Catalytic diastereo- and enantioselective additions of versatile allyl groups to N-H ketimines

Abstract

There are countless biologically active organic molecules that contain one or more N-containing moieties and broadly applicable and efficient catalytic transformations that deliver them diastereo- and/or enantioselectively are much sought after. Various methods for enantioselective synthesis of α-secondary amines are available (e.g., from additions to protected/activated aldimines), but those involving ketimines are much less common. There are no reported additions of carbon-based nucleophiles to unprotected/unactivated (or N-H) ketimines. Here, we report a catalytic, diastereo- and enantioselective three-component strategy for merging an N-H ketimine, a monosubstituted allene and B₂(pin)₂, affording products in up to 95% yield, >98% diastereoselectivity and >99:1 enantiomeric ratio. Utility of the approach is highlighted by synthesis of the tricyclic core of a class of compounds that have been shown to possess anti-Alzheimer activity. Stereochemical models, developed with the aid of DFT calculations, which account for the observed trends and levels of enantioselectivity are presented.

Graphical abstract

Catalytic enantioselective additions of carbanions to ketimines deliver products with a nitrogen-substituted quaternary stereogenic center (α-tertiary amine) but development of these transformations^1,2,3,4 is hardly straightforward. Ketimines are less reactive than aldimines and the reduced size difference between the substituents compared to aldimines makes enantiotopic face differentiation difficult. Catalytic enantioselective additions of allyl moities⁵ to ketimines, while much sought after, remain scarce. One study shows that reactions of allyl–B(pin) (pin = pinacolato) with acyclic N-benzyl ketimines may be promoted by a chiral bis-phosphine–Cu complex (Fig. 1a)⁶, and another deals with reactions of functionalized allylsilanes and tosyl-protected ketimines catalyzed by phosphoramidite–Pd complexes (Fig. 1a)⁷. Other disclosures cover highly activated ketimines, including cyclic sulfonylimines and their reaction with potassium allyltrifluoroborates (with Rh-based catalysts)^{8, 9} and isatin-derived N-Boc-ketimines and their reaction with allylsilanes (with Pd-based complexes and stoichiometric silver fluoride)¹⁰. Other approaches involve either enantiomerically pure ketimines^{11, 12} or enantiomerically pure allyl reagents^13,14,15. Our goal was to develop a method that would not require ketimine activation/protection and subsequent unmasking (Fig. 1b). The absence of a protecting group would bypass the intermediacy of E and Z mixtures of ketimine isomers, which can lead to lowering of enantioselectivity. Although preparation of N-H ketimines by condensation of ketones with ammonia followed by reaction with allylboron reagents is known¹⁶, as far as we are aware, diastereo- and/or enantioselective variants have not been introduced.

Figure 1. State-of-the-art in allyl additions to ketimines and goals of this study.

There are significant exisiting limitations and a number of compelling issues remain unaddressed. a, There are only a small number of methods for catalytic enantioselective addition of an allyl group to a ketimine. The substrate is typically equipped with an activating/protecting group, which might prove difficult to remove in the presence of similar functional groups within a product structure (e.g., another N-benzylamine). b, A direct approach to synthesis of α-tertiary amines may entail preparation of the requisite unprotected N-H ketimine through alkylation of readily available nitriles followed by catalytic site-, diastereo- and enantioselective multicomponent addition of 2-boryl-substituted allyl groups. One application relates to synthesis of the core tricyclic structure of a set of heterocyclic molecules that exhibit strong anti-Alzheimer activity. Bn, benzyl; Ts, tosyl; Ac, acyl; pin, pinacolato; G, R, L, functional groups; Pg, protecting group.

Based on the earlier investigations regarding enantioselective additions to aldehydes or ketones^17,18,19, which were recently extended to N-anisidyl aldimines^20,21, we envisioned the sequence in Fig. 1b. N-H ketimines would be accessed by addition of an organolithium or a Grignard reagent to a nitrile, many of which are commercially available^22,23. The ensuing catalytic process would combine an N-H ketimine, a monosubstituted allene and B₂(pin)₂ to generate homoallylic amines containing a pair of stereogenic centers and an alkenyl–B(pin) group. A number of biologically active organic molecules would thus become more readily accessible in enantiomerically enriched form; an example would be of the core structure of class of polycyclic compounds shown to possess the ability to reduce beta-amyloid production (Fig. 1b)^{24, 25}. Complications typically associated with site-selective removal of protecting/activating units would thus be obviated, particularly when relatively strong reducing (e.g., for an N-benzyl or an N-tosyl group, Fig. 1a) or oxidizing conditions²⁰ are needed and a stereogenic benzylic C–N bond is present.

