Naked d-orbital in a centrochiral Ni(II) complex as a catalyst for asymmetric [3+2] cycloaddition

Chiral metal catalysts have been widely applied to asymmetric transformations. However, the electronic structure of the catalyst and how it contributes to the activation of the substrate is seldom investigated. Here, we report an empirical approach for providing insights into the catalytic activation process in the distorted Ni(II)-catalysed asymmetric [3+2] cycloaddition of α-ketoesters. We quantitatively characterize the bonding nature of the catalyst by means of electron density distribution analysis, showing that the distortion around the Ni(II) centre makes the dz2 orbital partially ‘naked', wherein the labile acetate ligand is coordinated with electrostatic interaction. The electron-deficient dz2 orbital and the acetate act together to deprotonate the α-ketoester, generating the (Λ)-Ni(II)–enolate. The solid and solution state analyses, together with theoretical calculations, strongly link the electronic structure of the centrochiral octahedral Ni(II) complex and its catalytic activity, depicting a cooperative mechanism of enolate binding and outer sphere hydrogen-bonding activation.

O ptimization in asymmetric metal-based catalysis generally requires trial-and-error approaches, whereby a metal source and a 'privileged' chiral ligand 1 are combined without understanding of the three-dimensional (3D) and electronic structure of the active catalyst. Considering that synergistic activation of two reaction components is a key mechanistic strategy to attain high rate acceleration and selectivity 2,3 , it is particularly important to understand the electron density distribution (EDD) of the chiral catalyst, to gain insight into how the metal and ligand(s) cooperatively activate two distinct reaction components.
Over recent decades, the use of asymmetric nickel catalysts has substantially evolved both in acid-base and redox catalysis [4][5][6] . Several examples of chiral nickel complexes that act as asymmetric catalysts have been sporadically characterized by X-ray structure analysis, opening up a window of opportunity to discuss the stereo-discrimination process in asymmetric nickel catalysis [7][8][9][10][11][12][13][14][15][16][17] . Evans's transition-state model, which was proposed on the basis of X-ray analysis of the Ni(II)-diamine-enolate complex, is a remarkable example, explaining the stereochemical course in the catalytic asymmetric Michael reaction of nitroolefins with 1,3-dicarbonyl compounds 7,8 . However, despite the rapid progress in this field, the electronic structure of the asymmetric nickel catalyst is seldom investigated and exploited. The difficulties in understanding nickel asymmetric catalysts arise from several features: (i) They exhibit various oxidation states (0, þ 1, þ 2, þ 3 and þ 4) and electronic configurations (typically from 16e to 20e). (ii) The steric factor and the coordination numbers of the ligand(s) drastically influence the coordination geometry and the assembly-state of nickel complexes. (iii) The structure changes dynamically depend on solvents, substrates and additives due to the coordination equilibrium 18 .
Here, we describe the discovery and characterization of a distorted chiral Ni(II)-diamine-acetate complex and link the electronic structure to its catalytic activity (Fig. 1). In contrast to the classical symmetric octahedral Ni(II) catalysts (Fig. 1a), the Ni(II) catalyst reported here exhibits (L)-chirality at the Ni(II) center [19][20][21][22][23] , in which the coordination about the nickel atom is a distorted octahedron (Fig. 1b). The distortion in the octahedral Ni(II) complex makes the dz 2 orbital partially 'naked' due to the labile acetate ligand, which is weakly coordinated by electrostatic interaction. This unique electronic feature, involving both Lewis acidic and Brønsted basic natures, facilitates enolization of the a-ketoester. The distorted Ni(II) complex also exhibits hydrogen bond donor ability of the coordinated amine ligand, providing an outersphere binding site for the approaching electrophile 24 . With the bifunctional catalytic motif on the (L)-Ni(II) complex, we uncover the reactivity of the Ni(II)-enolate as a formal 1,3-dipolarophile that promotes [3 þ 2] nitrone cycloaddition. The regioselectivity for this transformation is distinct from that of classical [3 þ 2] cycloadditions 25 ; the vast majority of nitrone cycloadditions are predicated on the use of electron-deficient alkenes (Fig. 1c). This report provides insights into the elusive mechanistic basis for inverse-electron-demand (IED) [3 þ 2] cycloaddition 25-34 of a-ketoesters with (E)-nitrones (Fig. 1d).
We investigated the proposed formal [3 þ 2] cycloaddition using cyclic nitrones 2 to construct chiral tetrahydroisoquinolines, which are one of the privileged scaffolds for drug  [8][9][10][11]. The diamine (R,R)-4e, bearing cyclohexyl groups, was most effective, raising the e.e. value of 3aa to 83% e.e. Intriguingly, the diamine 4e has been overlooked as a chiral ligand for over a decade despite its potential utility and simplicity 42 . We finally found that diisopropylamine ( i Pr 2 NH: 10 mol%) can further improve not only the reactivity, but also the enantioselectivity 43  Structural determination of Ni(II) complexes in the solid state.
Electron density distribution analysis of I. To characterize the bonding mode around the distorted Ni(II) in I, we performed EDD analysis 47 using single-crystal X-ray diffraction data (Fig. 3, Supplementary Data 1). A 3D plot of the static deformation density of I highlights its valence electron density topology (Fig. 3a).

