H-Bonded Counterion-Directed Enantioselective Au(I) Catalysis.

A new strategy for enantioselective transition-metal catalysis is presented, wherein a H-bond donor placed on the ligand of a cationic complex allows precise positioning of the chiral counteranion responsible for asymmetric induction. The successful implementation of this paradigm is demonstrated in 5-exo-dig and 6-endo-dig cyclizations of 1,6-enynes, combining an achiral phosphinourea Au(I) chloride complex with a BINOL-derived phosphoramidate Ag(I) salt and thus allowing the first general use of chiral anions in Au(I)-catalyzed reactions of challenging alkyne substrates. Experiments with modified complexes and anions, 1H NMR titrations, kinetic data, and studies of solvent and nonlinear effects substantiate the key H-bonding interaction at the heart of the catalytic system. This conceptually novel approach, which lies at the intersection of metal catalysis, H-bond organocatalysis, and asymmetric counterion-directed catalysis, provides a blueprint for the development of supramolecularly assembled chiral ligands for metal complexes.

Introduction

Groundbreaking work by Toste and co-workers proved that it was possible to perform gold catalysis enantioselectively by using achiral ligands together with chiral anions, combining a dinuclear gold complex with a chiral silver phosphate to realize the asymmetric cyclization of allenes (Scheme A).[1] This contributed to the development of the field of “asymmetric-counteranion directed catalysis” (ACDC),[2] and chiral anions have since been employed as stereodirecting elements in other Au(I)-catalyzed transformations, in combination with both achiral[3] and chiral[4] gold complexes. However, nearly all examples of Au(I)-ACDC reported to date refer to reactions of prochiral allene substrates.[3,4] The only exceptions are the desymmetrization of diynes described by the group of Czekelius (Scheme B)[5] and cases of tandem Au/chiral acid catalysis wherein gold is generally not involved in the enantiodetermining Brønsted acid-catalyzed step.[6] To the best of our knowledge, all other attempts to leverage chiral counterions in more challenging asymmetric Au(I)-catalyzed reactions of alkynes,[7] which, unlike allenes, cannot be prochiral in themselves, were met with failure until now. In this regard, our group reported that mixing triphenylphosphinegold(I) chloride with BINOL-derived silver phosphates affords neutral gold complexes with an anionic phosphate ligand, which are not catalytically competent in enyne cycloisomerizations.[8] In fact, the counteranion has a profound impact on the reactivity and selectivity of all gold-catalyzed transformations, since properties such as its basicity, coordinating ability, and steric bulk influence the energy barriers for various elementary steps of the catalytic cycle.[9,10]

Scheme 1

Asymmetric Counterion-Directed Gold Catalysis

A further layer of complexity is added when the anion is chiral, as its position with respect to the reaction center, key for a successful ACDC, depends on difficult-to-predict electrostatic interactions with the cationic complex/intermediate or on the presence of a protic group on the substrate. Therefore, in the vast majority of cases, the noncovalent interactions required for enantioinduction[11] are not built into the catalytic system a priori but rather selected during optimization by a lengthy trial-and-error approach and rationalized only aposteriori.[12] Alternatively, the chiral anion can be precisely positioned by a rigid covalent linkage to the ligand, as in the “tethered counterion-directed catalysis” strategy recently disclosed by Marinetti, Guinchard, and co-workers for the enantioselective Au(I)-catalyzed tandem cycloisomerization–nucleophile addition to 2-alkynyl enones.[13] However, despite its elegance, this latter approach becomes akin to using a chiral ligand, so it is devoid of the flexibility and combinatorial potential offered by the original two-component ACDC.

Herein we detail a conceptually novel “H-bonded counterion-directed catalysis” strategy that enables the first general use of chiral anions in gold(I)-catalyzed reactions of alkyne substrates, as opposed to well-established allenes (Scheme c). Various bi- and tricyclic structures were accessed in good to excellent yield and enantioselectivity from 1,6-enynes, via 5-exo-dig and 6-endo-dig cyclizations with or without external nucleophiles.[14] To overcome the reactivity[8] and enantiocontrol[15] challenges described above, we envisioned to prepare gold(I) complexes with bifunctional phosphine ligands[16] incorporating dual H-bond donor groups[17] such as (thio)ureas[18] and combine them with BINOL-derived chiral anions[19] (Scheme ). The pendant H-bond donor would serve two purposes: (1) remove the chiral anionic ligand from the Au(I) coordination sphere, thus enabling substrate coordination and subsequent catalysis at the metal center; (2) place the chiral anion close to the substrate, ensuring good transmission of the stereochemical information.

