Ketones from Nickel-Catalyzed Decarboxylative, Non-Symmetric Cross-Electrophile Coupling of Carboxylic Acid Esters

Abstract

Synthesis of the C-C bonds of ketones relies upon one high-availability reagent (carboxylic acids) and one of low availability reagent (organometallic reagents or alkyl iodides). We demonstrate here a ketone synthesis that couples two different carboxylic acid esters, N-hydroxyphthalimide esters and S-2-pyridyl thioesters, to form aryl-alkyl and dialkyl ketones in high yields. The keys to this approach are the use of a nickel catalyst with an electron-poor bipyridine or terpyridine ligand, a THF/DMA mixed solvent system, and ZnCl₂ to enhance the reactivity of the NHP ester. The resulting reaction can be used to form ketones that have previously been difficult to access, such as hindered tertiary-tertiary ketones with strained rings and ketones with α-heteroatoms. The conditions can be employed in the coupling of complex fragments, including a 20-mer peptide fragment analog of Exendin(9–39) on solid support.

Graphical Abstract

While ketones are among the most versatile of functional groups in organic synthesis^[1] and commonly found in drugs and natural products, current methods of synthesis require starting materials of low abundance, such as organometallic reagents or organic halides (Scheme 1).^[2] We, and others, have developed new, functional-group tolerant methods for the synthesis of ketones, but these reactions remain limited to alkyl halides that have low commercial availability.^[3,4] A more attractive approach would be a decarboxylative coupling of two different carboxylic acid derivatives, as carboxylic acids are among the most abundant of alkyl substrates.^[5,6]

Scheme 1.

Strategies in Ketone Synthesis.

We hypothesized that this challenge could be addressed by the union of two recent developments in cross-electrophile coupling: the use of N-hydroxyphthalimide (NHP) esters as alkyl radical donors after decarboxylation^[7] and the finding that acylnickel(II) intermediates can capture radicals to form ketones in high yield.^[4e,8] Ketones could then be synthesized by the union of two activated esters – one tuned to form a metal acyl and one tuned to form a radical (Scheme 1).^[9]

At the outset, we foresaw several challenges related to the identity of the acyl donor. The ideal acyl donor should react to form an acylnickel(II) rapidly, but should not be so reactive as to limit functional-group compatibility, react with in-situ-generated zinc phthalimide,^[10] be prone to scrambling with the N-hydroxyphthalimide, or prevent isolation and purification. For example, our previous coupling of an acid chloride with an NHP ester formed the ketone product, but the high reactivity of the acid chloride limited utility and a large excess was required for high yields.^[7] We quickly found that S-2-pyridyl (SPy) thioesters^[11] fit all our criteria: they are easily synthesized directly from carboxylic acids and are stable to chromatography on silica gel, yet react rapidly with nickel.^[3,11] The resulting reaction is a new, remarkably general ketone synthesis.

Reaction optimization naturally started with the conditions we had previously reported for the coupling of activated esters with alkyl halides and NHP esters.^[3b,7] In contrast to these earlier studies, more electron-poor bipyridine and terpyridine ligands provided better results than electron-rich ligands (Table 1, entries 1–5). Reactions conducted with electron-rich ligands resulted in incomplete conversion of NHP ester and increased amounts of thioether product 1´ (entries 4 and 5). A mixture of THF and DMA (1:1 ratio) provided better results than either solvent alone by tuning the reactivity of the NHP ester (Table 1, entries 6–7) and reactions that contained a catalytic amount of ZnCl₂ improved the rate of conversion of NHP ester (Table 1, entry 8). While nickel is required for ketone formation, the reaction with zinc in the absence of nickel (entry 9) converted the NHP ester into decarboxylated alkane. This suggests that Zn might play a role in ketone-forming reactions by assisting in alkyl radical generation.^[12]

Table 1.

Optimized Reaction Conditions.