Results

Establishing feasibility

We began with the reaction of allene 2a with N-H ketimine 1a, obtained from the reaction of benzonitrile with methyllithium (MeLi; 88% yield; Table 1). Diastereoselectivity was complete in every case [>98:2 diastereomeric ratio (d.r.)] but efficiency was catalyst-dependent. There was 60–75% conversion to rac-3a with Cu complexes derived from triphenylphosphine, tricyclohexylphosphine or racemic 2,2’-bis(diphenylphosphino)-1,1’-binaphthalene (rac-binap) (entries 1–3). Evaluation of N-heterocyclic carbene (NHC) complexes (entries 4–6) showed that the combination of cyclohexyl-substituted imidazolium salt 4c and CuCl is the most effective: rac-3a was obtained in 90% yield and >98:2 d.r. (Table 1, entry 6).

Table 1.

Probing efficiency and diastereoselectivity with different (achiral) catalyst types.

Entry	Ligand	Conv. (%)*	Yield (%)†	d.r.*
1	PPh3	60	49	>98:2
2	PCy3	71	52	>98:2
3	rac-binap	75	72	>98:2
4	4a	80	66	>98:2
5	4b	55	53	>98:2
6	4c	>98	90	>98:2

Reactions were carried out under N₂ atmosphere; 1.2:1 ratio of ketimine:allene was used.

^*Conversion (based on allene consumption; (includes desired and decomposition products) and diastereomeric ratio (d.r.) values were measured by analysis of 400 MHz ¹H NMR spectra of unpurified mixtures; variance of values is estimated to be <±2%.

^†Yield of isolated and purified product (<±5%). See the Supplementary Information for details).

Abbreviations: TBS, tert-butyldimethylsilyl; pin, pinacolato; Mes, 2,4,6-Me₃C₆H₂; Cy, cyclohexyl; rac, racemic; binap, 2,2’-bis(diphenylphosphino)-1,1’-binaphthalene.

Identifying a chiral catalyst

Several types of Cu complexes were examined (Table 2). Enantioselectivity was minimal with binap [(2,2'-bis(diphenylphosphino)-1,1'-binaphthyl); entry 1, 44:56 enantiomeric ratio (e.r.)]. The desired product was isolated in appreciable yield and e.r. with segphos (5a, 51% yield, 18:82 e.r.; entry 2) or josiphos (5b, 51% yield, 19.5:80.5 e.r.; entry 3), but less so with the more conformationally flexible 5c (27%, 69:31, respectively; entry 4).

Table 2.

Studies to identify an effective chiral catalyst.

Entry	Ligand	Conv. (%)*	Yield (%)†	e.r.††
1	R-binap	70	65	44:56
2	5a	55	51	18:82
3	5b	65	51	19.5:80.5
4	5c	29	27	69:31
5	6	80	51	40:60
6	7a	>98	65	55:45
7	7b	54	51	95:5

Reactions were carried out under N₂ atmosphere; 1.2:1 ratio of ketimine:allene was used.

^*Conversion (consumption of allene; (includes desired and decomposition products) and d.r. was measured by analysis of 400 MHz ¹H NMR spectra of unpurified mixtures; variance of values is estimated to be <±2%.

^†Yield of isolated and purified products (<±5%).

^††Enantiomeric ratio (e.r.) values were determined by HPLC analysis (<±1%) (see the Supplementary Information for details).

Abbreviations: TBS, tert-butyldimethyl silyl; pin, pinacolato; Mes, 2,4,6-Me₃C₆H₂; Cy, cyclohexyl; binap, 2,2’-bis(diphenylphosphino)-1,1’-binaphthalene.

With imidazolinium salt 6 or sulfonate-bearing 7a enantioselectivity remained low (entries 5–6). However, when the mesityl (Mes) moiety of 7a was replaced by a 3,5-diaryl-substituted phenyl moiety (7b, entry 7) enantioselectivity increased dramatically: 3a was obtained in 95:5 e.r. (for X-ray structure of the derived alcohol, see Fig. 3a). The effectiveness of the 7b-derived catalyst was surprising for several reasons. Enantioselectivity was considerably higher compared to the closely related N-mesityl-substituted variant (7a; see below for mechanistic analysis). Additionally, while sulfonate-containing NHC–Cu catalysts have been used for enantioselective allylic substitutions^26,27 and conjugate additions^28,29,30, none emerged as optimal for a 1,2-addition.