Scheme 2

Design for Asymmetric H-Bonded Counterion-Directed Au(I) Catalysis

The feasibility of ligand abstraction via H-bonding interactions was supported by our previous studies on the self-activation of Au(I) chloride complexes featuring PPh3-based phosphinosquaramides and phosphinoureas.[18b] Regarding stereochemical considerations, we chose Buchwald-type phosphines functionalized at the meta or para position of the distal aryl ring due to both their easy electronic tuning and attractive geometric features in the context of linear dicoordinated Au(I) complexes (Scheme ).[16,20] In fact, the blocked rotation about the P–Caryl bond, enforced by the interlocking of the ortho-proton Ho between the two bulky alkyl groups, causes the P–Au–Cl and biphenyl axes to be parallel. This in turn projects the urea “anchor” for the chiral anion close to the substrate coordination site. This H-bonded ACDC approach features a modular and easy synthesis for both pieces of the catalytic system, which would click together based on predictable, well-precedented H-bond interactions between (thio)ureas and phosphate anions.[21] This strategy can thus benefit from all the typical advantages offered by supramolecularly assembled ligands for enantioselective metal catalysis,[22] most notably (i) simple preparation and tuning of the two components, which avoids long syntheses of new chiral ligands from scratch, and (ii) generation of a large library of catalysts through combinatorial and potentially automated methods, with the aim to speed up screening and optimization.

Results and Discussion

Synthesis of the Components of the Catalytic System

The bifunctional phosphino(thio)urea Au(I) chloride complexes were prepared with a high-yielding and modular three-step sequence from bromoanilines A1–3, in turn accessible in one or two steps from commercial building blocks (Scheme , see Supporting Information for details). A well-scalable Pd-catalyzed P–C coupling[23] allowed the introduction of different phosphine substituents on the biphenyl scaffold without the need to protect the free amine group. In this sense, the use of potassium phosphate in place of stronger bases such as sodium tert-butoxide was instrumental in suppressing undesired Buchwald–Hartwig coupling between the aryl bromide and the NH2 group. Phosphinoanilines B1–5 were then reacted with the iso(thio)cyanate of choice for the late-stage introduction of the key H-bond donor on ligands L1–13. Finally, ligand exchange with [(Me2S)AuCl] afforded Au(I) chloride complexes Au1–13. Thus, all complexes were prepared in four to five total steps and 42–60% overall yields except for Au12 featuring an N-aryl-N′-alkylurea, obtained in 31% overall yield. Previously described complexes Au14–16 with a triarylphosphine backbone were included for comparison.[18b]

Scheme 3

Synthesis of Phosphino(thio)urea Au(I) Chloride Complexes Au1–13

Selected X-ray structures displayed with ORTEP ellipsoid at 50% probability level; solvent molecules and selected H atoms omitted for clarity.

Representative complexes were characterized also by single crystal X-ray diffraction (see Supporting Information for a complete overview). In the solid state, all novel phosphinourea Au(I) complexes display an intact Au–Cl bond with the P–Au–Cl axis nearly parallel to the biphenyl axis, as expected. The urea moiety generally adopts an anti,anti-conformation (Au1, Au6, Au10, and Au13) but displays the much rarer anti,syn one[24] in Au4 and Au8 and shows various degrees of out-of-plane twisting of the aryl substituents (compare Au1 and Au10 in Scheme ). In the solid state, the urea group establishes H-bonds either with chloride ligands or with other urea units, depending on the ligand scaffold. As for phosphinothiourea Au(I) complexes, Au7 possessing a para-substituted ligand shows a classical [LAuCl] structure with intramolecular NH···Cl contacts. In the crystal structure of Au9, on the contrary, the S atom of the meta thiourea coordinates to gold with an almost linear P–Au–S axis, while the displaced chloride ion is stabilized by two H-bonds with the NH groups.

Regarding the other component of the catalytic system, we planned to introduce the chiral anion as a metal salt in order to simultaneously scavenge the chloride ligand from the Au(I) center. To this end, eight chiral salts (Ag1–6, Na6, and Cu6) were synthesized in 2–3 steps and 30–77% overall yield from commercially available (R)-configured binaphthols C1–4 (Scheme ). After treatment of the latter with phosphoryl chloride, the desired sulfonamide was added to the reaction mixture affording phosphoramidates[25]D1–6. Deprotonation of these Brønsted acids with silver carbonate delivered chiral silver salts Ag1–6. Crystals of Ag6 grown in toluene/pentane reveal a dimeric structure where the anion behaves as a bidentate O,O-ligand through the phosphoryl and sulfonyl O atoms (Scheme ). The Ag(I) centers are further stabilized by η2-interactions with the anthracenyl substituents, and one of them is also bound in a η1/η2 fashion to a molecule of toluene.

