A concise flow synthesis of indole-3-carboxylic ester and its derivatisation to an auxin mimic

An assembled suite of flow-based transformations have been used to rapidly scale-up the production of a novel auxin mimic-based herbicide which was required for preliminary field trials. The overall synthetic approach and optimisation studies are described along with a full description of the final reactor configurations employed for the synthesis as well as the downstream processing of the reaction streams.


Introduction
Indoles are amongst the most important bioactive heterocyclic structures being commonly encountered in the amino acid tryptophan (1), the related neurotransmitter serotonin (2) as well as numerous complex alkaloid natural products and pharmaceuticals [1][2][3][4]. Indoles also play a significant role as phytohormones that promote and regulate the growth and development of plants. Indeed, four of the five endogenously synthesised auxins produced by plants contain the indole motif ( Figure 1,  structures 3-7). As a consequence of their regulatory activity these structures have become prime targets for investigations into both enhancing plant growth as well as targeted plant growth inhibition generating new agrochemical herbicides [5]. A recent collaboration investigating the uptake and resulting distribution of synthetic indole-3-acetic acid analogues in broad leaf plant species (dicotyledons) required the preparation of structure 8 (Figure 1) at scale for extended field trials. Furthermore as weather patterns and environmental concerns impact significantly on the timing and ultimately the quantities of material required in such studies it was also deemed highly desirable to be able to produce material on demand using a flexible scale flow chemistry approach.
Since the report of the first synthetic access to indoles by Fischer in 1835 more than a dozen further unique indole syntheses have been reported showcasing the importance of developing new entries into these valuable structures [6]. However, common to the majority of these indole syntheses is the use of hazardous entities such as hydrazines (Fischer), diazonium species (Japp-Klingemann) or azides (Hemetsberger-Knittel) or the necessity to construct specifically functionalised precursors in a multistep sequence prior to indole ring formation [7]. In order to address these potential shortcomings we set out to develop a benign process relying on inexpensive substrates and non-toxic reagents that would rapidly deliver the desired indole in a readily scalable and continuous fashion.