Figure 3. Stereochemical models.

DFT calculations shed light on the origins of enantioselectivity. a, Transition states with a N→Na interaction account for high e.r.; I represents the preferred mode. b, The model suggests that disruption of the N→Na coordination by the long, flexible alkyl ketimine chain (3w,x Figure 2) in V might render mode VI competitive, leading to lower e.r. Free energy values relative to the major pathway refer to the MN12SX/Def2TZVPP_THF(PCM) level after geometry optimization performed with either MN12SX/Def2SVP_THF(PCM) (for a) or M06L/Def2SVP_THF(PCM) (for b and c). For details, see Sections 4 and 5 of the the Supplementary Information. Abbreviations: NHC, N-heterocyclic carbene; THF = tetrahydrofuran.

Applicability

The method has considerable scope (Table 3). At times higher catalyst loading was necessary for high conversion (e.g., 3a, entry 1 Table 3 vs. entry 7, Table 2) and ketimine:allene ratio was changed to 1:1.5 for better yield (from 1.2:1). Regardless of whether the N-H ketimine had an ortho aryl substituent that is electron donating (3b), electron withdrawing (3c) or relatively sizeable (3f–g), products were isolated in 59–95% yield, >98:2 d.r. and 98.5:1.5–99.5:0.5 e.r. Reactions with meta-substituted N-H ketimines were similarly efficient and selective (entries 8–10, Table 3). In reactions of ketimines with different para-substituted aryl units (see 3k–3m) e.r. ranged from 92.5:7.5 to 95:5. Products 3n–o (Table 3, entries 11–12), from reactions with the less electrophilic alkyl-substituted N-H ketimines, were isolated in 38% and 48% yield and 91:9 and 95:5 e.r., respectively. Fluoroaryl-substituted amines 3d–e (entries 4–5), 3h (entry 8), and 3l (entry 12) were obtained in 64–91% yield and 94:6–98:2 e.r. (>98:2 d.r.); the high yield and e.r. in these transformations, regardless of the position of the fluorine atom, shows that fluorine-metal interaction^31,32,33 does not exert an adverse influence. Diastereoselectivity was exceptional throughout (>98:2 d.r.).

Table 3.

Catalytic diastereo- and enantioselective additions to N-H ketimines.

Entry	G	Mol %; Yield (%)†	e.r.††
1	C6H5 (a)	7.5; 76	95:5
2	o-MeOC6H4 (b)	10; 95	98.5:1.5
3	o-ClC6H4 (c)	7.5; 91	99.5:0.5
4	o-FC6H4 (d)	7.5; 72	97.5:2.5
5	o,o-F2C6H3 (e)	7.5; 91	98:2
6	1-naphthyl (f)	10; 66	98.5:1.5
7	o-MeC6H4 (g)	7.5; 59	99.5:0.5
8	m-FC6H4 (h)	5.0; 64	94:6
9	m-MeOC6H4 (i)	10; 57	98.5:1.5
10	2-naphthyl (j)	7.5; 81	93:7
11	p-ClC6H4 (k)	7.5; 56	92.5:7.5
12	p-FC6H4 (l)	5.0; 71	95:5
13	p-MeOC6H4 (m)	7.5; 39	92.5:7.5
14	CyCH2 (n)	10; 38	91:9
15	Cy (o)	10; 48	95:5

Reactions were carried out under N₂ atmosphere; >98% disappearance of ketimine in all cases (includes desired and decomposition products).

^†Yield of isolated and purified products (<±5%).

^††Enantiomeric ratios determined by HPLC analysis (<±1%; see the Supplementary Information for details).

Abbreviations: TBS, tert-butyldimethylsilyl; pin, pinacolato; Cy, cyclohexyl.

Experiments were performed at least in triplicate. See the Supplementary Information for details and the results with achiral imidazolinium salt 4c.

Additions to heterocyclic N-H ketimines afforded products in high e.r. (cf. 3p–3r, Fig, 2a), although efficiency was slightly lower. Synthesis of amines 3s–3u shows that different monosubstituted allenes may be used, but the size of the allene substituent did impact efficiency (41–78% yield, >98:2 d.r., up to >99:1 e.r.).

