Decarboxylative Alkyl Coupling Promoted by NADH and Blue Light

Photoexcited dihydronicotinamides like NADH and analogues have been found to generate alkyl radicals upon reductive decarboxylation of redox-active esters without auxiliary photocatalysts. This principle allowed aliphatic photocoupling between redox-active carboxylate derivatives and electron-poor olefins, displaying surprising water and air-tolerance and unusually high coupling rates in dilute conditions. The orthogonality of the reaction in the presence of other carboxylic acids and its utility in the functionalization of DNA is presented, notably using visible light in combination with NADH, the ubiquitous reductant of life.


■ INTRODUCTION
Visible light is a prime stimulus to control the conformation of chemical bonds, 1 or their cleavage. 1a,2 The phototriggered formation of chemical bonds can enable frontier research in medicine and biology, 3 but their development is still a challenge in comparison to thermal click reactions 4 due to the slower rate and the need for UV-light and/or photocatalysts. 5 On one hand, photo-cross-linking methods still rely on unstable precursors like azirines or cyclopropanones. 5a,6 On the other hand, recent C−C coupling reactions using photobiocatalytic systems have shown great promise but these are still limited to activated substrates with auxiliary photosensitizers and electron donors. 7 As such, developments in self-sensitized, phototriggered, and fast C−C photocoupling between simple functionalities are still highly sought after (Scheme 1A). 8,9 Aliphatic linkages are particularly attractive due to their small size, robustness, and flexibility, which maximize the chances to obtain functional and metabolically stable conjugates. 8 Decarboxylative radical addition reactions (Scheme 1A) have recently emerged as prime tools to create aliphatic ligations in biomolecules. 8,9 These methods take advantage of the abundance of carboxylic acids 8,10 and the various technologies developed with Michael acceptors. 1b,11 Despite their success, radical addition reactions are slow (6−12 h) and require additional catalysts, inorganic reducing suspensions, and/or additives that are not native to biological systems. 8 The abundance of endogenous carboxylic acids in biomolecules or biomatrices poses a selectivity challenge for carboxylic acid substrates (1), due to their similar oxidation potentials. 8c−e In contrast, the N-hydroxyphthalimide (NHPI) esters (2) can be orthogonally activated in the presence of other carboxylates via single-electron reduction. 8a,b,f−l Recent methods based on desymmetrization 8n and late-stage carbene transfer 12 illustrate the potential of redox-active esters to be introduced through strategies unavailable to the parent carboxylic acids.
During our synthetic studies with redox-active carbenes, 12 we recognized that the coupling of redox-active esters and Michael acceptors 8a,b,f−l could significantly expand its capabilities with a suitable biocompatible reductant (Scheme 1B). The reduced nicotinamide adenine dinucleotide (NADH) would be ideal because it is a native component of biological systems.
The redox potential of NADH and its analogs (E ox {5} = 0.57 V vs Ag/Ag + ) is insufficient to activate redox-active esters (E red {2} ∼ − 1.1 ± 0.1 V vs Ag/Ag + ). 13 These dihydronicotinamides are potent single-electron reductants in the excited state (E ox *{5} = −2.60 V vs Ag/Ag + ), 14,15 but their short lifetimes in solution (τ{5*} ∼ 0.7 ns) 16 have limited their application as autonomous photoreductants. 14,17,18 At the onset of our work, these reagents required additional (photo)catalysts 8f−l, 18,19 or enzymes 20 under rigorously anhydrous and degassed conditions to drive reductive couplings. We reasoned that the short-lived excited states of these systems would have a minimal impact in photoinitiated reactions 21 and would avoid side-reactions in the presence of dioxygen derived from triplet-sensitization. The transient generation of the powerful photoreductant 5* would effectively circumvent the incompatibility with oxygen and moisture of other ground state super electron donors. 22 Importantly, the expected byproducts of the reaction would be biocompatible: the cofactor NAD + (or analogues thereof), CO 2 , and phthalimide (LD 50 {rat oral} > 5 g/kg). 23

