Total Synthesis of (±)-Paroxetine by Diastereoconvergent Cobalt-Catalysed Arylation

A total synthesis of paroxetine is reported, with a diastereoselective and diastereoconvergent cobalt-catalysed sp3–sp2 coupling reaction involving a 3-substituted 4-bromo-N-Boc-piperidine (Boc = tert-butoxycarbonyl) substrate as a key step. A 9:1 diastereoselectivity was obtained, while a control experiment involving a conformationally locked 3-substituted 4-bromo-tert-butyl cyclohexane ring proceeded with essentially complete stereoselectivity.


Introduction
Recent years have seen a marked rise in the use of cheaper transition metals for catalytic C-C bond formation. Iron and cobalt are particularly attractive for large-scale metal catalysis as they have far lower toxicities than nickel or palladium. Recent demonstrations of their ability to catalyse the union of aryl Grignard reagents with unactivated secondary alkyl halides represent a significant advance in the synthetic potential of these emerging protocols. [1,2] To date, investigations into the diastereoselectivity of such cross-couplings are limited. [3] With 3-and 4-substituted bromocyclohexane derivatives, a number of reports have demonstrated the preferential incorporation of the "nucleophilic" component into the less encumbered equatorial position in iron-catalysed cross-coupling reactions [e.g., Scheme 1 (i)]. [3a] Similar observations have been made for the analogous cobalt-catalysed reactions, with bicycloheptanes exo-3 and endo-3 both giving exo-4 when treated with PhMgBr [Scheme 1 (ii)]. [3b] Mechanistically, the stereochemical outcome has been explained by the involvement of a radical intermediate, as seen in the proposed mechanism shown in Scheme 2. [3c,4] Ferrate complexes (I) have been proposed as the reactive species when the Grignard reagent involved is unable to undergo β-hydride elimination. Indeed, such complexes have been prepared by Fürstner et al., who showed that they  most stable diastereoisomer (e.g., trans -2, exo-4). A late transition state for the reductive elimination step has also been invoked to explain the bias towards production of the thermodynamic product. [1,3] In this paper, we report a successful application of this diastereoconvergent cross-coupling methodology in a short synthesis of paroxetine 5 (Paxil ® , Scheme 3). [9] Paroxetine is a potent inhibitor of serotonin reuptake, and is widely prescribed for the treatment of depression, obsessive-compulsive disorder, panic disorder, social and anxiety disorder, and post-traumatic stress disorder. [10]