Scheme 4

Synthesis of Ag(I), Na(I), and Cu(II) Chiral Salts

X-ray structure displayed with ORTEP ellipsoid at 50% probability level; binaphthol scaffold and toluene in wireframe; selected solvent molecules and all H atoms omitted for clarity.

Optimization

Having built the library of [LAuCl] complexes and chiral anions, to validate our H-bonded ACDC design, we investigated the cycloisomerization of 1,6-enynes of type 1 (Table ),[26,27] a reaction that did not proceed at all employing [(Ph3P)Au(TRIP)].[8] We commenced by screening different silver salts in combination with phosphinourea gold(I) chloride Au1 at room temperature. This complex combined with AgTRIP was catalytically inactive, whereas together with less basic N-triflyl phosphoramidate[25] salt Ag1 afforded the desired product 2a in good yield and encouraging 79:21 er (Table , entries 1 and 2). The more basic N-mesyl and N-phenylsulfonyl analogues Ag2 and Ag3 induced comparable enantioselectivity but much lower reactivity (Table , entries 3 and 4). A clear trend between reactivity and basicity emerged: less basic counterions from stronger parent Brønsted acids[19,28] are essential for reactivity presumably because they coordinate less strongly to the cationic Au(I) center (which is isolobal to a proton)[29] and therefore can be abstracted more easily via H-bonds. Among N-triflyl phosphoramidate salts, Ag1 provided better enantiocontrol than Ag4 and Ag5 (Table , entries 5 and 6), which bear different groups at the 3,3′-positions of the binaphthol backbone. Overall, the results summarized in Table indicate that for a given anion, substitution at phosphorus influences basicity and coordinating ability, hence reactivity, while 3,3′ residues are responsible for enantioselectivity, as expected.[30]

Table 1

Screening of Silver Salts with Au1

entry	(R)-AgX	time (h)	yield (%)a	erb
1	AgTRIP	96	<5
2	Ag1	4.5	70	79:21
3	Ag2	4.5	7	81.5:18.5
4	Ag3	4.5	9	83.5:16.5
5	Ag4	4.5	42	57:43
6	Ag5	4.5	22	51:49

Determined by 1H NMR against internal standard.

Determined by HPLC on chiral stationary phase.

Next, the library of gold complexes was evaluated with the optimal silver salt (R)-Ag1. Only key data providing insight into the working mode of the catalytic system are shown in Table (see Supporting Information for comprehensive optimization studies). The presence of the urea group on the ligand is important for reactivity and essential for enantioselectivity, as [(JohnPhos)AuCl] with Ag1 delivered product 2a in lower yield and racemic form (Table , entries 1 and 2). Complex Au5, the thiourea analogue of Au1, was completely inactive, most likely because the S atom of the thiourea coordinates too strongly to the metal center, preventing substrate activation. This is consistent with the known thiophilicity of gold,[31] the Au–S bond observed in the crystal structure of related Au9, and previous studies on PPh3-based phosphinothourea Au(I) complexes.[18b] Moving the urea group from the para to the meta position of the biphenyl scaffold reversed the sense of enantioinduction (Table , entry 4 vs 1). Remarkably, either enantiomer of the product can thus be obtained preferentially using the same enantiomer of the chiral anion, in combination with a different achiral cocatalyst. Complexes Au14–16, equipped with urea or squaramide groups on triarylphosphine scaffolds,[18b] were poorly active and afforded (almost) racemic product (Table , entries 5–7).

Table 2

Screening of Selected Au(I) Complexes with Ag1

entry	LAuCl	time (h)	yield (%)a	erb
1	Au1	4.5	70	79:21
2	[(JohnPhos)AuCl]	6	40	50:50
3	Au5	44	0
4	Au8	4.5	70	27.5:72.5
5	Au14	44	32	50:50
6	Au15	44	10	45:55
7	Au16	24	46	56:44
8c	Au1	0.7	90	87:13

Determined by 1H NMR against internal standard.

Determined by HPLC on chiral stationary phase.

In benzene.

