Enantioselective synthesis of atropisomeric indoles via iron-catalysed oxidative cross-coupling

Heterobiaryl compounds that exhibit axial chirality are of increasing value and interest across several fields, but direct oxidative methods for their enantioselective synthesis remain elusive. Here we disclose that an iron catalyst in the presence of a chiral PyBOX ligand and an oxidant enables direct coupling between naphthols and indoles to yield atropisomeric heterobiaryl compounds with high levels of enantioselectivity. The reaction exhibits remarkable chemoselectivity and exclusively yields cross-coupled products without competing homocoupling. Mechanistic investigations enable us to postulate that an indole radical is generated in the reaction but that this is probably an off-cycle event, and that the reaction proceeds through formation of a chiral Fe-bound naphthoxy radical that is trapped by a nucleophilic indole. We envision that this simple, cheap and sustainable catalytic manifold will facilitate access to a range of heterobiaryl compounds and enable their application across the fields of materials science, medicinal chemistry and catalysis. Direct oxidative methods for the enantioselective synthesis of heterobiaryl compounds that exhibit axial chirality remain elusive. Now, the use of an iron catalyst in the presence of a chiral PyBOX ligand and an oxidant enables the direct coupling of naphthols and indoles with high levels of enantio- and cross-selectivity.

Heterobiaryl compounds that exhibit axial chirality are of increasing value and interest across several fields, but direct oxidative methods for their enantioselective synthesis remain elusive. Here we disclose that an iron catalyst in the presence of a chiral PyBOX ligand and an oxidant enables direct coupling between naphthols and indoles to yield atropisomeric heterobiaryl compounds with high levels of enantioselectivity. The reaction exhibits remarkable chemoselectivity and exclusively yields cross-coupled products without competing homocoupling. Mechanistic investigations enable us to postulate that an indole radical is generated in the reaction but that this is probably an off-cycle event, and that the reaction proceeds through formation of a chiral Fe-bound naphthoxy radical that is trapped by a nucleophilic indole. We envision that this simple, cheap and sustainable catalytic manifold will facilitate access to a range of heterobiaryl compounds and enable their application across the fields of materials science, medicinal chemistry and catalysis.
Atropisomeric biaryls comprise a privileged class of compounds whose applications span the fields of medicinal chemistry, catalysis and materials science; as such, a panoply of elegant and efficient methods have been developed for their synthesis 1 . The most convergent route to biaryls is generally the transition metal-mediated cross-coupling of two partners 2,3 (although notable advances in metal-free methods have been demonstrated recently) 4 . While this strategy generally results in cross-coupled products in good yields and predictable levels of chemo-and regioselectivity, these advantages may be offset by the requirement to synthesize two specifically functionalized coupling partners (Fig. 1a) 5 . In principle, oxidative coupling represents a more direct, atom economic and environmentally benign approach as it creates the desired aryl-aryl linkage from two C-H bonds 6,7 . This realization has led to a number of oxidative homocoupling procedures that can generate C 2 -symmetric BINOL (1,1'-bi-2-naphthol)-like structures in an enantioselective fashion. These include reactions mediated by transition metals, including copper [8][9][10] , iron 11 and vanadium 12 , among others. However, in the absence of specific functional groups, controlling the regio-, chemo-and enantioselectivity of the corresponding heterocouplings remains a formidable challenge, and successful examples have been limited to the synthesis of BINOL-or NOBIN (2-amino-2'-hydroxy-1,1'-binaphthyl)-type scaffolds [13][14][15][16] . In particular, Katsuki 17 demonstrated that iron salan complexes are effective in the enantioselective heterocoupling of naphthols, and Pappo 18,19 showed that chiral iron phosphate and disulfonate complexes act as effective precatalysts for the enantioselective synthesis of non C 2 -symmetric BINOLs and NOBINs (Fig. 1b). We considered whether this oxidative cross-coupling approach could be used in the development of a method for the enantioselective synthesis of axially chiral 3-aryl indoles, which are emerging as valuable members of the atropisomeric biaryl family 20 . The synthesis of atropisomeric aryl indoles has been a particular focus for enantioselective catalysis since the definitive reports of 3-aryl indoles from Li and Shi 21 , Gu 22 and Tan 23,24 . Article https://doi.org/10.1038/s41557-022-01095-9 be ΔG ‡ 353K = 26.0 kcal mol -1 ; the potential configurational lability of this material (a class 2 atropisomer) 33 motivated us to continue to investigate compounds with higher rotational barriers. We reasoned that a larger substituent at the C-2 position of the indole would substantially increase the barrier to rotation of the product. Hence, we subjected 2-tert-butylindole to the same reaction conditions; this afforded 39% of the desired heterocoupled product 3b in addition to a 48% yield of the homocoupled BINOL product (Table 1, entry 2). One strategy to mitigate homocoupling is to modulate the oxidation potential and nucleophilicity of the phenol component through the installation of an electron-withdrawing group [34][35][36] . When a naphthol bearing a C-3 methyl ester 1b was used in the cross-coupling reaction with 2-tert-butylindole without a large excess of either component, exclusive formation of the cross-coupled product 3c was observed in 89% yield (Table 1, entry 3). The rotational barrier of this molecule was determined to be ΔG ‡ 413K = 38.3 kcal mol -1 ; a barrier of this magnitude essentially precludes racemization unless forcing thermal conditions are employed.