■ RESULTS AND DISCUSSION
Toward this end, the reaction of the NADH model BNAH (5) with the redox-active ester 2a and the acrylate acceptor 3a was studied under blue light illumination (λ = 450 nm) without photocatalysts or additives (Scheme 2). 8f−l To our delight, the desired decarboxylative coupling product 4a was obtained in high yield using DMSO as solvent (entry 1). The reaction was found to be surprisingly fast, reaching 66% yield after 5 min of illumination (entry 2). Given the importance of maximizing the reaction rate for its implementation at higher dilution, 3,4,4c−i we explored related photoreductants. It was found that the dihydronicotinamide moiety is essential for high activity (entry 3) as well as the appropriate substitution at the heterocyclic nitrogen (entries 4,5). In line with seminal studies by Overman 8h and recent work by Shang, 24 the dihydropyridine 9 was found to promote the reaction, but it was slower and less efficient than the more biocompatible dihydronicotinamides (entry 6). 25 Interestingly, the N-butyl dihydronicotinamide BuNAH (10), which is the closest structural homologue to NADH among the photoreductants explored, was optimal both in terms of yield and rate (entry 7). This result can be rationalized by the slightly more reductive character of BuNAH (10) 26 than the N-benzyl-and N-aryldihydronicotinamides 5,8. Moreover, the performance of BuNAH (10) is only marginally affected by concentrations as low as 1 mM (entry 8). The system tolerates water as cosolvent (50% v/v; entry 9) and air atmosphere (entry 10). These are unique features that contrast with sensitive groundstate organic reductants 22a,4b,c,24,25 and other photocatalyzed reactions. 8f,h,18,19c,d Interestingly, BuNAH (10) can be prepared in multigram amounts, stored indefinitely as a solid, and handled for more than a week as a DMSO stock solution (see SI), thus enabling microdosing in high-throughput studies.
These results led us to explore the performance of NADH (11) due to its relevance as the native reductant in biological systems. The photophysics of NADH (11) have additional challenges due to its shorter excited state lifetime (τ{NADH} ∼ 0.4 ns) and the interaction between its dihydronicotinamide and adenine moieties. 27 To our delight, the commercial NADH disodium salt (11) promoted the coupling reaction in a dilute mixture of water and DMSO (1−10 mM; entries 11,12). Unlike that of BuNAH (10), it was found that the use of NADH (11) required inert atmosphere and larger amounts for optimal results, probably due to its higher sensitivity and/or less favorable photophysic properties.
We set out to explore the scope of the photocoupling using artificial BuNAH (10; conditions A) or natural NADH (11; conditions B) as photoreductants in aqueous (50% H 2 O in Scheme 1. Approach Towards Aliphatic Photo-Coupling with Native NADH Bio-Photoreductant a DMSO) and dilute conditions (1 mM; Scheme 3) most relevant in Chemical Biology. Alternatively, preparative scale reactions can be undertaken at higher concentrations in DMSO (see SI for details). Various Michael acceptors bearing electron-withdrawing groups such as ester (4a,b), amide (4c), 11c aldehyde (4d), ketone (4e), nitrile (4f), or sulfone (4g) were accommodated. Among these, acrolein was significantly less efficient as an acceptor, probably due to degradation of the sensitive aliphatic aldehyde product 4c. The maleimide scaffold (4h) that is common in bioconjugation reactions 1b,5e,11a,b was found to be very efficient. In stark contrast, no coupling product was obtained using the dihydropyridine 9, 24,25 thus illustrating the superior reactivity of BuNAH (10) or NADH (11) as photoreductants. 24,25 High yields and fast reactions also occur across a wide range of redox-active esters. Tertiary sites are coupled efficiently, thus allowing interesting structures to be functionalized, including bicyclic (4i), adamantyl (4j), piperidine (4k), cyclopropane (4l), and more complex scaffolds such as the drug gemfibrozil (4m,n). Secondary radical precursors are equally effective in the reaction (4o,q).
