Róbert Örkényi*[a] Gyula Beke,[b] Eszter Riethmüller,[b][‡] Zoltán Szakács,[b] János Kóti,[b]
Ferenc Faigl,[a] János Éles,[b] and István Greiner[b]
Abstract: Flow chemistry proved to be a valuable technique to improve the synthesis route to melanin-concentrating hormone receptor 1 (MCHr1) antagonists with the 1H,2H,3H,4H,5H-[1,4]di- azepino[1,7-a]indole scaffold. A one-step route for the hetero- geneous catalytic hydrogenation of ethyl 4-(2-nitrophenyl)-3- oxobutanoate for the synthesis of ethyl 2-(2,3-dihydro-1H-indol- 2-yl)acetate was developed, and the use of common reducing chemicals was avoided. N-Alkylation of the indoline nitrogen
Introduction
Many biologically active natural products contain the indole skeleton or the synthetically more challenging indoline scaffold.
These compounds serve as important and rich sources of phar- maceuticals.[1–3] Several 7-arylazepinoindolines were shown to have selective 5-HT2C receptor activities,[4,5] and polycyclic indoline derivatives were tested as sensitizers of bacteria againstβ-lactam antibacterial agents;[6]other derivatives were described as promising human protein kinase inhibitors.[7]Re- cently, some novel indoline derivatives were also described as apoptosis protein inhibitors[8]and monoacylglycerol acyltrans- ferase-2 inhibitors.[9]Accordingly, numerous synthetic methods leading to these structures can be found in the literature, and the frequently applied methods have been summarized in re- cent reviews.[10,11]
1H,2H,3H,4H,5H-[1,4]Diazepino[1,7-a]indole derivatives of general formula1are melanin-concentrating hormone recep- tor 1 (MCHr1) antagonists for the treatment of obesity through regulation of appetite.[12]Target molecules were first prepared by using the discovery chemistry route shown in Scheme 1, which provided material for in vitro and in vivo studies. During the preclinical development phase, revision and optimization of the original synthetic route was inevitable to deliver the re- quired quantity of selected derivative of1.
[a] Department of Organic Chemistry and Technology, Budapest University of Technology and Economics,
Budafoki út 8, 1111 Budapest, Hungary E-mail: orkenyi.robert@gmail.com
https://www.linkedin.com/in/róbert-örkényi-38122490/
[b] Gedeon Richter Plc.,
Gyömrői út 19-21, 1103 Budapest, Hungary
[‡] Current address: Department of Pharmacognosy, Semmelweis University, Üllői út 26., 1085 Budapest, Hungary
Supporting information and ORCID(s) from the author(s) for this article are available on the WWW under https://doi.org/10.1002/ejoc.201700849.
Eur. J. Org. Chem.2017, 6525–6532 6525 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
atom was also optimized by using a purpose-built flow reactor and by design of experiment (DoE). Applying an optimal set of parameters allowed us to decrease the amount of carcinogenic 1,2-dibromoethane used by a factor of 10. Additionally, nearly complete conversion was achieved in a fraction of the original reaction time (30 min vs. 4 d); therefore, the productivity (space-time yield) of the flow-reactor system was proven to be ca. 200 times higher than that of the batch process.
In the introductory section, we discuss the known possibili- ties of the synthesis of key intermediate7and its precursors in a retrosynthetic fashion. Then, the original route (Scheme 1) will be analyzed, and the investigation of the most problematic, suboptimal steps under flow-chemistry techniques is described.
We found only one published process for the preparation of diazepinoindolone 7 through a ring-closing reaction of bro- moethylindoline 5, which could be obtained by nucleophilic substitution of corresponding 1H-indoline4with 1,2-dibromo- ethane.[4,5,12]
Ethyl 2-(2,3-dihydro-1H-indole-2-yl)acetate (4) could be syn- thesized by two main routes according to the literature. One of them involvedN-protected aminophenyl derivatives containing ethyl but-2-enoates, from which target molecule 4 could be obtained by an aza-Michael reaction.[13–15]The other main pos- sibility was the reduction of corresponding indole3. According to the literature,3could be reduced by sodium [cyanotrihydri- doborate(III)] [Na(CN)BH3] in acetic acid[4,5]or with the borane trimethylamine complex (Me3N·BH3) in highly acidic media (tri- fluoroacetic acid).[16–18]
Considering the difficulties associated with the synthesis of ethyl but-2-enoates, we chose the second pathway involving indole3.