Figure 2. Further exploration of scope and illustration of utility.

A variety of desirable products can be synthesised. a, The method is applicable to a variety of heterocyclic substrates and allenes. Products derived from ketimines containing n-alkyl or iso-alkyl substituents (vs. methyl) can be obtained efficiently, in >98:2 d.r. and 85.5:14.5–96:4 e.r., depending on the substituent identity. For results with achiral imidazolinium salt 4c, see the Supplementary Information Table 1. b, Oxidation of the alkenylboronate moiety within the products derived from the NHC–Cu-catalyzed multicomponent reactions proceed efficiently to deliver the corresponding β-amino ketones (e.g., 9a), which represent the products of diastereo- and enantioselective Mannich-type additions. c, The method may be applied to the synthesis of the polycyclic core of compounds recently implicated in the fight against Alzheimer’s disease. Conversion of the C–B(pin) to a C–H bond promoted by a readily accessible NHC–Cu complex afforded 10. Formation of the derived thiourea and another NHC–Cu-catalyzed reaction generated the oxepane ring of 12. A two-step procedure involving oxidative cleavage/reduction and activation of the resulting primary alcohol delivered the desired aminothiazine ring and the strained tricyclic 13. Reactions were performed under N₂; there was >98% disappearance of ketimine in all cases (might include decompositiopn products). Yields correspond to isolated and purified products and represent an average of at least three runs (±5%). Diastereomeric ratios were determined by analysis of the 400 MHz ¹H NMR spectra of unpurified product mixtures (±2%). Enantiomeric ratios were determined by HPLC analysis (± 1%). See the Supplementary Information for experimental details and spectroscopic analyses.

The method extends beyond methyl-substituted substrates. Ketimines bearing an n-alkyl or unsaturated alkyl group (e.g., 3v–x, Fig. 2a) and the iso-propyl-containing ketimine precursor to 3y were converted to the desired amines efficiently and with exceptional diastereoselectivity. Nevertheless, e.r. varied depending on the substituent. Whereas 3v was formed in 94:6 e.r., there was gradual diminution in enantioselectivity as the side chain became longer (e.g., 88:12 e.r. for n-butyl-substituted 3w, 85.5:14.5 e.r. for pentenyl-substituted 3x). More enantioselective was the reaction that afforded iso-poropyl-containing 3y (96:4 e.r.); the X-ray structure secured for the N-acetyl derivative of 3y (Fig. 2a) indicates that there is no reversal in stereochemical induction (see Fig. 3a for corroborative X-ray data). A possible rationale for these selectivity trends will be provided below. Various compounds of interest, such as those derived from intramolecular hydroamination that afford heterocyclic derivatives^{34, 35}, may be accessed through functionalization of compounds such as 3x. Nonetheless, there are limitations. Reactions of ketimines that contain an α- or β-alkoxy or a benzyl group are inefficient, likely due to facile decomposition (enamine formation and β-elimination, respectively). The same applies to additions to trifluoromethyl-substituted ketimines (decomposition to unidentified products). There was no reaction with phenyl-tert-butyl N-H ketimine.

Utility

Catalytic enantioselective addition to ketimine 8 followed by oxidation of the C–B bond gave β-amino ketone 9a, the product of a Mannich-type addition, in 70% overall yield, yield and without any loss in d.r. (Fig. 2b). The efficiency with which 9a was obtained is higher than those shown in Table 3 and Fig. 2a, indicating that there might be some decomposition during purification and that yields may be improved if the alkenyl–B(pin) moiety is modified. The absolute stereochemistry of the product was confirmed through X-ray structure of primary alcohol 9b. It merits note that catalytic enantioselective enolate additions to N-activated ketimines (e.g., N-phosphinoylketimines^36,37 or those derived from α-ketoesters³⁸ or diethyl ketomalonate³⁹) are limited in scope (see the Supplementary Information for additional references).

