Benzo[7,8]indolizinoquinoline scaffolds based on Mg(ClO₄)₂-promoted regiospecific imide reduction and π-cyclization of N-acyliminium species. Analogues of the topo-1 poison rosettacin and 22-hydroxyacuminatine alkaloids

1 Introduction

The coupling process of N-acyliminium species, formed generally in situ by acidic treatment of amide or lactam bearing a leaving group at their α-position, constitute one of the power tools used in C—C bond forming reactions explored largely in contemporaneous synthetic organic chemistry (Hiemstra and Speckamp, 1991; DeKoning and Speckamp, 1996; Pilli and Russowsky, 1997; Speckamp and Moolenaar, 2000; Maryanoff et al., 2004; Yazici and Pyne, 2009a,b). The high stability of these cationic species due to the presence of a carbonyl function at the nitrogen atom, gives them unlimited reactivity and hence deliver a wide range of natural and unnatural polycyclic systems including alkaloids with, among others, broad biological interests (Hiemstra and Speckamp, 1998; Marson, 2001). While the formation of C–heteroatom bonds seems to be scarce but constitutes a novel and powerful strategy to access new complex compounds containing cyclic mixed N,O-, N,N-, N,S-, and N,Se-acetals as compiled recently (Daïch et al., 2014), the C—C bond formation, which has extensively been investigated, is still of interest and general uses for the exploration of new space of heterocyclic synthesis.

Among various nitrogen containing heterocyclic nuclei, pyrroloisoquinoline scaffolds have drawn tremendous interests in synthetic and medicinal chemistry due to their occurrence in many naturally occurring alkaloids and unnatural products with large spectrum of pharmacological activities. These systems can be exemplified by the antimicrobial, antioxidant and anti-inflammatory agents of general structure 1 including (-)-trolline (R⚌H, R₃⚌R₄⚌OH) (Moreno et al., 2012; Suyavaran et al., 2015), the corresponding amine such as (±)-crispine-A (2) which exhibits marked anticancer properties (Dhanasekaran et al., 2014; Mons et al., 2014), the first isoindoloisoquinoline alkaloid reported nuevamine (3) (Selvakumar and Ramanathan, 2011; Min and Kim, 2014; Maity and Roy, 2014) and the corresponding aza-analogues 4 (Hitchings and Vernon, 1988, 1990; Zamudio-Medina et al., 2010; Islas-Jácome et al., 2014) which curiously have not showed at this time any biological activity (Fig. 1).

Close

Fig. 1 Selected examples of bioactive natural 1–3 and unnatural 1,4 pyrroloisoquinolines fused or not, natural aromathecines 5 and our targets 6.

Export to PPT

2 Results and discussion

In our group, we have long been interested in the synthesis of analogues of such tetracyclic alkaloids containing in particular a pyrrolidone or isoindolone moiety. For that objective, we explored synthetic methodologies using π-cationic cyclization of thionium ions under Pummerer conditions (Bousquet et al., 2006), catalytic Lewis acid or Brönsted acid-promoted chiral π-cyclization versus aza-cyclization of N-acyliminium species (Fleury et al., 2011), aza-MIRC sequence in combination with diastereoselective arylation through cationic species (Saber et al., 2010, 2011; Hamid and Daïch, 2011) and an intramolecular catalytic cyclization using in particular Bi(OTf)₃ in very low catalyst loadings (1 mol%) (Pin et al., 2007). In parallel investigations some other heteroanalogues were provided by the tandem heterocyclisation/isomerisation/π-cationic cyclization (Daïch et al., 2014).