With an effective catalyst system in hand for the chemoselective production of the desired heterobiaryl, we focused on selection of an appropriate chiral ligand to facilitate an atropselective reaction. We discovered that ligands of the bis-oxazoline family were viable for an enantioselective transformation, with phenyl-substituted PyBOX ligand L1 affording the heterocoupled product in the presence of anhydrous iron(iii) chloride in a modest enantiomeric ratio (e.r.) of 60:40 (ref. 37 ). Both the yield and the e.r. were improved (80% yield Oxidative cross-couplings between indoles and phenols have been disclosed in the synthesis of benzofuranoindolines 25,26 , which are key components of complex natural products, including diazonamide and phalarine (Fig. 1c). A range of oxidants (including hypervalent iodine reagents 27 , iron(iii) salts 28 and electrochemistry 29 ) have been successfully employed in such cross-coupling reactions, some of which have shown exceptional levels of cross-coupling selectivity. We reasoned that if the steric bulk on the phenol-and particularly the indole component-was increased the overall process could favour rearomatization rather than [3 + 2] annulation to generate an atropisomeric heterobiaryl (Fig. 1d).

Results
We began by screening a range of oxidants for the reaction between a 1:1.1 mixture of 2-naphthol 1a and 2-methylindole. In a preliminary screen of conditions (see Supplementary Information), hypervalent iodine reagents and VOF 3 30 were poorly selective for the desired heterocoupling process, while [Cu(OH)•N,N,N',N'-tetramethylethylenediamine] 2 Cl 2 in air favoured formation of the homocoupled BINOL product. We were delighted to find that catalytic iron(iii) chloride in 1,1,1,3,3,3-hex afluoroisopropanol (HFIP) with di-tert-butylperoxide as co-oxidant 31,32 gave exclusively the cross-coupled heterobiaryl product 3a in 96% yield, as a single-indole C-3 regioisomer ( Table 1, entry 1). The remarkable selectivity of this reaction is noteworthy: we did not observe any trace of the potential homocoupled BINOL 4 or 3,3′-bisindole 5 side products. The rotational barrier of this product was determined to Classical cross-coupling Oxidative cross-coupling . This is probably due to the greater solubility of complexes derived from the hexahydrate salt versus anhydrous Fe salts. We recognized that the group at C-3 of the naphthol partner might also have an impact beyond enhancing chemoselectivity and found that the e.r. of 3e increased (to 84:16) with a larger (phenyl) ester group. We subsequently explored different PyBOX ligands in combination with changes at the C-3 ester, finding that the combination of 1-naphthyl-substituted PyBOX ligand L4 with a 2-isopropylphenyl ester substrate afforded the heterobiaryl product 3e in 63% yield and 92:8 e.r. (Table 1, entry 13). Modulation of the reaction time, an increase in the quantity of ligand (to 12 mol%) and indole (to 1.5 equiv.) and sonication of the Fe salt before ligand addition all continued the aggregation of marginal gains to ultimately afford the desired heterocoupled biaryl product in 91% yield and 94:6 e.r. These conditions were not effective with 3-cyano-substituted naphthol (1e), leading to <5% conversion, but the cross-coupling of 3-bromo 2-naphthol 1f to afford 3g proceeded with perfect chemoselectivity and good enantioselectivity (92% yield and 85:15 e.r.). With an optimized set of reaction conditions, we explored the substrate scope for the reaction (Table 2), focusing on naphthol 1d. In all cases, the mass balance was accounted for as unreacted naphthol starting material, and the 3,3′-bisindole product 5 was not observed. The size of the indole C-2 substituent is crucial in determining the rotational barrier of the biaryl products; hence, we explored substrates with sterically hindered groups in this position. An indole bearing a tert-amyl group couples effectively to afford 3h in 83% yield and high enantioselectivity (95:5 e.r.) with no trace of other products. Similarly, 2-adamantyl and bicyclo[2.2.2.]octane groups are also tolerated to afford 3i (71% yield; 91:9 e.r.) and 3j (80% yield; 86:14 e.r.) with good levels of selectivity. Other cycloalkyl groups, including 2-methylcyclopentyl and 2-methylcyclobutyl, are also effective in this reaction, affording biaryls 3k (75% yield; 94:6 e.r.) and 3l (90% yield; 96:4 e.r.), respectively. Changing to a smaller group at this position, such as iso-propyl, does not impact the cross-coupling efficiency or chemoselectivity (affording 3m in 93% yield as