The syntheses of ethyl 2-(1H-indol-2-yl) acetate (3) could also be divided into two main groups, excluding those reactions for which functional-group addition,[19,20] conversion,[21–23] and ring-contraction[24] reactions occurred or 3 was formed as a side product.[25,26]Several authors used 2-aminobenzyl alcohol as a starting material, from which a Wittig precursor could be obtained by the formation of the corresponding triphenylphos- phonium salt followed by anN-acylation reaction.[16,27–29]Indo- line3could be obtained in good yield in this way; nevertheless, the low atom efficiency of the used intramolecular Wittig reac- tion was not ideal for large-scale production.
Full Paper
Scheme 1. Original batchwise route for the synthesis of1. The highlighted suboptimal steps were optimized in flow. Reduction: see Table 1. The synthesis of 2can be found in the Supporting Information.
The other examples[5,30–34]involved reductive cyclization of 2, out of which the heterogeneous catalytic reduction of the nitro group followed by ring closure[30–32] seemed to be the most preferred way to prepare3.
Intermediate2could be obtained from 2-nitrophenylacetic acid (7), which was used in the acylation of malonic es- ter[30,31,35,36] or ethyl acetoacetate[35,37–39]in the presence of a strong base, and the resulting intermediate was then trans- formed into2in low to good yields. The most promising route involved the preparation of 5-[1-hydroxy-2-(2-nitrophenyl)- ethylidene]-2,2-dimethyl-1,3-dioxane-4,6-dione (11) as an inter- mediate from 2-(2-nitrophenyl)acetic acid (8).[32–34]It could be obtained from 2-(2-nitrophenyl)acetyl chloride (9) and de- protonated Meldrum's acid (10) under smooth conditions by using diisopropylethylamine (DIPEA) as the base. Heating of in- termediate11in ethanol provided2in excellent yield (see the Supporting Information).
In this article, we would like to highlight the numerous bene- fits that flow chemistry[40–42] can add to the development of the synthesis of lead compounds. One of them is the safe im- plementation of hazardous reactions,[43]because flow technol- ogy allows chemists to avoid direct contact with toxic reagents.
Heterogeneous continuous-flow hydrogenator equipment, such as the H-CubeTMreactor, allows the pyrophoric hydrogenation catalyst to be handled in a safe manner, without the need for a hydrogen cylinder in the laboratory.[40,44–48]Moreover, the use of flow chemistry for optimization studies can significantly de- crease the time required. The CatCartTM system used in this reactor system allows rapid catalyst screening, and working with small amounts of intermediates makes the optimization process more efficient than batchwise reactions from an eco- nomic point of view.
Results and Discussion
Our aim was to develop an environmentally friendly, scalable, and robust synthesis of 3,9,11-trisubstituated-1H,2H,3H,4H,5H-
Eur. J. Org. Chem.2017, 6525–6532 www.eurjoc.org 6526 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim [1,4]diazepino[1,7-a]indole derivatives 1. To achieve this, we used flow-chemistry technology only in those reactions for which it offered advantages over batch reactions. Systematic investigations were performed in our laboratory to find a con- tinuous heterogeneous catalytic method for nitro group reduc- tion and cyclization of2. In addition, we wished to develop a consecutive heterogeneous catalytic hydrogenation reaction that would make the one-step synthesis of4possible. We also investigated the N-alkylation reaction of 4 by 1,2-dibromo- ethane to decrease the reaction time, the dilution, and the large excess amount of the carcinogenic reagent.
The application of flow-chemistry techniques made this opti- mization procedure faster and safer and allowed exploration of novel synthetic possibilities.