In this context allied to our another ongoing pharmacological project directed to the development of short and effective synthetic approaches leading to the natural topoisomerase-1 (topo-1) poisons (Fox et al., 2003) such as rosettacin (5a, R⚌H), acuminatine (5b, R⚌Me) and 22-hydroxyacuminatine (5c, R⚌CH₂OH) as well as synthetic derivatives (Pin et al., 2008, 2009), we envisioned in this report the synthesis of a short library of the pentacyclic structures of type 6. These unknown compounds equipped with both quinoline fused via 2,3-positions to pyrrolo[2,1-a]isoquinolinone skeleton instead of pyrrolo[1,2-b]isoquinoline nuclei in the case of natural aromathecines 5, can be considered reasonably in point of view of structure-activity relationship studies (QSAR) as aromathecines analogues. Concisely, our investigations will be focused on the synthesis of compounds bearing isoquinoline skeleton with an alkoxy substituent or not, able to involves a hydrogen linkage often positive for the biological activity, and also quinoline substituted at its C₄- and C₆-positions. In addition, the position of R₁ and R₂ substituents will be favorable for further biological activities if any as inspired by the QSAR studies conducted on camptothecines as the most potent topo-1 inhibitors (Wu, 2003; Pommier, 2009; Pourquier and Lansiaux, 2011; Moukharskaya and Verschraegen, 2012).

Most of the hitherto reported syntheses leading to these types of compounds rely on an appreciable number of strategies. They can be divided into two main categories of reactions using radical or ionic process in the ultimate stage. Although the arylation through cationic species is largely documented in the literature in pyrrolidinone and isoindolinone series, the use of π-cationic cyclization of N-acyliminium ions seems to be particularly adapted in our case. In addition, problematic which can be raised during the feasibility of this strategy in pyrrolo[3,4-b]quinoline-1,3-dione ring, which is both unsymmetrical and possesses an additional alkaline quinoline nitrogen atom, are of high interests.

As highlighted in the retrosynthetic plan (Scheme 1), our strategy was based first on the exploration of Friedländer reaction (Boisse et al., 2009) starting from simple aromatic amines 7 and unconventional Friedländer substrate 8 to provide quinoline-2,3-diesters 9 already substituted at the expected C₄- and C₆-positions. The treatment of these derivatives or corresponding dicarboxylic acids by an appropriate amine 10 supplies the imides 11 which it is advisable to reduce regioselectively into the expected hydroxy lactams 12. The targeted products 6 bearing different substituents would then be reached in an ultimate stage of π-cationic cyclization (disconnection a in Scheme 1) in an acidic medium of the stable N-acyliminiums ions intermediate I stemming from 12. The viability of this strategy is based fundamentally on our capacity to control two sequence keys; the access to a wide variety of disubstituted diesters 9 and the imides 11 as well as the regiocontrol of the reduction of the latter 11 into the expected N-acyliminium precursors 12 (see Table 1).

Close

Scheme 1 Retrosynthetic scheme leading to pyrroloisoquinolines fused to quinoline ring as targets 6.

Export to PPT

Our study started with the synthesis of quinoline diesters 9 with different substituents R₁ and R₂ in adequate positions as outlined above for the QSAR study reasons (Scheme 2). This can serve also to measure the impact of these substituents first on both the reaction yields for the production of diesters 9 themselves and N-arylalkyl-imides 11. Thus, the Michael adduct 14 obtained by addition of aniline (13) to the activated acetylene DMAD 8 (80%) is obtained in ethanol instead of methanol as reported (Heindel et al., 1968; Jaen et al., 1995). The latter under Vilsmeier-Haak formulation (e.g., POCl₃, DMF, 0 °C for 1 h then 2 h at reflux) furnished the expected diester 9a in 76% yield. During this reaction, the formyl derivative J obtained regioselectively under the orientation effect of the amine function was never isolated but cyclizes directly into 9a after departure of one molecule of H₂O. The reaction is applicable to other anilines but show serious problems with anilines o- and m-substituted. Because of this approach is not generalizable for the synthesis of quinoline diesters 9 bearing in particular at their C₄-position substituent R₂ different to H, the reaction of the instable 2-amino-benzaldehyde (7a) was reacted with DMAD 8 according to the report of Heindel’s group (Eq. (2)) (Taylor and Heindel, 1967; Bryce et al., 1983; Rao et al., 1994). This provided in one-pot (two steps) procedure the expected product 9a in 75% yield. We were interested by an easy approach in one-pot protocol, without using benzene as solvent, and applicable in large scale.

Close

Scheme 2 Different schemes leading to substituted quinolie-2,3-diesters 9a-d.