the sole product) but does lead to a reduction in enantioselectivity (to 60:40 e.r.). Enantioselectivity is restored with an indole bearing an α,α-dimethylbenzyl group (to afford 3n in 80% yield and 96:4 e.r.). The combination of a phenyl ester on the naphthol with the α,α-dimethylbenzyl indole was also viable and more selective in this reaction, generating 3o in 85% yield and 98:2 e.r. We subsequently explored substitution around the indole ring, and indoles bearing halogens such as fluorine or chlorine both undergo oxidation without incident to afford biaryls 3p (80% yield; 90:10 e.r.) and 3q (82% yield; 94:6 e.r.), respectively. Next, we examined C-5 substitution on the indole reactant. Electron-donating groups such as methoxy are highly effective, affording biaryl 3r in 97% yield and 93:7 e.r. 5-Alkyl groups are also well tolerated, affording biaryl 3s in 78% yield and 93:7 e.r. We observed that electron-withdrawing groups such as fluorine in this position led to lower conversions. as in 3t (60% yield; 92:8 e.r.), and considered that this may provide some insight into the mechanism of this transformation. Consequently, we    decided to study a series of different electron-withdrawing groups in this position to evaluate the impact on the reaction. A 5-bromo substituent led to biaryl 3u in only 40% yield, but with a relatively high enantioselectivity (90:10 e.r.). More powerful electron-withdrawing groups on the indole coupling partner led to the generation of biaryls bearing an ester 3v (40% yield; 85:15 e.r.), a trifluoromethyl group 3w (28% yield; 72:28 e.r.) or a cyano group 3x (13% yield; 85:15 e.r.) in lower yields and enantioselectivities; we also saw a reduction in chemoselectivity as manifested by the competitive formation of small quantities of the C 2 -symmetric BINOL product. It is clear that the electronic nature of the substituents on the indole has an impact on conversion and selectivity. C-6 substitution is tolerated, albeit with slightly lower enantioselectivity: 6-methyl 3y (79% yield; 88:12 e.r.), 6-fluoro 3z (63% yield; 84:16 e.r.) and 6-chloro 3aa (56% yield; 89:11 e.r.) are all effectively produced. An indole bearing a C-7 methyl group is also a competent partner in this reaction, leading to the corresponding biaryl 3ab in 85% yield and 86:14 e.r. Next, we examined whether the introduction of different groups on the naphthol coupling partner was possible. A 6-bromo substituent coupled effectively to afford biaryl 3ac in 66% yield and 93:7 e.r.; a more conjugating group in 6-phenyl was also successful to afford 3ad with lower conversion (52% yield) and 90:10 e.r. We were able to accommodate groups on both coupling partners: a 6-methylindole coupled with a 6-bromo naphthol to exclusively afford the heterocoupled biaryl 3ae (67% yield; 88:12 e.r.). This principle can be extended to the formation of different biaryls such as 3af (76% yield; 91:9 e.r.). Different substituents on the naphthol component can also be combined with different C-2 substituents on the indole component to afford an array of different products; these are exemplified by the formation of biaryl compounds bearing bromo 3ag (54% yield; 95:5 e.r.), methoxy 3ah (82% yield; 92:8 e.r.), aryl 3ai (45% yield; 95:5 e.r.) and alkyl 3aj (66% yield; 93:7 e.r.) groups.
The atropisomeric biaryl 3o contains a number of different functional groups; to demonstrate their orthogonality, chemoselective derivatizations were implemented (Fig. 2). The C-3 ester on the naphthol 3o (97:3 e.r.), which is implicated in the observed selectivity in the cross-coupling process, can be transformed into tertiary alcohol 6 by O-functionalization followed by the addition of an excess of Grignard reagent. The C-3 ester group can also be conveniently removed by a palladium(ii)-catalysed reductive decarboxylation in the presence of stoichiometric triethylsilane to afford 7. Although this requires high temperatures, the magnitude of the barrier to rotation enables this to be performed without compromising the enantiointegrity. The phenol in 7 can simply be transformed into triflate 8; the absolute configuration of this compound was confirmed by X-ray crystallography. This compound can function as a divergent intermediate for a range of cross-coupling reactions, as exemplified by the formation of 9, through a palladium(ii) coupling with diphenylphosphine oxide. 3-Aryl indole phosphine oxides such as this have been directly reduced to phosphines, which have been employed in catalysts for organocatalytic enantioselective reactions 38,39 .