This data set highlights that a H-bond donor on the ligand is crucial for both reactivity and enantiocontrol. The mere presence of a chiral anion in the reaction mixture is not sufficient to transfer effectively the stereochemical information, as its proximity to the reaction center cannot be guaranteed in the absence of a suitably placed H-bond donor. These observations lend credibility to the original design, wherein the pendant urea was envisioned to enable both anion abstraction and precise positioning of the source of chirality (Scheme ). The presence of H-bonding at the heart of the catalytic system can be inferred also from solvent effects. For instance, the performance of the optimal Au1/Ag1 combination improved by replacing dichloromethane with benzene (Table , entry 1 vs 8, 79:21 vs 87:13 er). The enantiocontrol generally increased at −20 °C (Table ). A qualitative correlation between solvent polarity (defined by ENT values)[32] and enantioselectivity levels was observed, in agreement with H-bonding interactions between the urea and the anion being favored in apolar solvents. The least polar solvent for either the chlorinated or aromatic series (chloroform for entries 1–3 and toluene for entries 4–8 in Table ) delivered the product with the highest enantiomeric ratio, up to 94.5:5.5 er.

Table 3

Solvent Effect with Au1/Ag1 Catalytic Systema

entry	solvent	ENT (ref [32])	yield (%)a	erb
1	CHCl3	0.259	39	93:7
2	CH2Cl2	0.309	85	92:8
3	ClCH2CH2Cl	0.327	60	88:12
4	C6H5CH3	0.099	>95	94.5:5.5
5	1,2-(CH3)2C6H4	na	86	93:7
6	C6H5Cl	0.188	>95	91:9
7	C6H5F	0.194	75	90:10
8	C6H5CF3	na	89	90.5:9.5

Determined by 1H NMR against internal standard.

Determined by HPLC on chiral stationary phase.

Once identified toluene as the best solvent, a final fine-tuning of the catalytic system was performed (Table ). Replacing Ag1 with Ag6, i.e., swapping the triisopropylphenyl groups for 9-anthracenyl substituents at the 3,3′-binaphthol positions, led to a considerable improvement in enantioselectivity, although at the expense of conversion (Table , entries 1 and 2). In order to increase reactivity, more electrophilic Au(I) complex Au10 with a trifluoromethyl group meta to the phosphine was employed, and an excellent yield of 2a could be obtained at −10 °C while maintaining the high enantioselectivity (Table , entries 3 and 4). Complexes Au2–4, Au11, and Au12 with different substituents on the urea were also evaluated, since electronic properties influence H-bonding ability.[17,21] However, none of them surpassed the performance of complex Au10 possessing a simple phenyl urea, which offered the right balance between electronic activation and steric hindrance (see Supporting Information for details).

Table 4

Fine Tuning of the Catalytic Systema

entry	LAuCl	(R)-AgX	T (°C)	yield (%)a	erb
1	Au1	Ag1	–20	>95	94.5:5.5
2	Au1	Ag6	–20	50	98:2
3	Au10	Ag6	–20	78	98:2
4	Au10	Ag6	–10	>95	98:2

Determined by 1H NMR against internal standard.

Determined by HPLC on chiral stationary phase.

Generality of the Enantioselective 5-exo-dig and 6-exo-dig Cyclizations

The substrate scope of the formal [4 + 2] cycloaddition of enynes 1 catalyzed by the Au10/Ag6 system was then assessed (Scheme ). Enynes 1a–h bearing various linkers underwent cyclization to deliver the corresponding products 2a–h in high yield (84–95%) and with excellent enantioselectivity (96:4 to 98:2 er). The standard reaction to 2a could be performed on 2-mmol scale in 24 h under air and in technical-grade toluene, with Au(I) and Ag(I) loading reduced to 1 mol %, with a 99% yield and 98:2 er for the product. This catalyst loading is the lowest reported so far in asymmetric cycloisomerizations of enynes 1, which is remarkable considering that all previous methods relied on stereochemical information covalently embedded in a chiral ligand.[27] Apart from ethers (2a, 2b), also allyl (2c), ester (2d, 2h), carbamate (2e), tosylate (2f), and acetal groups (2g) were tolerated. As for variations on the alkyne moiety, products with electron-poor as well as electron-rich groups at the para (2i–l), ortho (2m,n), and meta position (2o,p) of the aromatic ring were obtained with good to excellent yield and enantiocontrol (85–98%, 89.5:10.5 to 98.5:1.5 er). Also substrates 1q and 1r, possessing respectively 1-naphthyl and benzothiophenyl substituents at the alkyne terminus, underwent reaction smoothly and enantioselectively. The attainment of such high levels of enantiocontrol across a broad scope is unprecedented in this transformation.[27]

Scheme 5

Enantioselective Formal [4 + 2] Cycloadditions of 1,6-Enynes Based on 5-exo-dig Cyclizations

Reactions performed under Ar or N2 in anhydrous toluene (0.1 or 0.2 M), unless otherwise stated. Yields of material isolated after purification, er determined by HPLC or SFC on chiral stationary phase.