Export to PPT

Thus, we explored this reaction with same ingredients by screening other solvents and different reaction conditions. Gratifyingly, the best formulation seems to be the use of glacial acetic acid at reflux for 3 h (Eq. (3)). Under these conditions, diester 9a was obtained in 88% yield better than the one obtained in last both cases (Scheme 2). Finally, the application of this new and simple protocol to an array of aromatic amino-ketones 7b-d resulted in the formation of the expected quinoline-2,3-diesters 9b-d via the intermediacy of K similar to J. The products were isolated in yields superior to 88% in all cases rendering this ‘one-pot’ two-step procedure more practical and competitive than the ones published more recently (Madhav et al., 2010; Guerrini et al., 2011; Tanmoy et al., 2012).

With high quantity of desirable quinoline-2,3-diesters 9b-d in hand (up to 10 g), we then targeted the N-substituted imides 11 precursors of the expected α-hydroxy lactams 12. Thus, as shown in Scheme 3, saponification of the diesters 9b-d (e.g., NaOH, MeOH/H₂O, reflux 12 h than AcOH) (Kim et al., 1993; Brown et al., 1994) resulted in the formation of the diacids 15b-d after acid treatment. It is worth of mentioning that the use of HCl_aq in high concentration or other strong acids provide the expected diacids 15b-d but with low up to marginal yields. This is due to the protonation of quinoline nuclei which has a pKa value of 9.50 (pKa of hyrdoquinolium is of 4.94). Interestingly acetic acid with pKa value of 4.75 seems to be particularly adapted to resolve this problem since the diacids 15b-d were isolated in high yields (82–91%; Table 2).

Close

Scheme 3 Production of imides 11 by imidation of the diacids 15b-d.

Export to PPT

Carboxylic acids have been reported to react with amines under heating solvent free (or in the presence of solvent) or under azeotropic removal of water in Dean-Stark apparatus (Kato et al., 1999; Dahr et al., 2003). In the latter case, aromatic hydrocarbons were used as solvent and the addition of catalytic amount of dry NEt₃ is needed in certain cases in racemic or chiral series (Polniaszek et al., 1990; Lee et al., 1995, 1997a,b). According to the above both procedures, all attempts of reaction of o-dicarboxylic acids 15 with primary amines 10 solvent free, in toluene or xylene in the absence or presence of catalytic amounts of NEt₃ failed to deliver the expected diacid products 11.

During these attempts, in addition to the starting dicarboxylic acids 15, we observe at a minimum three compounds corresponding to 16, 17 and 18 which resulted from the reaction of amine with carboxylic acid function. Frustratingly, only traces of the expected imides 11 were observed rendering this protocol inefficient in our cases.

From these results, attempts were then made on the basis of known peptide coupling procedures (Karmali et al., 2004) by using DCC as peptidic coupling agent, DMAP as base, phenethylamine (10a) and quinoline-2,3-dicarboxylic acid (15a). The purification of the crude reaction mixture provides in all cases the expected imide 11aa which is contaminated with small amounts of DCC and the amide-acid 16aa. Heating the reaction mixtures at different temperatures, in different solvents with different DCC/DMAP ratio, etc. did not bring significant improvement to deliver the expected imide 11aa as the sole reaction product (Scheme 3 and Table 2).

The latter protocol is better in regard to the precedent in terme of chemoselectivity but the reaction is incomplete since near 11 the amide-acid 16 was formed in variable proportions always in disfavor of 11 (Table 2). Elsewhere, treatment of the mixture under cyclodehydration conditions (Mehta et al., 1960) by using Ac₂O at reflux in the presence of AcONa resulted in the formation of imide 11 with however average yields not exceeding 50%.