Entry [a]	Change from Scheme	1 (%)[b]	1’ (%)[b]
1	None	78 (78)[d]	0
2	L7 instead of L1	65	7
3	L9 instead of L1	64	0
4	L8 instead of L1	19	28
5	L3 instead of L1	38[c]	1
6	THF instead of 1:1 THF/DMA	29	0
7	DMA instead of 1:1 THF/DMA	38	5
8	No added ZnCl2	53	0
9	No nickel	0[e]	0
10	No ligand	19	3
11	No Zn reductant	0[f]	0

^[a]A mixture of NHP ester (0.25 mmol), thioester (0.25 mmol), NiBr₂(dme) (2 mol %), ligand (2 mol %), Zn (0.5 mmol) and ZnCl₂ (0.05 mmol) was stirred in THF/DMA (1:1, 0.5 mL) at rt for 24 h. For additional data on reactions with L2 and L4-L6, see the Supporting Information.

^[b]corrected GC yield using n-dodecane as internal standard.

^[c]Significant NHP ester remained.

^[d]Isolated yield after column chromatography on a 0.5 mmol scale reaction.

^[e]NHP ester consumed to n-propylbenzene quantitatively.

^[f]Both starting materials recovered.

Applying these optimized conditions to a variety of different carboxylic acid pairs demonstrated that the chemistry was useful for the synthesis of dialkyl ketones and aryl-alkyl ketones (Scheme 2). Functional group compatibility included acidic N-H bonds (14, 15, 16, 35), protected nitrogens (6, 13, 23, 32) and protected alcohols (37, 39). Notably, a primaryl alkyl bromide (9) did not react with nickel or the nucleophiles present. Generally, primary and secondary carboxylic acids could be coupled as either the acyl donor (thioester) or the radical donor (NHP ester), allowing for the easy synthesis of otherwise rare ketones, such as diheterocyclic ketones (22-24) and ketones with two strained rings (30-34).^[13] The coupling could be used to join two larger fragments, such as biotin or isonipecotic acid with a steroid core (35, 39). Because carboxylic acids are often found in natural products and in intermediates towards medicinal compounds, complex ketones can be made in a single step starting from these unusual starting materials, such as broadleaf herbicide 2,4-D (38) or the natural product abietic acid (40), a component of rosin. This method could also be used to make an α-fluoroketone (37, from 2-fluoropropanoic acid) and aryl-alkyl ketones from aryl thioesters (20, 21).^[14]

While these conditions are general, two limitations should be noted. First, while most thioester substrates tested coupled in reasonable yield, reactions with thioesters derived from α-amino acids and 2-picolinic acid were low yielding. Second, most N-hydroxyphthalimide esters coupled in high yield, but reactions with esters derived from secondary cyclopropyl carboxylic acids, difluoroacetic acid, and unstrained tertiary carboxylic acids (such as pivalic acid) were low yielding. The secondary cyclopropane limitation could be overcome with the use of an α-silyl group (29), which is easily introduced and removed.^[15]

The use of two different carboxylic acids instead of a carboxylic acid and an alkyl halide or organometallic reagent has several notable advantages. First, the large number of carboxylic acids makes new chemical space available: (−)-menthyloxyacetic acid (12), α-amino acids (13, 14, 15), lithocholic acid (35) and 2,4-D (38) are readily available as carboxylic acids, but their corresponding alkyl halides are not commercially available. In the cases where the corresponding alkyl bromides are available, the cost difference can be significant (greater than 10-fold for 32, 34, see SI). Second, because two acids are used, the mode of activation can be swapped to improve yields with little extra synthetic effort (Scheme 3): cyclopropane and adamantane carboxylic acids were best coupled as thioesters (25, 33) and α-amino acid NHP esters provided the best yields (13-15). This strategy is facilitated by the use of crude esters, formed from the corresponding uronium reagents,^[11c,16] instead of purified materials (Scheme 3).

Scheme 3.

Flexibility in substrate activation has strategic advantages.

[a] As in Scheme 2. [b] Ŕ = 4-phenylbutyric acid (0.5 mmol, 1 equiv) and R = ethyl succinate (0.75 mmol, 1.5 equiv). Yield is corrected GC yield. See Supporting Information for full details.