Carried out at 2 mmol scale, with 1 mol % Au10 and 1 mol % Ag6, in technical-grade toluene (0.6 M) under air for 24 h.

At 23 °C.

At 0 °C.

With 10 mol % Au10 and 10 mol % Ag6.

Including 5% of inseparable 6-endo-isomer.

Reaction time: 96 h.

Compared to linker and alkyne variations, the catalytic system showed higher sensitivity to changes to the alkene part (compare 2a with 2s–w in Scheme ), in line with the fact that the chiral catalytic ensemble needs to discriminate the two enantiotopic faces of the double bond. Thus, enynes 1s and 1t possessing geranyl or trans-cinnamyl groups cyclized to products 2s and 2t in less than 48 h (90–99% yield, 89:11 to 95.5:4.5 er), even though higher temperatures and/or catalyst loadings were required. Nevertheless, this marks a significant improvement with respect to previous chiral gold catalysts, which required 7–14 days at room temperature.[27e] Enyne 1u possessing a cis-cinnamyl substituent afforded product 2u in lower yield (48%, due to a competing cycloisomerization to an achiral diene product) and enantioselectivity (82:18 er). Formation of epimeric products 2t and 2u highlights the stereospecificity of the reaction with respect to the alkene configuration. More hindered substrates such as 1v and 1w, bearing respectively a cyclohexene or a tetrasubstituted alkene, delivered products 2v and 2w in good yield (61–95%) and moderate but encouraging enantioselectivity (80:20 to 85:15 er), considering that such bulky enynes were never engaged in this cycloisomerization, let alone asymmetrically.

Judging by the enantioselectivity of the 5-exo-dig cyclizations presented in Scheme , the key interaction between the urea and the anion is not disrupted by H-bond acceptors on the substrates (such as ester, carbamate, tosylate, and nitro groups in 1d–f, 1h, 1l), even though they are present in up to 40-fold excess with respect to the chiral anion. Protic additives with H-bond donor ability were also tolerated, as spiking the reaction of model enyne 1a with 5 equiv of methanol still delivered product 2a in 80% NMR yield and 97:3 er. These observations suggest that H-bonding between the urea and the anion is strong enough to make for an effective and robust catalytic system, even if based on noncovalent interactions. To further prove this point, the Au10/Ag6 catalytic system was applied also to the 6-endo-dig cyclization of enynes, with and without the addition of protic exogenous nucleophiles (Scheme ).

Scheme 6

Enantioselective 6-endo-dig Cyclizations of 1,6-Enynes without (A) or with (B) Nucleophile Addition

Yields of material isolated after purification, er determined by HPLC or SFC on chiral stationary phase.

For 48 h.

At 23 °C.

Thus, O-tethered enyne 3 was converted to oxabicyclo[4.1.0]hept-4-ene 4 in 70% yield and 92.5:7.5 er.[33] Cyclization of benzene-tethered enyne 5 followed by the addition of O-based nucleophiles[27e,34] delivered compounds 6a–e in moderate to good yield and er (55–77%, 88.5:11.5 to 92.5:7.5 er).[35] The addition of a N-centered nucleophile to enynes of type 5 was demonstrated for the first time, obtaining azide 6f in 62% yield and 90:10 er. Instead, fluoride addition afforded product 6g in 69% yield and only 61:39 er. The low enantiocontrol in the formation of 6g can be explained by the strong tendency of the fluoride ion to H-bond with the urea,[36] thus preventing the key interaction that keeps the chiral anion in place.