Given the efficiency of glacial acetic acid in the previous transformation (7a-d + 8 → 9a-d), as well as that shown in the literature including ours in a close domain (Chihab-Eddine et al., 2001; Cul et al., 2004), we envisioned the evaluation of o-dicarboxylic acids 15a-d under these conditions. Thus heating diacid 15 and amine 10 at reflux of glacial AcOH (Method B) provided pleasingly the expected imide 11 as the sole reaction product generally in appreciable yields (Table 2). From these results, some remarks deserve to be underlined:

1. the reaction yields are appreciable and ranged from 62% up to 90% except for imides 11ab (34%) and 11ac (39%);

2. the reaction is very easy to use, proceeds cleanly and occurs in one-pot ‘two-steps’ procedure via the tandem intermolecular amidation/cyclodehydration without isolation of the intermediate amide-acid 16 and/or 17;

3. the reaction is reproducible and even at large-scale (up to 15 g); and finally

4. the reaction, in particular its yield, seems to be very sensitive to the analytical quality of AcOH which has a pivotal role as a solvent and probably as a catalyst.

Reactions of unsymmetrical imides with hydride agents (Hitchings and Vernon, 1988, 1990; Wijnberg et al., 1975; Kim et al., 1981; Goto et al., 1989; Mamouni et al., 1996; Chihab-Eddine et al., 2000; Ding et al., 2015) and organometallic reagents (Bahajaj et al., 1994, 2001; Saito et al., 2001) are very scarce and resulted often in the formation of two regioisomers depending strongly on the reaction conditions. In spite of the importance of quinoline and derivatives in medicinal chemistry notably, only one paper (Chiurato et al., 2010), in addition to our contribution (Pin et al., 2007), is dedicated to the reduction reaction of N-substituted pyrrolo-quinolinediones. These investigations were inspired also from the investigations given in pyrrolo[3,4-b]pyridinediones series as outlined previously (Hitchings and Vernon, 1988, 1990) and especially from the ones given by Goto’s group in 1989 (Goto et al., 1989).

In this context, attempts to reduce regioselectively one carbonyl function of the model imide 11aa was performed in a mixture of MeOH/CHCl₃ in a 1/1 ratio (V/V) by using 1.5 equivalents of NaBH₄ in the presence of Mg(ClO₄)₂ at –20 °C (Goto et al., 1989; Chiurato et al., 2010). The optimal conditions were reached with the use of 2 equivalents of Mg(ClO₄)₂. Thus under these conditions, N-phenylethyl imide 11aa deliver after 2 h of reaction the expected hydroxy lactam 12aa in 85% yield after purification by chromatography on silica gel column using cyclohexane/EtOAc (1/2) as the eluent (Pin et al., 2007). Importantly, no traces of the hydroxy lactam regioisomer were detected in the crude reaction mixture.

Having demonstrated the efficiency of the quinoline-imide complexation by Mg(ClO₄)₂, which resulted in the activation of one C⚌O function as shown in the scheme of Table 3, this reaction thus was extended successfully to other imides 11 and the results are summarized in Table 3. Interestingly, the reduction occurred regioselectively providing the expected corresponding hydroxyl lactams 12 in yields ranging from 69% up to 88% after purification. Advantageously, the nature and the position of the substituents, particularly on the pyrroloquinolinedione scaffold bearing the imide functionality, did not exert any influence on the reduction process.

Table 3 Reduction of imides 11 into hydroxyl lactams 12.

Entry	Imide	R1	R2	R3	R4	Hydroxy lactam	Yield (%)a
1	11aa	H	H	H	H	12aa	85
2	11ab	H	H	MeO	H	12ab	77
3	11ac	H	H	MeO	MeO	12ac	69
4	11ba	H	Me	H	H	12ba	87
5	11bb	H	Me	MeO	H	12bb	83
6	11bc	H	Me	MeO	MeO	12bc	88
7	11ca	H	Ph	H	H	12ca	79
8	11cb	H	Ph	MeO	H	12cb	70
9	11cc	H	Ph	MeO	MeO	12cc	81
10	11da	Cl	Ph	H	H	12da	79
11	11db	Cl	Ph	MeO	H	12db	85
12	11dc	Cl	Ph	MeO	MeO	12dc	80

a Yields obtained for isolated 12 after purification by chromatography.