We then investigated the possibility of application to enantioselective synthesis of the core structure of the aforementioned anti-Alzheimer compounds (Fig. 2c). NHC–Cu-catalyzed protodeboration¹⁷ of enantiomerically enriched alkenyl–B(pin) amine 3c furnished 10 in 72% yield. Synthesis of α-olefin 10 by a related route and with a sterically less demanding Cu–H complex [vs. Cu–B(pin) addition/protodeboration] would present a chemoselectivity issue (competitive reaction with ketimine⁴⁰); moreover, we find that the presence of a B(pin) group is critical to high enantioselectivity (see Fig. 3). Thiourea generation and removal of the silyl group afforded alcohol 11 in 71% overall yield. The cyclic ether was formed by treatment of 11 with 10 mol % CuI and 20 mol % 8-hydroxyquinoline (110 °C, 24 h)⁴¹, affording oxepane 12 in 78% yield. Oxidative cleavage of the vinyl group, reduction of the resulting aldehyde and subjection of the resulting primary alcohol to triflic anhydride (−20 °C, 2 h)²⁵ delivered 13 in 75% yield after recrystallization (this compound is unstable towards a variety of chromatography procedures). The aryl ring in 13 may be functionalized site selectively according to formerly reported procedures^42,43,44 (see the Supplementary Information for extended bibliography).

Stereochemical models

The results of DFT calculations are in agreement with the high diastereoselectivities (see the Supplementary Information for details). We then evaluated the role of the chiral NHC ligand that contains a pendant sulfonate moiety on enantioselectivity (Fig. 3a). The computational errors for modeling a charged species notwithstanding, we propose a similar steric and electronic environment as suggested formerly vis-à-vis enantioselective allylic substitutions effected by the same catalyst class²⁷. The sulfonate group is probably situated in the rear (I–II); this would allow for the large 3,5-bis-(2,4,6-(i-Pr)₃-phenyl)phenyl moiety to obstruct the right side of the complex, and causes the sizeable B(pin) moiety to be situated in the less occupied left/front quadrant in I (Fig. 3a). In II, which would lead to the minor enantiomer, there is steric repulsion between the B(pin) group and the chiral ligand’s N-aryl group, and thus the energy barrier would be higher (7.4 kcal/mol less favored than I). Consistent with the above analysis (Fig. 3a) the high energy of II and the steric pressure involving the B(pin) moiety is reflected in a considerably widened C^NHC–Cu–C₁–C₂ dihedral angle (177.2° vs. 151.3° in I). Calculations on the system containing the more diminutive NHC–Cu complex derived from 7a point to energetically similar processes (energy difference of 0.6 kcal/mol between III and IV; Fig. 3b), which is in agreement with the lower e.r. obtained when the NHC–Cu complex derived from 7a is involved (55:45 vs. 95:5 e.r., Table 2).

The stereochemical model offers a rationale for why reactions with imines containing longer linear alkyl groups (e.g., n-butyl or 4-pentenyl) are less enantioselective (Fig. 2c); these lower e.r. might partly arise from an increase in attractive London dispersion forces between the 3,5-bis-(2,4,6-(i-Pr)₃-phenyl)phenyl group and the substrate’s alkyl chain^45,46,47. It is however more plausible that higher conformational mobility of the alkyl chains disrupt N→Na chelation [less optimal C^NHC-Cu-C₁-C₂ dihedral angle in I (151.3°) vs. in I (131.8°)]. The smaller energy difference between anionic structures V and VI (2.2 kcal/mol, Fig. 2a) supports the notion that enantioselectivity would probably be lower without a sodium bridge, and that reaction via VI is likely the most competitive pathway versus the involvement of the most favored I. The C^NHC–Cu–C₁–C₂ and N–Cu–C₁–C^NHC dihedral angles of V and VI are close to the optimal parameters in I, implying that some strain induced by N→Na association in I is released in V. With the less flexible isopropyl substituent in 3y the aforementioned chelation may remain intact, allowing for higher enantioselectivity (96:4 e.r.).

Conclusions

The catalytic method introduced here puts forth an expeditious strategy for synthesis of α-tertiary amines in high diastereo- and enantiomeric purity, thus providing a solution to an important and persisting problem in catalytic enantioselective synthesis. There are no more than a small number of catalytic enantioselective protocols that allow access to such coveted N-containing compounds, yet none involves an unprotected/unactivated imine. These investigations provide the first step towards development of a series of catalytic enantioselective reactions involving N-H ketimines and other types of readily available and versatile carbon-based nucleophiles, protocols that render a range of chiral drug candidates with one or more α-tertiary amine moieties much more accessible. Finally, this study further expands the utility of sulfonate-containing chiral NHC ligands, previously utilized in catalytic enantioselective conjugate additions⁴⁸, allylic substitutions⁴⁹ as well as copper-boryl additions to alkenes⁵⁰ and allenes⁵¹ and copper–hydride additions to alkenes⁵², to include allyl additions to ketimines.