Results and Discussion
Our retrosynthetic analysis is shown in Scheme 3, with secondary bromide 6 as a key intermediate for the introduction of the aryl residue. As the coupling is expected to be diastereoconvergent, our strategy allows for its production as a mixture of diastereomers. Consequently, the synthesis of 6 can be envisioned from commercially available 8 using standard chemistry.
A literature survey revealed that N-protected 4-bromopiperidine derivatives have been used as substrates in iron/ cobalt-catalysed coupling reactions, [11] yet none of the examples reported featured α-alkyl substitution. Moreover, cross-couplings on six-membered rings with α-alkyl substituents have little precedent. [12,13] Hence, both the reactivity and the stereochemical outcome of our key step would be instructive, given that substituted piperidines are ubiquitous in natural products and medicines. [9j,14,15] The synthesis of bromide precursor 6 was readily accomplished in three steps from known diol 7 (Scheme 4). [16] Regioselective tosylation of 7 (dr 58:42) proceeded in good yield using triethylamine (2.1 equiv.) as the base. Introduction of the sesamol group with Cs 2 CO 3 in DMF gave adduct 10 in 51 % yield. This yield improved to 76 % when a toluene solution of 9 and sesamol was exposed to aqueous NaOH using tetrabutylammonium hydroxide as phasetransfer catalyst. Finally, conversion of alcohol 10 into bromide 6 was achieved in 65 % yield through the action of bromotriphenylphosphonium bromide. Bromination of the electron-rich sesamol ring was never observed with this reagent, in contrast to related procedures using triphenylphosphine and bromine, where it proved to be a minor sidereaction (10 % yield).
We were now in a position to examine our key crosscoupling step with p-fluorophenylmagnesium bromide. For completeness, we decided to separate the diastereoisomers of 6 in order to rigorously establish diastereoconvergence for each stereoisomer. To that end, the cis and trans diastereomers of alcohol 10 were separated by chromatography, then each was brominated to give trans-and cis-6, respectively. However, it proved more convenient to separate cis and trans bromides 6 by selective precipitation from hexane/Et 2 O.
Increasing the temperature and the amount of Grignard reagent used led to a modest increase in yield (Table 1, entry 6), but the product mixture now contained significant levels of elimination product 12. It is unclear whether 12 was formed by a reductive elimination process, or by degradation of the starting material. The formation of thermody- namically more stable alkene 12 as the only observed elimination product suggests that the latter process predominates. In the iron-catalysed coupling reaction between tolylmagnesium bromide and 5-phenyl-1-bromopentane, Nagano and Hayashi described that the reaction was improved by using diethyl ether as the solvent instead of THF. [17] Unfortunately, in our case, switching the solvent to diethyl ether (Table 1, entry 7), lowered both the yield and the selectivity, as did the use of N-methyl-2-pyrrolidone (NMP) in combination with THF (Table 1, entry 8). [18] The procedure of Bica and Gaertner, i.e., the use of the ionic liquid bmim-FeCl 4 as an iron source, was also investigated, [11b] but this too gave low yields and poor selectivities ( Table 1, entry 9).
At this juncture we decided to examine the use of more reactive Co III catalysts. Pleasingly, our first reaction with Co(acac) 3 /TMEDA, [19] gave 11 in 20 % yield (Table 1, entry 10) with an improved selectivity for trans-11. The yield was elevated to 31 % by using a molar equivalent of TMEDA and 2.1 equiv. of ArMgBr ( A step change in performance was noted when we employed 10 mol-% of Co(acac) 3 with a combination of TMEDA and HMTA as additives in MeTHF (Table 1, entry 14). [4] Under these conditions, the desired product (i.e., 11) was obtained in 77 % yield with a trans/cis ratio of 90:10 ( Table 1, entry 14). The yield dropped significantly when the reaction was carried out at lower temperature (Table 1, entry 15), though the same product ratio was obtained. Attempts to increase the selectivity by adding a more dilute solution of the Grignard reagent ( Although the diastereoselectivity was satisfactory, it did not match the levels obtained with locked cyclohexyl derivative cis-1 (i.e., 96:4; Scheme 1). As a control experiment, we decided to prepare cyclohexyl analogue cis-13 and test it, using our optimised conditions, in a cross-coupling reaction with (4-fluorophenyl)magnesium bromide (Scheme 5). Surprisingly, it gave trans-14 as the sole reaction product, as judged by 1 H and 19 F NMR spectroscopic analysis. The strong conformational lock imposed by the tert-butyl group provides a possible explanation for the improved diastereoselectivity obtained in this system compared to the N-Bocpiperidine substrates (see Table 1). However, it is also plausible that N-Boc chelation prior to the reductive elimination provides additional stabilisation for an axial organocobalt intermediate. Finally, the synthesis of (Ϯ)-paroxetine was completed by removal of the Boc protecting group (Scheme 6). Following Jacobsen's conditions, [20a] our target 5·HCl was obtained in Scheme

Conclusions
In conclusion, we have developed a short route to (Ϯ)paroxetine using a cobalt-mediated cross-coupling reaction to construct the 3,4-disubstituted piperidine scaffold. The key step is notable for being diastereoconvergent, consistent with reported mechanistic studies. Importantly, for bromocyclohexane 13, the diastereoselectivity was essentially complete, whereas for N-Boc-piperidine 6 it dropped to 9:1. Notably, our synthesis of (Ϯ)-paroxetine 5 is unique in that the p-fluorophenyl ring is introduced in the penultimate step. An enantioselective total synthesis is currently under investigation.