As a further test of the utility of this ketone synthesis, we explored couplings with polypeptides on solid support (Figure 1). Solid-supported synthesis is the most common format for amide bond formation, but only a limited set of other reactions have been employed in this context,^[17,18] and cross-electrophile coupling has never been employed in solid-phase peptide synthesis. Ketones in polypeptides are useful in peptidomimetic backbones as ketomethylene isosteres,^[19] and ketones in side chains can undergo chemoselective functionalization via reductive amination or oxime/hydrazone ligation.^[1,20] While ketones can be introduced via incorporation of non-natural amino acids into the growing chain, an approach that could form diverse ketone products from a single peptide would be useful in the discovery process.

Figure 1.

(A) Ketones synthesized on solid support. Cross-electrophile coupling using (20 equiv) NHP-ester, (50 equiv) Zn, (20 equiv) ZnCl₂, (2 equiv) NiBr₂dme, (2 equiv) L1, in DMA at 20 °C for 16–24 h. (B) Hexapeptide ketones. Xxx indicates an L-glutamic acid residue that has been cross-coupled. (C) Cross-electrophile coupling on an Exendin (9–39) analogue. L^N indicates L-norleucine. (D) Antagonism of GLP-1 (4.5 nM; EC₅₀ ~ 0.05 nM) activation of GLP-1R. Receptor activation was assessed by monitoring 3’,5’-cyclic adenosine monophosphate (cAMP) production in GLP-1R expressing HEK293 cells. The IC₅₀ values for P3 and P4 are 560 nM and 1300 nM, respectively. N = 4, mean ± standard deviation.

We prepared several peptides bearing ketone side chain functionality via a combination of standard solid-phase peptide synthesis and cross-electrophile coupling (Figure 1). The site of ketone installation was defined by incorporating L-glutamic acid (γ−allyl ester) at this position; any other Glu residues were incorporated with t-butyl ester-protected side chains. The allyl ester could be selectively converted to the acid^[21] and converted to the SPy ester with HPPT.^[11^c] In addition to two short model

peptides, we prepared a 20-mer (P3) and a ketone-bearing analog (P4) derived from Exendin(9–39), an antagonist of the glucagon-like peptide-1 receptor (GLP-1R) that is currently in clinical trials to treat postbariatric hypoglycemia.^[22] Tandem mass-spectrometry established ketone incorporation at the desired site in P4. Both P3 and P4 displayed dose-dependent inhibition of GLP-1 activity at the GLP-1R, which demonstrates that the new method can be used to modify complex, medicinally important molecules.^[1]

While this work was being completed, Baran et al. reported a related approach to the coupling of two different carboxylic acids that utilizes in-situ generated mixed anhydrides as the acyl donor instead of SPy esters.^[9] The methods are complementary in many ways: while the Baran method avoids pre-formation of one of the activated esters, our approach allows for greater selectivity (1:1 substrate ratio instead of 2:1 ratio); our different catalyst can be used at a lower catalyst loading (2 mol% instead of 20 mol%); and, the use of thioesters appears to provide better scope with hindered carboxylic acids (34, 40).

Together, these two reports greatly expand the number and types of ketones that can be easily accessed.^[23] The ability of a single nickel catalyst to differentiate between the two activated esters – thioester by oxidative addition, NHP ester by single electron transfer – is crucial to this unusual, non-symmetric coupling of two carboxylic acid derivatives. Further improvements are underway, including one-pot methods, alternative substrate pairs, and expanding the types of ketones that can be synthesized.

Table 2.

Substrate Scope for the Decarboxylative Coupling of NHP Esters with Thioesters.^[a]

^[a]Reactions were conducted on 0.5 mmol scale in 1 mL of solvent for 24 h. Yields are isolated yields after purification unless otherwise noted.

^[b]1.5 equiv of NHP ester used.

^[c]10 mol% NiBr₂(dme) and ligand were used.

^[d]Lower concentration (0.5 mmol in 2 mL).

^[e]Reaction run on 0.2 mmol scale at standard concentration. Boc = tert-butoxylcarbonyl, Cbz = carboxybenzyl, TBS = tert-butyldimethylsilyl