This paper is organized into two main sections as follows:
investigation of the reductive ring-closing reaction of 2 to- gether with the reduction of indole 3to 2,3-dihydroindole 4 and the optimization of theN-alkylation reaction of4.
Investigation of the Continuous Catalytic Hydrogenation in a Heterogeneous Continuous-Flow Hydrogenator Next, we investigated the reductive cyclization reaction of 2 to ethyl (1H-indole-2-yl)acetate (3) that was described in the literature (shown in Table 1).
Reduction with zinc in acidified reaction media (Table 1, en- try 1) was neglected because of the fact that the use of a huge excess amount of the heavy metal was required, and it could not be scaled up because of stirring issues. Titanium trichloride worked well as a reducing agent (Table 1, entry 2), but this reaction required a large amount of buffer solution, which de- creased the environmental friendliness of the reaction. Never- theless, the palladium-catalyzed transfer-hydrogenation reac- tion (Table 1, entry 3) gave the best yield, whereas the catalytic hydrogenation reaction with hydrogen gas (Table 1, entry 4)
Full Paper
Table 1. Comparison of the known reductive cyclization reactions of2into3.
Entry Reaction parameters Yield[a] Yield[ref.]
[%] [%]
1 Zn (89 equiv.), sat. NH4Cl/H2O, THF, 2 h, r.t. –[b] 95[34]
2 10 % TiCl3(7 equiv.) in HCl/H2O (≈ 10–20 %) 75 75[33]
NH4OAc/H2O (4M), acetone, 1 h, r.t.
3 10 % Pd/C, HCOONH4(11 equiv.), EtOH 96 88[30]
1 h, r.t.
4 10 % Pd/C, H2(atm. pressure), EtOH, 20 h 91 –[c][32]
[a] Yield of isolated product; for procedures, see the Supporting Information. [b] The literature method was not tested. [c] No yield was given. The reaction was performed under H2pressure (3.5 bar) for 96 h.
also provided the product in excellent yield, with the benefit of an easier workup procedure: there was no need to remove the excess amount of the HCOONH4reagent by extraction.
The heterogeneous catalytic hydrogenation reaction with a continuous-flow reactor is widely used as a result of its advanta- ges.[45,48,49] In the beginning of our work, we screened three different catalysts under the same conditions in the heteroge- neous continuous-flow hydrogenator (Scheme 2), and the re- sults are shown in Table 2.
Scheme 2. Heterogeneous continuous-flow reduction with an H-CubeTMreac- tor. Catalyst screening and optimization of consecutive hydrogenation reac- tions for the synthesis of indolines.
Table 2. Results of the catalyst screening.[a]
Entry Catalyst Conversion Selectivity to3 Selectivity to4
[%][b] [%][b] [%][b]
1 Ra-Ni 65.6 33.0 0.0
2 10 % Pd/C 100.0 91.4 0.7
3 10 % Pd/Al2O3 100.0 98.5 0.1
[a] All experiments were performed in the H-CubeTMdevice with a 0.05M
solution of2in ethanol with a flow rate of 0.5 mL min–1at room temperature using the Full-H2mode (atmospheric pressure of hydrogen). [b] Determined by LC–MS (DAD: diode-array detector).
Raney-nickel catalyst (Table 2, entry 1) appeared to be less active and selective than palladium catalysts. Both batch (Table 1, entry 4) and flow results (Table 2, entry 2) were consist- ent with each other upon using charcoal-supported palladium and good conversions were reached, but the selectivity values were not optimal. Almost quantitative yield was achieved with
Eur. J. Org. Chem.2017, 6525–6532 www.eurjoc.org 6527 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim an aluminum oxide supported palladium catalyst (Table 2, en- try 3), in which case the workup of the reaction could be simpli- fied to evaporation of the reaction mixture.
We also observed the formation of a small amount of over- reduced product 4upon using charcoal-supported palladium;
therefore, it seemed logical to develop a consecutive catalytic hydrogenation method to produce compound4from2without the isolation of indole3.