Discussion
The majority of mechanisms proposed for Fe-catalysed oxidative cross-couplings are based on an Fe(iii)/Fe(iv) cycle and broadly parallel the accepted mechanisms for the operation of haem-containing enzymes 40,41 . Cross-coupling selectivity in non-haem systems is usually determined by differences in oxidation potentials that control which cross-coupling partner is oxidized preferentially, in conjunction with other parameters that influence nucleophilicity and acidity 11,42 . To probe the determinants of reactivity and selectivity in our system, we measured the oxidation potentials of 2-tert-butylindole (0.71 V versus Ag/AgCl) and 2-isopropylphenyl 3-hydroxy-2-naphthoate (1.42 V versus Ag/AgCl) in HFIP. These measurements clearly showed that under these conditions the indole has a lower oxidation potential than the naphthol component. To determine whether our oxidation state measurements were reflected by the presence of radical species in solution, we employed electron paramagnetic resonance (EPR) spectroscopy. By stirring the indole in HFIP/DCE (1,2-dichloroethane) without the exclusion of oxygen, we were able to observe a species (g = 2.0051) in low spin concentration that we identified as the indole radical (by virtue of its characteristic 14 N hyperfine signature; Fig. 3a). In the presence of the Fe(iii) catalyst, a different species also consistent with an indole radical 43,44 could be observed (g = 2.00265; Fig. 3b). This lacked the 14 N hyperfine structure observed previously, probably due to reduced nitrogen character in the wavefunction and rapid relaxation as a consequence of being proximal to the metal centre; no other radical species apart from Fe(iii) were observable (see Supplementary Information). We were also able to capture the adducts of the proposed indole radical species (by high resolution mass spectrometry) by addition of the trapping agents triethylphosphite and 5,5-dimethylpyrroline-N-oxide in low yields (see Supplementary Information). We determined the oxidation potentials of the indoles used in the synthesis of 3t-3x (Fig. 3c) and found that the presence of electron-withdrawing groups at the C-5 position had a substantial impact: the oxidation potential of the unsubstituted indole was 0.71 V, whereas this value increased to 1.22 V for the 5-cyano derivative. This is consistent with the electronic nature of the 5-substituent on the indole limiting the ease of oxidation, which would impact the rate of formation of indole radicals. However, as the spin concentration was very low throughout our electron spin resonance heterocoupling experiments, we considered whether the observation of the indolyl radical was potentially an off-cycle event occurring independent of the cross-coupling. To probe the significance of this species further, we constructed indole 2w, which contains an internal alkene radical trap at the C2 position. We reasoned that if   c, The oxidation potentials of 5-substituted indoles and yields in cross-coupling reaction with 2-isopropylphenyl 3-hydroxy-2-naphthoate 1d indicate that less electron-rich π-nucleophiles give lower yields and lower selectivity. The oxidation potentials were measured versus Ag/AgCl. d, Outline of the catalytic cycle. We believe that the generation of indole radical 13 is an off-cycle event and that the catalytic cycle involves the formation of Fe(iv) complex 11, which undergoes single-electron transfer (SET) to give a naphthyl radical 12 that is trapped by the indole as a π-nucleophile (14). Hydrogen atom abstraction to the indolyl radical was playing a major role in the C-C bond-forming event, this species would be intercepted through an intramolecular radical cyclization rather than undergoing competitive intermolecular cross-coupling in the presence of 1c. When a mixture of naphthol 1c and indole 2w was treated with catalytic iron(iii) chloride in HFIP/DCE with di-tert-butylperoxide as the co-oxidant, this led to formation of the cross-coupled product 3ak in 69% yield, alongside a small quantity of oxidative cycliszation product of the indole 2x (<5%) and some recovered naphthol (10%; Fig. 3e). However, when indole 2w alone was subjected to identical conditions (Fig. 3f), oxidative cyclization product 2x was isolated in 34% yield (alongside 21% recovered starting material; see Supplementary Information for a discussion of the mechanism). These observations suggest that despite the lower oxidation potential of the indoles, oxidation of the heterocyclic component does not play a productive role within the catalytic cycle. The EPR observation of a very low concentration of an Fe-bound indole radical in the cross-coupling reaction thus probably reflects slow indole oxidation, which does not lead to isolable products unless it may be trapped by an appropriate (intramolecular) acceptor. This is also consistent with the lack of formation of the homocoupled indole product under the reaction conditions in the absence of the naphthol