Importantly, derivatization of the water- and azide-addition products gave access to O- and N-derivatives equivalent to the formal addition of noncompetent nucleophiles, thus expanding the scope and underlying the usefulness of this enantioselective protocol. For example, alcohol 6a was transformed into benzoate 6h, as well as into diene 7, which displays the core of the carexane natural products (Scheme A).[27e,37] Triazole 6i, primary amine 6j, and amide 6k were easily obtained from azide 6f without erosion of the enantiopurity (Scheme B). Cyclizations of 1,6-enynes bearing a terminal alkyne generally proceeded with lower enantioselectivity than those of internal aryl alkyne substrates (Scheme C). The methoxycyclization of benzene-tethered enyne 8 delivered product 9 in 94% yield and 78.5:21.5 er, which could be increased to 87:13 er at the expense of yield (see Supporting Information for details).[38] In order to preserve the stereocenter, the cyclopropyl gold carbene generated upon 5-exo-dig cycloisomerization of O-tethered enyne 10 was trapped in situ with diphenyl sulfoxide.[39] Under unoptimized conditions, cyclopropyl aldehyde 11 was obtained in 42% yield and 83:17 er, which compares favorably with the only other enantioselective preparation reported so far (3% ee).[40]

Scheme 7

Derivatization and Scope Extension

Prepared from a batch of 6f with 93:7 er (see Supporting Information).

Finally, we sought to extend this H-bonded counterion-directed catalysis strategy to other types of reactions. At 1 mol % loading, phosphinosquaramide complex Au16 in combination with (R)-Ag6 catalyzes the tandem cyclization–indole addition to 2-alkynyl enones 12, affording furans 14a–d in good yield and enantioselectivity (Scheme ).[41,13] In this case, the stereocenter is not created during the cycloisomerization but forms in the subsequent intermolecular nucleophilic attack to a carbocation intermediate. Chiral salt (R)-Ag6 alone affords predominantly the opposite enantiomer of product 14a, and the Au10/(R)-Ag6 combination employed for 1,6-enynes gives slightly lower enantioselectivity. These results indicate that tuning of the achiral catalytic component to get high enantioselectivity is required for each reaction class (see Supporting Information for more examples, including an allenol[1] cyclization).

Scheme 8

Enantioselective Cycloisomerization–Indole Addition to 2-Alkynyl Enones

Yields of material isolated after purification, er determined by HPLC or SFC on chiral stationary phase.

Mechanistic Studies

Additional control experiments were conducted to shed light on the working mode of the final, optimized Au10/Ag6 catalytic system (Scheme and Table , see Supporting Information for further tests). To this end, cationic complex Au17 was prepared by treatment of Au10 with equimolar AgSbF6 in the presence of acetonitrile. Its structure was confirmed by X-ray diffraction, showing that in the solid state the urea H-bonds to the fluoride of the counteranion. When AgSbF6 was replaced by (R)-Ag6, neutral complex Au18 was isolated instead, with no incorporation of acetonitrile as indicated by NMR.

Scheme 9

Synthesis of Complexes with Ureaphosphine L10

Table 5

Control Experiments

entry	[Au]	[Na or Ag]	yield (%)a	erb
1	Au10	(R)-Ag6	>95	98:2
2	Au18		>95	93:7
3	[(JohnPhos)AuCl]	(R)-Ag6	0
4	Au10	(R)-Na6	0
5	[(JohnPhos)Au(NCMe)]SbF6	(R)-Ag6	24	50:50
6	[(JohnPhos)Au(NCMe)]SbF6	(R)-Na6	3	50:50
7	Au17	(R)-Ag6	>95	50:50
8	Au17	(R)-Na6	>95	79:21
9c	Au17	(R)-Na6	>95	90:10

Determined by 1H NMR against internal standard.

Determined by HPLC on chiral stationary phase.

With 15-crown-5 (100 mol %).

These complexes were then used in a series of informative control experiments, which highlighted how the chloride ligand and the countercation of the chiral salt play a role too (Table ). Preformed complex Au18 delivered product 2a with yield and enantioselectivity comparable to the in situ combination of Au10 and (R)-Ag6 (93:7 er vs 98:2 er, Table , entries 1 and 2). No reactivity was detected employing either [(JohnPhos)AuCl] with (R)-Ag6 or Au10 with (R)-Na6 (Table , entries 3 and 4), indicating respectively that the urea is required to remove the chiral anion from Au and that sodium, unlike silver, is not able to scavenge the chloride ligand. Experiments with cationic complexes Au17 and its urea-free counterpart [(JohnPhos)Au(NCMe)]SbF6 were then performed. Combining [(JohnPhos)Au(NCMe)]SbF6 with (R)-Ag6 and (R)-Na6 delivered racemic material in low yield (Table , entries 5 and 6), thus emphasizing the importance of the tethered urea not only for reactivity but also for enantioselectivity, achieved through precise positioning of the chiral anion via H-bonding. In an apparently surprising outcome, also cationic complex Au17 combined with (R)-Ag6 yielded racemic product (Table , entry 7). This can actually be explained by catalysis carried out by achiral cationic species Au17 on its own, because the chiral anion associates preferentially to Ag+ over Au+ (or the urea). In agreement with this picture, product 2a was obtained with 79:21 er when (R)-Na6 was used (Table , entry 8). Moreover, when 1 equiv of 15-crown-5 ether was added in order to chelate Na+ and thus direct the anion to Au+, the enantiomeric ratio of the product further improved to 90:10 (Table , entry 9). Therefore, in order to attain high enantioselectivity using this H-bonded system, it is crucial to tie the generation of the catalytically competent cationic Au(I) center to the removal of other cations (Ag+ and to a lesser extent Na+), which would otherwise “sequester” the chiral anion, leading to racemic background reactivity. In this sense, when combining Au10 and (R)-Ag6in situ, precipitation of AgCl not only frees up a coordination site on gold but also ensures that Ag+ is removed from deleterious solution equilibria with the anion.