Results and Discussion
In order to generate the core indole unit through a robust synthetic sequence we decided to investigate the treatment of a 2-chloronitrobenzene 9 with ethyl cyanoacetate (10) as the nucleophile in a base-mediated S N Ar reaction (Scheme 1). The resulting adduct 11 would then be subjected to heterogeneous hydrogenation conditions to produce the indole product 12 through a reductive cyclisation sequence. From the corresponding ester functionalised indole 12 we anticipated that condensation with hydrazine would furnish the corresponding acyl hydrazine 13 which could be cyclised to the desired product 8 through the action of a reactive carbonyl donor such as CDI (1,1′-carbonyldiimidazole) or triphosgene (bis(trichloromethyl)carbonate). This synthetic strategy presented various advantages as it relies on readily available substrates and reagents, creates small amounts of non-toxic byproducts (base·HCl, H 2 O) and uses industrially favourable hydrogenation protocols in the key cyclisation step.
To commence the study we first conducted a comprehensive screening program to determine flow compatible conditions for the formation of compound 11 optimising for solvent, base, temperature and reagent stoichiometry -selected results are presented in Table 1.
Although DMF and MeCN were shown to be excellent solvents for the reaction ( Table 1, entries 3-6, 10, 15-23, and 26) we encountered difficulties in efficiently extracting the product upon quenching the reactions (1 M HCl). EtOAc was promising and made extraction very easy but during the reactions small quantities of a dark red, sticky, precipitate were observed (Table 1, entry 12). This was considered problematic for processing in a meso flow reactor due to the potential for causing blockages. It was however found that the addition of between 10-20% v/v MeCN ensure a fully homogeneous solution and also allowed for simple aqueous extraction with good recovery (>90%). However, from the provisional results DCM stood out as the most viable solvent (Table 1, entries 7, 24 and 25) and allowed a 0.25 M solution of substrate 9 to be processed with TMG or DBU (2.5 equiv) and ethyl cyanoacetate (10, 1.1 equiv) at 50 °C in quantitative conversion.
Based upon these screening results we devised a simple flow set-up where two stock solutions were united at a simple T-mixing piece and subsequently directed into a heated flow coil reactor maintained at 50 °C (Scheme 2). The intensely red coloured solution (anion of the S N Ar adduct) [8] which quickly formed was quenched after the incubation period (35-108 min) Scheme 2: Initial flow reactor setup for the synthesis of intermediate 11.
using a third flow stream of hydrochloric acid (1 M) blended via a dedicated mixer chip before the combined mixture was phase separated.
To enable the phase separation we utilised a passive membrane system based upon a modified Biotage universal separator [9]. This enabled the heavier chlorinated phase to be removed from the lower connection and for the lighter aqueous phase to be decanted from an overflow positioned run-off. At flow rates of 0.5-1.2 mL/min emanating from the main reactor this unit performed reliably giving excellent quenching and separation. However, at higher flow rates issues were encountered with incomplete partitioning (some emulsion formation) of the biphasic mixture resulting in the loss of product containing material to the aqueous run off. This was determined to be a result of the high shear generated in the in-line mixing chip at the higher flow rates and the need for longer settling times of the heavily segmented flow. This issue could be overcome by directing the aqueous run off from the first separator into a second equivalent unit or by simply splitting the original quenched flow over two parallel separators. Although functional these approaches were not the optimum in design or utilisation of the separator components. We therefore also investigated the replacement of the problematic mixing chip with various configured T-and Y-connectors but this immediately gave other issues due to incomplete quenching which resulted in poor product recovery and associated contamination. A more straightforward approach proved to be to introduce a flow stratification zone prior to the separator which was achieved through the expedient introduction of a section of wider bore tubing (expanded from the reactor i.d. 1.6 mm → 4.0 mm × 20 cm) [10]. This enabled reactor throughput flow rates of up to 2.0 mL (for flow rates above 1.0 mL a second 52 mL flow coil was added to the system) to be successfully quenched inline and then successively handled by a single Biotage universal separator. In all cases the output from the main reactor showed quantitative conversion with only the product being isolated following solvent evaporation. At a flow rate of 1.8 mL/min from the reactor this gave a maximum throughput of 27 mmol/h operating at steady state.
Despite the versatility of the reactor its productivity was lower than we ideally wanted and unfortunately DCM proved to be an incompatible solvent with the following hydrogenation step. Although solvent swapping would have been possible we determined that when EtOAc was used as the solvent and diluted with EtOH in the presence of acetic acid as an additive this allowed for the successful reductive cyclisation to the indole. As a result we investigated the scaled synthesis of intermediate 11 in EtOAc. Having had previous success with handling slurries in flow using the Coflore, AM technology ACR device [11], we decided to utilise this equipment in the synthesis to overcome the issues encountered with solubility [12][13][14].
The starting materials were prepared as individual stock solutions in EtOAc and mixed sequentially, first the base and the ethyl cyanoacetate (10) being combined (note this was an exothermic process) before meeting a solution of the aryl chloride 9 and entering the ACR which was agitated at 8 Hz (Scheme 3).
In an attempt to intensify the process and with the specific aim of reducing the amount of solvent in order to maintain a viable working concentration in the subsequent reduction step (after dilution with EtOH) we undertook a series of concentration and flow rate optimisations. It was ultimately found that at a combined flow rate of 3 mL which equated to a residence time of approximately 30 min and a reactor temperature of 65 °C we were able to achieve a quantitative conversion of a 1.5 M solu-Scheme 3: Coflore ACR setup for the synthesis of intermediate 11.

