Radical coupling of β-ketoesters and amides promoted by Brønsted/Lewis acids

Abstract

Recent advances in photocatalysis have enabled radical methods with complementary chemoselectivity to established two electron bond forming approaches. While this radical strategy has previously been limited to substrates with favorable redox potentials, Brønsted/Lewis acid activation has emerged as a means of facilitating otherwise difficult reductions. We report herein our investigations into the Lewis acid-promoted redox activation of β-ketocarbonyls in a model photocatalytic radical alkylation reaction. Rapid evaluation of substrates and reactions conditions was achieved by high throughput experimentation using 96-well plate photoreactors.

Over the past decade, visible light-mediated photoredox chemistry has received significant attention as a selective, accessible, and ecologically-friendly complement to traditional two-electron approaches [1]. Radical based bond-forming strategies that would otherwise require polarity reversal, or “radical umpolung”, present a particularly enabling means of expanding synthetic organic reactivity. Despite this promise, however, innate reactivity in photochemical modes is typically limited by the inherent redox potentials of the reaction substrates. Notable contributions by Yoon [2] and Xia [3] have demonstrated that the LUMO lowering effect of Lewis acid carbonyl activation affects not only electrophilicity, but also reduction potential. Inspired by these early examples of Lewis acid/photoredox catalytic systems, we sought to combine our longstanding interests in Lewis acid catalysis [4], homoenolate reactivity [5], and photoredox [6] catalysis to achieve challenging/umpolung C–C bond disconnections.

Tertiary β-hydroxyamides are valuable synthetic precursors to a number of pharmaceutically-relevant γ-hydroxyamines (Fig. 1A) [7]. Traditional two-electron approaches to accessing the β-hydroxyamide scaffold include the Reformatsky reaction and the addition of amide enolates to acyl silanes featuring a 1,2-Brook rearrangement (Fig. 1B) [8]. While these methods have proven capable in providing the desired compounds, the direct conversion of β-ketocarbonyls to tertiary β-hydroxyamides via reductive strategies remains elusive. Nucleophilic approaches to this transformation have been complicated by poor regioselectivity and by the facile keto-enol tautomerism of the 1,3-dicarbonyl system in the presence of basic nucleophiles. A radical umpolung strategy therefore presents an appealing opportunity to overcome the limitations of two-electron reaction alternatives. Using the previously discussed Lewis acid/photoredox dual catalytic strategy, we investigated the radical Umpolung alkylation of β-ketocarbonyls (Fig. 1C).

Fig. 1.

(A) Tertiary γ-hydroxyamine-derived pharmaceutical compounds. (B) Synthesis of β-hydroxyamides from acyl silanes and amide enolates. (C) Design of an acid-promoted photocatalytic alkylation of β-ketocarbonyls.

Previously, our group developed radical methods wherein the reductive alkylation and arylation of arylidene malonates was enabled by Lewis acidic ammonium or lanthanide salts (Fig. 2) [9]. These two classes of Lewis acids shifted the reduction potential of arylidene malonates by as much as 0.8 V, enabling generation of highly stabilized β-radicals (Fig. 2A) [9c]. Interested in expanding the substrate scope of the arylidene malonate radical-radical coupling, we employed cyclic voltammetry (CV) to determine the reduction potentials of other bidentate Lewis basic compounds. The resulting shifts in reduction potential with and without equimolar Lewis acid were used to guide our choice of substrate. From this survey of a set of diverse Lewis basic scaffolds, we identified three general types of reduction potential responses to the addition of Lewis acid, typified by representative substrates isatin 1, ethyl benzoylformate 2, and ethyl benzoylacetate 3 (Fig. 2B). In response to the addition of equimolar Sc(OTf)₃, the first reduction potential of isatin was rendered more cathodic, making it a poor substrate for the envisioned Lewis acid-promoted reduction. Upon treatment with Sc(OTf)₃, the highly reversible reduction of ethyl benzoylformate was made approximately 0.25 V less cathodic, suggesting only mild Lewis acid redox activation. Finally, ethyl benzoylacetate exhibited a strong current response, as well as the largest observed shift in reduction potential (0.9 V) upon addition of Lewis acid. The unactivated reduction potential of this substrate (–2.2 V) places free ester 3 outside the reducing power of common iridium and organic photoredox catalysts (e.g., Mes-Acr⁺: E_red = –0.57 V, Ru(bpy)₃²⁺: E_red = –1.33 V, and 4CzIPN: E_red = –1.21 V) [10], indicating that it should only be capable of photoredox-catalyzed single electron reduction in the presence of Lewis acid [11]. Furthermore, while 3 did not possess a reversible CV curve on its own, the addition of scandium(III) triflate led to a more reversible curve, potentially indicating an increase in persistence of the ketyl radical. Based on these findings we selected benzoylacetate 3 and β-ketoamide 4 as the model substrates for our investigations into the dual photoredox/Lewis acid catalytic methods.