component, where indole starting material can be recovered. In contrast, we were able to isolate the homocoupled BINOL derivative in 68% yield when naphthol 1d was treated under the reaction conditions in the absence of an indole. The formation of this product probably occurs via the reaction of a ligated naphthoxy radical, which is trapped by a naphthol as a π-nucleophile 45 , and we hence considered whether this mechanism could be operative for our observed heterocoupling. Mass spectrometry of the reaction between mixture naphthol 1c and indole 2b (R 2 = t Bu) under standard conditions allowed us to identify a species consistent with an Fe-bound naphthoxide complex (see Supplementary Information for mass spectrometry evidence for this species). We considered whether the divergent reactivity of 5-substituted indoles observed previously might be explained by the relative nucleophilicity of these substrates. Mayr 46 determined nucleophilicity parameters for indoles, which demonstrate that electron-withdrawing groups in the 5 position lead to a reduction in the rate of attack on a standardized electrophile. This is coherent with observations from Baran 47 , who showed that reactions between indoles and ketone-derived radicals are less efficient with electron-deficient indoles. To probe this further, we performed a Hammett analysis of the coupling reactions that yield 3t-3x. The Hammett plot (the ratio of the initial reaction rate (k s /k u , where k s = rate constant of reaction with substituted indoles and k u = rate constant of reaction with indole) versus substituent constant σ p parameters) gave a linear graph with a negative slope (reaction constant ρ = −0.49; R 2 = 0.99; see Supplementary Information). This is indicative of the build-up of positive charge on the indole during the rate-determining transition state and is consistent with its proposed role as a π-nucleophile. In cases where the conversion to the heterobiaryl is low (3u-3x), we were able to recover both unreacted indole and naphthol. This is consistent with slow trapping of the ligated naphthoxy radical with indole limiting the rate of reaction where the nucleophilicity is relatively low and is also reflected in the (incrementally) lower ratio of heterocoupled:homocoupled products when electron-deficient indoles are used.
We propose that the Fe(iii) salt forms octahedral PyBOX complex 10 in the presence of L4, which can undergo ligand exchange to form a complex in which the naphthol binds in a bidentate fashion (see Supplementary Information for mass spectrometry data consistent with these complexes). Oxidation to Fe(iv) complex 11 occurs with di-tert-butylperoxide, (which also liberates a tert-butoxy radical); subsequent reversible single-electron transfer generates an Fe(iii)-ligated naphthoxy radical 12 (ref. 48 ). We propose that indole radical 13 can be generated from indoles in the presence of Fe(iii) complex 10 and an external oxidant (Fig. 3d) or by single-electron transfer from Fe(iv) complex 11. This radical may be complexed (reversibly) to the Fe(iii) centre 49,50 , which would confer extra stability to this species and potentially render it persistent 51 ; this, in conjunction with the extremely low concentration of this species, is consistent with our observation that the homocoupled 3,3′-bisindole is not a product of this reaction. We believe it is likely that this species does not play an important role in the cross-coupling reaction.
Katsuki 52 previously noted the necessity of two cis sites on the Fe centre being available to enable the binding of two naphthols for cross-coupling to generate BINOLs. In our system, binding both an indole and a bidentate naphthol on the same metal centre makes the requirement for their close approach extremely challenging from a geometric perspective. As such, we tentatively propose that addition of the π-nucleophilic indole could occur via an outer sphere mechanism in which the key facially selective addition to the Fe(iii) naphthoxy radical species is directed by the C 2 -symmetric ligand as in 14. The resultant radical could subsequently undergo hydrogen atom abstraction 53 or oxidation to afford 15 (refs. 54,55 ), followed by ligand exchange to enable release of the enantioenriched heterobiaryl system 3 and an Fe(iii) complex able to continue the catalytic cycle.

Conclusion
We have described a process for the enantioselective synthesis of atropisomeric heterobiaryl derivatives via a direct oxidative cross-coupling that constructs the key biaryl linkage from two C-H bonds. This reaction utilizes a cheap and abundant Fe catalyst in the presence of a readily available chiral PyBOX ligand to enable a remarkably chemoselective cross-coupling between indoles and phenols. We envision that this process will enable the application of these and similar heterobiaryl compounds across the fields of materials science, catalysis and medicine.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41557-022-01095-9.