst [8]. To avoid polymerization, 10 equivalents of acid was used as the default. The chlorinated solvents (entries 1–3) gave 5 [9] in very high yields by ¹H NMR, with dichloromethane (entry 1) and dichloroethane (entry 2) giving the highest yields. Toluene (entry 4), a nonpolar solvent that would allow for high temperature refluxes, also gave a high yield. Polar solvents and protic solvents were not suitable for this reaction. No desired product was formed in acetonitrile (entry 5). THF (entry 6), 1,4-dioxane (entry 7), MeOH (entry 8), and DMF (entry 9) effected decomposition of the product. The optimal protic acid catalyst was TFA (entry 1). For the acids stronger than TFA, MsOH (entry 11) and HCl (entry 10), the percent yields decreased with increasing acid strength. Weaker acids, such as trichloroacetic acid (entry 12) and AcOH (entry 13), also resulted in lower yields. No product was obtained using AcOH, the weakest acid screened, even with heating. As anticipated, a superstoichiometric amount of acid was needed, as no product was obtained when a single equivalent was added. Increasing the number of equivalents increased the yield dramatically, up to 5.0 equivalents (entries 14–16). Beyond 5.0 equivalents, only a slight increase in the yield was observed (entry 1). For the multistep procedure originally employed, 10 equivalents were optimal. A one-pot procedure was further developed for the ease of the reaction. When 10 equivalents of TFA were added in a one-pot procedure, the yield was substantially lower than for the multistep procedure (entry 17). We hypothesized that residual triethylamine from the isoindole formation consumed some of the acid, forming triethylammonium and decreasing the yield. Increasing the amount of TFA added from 10 to 20 equivalents gave the highest yield (entry 18).

Table 1.

Optimization of reaction conditions.

Entry	Solvent[a]	Acid	Equiv.	Yield (%)[b]
1	CH2Cl2	CF3CO2H	10	>98 (84)
2	ClCH2CH2Cl	CF3CO2H	10	98 (81)
3	CHCl3	CF3CO2H	10	91
4	toluene	CF3CO2H	10	95 (78)
5	Acetonitrile	CF3CO2H	10	0
6	THF	CF3CO2H	10	decomposed
7	1,4-dioxane	CF3CO2H	10	decomposed
8	MeOH	CF3CO2H	10	decomposed
9	DMF	CF3CO2H	10	decomposed
10	CH2Cl2	HCl	10	48 (18)
11	CH2Cl2	MsOH	10	48[c]
12	CH2Cl2	CCl3CO2H	10	80 (77)
13	CH2Cl2	CH3CO2H	10	0[d]
14	CH2Cl2	CF3CO2H	1.0	0
15	CH2Cl2	CF3CO2H	2.5	23
16	CH2Cl2	CF3CO2H	5.0	94
17[e]	CH2Cl2	CF3CO2H	10	67 (63)
18[e]	CH2Cl2	CF3CO2H	20	100 (91)

^[a]For reactions in which 1b was isolated, this refers to the cyclization step.

^[b]Yields were determined by ¹H NMR with 1,3,5-Trimethoxybenzene as the internal standard. Isolated yields are shown in parentheses.

^[c]Purified product was a complex mixture. Isolated yield not determined.

^[d]No cyclization observed at 23 °C or 50 °C.

^[e]A one-pot procedure was used in which the isoindole was formed by the addition of triethylamine (1.2 equiv) to tryptamine (1.2 equiv) and 2-bromomethyl benzaldehyde (1.0 equiv, 0.10 M in CH₂Cl₂) at 23 °C. After 2 h, sufficient DCM was added to bring the concentration to 0.02 M, and TFA was added at −40 °C. The reaction mixture was subsequently allowed to warm to 23 °C and stirred for 16 h.

Our optimized conditions were employed to evaluate various β-arylalkylamine reacting partners (Scheme 3). First, substituted tryptamines were investigated. To establish the relevance of the β-arylalkylamine electronics, electron-rich and electron-poor tryptamines were investigated. Two tryptamines bearing an electron-donating substituent were investigated: 5-methoxytryptamine [10] and 5-methyltryptamine [11], which gave the desired products 6 and 7 in yields comparable to that of 5. Thus, the presence of an electron-donating group did not significantly affect the yield. Increasing the electron-donating group strength, from methyl to methoxy, was associated with a slight decrease in yield. However, the presence of an electron-withdrawing substituent substantially reduced the yield, as can be seen with the 5-chloro-tryptamine [12] cyclization product (8). To effect complete cyclization, a reaction time of three days at 23 °C or 16 h at 60 °C was required. To probe the effect of indole N-alkylation, 1-methyltryptamine [8b] was cyclized to give 9. As anticipated, the N-alkyl group slightly decreased the yield. To extend the scope of this work beyond tryptamines, 4-(aminomethyl)indole, an established Pictet-Spengler substrate [13], was investigated. This substrate was of interest due to the strained geometry of the cyclization product. To reduce the time required for complete cyclization from three days to 16 h, the reaction mixture was heated to 60 °C to give 10.