Fig. 2.

(A) Dual catalytic radical reactions of arylidene malonates. (B) Cyclic voltammograms of dicarbonyls with and without Sc(OTf)₃.

To perform a systematic evaluation of reaction conditions compatible with the envisioned dual catalytic system efficiently, semi-quantitative high-throughput experimentation (HTE) with 96-well plate photoreactors was conducted using a wide range of variables (Fig. 3). HTE has attracted increased attention in the past decade from both academic and industrial investigators in a variety of applications, including photoredox catalytic method development [12]. In comparison to traditional screening, HTE allows for much higher efficiency when evaluating reaction conditions, both in terms of reagent consumption as well as time. For our initial 96-well plate, we chose to simultaneously screen four literature-precedented benzyl radical precursors (potassium trifluoroborate [13], silane [14] silicate [15] and Hantzsch ester [16]), three different photocatalysts, two solvents, two acids, and the two previously-selected substrates. A qualitative triage of results from this screen indicated that desired tertiary β-hydroxycarbonyls were indeed detected by UPLC-MS from reactions of either ketoester 3 or ketoamide 4 with benzyl trifluoroborate salt 5 or benzyl Hantzsch dihydropyridine 6. These coupling partners were therefore selected for further study to validate the reactivity observed in HTE and probe the mechanism of the dual catalytic system. Having seen no reactivity from the wells using alkyl silane and alkyl silicates, however, we opted to forgo further investigations into their reactivity.

Fig. 3.

Initial 96-well screening of coupling partners and reaction parameters.

After confirming the reactivity observed in HTE at larger scale, we then evaluated the reactions of β-ketocarbonyls 3 and 4 with potassium trifluoroborate salts. The optimization of these reaction conditions revealed increased yield of hydroxyester 7 upon the addition of benzoic acid and the necessity of light and scandium, but interestingly, no loss of reactivity in the absence of photocatalyst (Eq. 1).

(1)

With this evidence of photocatalyst-free, visible light-mediated reactivity, we set out to determine which reaction components were responsible for visible light absorption. Drawing on precedented transformations of 1,3-dicarbonyl species with trifluoroborate salts from Vedejs and coworkers, we initially proposed boronate 8 as the photoactive component in the alkylation of 4 [17]. While the UV-vis spectrum of ketoamide 4 displayed no absorption in the visible range and only a slight red shift upon the addition of trifluoroborate 5, addition of fluorophilic species like TMSCl or silica gel [18] resulted in a significant shift in absorbance into the visible range, which roughly corresponded to the absorbance of isolable boronate 8 (Fig. 4).

Fig. 4.

UV-vis spectroscopic analysis.

A series of control experiments employing these fluorophile-promoted conditions revealed a significant decrease in tertiary hydroxyamide 9 yield upon deviation from standard reaction conditions (Table 1). Only minor consumption of ketoamide 4 was observed in the absence of fluorophile or proton source (Table 1, entries 2 and 3), and efficient conversion to boronate 8 was seen in the absence of light (entry 4). In the presence of blue light and the radical trap TEMPO, ketoamide 4 was almost fully consumed, but formed no detectable boronate 8 or hydroxyamide 9, instead providing only a complex, inseparable mixture (entry 5). Reactions with TEMPO in the dark, however, provided boronate 8 in high yield (entry 6). Taken together, these results suggested a visible light-free, fluorophile-promoted, two-electron process to form an intermediate electron donor-acceptor (EDA) complex capable of intramolecular SET upon visible light irradiation.

Table 1.

