New Approaches to Ondansetron and Alosetron Inspire a Versatile, Flow Photochemical Method for Indole Synthesis. †‡

An oxidative photocyclisation of N -arylenaminones to indoles is described, that mirrors the Fischer indole synthesis but uses anilines in place of arylhydrazines. Its value is exemplified with new approaches to the WHO-listed APIs ondansetron and alosetron. Tetrahydrocarbazolones and related heterocyclic ring systems are privileged structures in medicinal chemistry with a number of active pharmaceutical ingredients bearing such motifs. 1 For example, alosetron 1 is a 5-HT 3 antagonist used in the management of chronic irritable bowel syndrome (IBS), 2 while ondansetron 2 is a WHO listed essential medicine used to treat

Tetrahydrocarbazolones and related heterocyclic ring systems are privileged structures in medicinal chemistry with a number of active pharmaceutical ingredients bearing such motifs. 1 For example, alosetron 1 is a 5-HT 3 antagonist used in the management of chronic irritable bowel syndrome (IBS), 2 while ondansetron 2 is a WHO listed essential medicine used to treat 5a-e, X = H, 9,10 , F, 11 Cl, 12,13 Br, 13   nausea in cancer and COVID-19 patients. 3, 4 The pandemic has given impetus to calls for the on-shoring of the manufacture of such essential medicines and in that context, we sought to develop a synthesis of the tetrahydrocarbazolone core 4 of ondansetron 2 that was cheaper and more benign than the published strategies ( Figure 1). [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] The approach we envisioned (Scheme 1) used the known condensation of bulk chemicals Nmethylaniline and 1,3-cyclohexadione 12, followed by photocyclization of the resulting adduct 5a under oxidative conditions. 21  An attractive feature of the strategy was its potential to effect the synthesis without recourse to stoichiometric reagents, metal catalysts or harsh reaction conditions. It also offered high atom-economy with a reduced and more benign waste-stream compared to the current industrial synthesis involving a palladium catalysed cyclisation of aryl bromide 5d (R = Me). 13 The conversion of adduct 5a to ondansetron precursor 4 has been described previously using catalytic palladium acetate in acetic acid at 100 °C under an oxygen atmosphere. 9 Its cyclization to dihydroindole 13 has also been reported to proceed in benzene using a 400 W medium pressure mercury lamp, with subsequent oxidation to 4 accomplished using DDQ or Mn(OAc) 3 . 10 Thus, our initial focus was to establish conditions to achieve the direct conversion of 5a to 4, photochemically under flow. A viable method was established when a 0.02 M solution of 5a in Please do not adjust margins Please do not adjust margins acetonitrile containing 5 mol% iodine underwent conversion to 4 in 85% yield following irradiation with a 36 W Philips UVC lamp for 1 h. 22-24 Analogous reactions with UVA and UVB lamps had proven ineffective, largely returning the starting material with trace by-products. Similarly, the reaction was slowed when the concentration of iodine was increased, suggesting that it competed for photons with other components in the reaction mixture. 25 Pleasingly, air was also effective as an oxidant when introduced as bubbles into the flow stream. Though crude product mixtures were not as clean as those attained with catalytic iodine, the isolated yield of 4 was comparable.
Buoyed by this success, we next sought to apply the method in the preparation of alosetron precursor 14 (Scheme 2). 2,26 We were disappointed to find that precursor 16 failed to give the anticipated product 14 on irradiation with UVC light under the aforementioned oxidative cyclization sequences, delivering instead an intractable product mixture. The failure was traced to a photochemical scission of cyclic amide 14, 23 as evidenced by the formation of indole 15 as the primary product when methanolic solutions of 14 or 16 were irradiated with UVC. Pleasingly, the analogous Boc protected precursor 17 gave the reaction smoothly to deliver alosetron precursor 14 in 63% yield after deprotection with TFA. Development of the procedure into a general method for the conversion of N-arylenaminones to indoles became our next goal. While similar 6-photocyclization reactions have been described for the synthesis of indolines, these are characterised by poor productivity, low energy efficiency and limited substrate scope. 10,27 Recent advances in photocatalysis have allowed cyclisations to be performed using blue light but the need for long irradiation times and expensive iridium catalysts limits its appeal. 28 The scope of the reaction has been shown to be good for carbamate protected N-arylenaminones under UVA irradiation, 29 but for API precursors 4 and 14 this would require subsequent deprotection, indoline oxidation and N-methylation steps. Thus, the UVC-induced oxidative cyclisation procedure had the potential to provide a complimentary method for substrates bearing other nitrogen substituents.