Heterogeneous catalytic hydrogenation of N-unprotected indoles can be challenging owing to many difficulties, such as a highly resonance-stabilized aromatic core, which requires harsh reaction conditions. Furthermore, these compounds and the product as bases are able to poison the catalyst metal. The most promising results were achieved by Kulkarni et al.,[50]who used an acid catalyst to protonate the indoles at the C3 position to disrupt the aromaticity and to generate an iminium ion, which could be hydrogenated under less harsh conditions. Nev- ertheless, the use of an acid as a catalyst[37,51–57]raises further difficulties, such as polymerization caused by even weak elec- trophiles, as it is known that indoles are highly activated aro- matic compounds.[58,59] Unfortunately, overhydrogenation also remained an issue in acidic media, and this resulted in the formation of significant amounts of byproducts (mainly octa- hydroindoles).[50]
During the optimization, multiple parameters were consid- ered such as the solvent, the quality and the quantity of the acid catalyst, and the temperature. Although, the closest state- of-the-art method for the selective reduction of indoles was performed at 30 bar hydrogen pressure with a platinum cata- lyst,[50] we chose a cheaper palladium catalyst. We wished to achieve indole reduction at atmospheric pressure, which would be beneficial from the perspective of scale-up, as an autoclave would not be needed to perform the reaction. Consequently, all experiments were performed at atmospheric pressure (Full- H2mode). The main optimization steps are shown in Scheme 3.
In this case, water, which was used previously, could not be used as the solvent because in acidic media the ester function would be hydrolyzed. Of the three solvents employed (EtOH, EtOAc, and AcOH), acetic acid proved to be the best (Scheme 3, entry 2). This correlates with previous findings[50]that the se- lectivity towards4is increased upon increasing the nucleophil- icity of the solvent. Consequently, a less polymerized byproduct is formed along with a lower amount of theN-hydroxy deriva- tive[37] of 3. This is because in more apolar solvents, aniline andN-hydroxyaniline compounds are more nucleophilic. In the
Full Paper
Scheme 3. Main steps in optimizing the heterogeneous continuous-flow reduction with the H-CubeTMreactor. All experiments were performed at atmospheric pressure.
latter case, the ring-closing reaction followed by the loss of water in the geminal dialcohol intermediate is faster than hydrogenation (see Table S2 in the Supporting Information). As the reduction of the N-hydroxyindole derivative requires a longer reaction time, or harsher conditions, the use of EtOAc as the solvent results in a main product withm/z= 220, which is ethyl 2-(1-hydroxy-1H-indol-2-yl)acetate, and this is the N- hydroxy derivative of indole3(based on the literature[37]and LC–MS).
Nevertheless, screening of the acid catalyst (i.e., trifluoro- acetic acid, TFA; methanesulfonic acid, MsOH; trifluoromethane- sulfonic acid, TfOH) in the previously mentioned three solvents (Table S2) showed that the combination of methanesulfonic acid in acetic acid provided the best results (Scheme 3, entry 3), and all other selectivities towards4were less than 6 % in com- binations of these three solvents and the three acid catalyst.
Next, we tested the effect of the quantity of the acid catalyst (1, 3, and 6 equiv.) along with the concentration (0.005–0.05M, Table S3) on the reaction. Dilute solutions are favorable for flow hydrogenation reactions to ensure that the excess amount of hydrogen is not the limiting factor, and in this case, dilute solu- tions also decrease the possibility of the polymerization side reaction that occurs: protonated 3(indolenium cation) reacts through electrophilic aromatic substitution with3.
The best result still had medium selectivity and low conver- sion in the second step (Scheme 3, entry 4), so we elevated the reaction temperature to accelerate the reaction by using the same flow rate and the same residence time, which gave prod- uct4with good conversion and selectivity (Scheme 3, entry 5).