Structural analysis of I in solution.
We then characterized the structure of mononuclear Ni(II) complex I in THF (Fig. 4), at the same catalyst concentration as used for the IED [3 þ 2] cycloaddition reaction. The unique shift for carboxylates (1,598 and 1,558 cm À 1 ) in the infrared spectrum supports the idea that the two distinct acetates coordinate to the Ni-centre in a bidentate manner (Fig. 4a) 52 . The band observed at 3,225 cm À 1 also suggests that the N-H functionality on the ligand can act as a proton donor in THF 53 . The electronic absorption spectrum of I, which shows broad bands at 408, 753 and 1,182 nm, is typical of pseudoctahedral geometry of Ni(d 8 ) (Fig. 4b, bottom) 52 . The electronic circular dichroism (ECD) signal of I was observed not only in the ultraviolet-vis region, but also in the near-infrared region (Fig. 4b, top). The ECD spectrum observed in the d-d transition region demonstrated that the nickel centre in I is chiral even in solution. A feature of the observed ECD spectra is that the intensity of the near-infrared bands is significantly stronger than that of the visible absorption. Density functional theory (DFT) calculations on the noncentrosymmetric I at the level of UM06/6-311G(d,p) (SDD for Ni) were also performed ( Supplementary Figs 9-14, Supplementary Tables 4 and 5) 54,55 . The simulated infrared and ECD spectra fitted reasonably well with the experimental data. These results suggest that (i) Ni(II)diamine-acetate I retains pseudo-octahedral structure in the solution state, in which the two structurally distinct acetates coordinate to the Ni(II) centre in bidentate manner and (ii) chirality-at-metal in I is substitutionally and configurationally inert even in the presence of an excess of coordinative THF.
A plausible catalytic cycle. Structural analyses of I in the solid and solution states (Figs 2-4), as well as their catalytic activity differences ( Table 2), suggested that the monomeric species should predominantly control the stereo-discrimination process (Fig. 5). We assume that the exposed nature of the d-orbital allows it to interact with a-ketoester 1a as a Lewis acid, making the ketoester susceptible to deprotonation by the acetate ligand acting as a Brønsted base, leading to the formation of (Z)-Ni(II)-enolate that contains the five-membered chelating ring (Fig. 5, step (i)). Based on the X-ray (Fig. 2b) and EDD analyses of I (Fig. 3), we propose a Ni(II)-enolate, in which the carbonyl group in the ester in 1 coordinates to Ni(II) at the pseudoapical position. A new perspective in the transient Ni(II)-enolate is its coordination pattern in the octahedral structure; the Ni(II)-enolate reported by Evans 7,8 occupies the same plane with the chiral diamine. The crystallographic evidence that the N-H functionality at the equatorial position in Ni(II) complex I can contribute to activating the Lewis base (Fig. 2b) suggests that the subsequent H-bonding activation of the nitrone 2a (refs 34,44,45) would be a key driving force for fixing the two reaction components in close proximity (Fig. 5, step (ii)), thereby enhancing the reaction rate with high diastereo-and enantioselectivity (Fig. 5, step (iii)). In the proposed model, the centrochiral Ni(II) directly activates the a-ketoester 1a with the ligand-enabled H-bonding activation of the nitrone 2a.
The model presented herein can explain the obtained absolute stereochemistry of 3aa.
Scope and chemoselectivity of Ni(II) catalysis. The scope of the formal IED [3 þ 2] cycloaddition using the catalytic triad of Ni(OAc) 2 , (R,R)-4e and i PrNH 2 was examined (Fig. 6). a-Ketoesters 1 bearing various substituents on the aromatic moiety as well as longer alkyl chain substrates served as substrates, giving the corresponding [3 þ 2] adducts 3. Substrate 1h, bearing the sterically demanding 1-adamantyl group can also participate in the catalytic reaction, affording the corresponding isoxazolidine 3ha in 74% yield, with 25/1 d.r. and 94% e.e. The terminal olefin in 1i remains intact in the reaction of 2a, giving 3ia in 77% yield, 35/1 d.r. and 88% e.e. Substituted (E)-nitrones (3ab, 3ac and 3ad) were also applicable in the developed catalytic system, and comparable reactivity and selectivity were obtained by slightly tuning the reaction conditions. A characteristic feature in this Ni(II)-catalysis is its chemoselective recognition of a-ketoesters and (E)-nitrones. When the reaction was performed using 2a with vinyl ether 5 instead of 1a, no reaction took place (Fig. 7a); this is complementary to the reported IED [3 þ 2] cycloadditions [25][26][27][28][29][30][31][32][33][34] , and supports the validity of our working model, which involves a Ni(II)-enolate as the active dipolarophile. The observation that no reaction occurred when we applied the catalytic system to the reaction of (Z)-6 and 1a (Fig. 7b) represents additional evidence that the present catalytic system selectively activates (E)-nitrones. Another issue upon which we focused is the use of E/Z-isomerizable nitrone 7 (ref. 56) in our catalytic system (Fig. 7c), because controlling the isomerization is an issue of contemporary interest in organic synthesis 57 as well as in biology 58 . The formal [3 þ 2] cycloaddition of E/Z isomerizable ester-conjugated nitrone 7 with 1a selectively afforded (3S,4R,5R)anti-anti-8 as a single diastereoisomer in 90% yield with 64% e.e. The geometry selection of nitrone in the present catalytic system is complementary to that reported in (Z)-nitrone-selective reactions using chiral Cu(II) catalyst 31 . All these considerations indicate that the structure and geometry of both a-ketoester 1 and (E)-nitrone 2 can be discriminated in the present Ni(II)-catalyst system, facilitating the formation of stereochemically complex isoxazolidines 3 and 8.  Step i Step ii