To that end, we first targeted tetrahydrocarbazolones 5b-n bearing substituents on nitrogen and the cyclohexene ring (Table 1). Pleasingly, each proceeded smoothly in yields of ~70%. Notably, substrates bearing an N-benzyl substituent required shorter residence times than those with N-methyl and N-ethyl residues. Though secondary enaminone 5a (R = R' = H) failed to give tetrahydrocarbazolone 3, that product could be accessed from the N-Boc-indole 18c, which was delivered in high yield when using catalytic I 2 as the oxidant. Attention next turned to the influence of aromatic substituents on the course of the reaction ( Table 2). The ortho-, metaand para-tolyl and anisole derivatives 19a-f were each prepared and converted into indoles 20a-f by the aforementioned protocols. In most cases, reactions proceeded well using air as the oxidant, with hydrolysis of the starting material accounting for much of the outstanding mass balance. Indeed, this proved especially significant for the ortho-tolyl derivative 19a  20a, leading to a modest yield. In this case, switching the oxidant to catalytic I 2 had little impact on the efficiency of the reaction, in stark contrast to the ortho-anisole derivative 19d where the switch elevated the yield of indole 20d from 25% to 81%. Competitive hydrolysis also contributed to the modest 37% yield attained for indole 20g while for the para-chloro analogue 20h (49%), other side reactions reduced the yield significantly. The reaction failed to give indoles 20i or 20j from secondary enaminones 19i and 19j (R = H) but was extended successfully to the fused polycycles 20l and 20m.
Changing the size of the cyclic enone had a dramatic effect on the reactions' efficiency (Table 3). 29 Indeed, for cyclopentenone 21a (n = 0) only traces of the expected indole 22a were observed using the aforementioned procedures. Optimization studies were rewarded with the production of indole 22a in 35% isolated yield when cat. I 2 was used in conjunction with a 60 W lamp. Under these conditions, cyclisation was able to compete with hydrolysis of the starting material. The same conditions also allowed cycloheptenone 21b to be transformed into indole 22b in 56% yield, but proved ineffective with cyclooctenone 21c, which returned N-methylaniline as the major product.   (Table 3). While the dihydroindole and dihydroquinoline derivatives failed to give indoles 22d and 22e, the corresponding tetrahydrobenzoazepine was successfully converted into tetracyclic indole 22f. These observations indicate that the size of the ring fused to the N-arylenaminone determines whether a reactive conformer with good orbital overlap can be achieved. 24 Table 3. Extension to other fused ring systems.  Initial attempts to induce cyclisations of acyclic N-arylenaminones (Table 4) were beset by low yields and the recovery of starting material. 30 A breakthrough came with the observation that reactions could be induced under conditions of high dilution. Thus, at 0.002 M in acetonitrile using air as the oxidant, we were able to realize each of the oxidative cyclisation reactions of enamines 23a-l to indoles 24a-l. A notable anomaly was enamine 23m (R=H, R 1 =Me, R 2 =R 3 =CO 2 Me), which was returned under these conditions but gave indole 24m in 63% when acetic acid was added to the reaction mixture. We presume that acetic acid catalyses the proton transfer step, 25  26, that follows conrotatory 6-electrocyclisation of 23 via its singlet excited state (Scheme 3). Another point of note is that the substitution pattern for indole 24e mirrors that of the nonsteroidal anti-inflammatory and analgesic, pravadoline. 31 Scheme 3. Mechanistic course of the reaction.
Finally, the method has been extended to N-arylenaminones where the amine is bound to the -carbon of the enone, and bond formation occurs to its -carbon (Table 5). 32 As with the aforementioned examples, the cyclohexanone derivatives 27ac each gave the corresponding indoles 28a-c in good yield after a residence time of 1 h, while the yield for cyclisation of the acyclic enone 27d to indole 28d was modest. In conclusion, we have developed a cheap, efficient and atom economic route to indoles that has wide applicability. Strategically, the method mirrors the Fischer indole synthesis but replaces the arylhydrazine component with an aniline which are generally cheaper, less toxic and more widely available. The method allows products such as the ondansetron precursor 4 to be synthesized on a gram/hour basis using a standard laboratory set-up (see ESI