Scheme 3.

Reaction scope and regioselectivity.^a (a) Isolated yields, followed by product distribution for regioisomeric mixtures. (b) Heated reaction mixture in DCE to 60 °C after TFA addition to promote cyclization. (c) Characterized as TFA salt. (d) Characterized as HCl salt. (e) Determined by ¹H NMR. (f) From crude mixture. (g) From lyophilized pure mixture. (h) Expected major and minor products. Identities were not confirmed. (i) Determined by HPLC from acidified crude mixture.

To expand the scope of this work beyond indole nucleophiles, 1-(2-aminoethyl)pyrrole and substituted phenethylamine substrates were investigated. The pyrrole ring is a privileged scaffold found in many bioactive molecules [14]. Furthermore, Pictet-Spengler reactions of 1-(2-aminoethyl)pyrrole have been reported [15]. Purification attempts were unsuccessful, even following acidification with AcOH or TFA, as the product oxidized readily. Impurities arising from the viscosity of the product were excluded using procedural modifications, giving 11 [16]. Attempts to cyclize homopiperonylamine [17], dopamine [18], and 3-methoxyphenethylamine [18,19] substrates from the seminal Pictet-Spengler works [5,19] were unsuccessful. The isoindoles were formed cleanly, but homopiperonylamine remained uncyclized, 3-methoxyphenethylamine polymerized, and dopamine [20] gave a complex mixture. Additionally, 2-(aminomethyl)indole and 3-(aminomethyl)indole remained uncyclized, both at room temperature and 60 °C.

Next, substituted 2-bromomethylbenzaldehydes were screened. Reactions of 2-bromomethyl-3-chlorobenzaldehyde, 2-bromomethyl-4-chlorobenzaldehyde, and 2-bromomethyl-4-fluorobenzaldehyde with tryptamine were used to probe the electronics and regioselectivity of the reaction, giving regioisomeric mixtures 12–14, respectively. Diminished reaction yields were observed for the electron-deficient halogen-substituted aldehydes. Chlorine substitution of the aldehyde decreased the yield by 11–15%, and fluorine substitution decreased the yield by 21%. These data suggest an inverse relationship between the electron deficiency of the substrate and the yield. Chlorine substitution in the 3-position gave over twofold greater regioselectivity than the 4-position. As anticipated, fluorine substitution in the 4-position gave approximately the same regioselectivity as chlorine substitution in the 4-position. Using 1D ¹H NOE experiments, 12a was identified as the major product of 12 (Fig. S2). Consistent with our observations for 17 (Scheme 4), described infra, the more stable isoindolium regioisomer cyclizes to form the major product. The isoindolium that leads to 12a is more stable than the isoindolium leading to 12b because in the former, chlorine is meta to the iminium and stabilizes it via conjugation, whereas in the latter, chlorine is ortho and thus, not in conjugation with the iminium. Regioisomers of 13 and 14 were not separable by TLC or HPLC.

Scheme 4.

Regioselectivity of isoindolium formation.

Attempts to investigate the effects of substitution at the isoindolium 1-/3-position, using 2-bromoethylbenzaldehyde, were unsuccessful. Cyclization was not observed, even with heating. However, the regioselectivity of the isoindolium formation (17) was determined (Scheme 4). The isoindole and isoindolium were formed in situ in CDCl₃. Following acidification, a ¹H NMR spectrum of the isoindolium was obtained (Fig. S4) [4]. A peak at 5.15 ppm integrating to two hydrogens was observed, which is close to the chemical shift of the benzylic isoindolium protons from the preliminary studies. The isoindolium C—H singlet was also absent. These findings establish that for 17, the more substituted isoindolium 17a forms exclusively and support our hypothesis that steric hinderance from the methyl group prevented cyclization.

Finally, we investigated the effects of combining 2-bromomethyl-3-chlorobenzaldehyde with 4-(aminomethyl)indole (15) and 5-methoxytryptamine (16) on the yield and regioselectivity. Interestingly, using an electron-poor aldehyde with 4-(amino-methyl)indole improved the yield considerably. However, the regioselectivity of 15 was about half that of 12. The increase in yield with 5-methoxytryptamine, for 16 versus 6, was negligible, and the regioselectivity of 16 was also substantially lower than 12.

Conclusion

We have developed an efficient synthesis of polycyclic isoindolines in good yields using a one-pot procedure. The in situ-generated nucleophilic isoindoles were converted to electrophilic isoindoliums via protonation, which underwent Pictet-Spengler-type cyclizations to give the polycyclic isoindolines. These results open the door to other reactions that can utilize isoindoliums as electrophiles via isoindole umpolung and extend the scope of the Pictet-Spengler reaction to include isoindolium electrophiles. Further development of addition reactions using the isoindole umpolung strategy and biological evaluation of the polycyclic isoindolines are ongoing and will be reported in due course.