The best reaction parameters were also used in a batch reac- tion, from which4was isolated in 75 % yield. By this, we devel- oped a route for the synthesis of indoline4from2that gives 4in higher yield than any known two-step synthesis. Moreover, this procedure replaces all chemical reducing agents (e.g.,
Eur. J. Org. Chem.2017, 6525–6532 www.eurjoc.org 6528 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim heavy metal zinc, or cyanide-containing NaCNBH3) by environ- mentally friendly catalytic hydrogenation.
Investigation of theN-Alkylation Reaction under Flow Conditions using a Packed-Bed Reactor
First, we investigated the N-alkylation of 4 by 1,2-dibromo- ethane. The major drawbacks of this step are the long reaction time (up to 4 d), the use of a large excess amount (20–60 equiv.) of a carcinogenic reagent, and the necessity to use high-dilu- tion techniques. The latter two work against the formation of bisadduct6, which is formed as a byproduct.
Higher temperatures under pressurized conditions in over- heated solvents, which are possible in flow chemistry, could decrease the reaction time; we also wished to find an optimal set of reaction parameters for which a lower amount of the reagent was sufficient. Therefore, we investigated theN-alkyl- ation of4by 1,2-dibromoethane by using both heterogeneous (packed-bed reactor) and homogeneous conditions (loop reac- tor).
After screening runs under pressurized microwave batch conditions (see the Supporting Information), in which four solid inorganic bases (M2CO3, M = Na, K, Cs; KOH) were tested, with or without a phase-transfer catalyst (PTC), Cs2CO3without a PTC was found to be an appropriate choice. A Syrris Asia Heater Module® column was used in the heterogeneous continuous- flow alkylation. As no reaction occurred at room temperature, we could simplify the flow system. There was no need for two separate pumps or a mixer unit to combine a solution of indoline substrate 4 with the 1,2-dibromoethane reagent (Scheme 4).
Scheme 4. Heterogeneous continuous-flowN-alkylation reaction with the use of a solid inorganic base or a solid-supported base. The Omnifit® column (10 mm i.d. × 150 mm) was filled with either a mixture of Cs2CO3/Na2SO4
[1:1, Cs2CO3(2.25 g, 28 equiv.) and Na2SO4(2.25 g, 65 equiv.)] or Si-DMAP (SiliaBond®; 5.0 g, 4.65 mmol, 19 equiv., 0.93 mmol g–1). BPR = back-pressure regulator.
The premixed reaction solution was simply pumped through the packed-bed reactor filled with cesium carbonate. Although the conversion was almost complete, the selectivity towards5 was fairly low (Table 3). On the basis of literature data[60,61]and
Full Paper
the mass spectrum obtained from LC–MS (DAD) measurements of the crude reaction mixture, the main monobromo compound withm/z= 355/357 was assumed to contain the 2-bromoethyl group with aN-carbamate linker, which was presumably formed upon the reaction of4with CO2. Owing to the fact that this approach was not successful, we tried to use a solid-phase- supported basic reagent instead of the solid inorganic base, which could not serve as a source of carbon dioxide. 4-(Dimeth- ylamino)pyridine (DMAP) bound to silica (Silia-DMAP; Silia- Bond®) was the chosen base. In this case, both the conversion and selectivity depended on the residence time (Table 4, for details see the Supporting Information).
Table 3. Results of the heterogeneous continuous-flowN-alkylation reactions by using Cs2CO3as a solid base.[a]
Entry Flow rate Conversion Selectivity Selectivity
[mL min–1] [%][b] to5[%][b] (m/z= 355/357)[c][%][b]
1 0.5 86 16 57
2 0.25 86 21 63
[a] All experiments were performed as shown in Scheme 4. [b] Determined by LC–MS (DAD) measurements. [c] Based on the literature[60,61]and the mass spectra obtained from the LC–MS (DAD) measurements; the monobromo compound withm/z= 355/357 is presumably 2-bromoethyl 2-(2-ethoxy-2- oxoethyl)-2,3-dihydro-1H-indole-1-carboxylate.