The hydroxyl lactams 12 being obtained could allow reasonably a facile approach via an intramolecular catalytic π-cationic cyclization to new pentacyclic compounds 6 analogues to the active natural aromathecines 5 as potent topo-1 inhibitors. To the best of our knowledge, the use of the present N-acyliminium precursors 12 to form a central six-membered ring as a piperidine ring fused to other nucleus represents a novel illustration of the α-amidoalkylation chemistry. In addition, no examples of intramolecular trapping of these π-cationic species was reported. But just now, only one example of such C—C bond forming reaction, using an enol function as internal nucleophile for trapping the N-acyliminium ion, is reported (Chiurato et al., 2010).

As shown in Scheme 4, treatment of hydroxy lactam 12aa at the outset with BF₃. Et₂O in dichloromethane for 24 h was unsuccessful and even if the reaction time was prolonged for additional 48 h and the quantity of Lewis acid increased up to 6 equivalents relative the reactant 12aa. In all these exploratory attempts, the starting hydroxyl lactam 12aa was recovered intact after the reaction hydrolysis.

Close

Scheme 4 TFA-catalyzed intramolecular α-amidoalkylation of hydroxy lactams 12 into the targeted pentacyclic systems 6 analogous of aromathecines.

Export to PPT

Also, the use of catalysts traditionally useful in this type of reaction such as TiCl₄, SnCl₄, SnCl₂, AcOH, etc. (See for example: Yazici and Pyne, 2009a,b) was similarly ineffectual. However, the use of AlCl₃, with high analytical purity of 99.99%, at reflux of dichloromethane resulted in the formation of only traces of the expected cyclized product 6aa. Besides, when the hydroxy lactam 12aa is reacted in neat TFA (5 mL of TFA for 1 mmol of hydroxy lactam 12aa) at room temperature for 48 h according to our precedents reports (Saber et al., 2010; Fleury et al., 2011; Hucher et al., 2001, 2005; Pesquet et al., 2005), in our satisfaction, we isolated after a classical work up, a unique product identified to be the pentacyclic compound 6aa in a good yield of 82% (Scheme 4). Given the slowness of this transformation, we thus carried out the reaction with same quantity of TFA but at reflux hopping to increase sensibly the reaction kinetic. Surprisingly, we isolate another cyclized product identified to be the pentacyclic compound 19aa bearing a trifluoroacetate group at the angular carbon in appreciable yield of 66% after purification by chromatography on silica gel column.

Having demonstrated the power of TFA (in precise ratio to the reactant 12), among others Brönsted or Lewis acids, to promote the intramolecular α-amidoalkylation reaction via the formal N-acyliminium cation I generated in situ from hydroxyl lactam 12aa, we envisioned then to extend this reaction to hydroxy lactams bearing other aryl π-nucleophiles (Scheme 4). Thus, the use of both protocols different only by their reaction temperature and time (Method A and Method B) as in Scheme 4, demonstrates the effectiveness of the Method B in all cases. Under these conditions, hydroxy lactams 12 undergo the π-cationic cyclization again via the intermediate of type I to form the cyclized products 6, surprisingly, in yields ranging from 59% up to 73% after purification except in the case of the hydroxy lactam 12ca. In the latter case, the reaction delivers as for N-acyliminium precursor 12aa, the angular substituted pentacyclic product 19ca in 65% yield comparable to the one obtained for 19aa (66%; See structures in last line of Scheme 4). It should be noticed here that only the hydroxy lactam 12ca has required the use of the method A to reach the cyclized product 6ca (72%) in which only the π-cationic cyclization was taken in consideration. Ultimately, from these results and as in the case of previous ‘imidation’ and reduction reactions, it seems that the nature and the position of the activated and/or deactivated substituents in left (electrophile) and right (nucleophile) parts of the hydroxyl lactams precursors, does not influence significantly the cyclization process.

Taking into account the less reactivity of hydroxy lactams 12, their intramolecular arylation through cation intermediate into corresponding pentacyclic systems 6 require generally drastic conditions. The latter are generally justified with regard to the results obtained in pyridine series by Vernon’s group (Hitchings and Vernon, 1988, 1990; Bahajaj et al., 1994, 2001). In these cases, PPA at high temperature (100–125 °C) or TFA at reflux were necessary to complete the cyclization reaction which accompanied in most cases with partial decomposition of the starting materials.