Fluorophile-promoted, visible light-mediated alkylation control reactions.

Entry	Deviation from std. cond.	Yield (%)a
		4	8	9
1	None	0	0	30
2	No SiO2	85	8	4
3	No PhCO2H	63	3	8
4	No light	0	90	0
5b	With TEMPO	7	0	1
6	No light, with TEMPO	1	74	0

^aYield determined by 1H NMR spectroscopy of crude reaction mixture.

^bComplex, inseparable mixture of unknowns observed.

Unfortunately, attempts at visible light-mediated conversion of putative intermediate boronates 8 and 10 to hydroxyamide 9 were met with only limited success (Scheme 1). The addition of boronate 8 to modified reaction conditions with benzoic acid and fluorophile under visible light irradiation provided 9 in only 14% yield. With these observations, it is currently unclear whether the conversion of 8 to 9 proceeds via a bimolecular process, wherein one benzyl group is transferred from one molecule of 8 to another, or a unimolecular process from another species generated in situ. Further mechanistic studies into this Lewis acid-promoted radical Umpolung reaction are underway.

Scheme 1.

Visible light-mediated reactions of putative intermediates 8 and 10.

Concluding our investigation into the reaction of β-ketocarbonyls with potassium trifluoroborate salt 5, we turned our attention to the Lewis acid-promoted photocatalytic reactions of Hantzsch esters. A second round of HTE (details see Supporting information) revealed significant hydroxyamide mass hits in reaction wells without any exogenous Lewis or Brønsted acid. Subsequent reaction optimization led to conditions providing hydroxyester product 7 in 53% yield (Eq. 2).

(2)

Given the reduction potentials of unactivated ketoester 3 (E_red = –2.24 V) and benzyl Hantzsch ester 6 (E_red = +0.995 V) [19], however, a complete photoredox catalytic cycle using fac-Ir(ppy)₃ should not be feasible in the absence of acidic activation [10a]. Based on these observations and on our group’s previous work involving ammonium salt activation of arylidene malonates [9a], we proposed the following mechanism (Scheme 2). Single electron oxidation of benzyl Hantzsch ester 6 by the photoexcited iridium(III) catalyst provides benzyl radical A and Hantzsch pyridinium B. Brønsted acid activation of ketoester 3 by pyridinium B results in complex C, which can then be reduced by the iridium(II) complex to form ketyl radical D. Radical-radical coupling of intermediate D with benzyl radical A results in the hydroxyester 8.

Scheme 2.

Proposed reaction mechanism.

Although these conditions provided full consumption of ketoester 3, conversion to product remained lower than desired. Given recent work by Wu and coworkers [20], along with UPLC-MS data of unpurified reaction mixtures, we believe conversion is currently hampered by a competitive ketyl radical dimerization. The ratio of radical dimerization to cross-coupling can be rationalized by the persistent radical effect [21]. Assuming idenital rates of radical generation, the rapid dimerization of a transient radical results in the relative excess of a more persistent radical in solution. As the persistent radical accumulates, it preferentially cross-couples with the transient radical. In contrast to our earlier work with the highly stabilized arylidene malonate-derived radicals that did not dimerize, these less-stabilized ketoester-derived radicals demonstrated less persistence despite a similar method of acid activation. Investigations to modulate this reactivity are underway.

In conclusion, we have applied cyclic voltammetry and high-throughput experimentation to develop two modes of photochemical reactivity. We can induce a nontraditional mode of Lewis acid activation where 1,3-dicarbonyl species undergo a photoinduced single-electron transfer, potentially within a dialkyl boronate complex formed in-situ. This photocatalyst-free transformation leads directly to sterically-congested tertiary β-hydroxyesters and amides via a disconnection unsuited for two-electron polar chemistries. We also investigated the direct radical-radical coupling of β-ketoester-derived ketyl radicals with alkyl radicals from substituted Hantzsch dihydropyridines. Even though the Hantzsch ester provides Brønsted acid activation and enables the photocatalytic reduction of the β-ketoester, it does not significantly increase the persistence of the resultant ketyl radical, as evidenced by observed ketoester dimerization. Despite these initial shortcomings, these investigations show promise as gateways to achieving this unique and difficult disconnection, and we intend to continue exploring novel acid-activated/photocatalytic transformations.