Table 4. Results of the heterogeneous continuous-flowN-alkylation reactions by using a solid-supported reagent (Silia-DMAP) as a solid base.[a]
Entry Flow rate Conversion Selectivity to5
[mL min–1] [%][b] [%][b]
1 0.5 44 89
2 0.25 92 81
3 0.125 100 32
[a] All experiments were performed as shown in Scheme 4. [b] Determined by LC–MS (DAD).
Acceptable conversion and selectivity values could be achieved by this technique by using a flow rate of 0.25 mL min–1 (Table 4, entry 2). Owing to the fact that this reaction requires a stoichiometric amount of base, the column filled with Silia- DMAP exhausts rapidly. In the case of a well-established rea- gent recovery system using intelligent pumping[62] between multiple columns,[63] scaled-up manufacturing can be man- aged, albeit not economically. Furthermore, regeneration of the bed and the use of solid-supported reagents are not common practice in industry.
Investigation of theN-Alkylation Reaction under Flow Conditions using a Loop Reactor
Truly continuous production can be achieved by homogeneous N-alkylation in a loop reactor, and this is why we also investi- gated this option.
Design of experiments (DoE) is a widely used tool to opti- mize reaction parameters.[64–66] In this section, we would like to present further justification of the benefits of DoE and its combination with flow chemistry.
This section is divided into four parts: first, we chose appro- priate reagents (base, solvent) in the batch test reactions, sec-
Eur. J. Org. Chem.2017, 6525–6532 www.eurjoc.org 6529 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ond, reaction parameters that needed to be optimized were used in the DoE as independent variables; third, the flow-reac- tor system was assembled and the designed experiments were performed; fourth, further optimization of the reaction parame- ters was undertaken.
First, a few batchwise test reactions were performed to iden- tify a suitable organic base. DIPEA seemed to be applicable, as Et3N formed an insoluble hydrobromide salt as a co-product in acetonitrile during the reaction. Although these pre-experi- ments were necessary to perform the optimization in flow, we also found that if a larger amount of base was used, the reac- tion was less selective. Therefore, we gained valuable informa- tion about this one reaction parameter, and we established that the amount of DIPEA should be minimal (1.15 equiv.). Thus, we decreased the number of independent variables, and by that, we also decreased the number of experiments needed. With completion of these test reactions, the optimization could be started by the DoE.
A four-factor experimental design was proposed, in which the independent variables were the temperature (100 < T <
140; [°C]), the residence time (15 <tRes< 45; [min]), the amount of reagent (10 < 1,2-dibromoethane < 30; [equiv.]), and the dilu- tion (50 <c< 100, [mL g–1]).
Our design of experiment contained 2n–1+3 cases (in which nis the number of factors), which included eight experiments for each of the four independent variables (corners of the cube) and three more in the center point (see the Supporting Infor- mation) to identify deviation of the results. The reaction mix- tures were analyzed by LC–MS (DAD), and the effect of the inde- pendent variable on the dependent variables was examined, including the conversion, selectivity, and their geometric mean [reduced dependent variable: sq(C×S)].
A purpose-built flow-reactor system was assembled (Scheme 5a), in which the reactions were performed according to the designed experimental procedure. Similarly to the heterogeneous N-alkylation, a one-pump setup was used, be- cause no reaction occurred in the presence of base at room temperature.
The impact of all the reaction parameters was assessed by using Statistica software (see the Supporting Information). The results showed that the temperature had the greatest impact on the reduced dependent variable. The most interesting and valuable effect was that better results could be achieved by using the same amount of reagent at a higher temperature (Scheme 5b). In other words, we were able to decrease the necessary amount of the carcinogenic reagent by elevating the temperature.
Using these findings, we lowered the amount of 1,2-di- bromoethane to 7 equivalents, which was almost one tenth of the amount used in the original batchwise procedure (60 equiv.) with a conversion of 95 % and selectivity of 91 % (see the Supporting Information). A further decrease in the amount of this reagent was not feasible if we wanted to keep the value of the reduced dependent variable over 93 % (see the Supporting Information).
Moreover, almost complete conversion (95 %) was achieved with a residence time less than 30 min, which was significantly