Scheme 4:
Quenching and work-up of the reaction stream from the Coflore ACR for the synthesis intermediate 11.
tion of substrate 9. Under continuous operation the reactor output was a dark red suspension comprising approx. 10% solid by volume but was easily processed through the system even over extended periods of time (>14 h). This set-up gave a theoretical working throughput of 0.108 mol/h. Next, to facilitate the integrated quenching and work-up we added a static mixing element [15] at the confluence point of an aqueous solution of 2 M HCl delivered at a flow rate of 6 mL/min (Scheme 4). Evidence for effective quenching was immediately observed by the transformation of the dark red reaction mixture to a pale yellow biphasic solution which quickly phase separated upon collection. Confirmation of the successful quenching was obtained by 1 H NMR and LC analysis of the organic phase which indicated only the product 11 and trace amounts of ethyl cyanoacetate (10). Finally, manual separation followed by drying over Na 2 SO 4 and solvent evaporation gave the desired product 11 in 94-96% yield, based upon 6 sampled aliquots of 20 min each processing time.
Several laboratory approaches to the automation of batch separation have been reported using machine vision systems [16,17], and inline detection devices employing optics [18] or inductive conductivity (impedance measurements) [19] to determine phase partitions. However, due to specific project time limitations we were constrained to use a more manual approach but would certainly incorporate such a labour saving device into a future development version. Our basic system used a simple batch collection vessel with two HPLC pumps with appropriately positioned inputs to remove independently the aqueous and organic phases. Occasional manual intervention allowed intermittent recalibration of the pumps to maintain working volumes and create a suitable phase segment for removal. Again, following evaporation of the solvent a high isolated yield of the target molecule 11 was attained in 93-97%.
The collected solution generated following aqueous quenching and separation contained 0.4 M product 11 which was further diluted with EtOH to furnish a 0.2 M stock solution for use in the next reductive cyclisation step. The reduction was performed using a ThalesNano H-cube system [20,21] operating in full hydrogen mode with a 10 mol % Pd/C catalyst cartridge. At a flow rate of 0.4 mL/min and a temperature of 50 °C full conversion to the indole product 12 was achieved ( Table 2, entry 1). When the flow rate was raised to 0.8-1 mL/min complete consumption of the starting material was observed but a mixture of products was isolated ( Table 2, entries 2 and 3). Careful analysis of the composition indicated the presence of the desired product (12, 62%) along with two other compounds (Figure 2), which following chromatographic isolation, were assigned as the amino indole (14, 29%) and the N-hydroxyamino indole (15, 9%). As expected these latter two products resulting from incomplete reduction were obtained in greater amounts as the flow rate was further increased along with an increasing proportion of the hydroxyaminoindole 15. This correlates with previous literature findings employing catalysed hydrogenation [22][23][24][25][26] or stoichiometric metal (zinc or indium) mediated reduction and is consistent with a stepwise reduction mechanism (Scheme 5) [27][28][29]. It is evident from the sequence that several equivalents of hydrogen are necessary for full reduction and as a result a higher concentration (pressure) of hydrogen would be beneficial ( Table 2, entries 4-10). We also noted that according to the mechanism protonation of the anilino nitrogen could be beneficial in promoting the reduction steps as well as aiding in the loss of water (16 and 18) and ammonia (20). We therefore screened a series of acid catalysts which highlighted acetic acid (10-30 mol %) as the optimum additive ( Table 2, entries [11][12][13][14]. Stronger acids or higher loading of acid often resulted in the generation of high internal pressures and in certain cases the formation of precipitates was noted requiring premature termination of the run and a safe shutdown. Using acetic acid and limiting the acid concentration (10 mol %) whilst working at a higher internal pressure (15 bar) enabled stable and continuous operation permitting the flow rate to be raised to 1.3 mL/min whilst ensuring a >98% conversion to the desired indole 12 ( Table 2, entries 15-18) [30,31]. This translated into a throughput of 15.6 mmol/h (3.7 g/h product). The final product could be easily isolated in 93% yield as an off white solid by solvent evaporation and trituration with 9:1 hexane/Et 2 O. This purification removed both residual acetic acid and, if present, small traces of byproducts.
In the next sequence we looked at telescoping the final two steps of the process; substitution of the ethoxy group by hydrazine and then ring formation to the 3H-[1,3,4]oxadiazol-2one unit [32,33].
In this procedure a 0.95 M THF solution of compound 12 was united with a solution of hydrazine (1.0 M in THF) and directed into a heated flow coil to be superheated at 100 °C (Scheme 6). through a scavenging cartridge of QP-SA (a sulfonic acid functionalised polymer). The product 8 was obtained after solvent evaporation as a yellow solid (94%) but required recrystallization from DCM to give a white amorphous powder of high purity in 82% yield.
A series of attempts to employ dimethyl carbonate as a replacement reagent for CDI in the final step failed under a range of conditions as did trials to directly utilise (methoxycarbonyl)hydrazide (CAS 6294-89-9) in the previous acyl hydrazine forming step. However, we found that triphosgene (0.4 equiv) could be successfully used in the latter cyclisation process [34][35][36][37]. Using a similar reactor assembly as per Scheme 6 the triphosgene (0.8 M in CHCl 3 ) was combined with the solution of intermediate 13 and passed through a heated (55 °C) flow coil and then through a packed bed scavenging cartridge containing QP-DMA (N,N-dimethylbenzylamine polystyrene). Following solvent evaporation this gave the product as an off white solid in 91% isolated yield. In practice this ap-proach proved far superior to the previously employed CDI cyclisation providing higher overall yields and improved quality product direct from the reactor. Indeed, the final product 8 was of sufficient purity for direct application allowing on demand and scalable production from a stable intermediate 12 in less than 2 h from reactor start-up.

Conclusion
Using the procedures outlined we have been able to rapidly assembly the desired target molecule 8 through a flexible workflow generating sufficient material for on-going field trials. Obviously in the context of our currently assembled route the reductive cyclisation to the indole unit 12 is the process limiting step (Scheme 7

Supporting Information
Supporting Information File 1 Reproductions of 1 H and 13 C NMR spectra for the reported compounds.