Mononuclear catalyst
AcOH· i Pr 2 NH   ARTICLE Here, we not only showcase the first catalytic, asymmetric and direct formal [3 þ 2] cycloaddition of transient a-ketoester enolates, but also broaden the concepts underlying the design of asymmetric catalysts. A series of spectroscopic analyses in both the solid and solution states, supported by DFT calculations, were able to relate the electronic structural features of the distorted, desymmetrized Ni(II) complex to its catalytic activity. Specifically, we demonstrated that the EDD analysis of the chiral catalyst is a powerful methodology to experimentally visualize its electron density topology, and provides a quantitative insight into how the catalyst initiates the reaction. The presence of naked d-orbital interacting with the labile acetate ligand allows this acetate ligand to act as a Brønsted base to generate the (L)-Ni(II)-enolate. Furthermore, the bifunctional catalytic motif that enables the merger of H-bonding activation with amine ligand and enolate formation provides a mechanistic basis to expand the potential utility of ligand-induced octahedral metal centrochirality [19][20][21][22][23] . The operational simplicity and the easy tunability of the present catalytic system holds a vast potential for designing chiral chemospecific spaces with high predictability.

Methods
General. Detailed experimental procedures, characterization of compounds, Cartesian coordinates and the computational details can be found in Supplementary Figs 1 Catalytic formal [3 þ 2] cycloaddition of 1a with 2a. MS 4A (100 mg, powder purchased from Nacalai Tesque) in a Schlenk flask equipped with a magnetic stirring bar was flame-dried under reduced pressure for 5 min. Upon cooling to room temperature, the flask was refilled with N 2 , and a-ketoester 1a (23.4 mg, 0.10 mmol) and (E)-nitrone 2a (17.7 mg, 0.12 mmol) were added and dried under vacuum. The flask was backfilled with N 2 , THF (250 ml) was added at room temperature, and the flask was cooled to -30°C. To the resulting solution was added the prepared catalyst solution (5 mol%: 25 ml, 0.2 M in THF) and i Pr 2 NH (10 mol%: 25 ml, 0.4 M in THF). The reaction mixture was stirred for 24 h at -30°C. Aluminium oxide 60 (B20 mg, Merck) was added, and the mixture was diluted with EtOAc cooled at -30°C. The solution was passed through a pad of Aluminium oxide 60, to remove the nickel catalyst, and then eluted with EtOAc and concentrated under reduced pressure. The d.r. (450/1) was determined from the 1 H NMR spectrum of the crude sample. The residue was purified by column chromatography [CHROMATOREX NH (NH-DM1020, 100-200 mesh, Fuji Silysia Chemical) to give 3aa in 74% yield (28.2 mg, 0.0740 mmol). The e.e. of (-)-(1R,2R,10bS)-3aa (91% e.e.) was determined by means of chiral HPLC analysis (CHIRALCEL OD-H, 0.46 cm (f) Â 25 cm (L), n-hexane/2-propanol ¼ 95/5, 1.0 ml min À 1 , major; 9.8 min, minor; 17.7 min).
Electron density distribution analyses of I . THF. The diffraction data were collected using a RIGAKU AFC-8 diffractometer equipped with a Saturn70 CCD detector with MoKa radiation by an oscillation method at 90 K. X-rays were monochromated and focused by a confocal mirror. Sixteen data sets were measured with different crystal orientations and detector positions. For all data sets, camera distance was 40 mm. Bragg spots were integrated, scaled and averaged up to sin y/l ¼ 1.22 Å À 1 by the programme HKL2000 (ref. 59) Lorentz and polarization corrections were applied during the scaling processes. Analytical absorption corrections 60 were applied. The initial structure of I . THF was solved by a direct method using the programs SIR2004 (ref. 61), and refined by a full matrix least-squares method on F 2 using the programme SHELXL2014 (ref. 62). Refinements with a multipole expansion method using the Hansen-Coppens multipole formalism 63 and topological analyses based on the resulting parameters were performed with the XD2006 package 64 . Crystal data of I for EDD analysis is provided in Supplementary Data 1.