Further insights into the catalytic system are presented in Scheme . First of all, single-crystal X-ray diffraction of products 2h and 6a and comparison of optical rotations with available literature values indicate that the newly created stereocenter is (R)-configured in both 5-exo-dig and 6-endo-dig reactions. This implies that in the enantiodetermining step of both cyclization modes, the Si face of the alkene attacks the Au(I)-activated alkyne, delivering the corresponding cyclopropyl Au(I) carbene intermediates (Scheme A). The absence of nonlinear effects (Scheme B) suggests that only one chiral anion is involved in the enantiodetermining step, in line with formation of the expected 1:1 urea:anion complex.[21b,21c]

Scheme 10

Mechanistic Investigations: (A) Enantiofacial Selectivity, (B) Study of Nonlinear Effects, (C) Use of Complex Au4 as Au10 Surrogate, (D) 1H NMR Titration of Au4 with (R)-Na6 (298 K, CD2Cl2), (E–J) Kinetic Studies

Spectroscopic observation of the H-bonding interaction between the urea of complex Au10 and the chiral anion proved to be nontrivial. Mixing of Au10 and (R)-Ag6, in various ratios and in the absence of substrate, invariably resulted in at least partial chloride abstraction. In turn this led to poorly resolved NMR spectra, where species tentatively identified as chiral [LAuX] Au18 with the anion behaving as anionic ligand, and achiral chloride-bridged dinuclear complexes predominated. On the other hand, when Na6 was used instead of Ag6 to circumvent chloride scavenging, no changes were detected by NMR. In this last case, the chiral anion presumably remained associated with Na+ without interacting at all with the neutral urea. We resorted to study the combination of (R)-Na6 and complex Au4 in the presence of a constant excess of 15-crown-5, to force dissociation of the sodium cation, while at the same time avoiding undesired chloride scavenging. Au(I) chloride complex Au4 was chosen as a convenient surrogate for Au10, since the NH signals of its bis(trifluoromethyl)phenylurea resonate in a free region of the 1H NMR spectrum. The performance of Au4 in the standard asymmetric reaction is comparable to that of Au10, even if with marginally lower reactivity (consistent with the absence of the meta CF3 group) and enantioselectivity (Scheme C). Gratifyingly, 1H NMR titration of complex Au4 (10 mM in CD2Cl2 at 25 °C) with increasing amounts of (R)-Na6 in the presence of 20 equiv of 15-crown-5 resulted in clear deshielding of both NH signals of the urea, indicating their engagement in H-bonds with the anion (Scheme D).[42] The titration was recorded at the same catalyst concentration present in the reaction mixtures (5–10 mM), and the establishment of such H-bonds during the reaction is expected to be entropically even more facile because after AgCl precipitation the Au(I)-bound chiral anion should already be in close proximity to the H-bond donor.

Finally, kinetic studies on the cycloisomerization of 1a to 2a catalyzed by the Au10/(R)-Ag6 system in toluene-d8 at −10 °C were undertaken. By use of the variable time normalization analysis (VTNA) introduced by Burés,[43] the reaction was found to be approximately first order in catalyst (0.9) and 0.5 order in substrate (Scheme E,F). A first order in catalyst is common to most catalyzed transformations, provided that the catalyst does not decompose or aggregate into off-cycle species. A partial, noninteger order in substrate is expected for unimolecular catalyzed reactions that follow Briggs–Haldane kinetics[44] and possess a Michaelis–Menten constant[45] (KM) similar to substrate concentration.[18b] To verify that this was indeed the case for the cycloisomerization under study, the reaction progress kinetic analysis (RPKA) popularized by Blackmond[46] was carried out. 1H NMR monitoring of the reaction indicated clean conversion of enyne 1a to product 2a (Scheme G), but rate analysis revealed an initial induction period (Scheme H), most likely related to a noninstantaneous chloride abstraction.[47] The double reciprocal Lineweaver–Burk plot was thus constructed using data in the 1.5–10 h time range (Scheme I),[48] obtaining a Michaelis–Menten constant (KM) of 74 ± 4 mM.[49] Given the 0–100 mM substrate concentration present during the reaction, this intermediate KM value justifies the partial order in substrate determined by VTNA. The experimental 0.5 order found for the entire reaction course falls within the range of the calculated elasticity coefficient ε,[50] which predicts the changing order in substrate for Briggs–Haldane kinetic regimes from KM and substrate concentration (Scheme J).