Besides compounds 6, the formation of products 19aa and 19ca, in particular, can be explained if we consider the peroxidation in acidic medium with oxygen from air at the angular carbon of the pentacyclic systems 6aa,ca to furnish the peroxide derivative A (Scheme 5). The latter, is transformed directly into D via B following protonation and loss of hydrogen peroxide. Alternatively, B can lead to C after H₂O₂ displacement with H₂O (generated from the intramolecular arylation leading to the cyclic system 6) followed by dehydration of C in acid media. At this stage, an ultimate amidolakylation of the stable cation D using oxygen atom from CF₃CO₂⁻ group as external nucleophile occurred and delivered the cyclic trifluoroacetate derivatives 19aa and 19ca in comparable yields. This hypothesis can be corroborated also with the observations made by numerous authors, including members from our collaborators. The latter claim an easy peroxidation reaction in the presence of oxygen (Murahashi et al., 1991; Sánchez et al., 2001) of bicyclic keto-amides (Haeusler et al., 1978; Yates and McLachland, 1978), tricyclic keto-amides (Bourry et al., 2004, 2006; Chang et al., 2013), as well as tricyclic bis-keto-amides related to our systems (Godard et al., 1971).

Close

Scheme 5 Plausible mechanism for the formation of the pentacyclic products 19aa and 19ca.

Export to PPT

The structure elucidation of the cyclized products 6 and 19 as well as all intermediates was based on their spectroscopic data (IR, ¹H NMR and ¹³C NMR including NOE Difference and DEPT experiments) as well as their elemental analyses.

Indeed, in the ¹H NMR spectra of compounds 6 the angular proton signals absorb downfield compared to the same protons of their hydroxyl lactams congeners 12 when R₃ and R₄ are H but absorb upfield, when at minimum, one of both groups R₃ and R₄ is equal to MeO group. Interestingly, these correlations are very good except for the sole pentacyclic product 6aa while we observe no angular proton signals for compounds 19. On the other hand, in the ¹³C NMR spectra of 6, we observe the appearance of an additional quaternary carbon in the aromatic region which disappears in the DEPT program spectra. In the case of trifluoroacetate derivatives 19, we observe similarly two quaternary carbons which also disappear in the aromatic region in the DEPT program spectra. These facts are in agreement with these observed for related compounds thus attesting that the π-cationic cyclization occurs alone for the production of derivatives 6. In both cases, this is associated with an oxidation reaction of the angular carbon followed by an intermolecular O-amidoalkylation via the trifluoroacetate anion present in the acid medium for the production of compounds 19aa and 19ca.

3 Conclusions

In summary, we have documented successfully a general five-step synthesis of a short library of benzo[7,8]indolizino-quinolines analogues of the potent topoisomerase-1 (topo-1) poison rosettacin (5a) and 22-hydroxyacuminatine (5c) alkaloids. The sequence starts from readily available ortho-ketoanilines and dimethyl acetylenedicarboxylate (DMAD). Thus, methyl quinoline-2,3-dicarboxylates were generated by the effective tandem aza-Michael addition/cyclodehydration in appreciable yields and their hydrolysis provide corresponding o-dicarboxylic acids. The latter with various amines deliver N-substituted imides via the tandem intermolecular amidation/cyclodehydration reaction promoted by glacial acetic acid. The hydroxy lactams, obtained by the regioselective reduction using sodium borohydride in methanol under the promotion of Mg(ClO₄)₂ as Lewis acid, in turn, furnished in neat TFA an assembly of new benzo[7,8]indolizinoquinolines in acceptable to good yields. Among Lewis and Bönsted acids tested, only TFA promotes in quinoline series the π-cationic cyclization via stable N-acyliminium species as an important key step.

Ultimately, the power of this general practical approach was demonstrated by:

1. the four points of the diversification made,

2. the generation of trifluoroacetate group at the angular position of certain pentacyclic systems could constitute a valuable N-acyliminiun precursor for trapping further various nucleophiles, and

3. many steps of the ‘five-step’ sequence we used could be conducted in large scale. These facts outline in part its potential applications in chemical synthesis necessary for the future QSAR studies.

4 Experimental section