Scheme presents a tentative mechanism for the enantioselective formal [4 + 2] cycloaddition of enyne 1a catalyzed by Au10/(R)-Ag6, which takes into account all the experimental, spectroscopic, and kinetic evidence discussed above, as well as previous studies on this cycloisomerization.[26b,27e] Upon mixing Au(I) chloride complex Au10 with (R)-Ag6, neutral complex I possessing a phosphoramidate anionic ligand forms. The expected chloride scavenging accompanied by precipitation of AgCl matches the observed formation of a solution (with very few solid grains) upon addition of a solution of (R)-Ag6 to a thick whitish suspension of Au10 and enyne in toluene. Species I represents the entry to the catalytic cycle and coincides with Au18, prepared ex situ (see Scheme ) and catalytically competent (see Table ). In the presence of enyne 1, neutral complex I is proposed to be in equilibrium with cationic complex II, wherein the anion is H-bonded to the pendant urea and the alkyne coordinates to Au. The presence of this equilibrium is consistent with the partial order in substrate observed in the Briggs–Haldane kinetics. Additionally, H-bonding interactions between the anion and both NH groups of the urea were observed spectroscopically in a model system (see Scheme D). Regarding their precise geometrical arrangement, we propose that the NH residues establish two H-bonds with the phosphoryl O atom. This speculation is based on solid state considerations and DFT calculations with implicit solvent models.[51] In the crystal structures of (R)-Ag6 and related triethylammonium salt (R)-EtNH6, the P–O and S–O bonds are almost parallel with the two O atoms forming a ∼3 Å wide “pincer” (see inset in Scheme ), while Au10 has a 2.0 Å H–H distance between the urea NH groups. The most stable conformer computed by DFT for the model diphenylurea–chiral phosphoramidate couple shows two H-bonds to the phosphoryl O atom; however, a conformer where the NH groups H-bond to both the phosphoryl and sulfonyl O atoms is only 0.8 kcal/mol higher in energy.[51]

Scheme 11

Proposed Mechanism for the Cyclization of Enyne 1 under H-Bonded Counterion-Directed Au(I) Catalysis

The enyne is expected to be oriented as depicted for complex II (Scheme ), with the arene pointing toward the more crowded BINOL region and the linker in an unencumbered zone. This would be consistent with the lower er (92.5:7.5 to 95:5) observed in the cyclization of ortho-substituted substrates 1m, 1n, and 1q, which are not so well accommodated, and with the catalyst ability to tolerate instead even very bulky linkers on substrates 1a–g. At this point, the enantiodetermining C–C bond formation takes place, i.e., the attack of the Si face of the alkene to the activated alkyne, leading to cyclopropyl Au(I) carbene III.[52] Friedel–Crafts-type ring expansion, deprotonation of the Wheland intermediate, and protodeauration then afford species IV. Product/substrate ligand exchange is likely mediated by the anion[18b] via the intermediacy of species I, thus closing the cycle and releasing product 2.

Conclusions

We describe the concept of asymmetric H-bonded counterion-directed catalysis, based on H-bonding interactions between a chiral anion and a suitably positioned H-bond donor group on JohnPhos-type ligands for Au(I). The presence of such interactions was substantiated by 1H NMR titrations and structure–activity studies with modified ligands and chiral salts, as well as by the observed solvent and fluoride effects. For the first time, a broad range of alkyne substrates was engaged in challenging enantioselective gold-catalyzed reactions using chiral anions as the source of the stereochemical information, at catalyst loading down to 1 mol % and with excellent functional group tolerance.

This new paradigm, with a modular, short, and tunable synthesis of the two catalytic components, has the potential to speed up the development of enantioselective versions of various transition-metal catalyzed reactions, provided that the ligand for the metal of choice is equipped with a suitably placed H-bond donor group for a chiral anion.