Interestingly, the products 4q,q' demonstrate that the norbornenyl−nortricyclyl radical equilibrium 28 can be established before their capture by the Michael acceptor. Primary Scheme 3. Scope Study ⧧ ⧧ Yields were determined by 1 H NMR using an appropriate internal standard; for isolated yields at preparative concentrations, see SI. HE; Hantzsch ester (9). a Ar atmosphere. b 100 mM concentration. c DMSO was used as solvent. d Dihydropyridine 9 was used instead of BuNAH (10) for comparison. e 20 mM concentration. carboxylate derivatives led to the products (4r−u) featuring robust and flexible alkyl-ligations. These include the crosscoupling of indole (4r), D-biotin (4s), a fatty acid (4t), and pyridine (4u) derivatives. Among those, the biotinylated product 4s displays an easily oxidizable thioether, a polar urea, a secondary amide, and an anomerically activated sugar. 29 Moreover, the reaction was proven to be useful in the latestage functionalization of natural products, including the peptide model derived from alanine (4w), and various densely functionalized terpenes with unprotected ketone, enone, olefin, diene, alcohol, and ester functions (4v,x-aa). The orthogonality between redox-active esters and unprotected carboxylic acids is demonstrated on the products 4n,aa. These substrates would lead to mixtures of products and/or polymers through existing coupling reactions based on oxidative decarboxylation. 8c−e Furthermore, the coupling reactions were complete in 10−75 min. This is substantially faster than previous methods despite the dilute conditions. Particularly sensitive or apolar substrates were understandably less efficient in the standard dilute aqueous media of the reaction (i.e., 4d,s,u,v,y,z,aa). In these cases, coupling efficiencies are enhanced simply by using higher concentration, inert atmosphere, and DMSO as solvent. However, in more favorable substrates, the reaction could operate even in pure water as solvent with similar results (4r,w).
The features of this system in terms of rate, concentration, water tolerance, and solubility of its components made it an ideal candidate for the in vitro alkyl photocoupling on polar biomacromolecules. To benchmark the performances of BuNAH (10) and NADH (11) in this context against comparable decarboxylative coupling methods, we set out to explore coupling reactions on DNA. 8b,d These are important in the synthesis of DNA-encoded libraries (DEL) 8b,d,9e yet challenging due to the complex functionality of the substrates and low scale at which these reactions need to occur. To our delight, DEL headpieces 12a,b were coupled efficiently using either BuNAH (10) or NADH (11) and blue light at 10 nmolscale to deliver "on-DNA"-functionalized products 13aa-ae,babe (Scheme 4). These reactions are generally completed in 1 h with excellent yields despite the micromolar concentrations (100 μM) in buffered media. Importantly, all the components in this system can be handled as dilute solutions, thus facilitating mixing in small reaction volumes (<300 μL). These features are characteristic of this system and can facilitate the future implementation of this reaction in automatic platforms.
The kinetic time-profile of the reaction with BuNAH (10) was obtained using in situ no-D NMR monitoring. 30 Nondeuterated DMSO was used to prevent any potential artifacts due to solvent isotopic effects in the propagation of the radical chain. However, it was found that the reaction proceeds similarly in DMSO and DMSO-d 6 , without any solvent-derived byproducts (see SI). This way it was possible to confirm that the reaction can be completed in 4.3 min at 100 mM concentration (Scheme 5A; gray). 8f−l Moreover, 10- fold (10 mM) and even 100-fold (1 mM) dilutions resulted in a surprising rate acceleration (Scheme 5A; blue traces). The reaction is completed in just 80 s of illumination at 1−10 mM with identical efficiency. To discern the origin of the acceleration, a control experiment was run in the least favorable concentration (100 mM) using a thinner reactor tube (1.25 mm diameter) to minimize the light path on the system. This resulted in significantly faster kinetics (Scheme 5A; red). This result demonstrates that the acceleration observed upon dilution stems from the attenuation of the inner filter effect. 31 With the usage of NADH (11) as photoreductant, the reaction is slower (Scheme 5B) but faster than previous decarboxylative coupling reactions. 8,24 The reaction is completed in 60 min (>80% in less than 25 min). These results are remarkable considering the dilute conditions (20 mM) in the presence of only 1.5 equiv of the acceptor 3a and NADH (11). Importantly, the system is stable in the absence of light (Scheme 5C). After a long dark period, the system was illuminated obtaining an identical kinetic profile to that of a standard experiment, as evidenced by the time-shifted overlay (Scheme 5D). This demonstrates the absence of static deactivation in the dark, which may be relevant in cases where other equilibria need to be established before the C−C coupling event is phototriggered. 3,5,6 Absorption spectroscopy revealed that the light absorption of BuNAH (10) is similar to those of other dihydronicotinamides, 16 featuring a strong band at 350 nm that extends into the visible region (Scheme 6A, left). In the presence of the redox-active ester 2a, which only absorbs below 350 nm, the absorption increases marginally at 450 nm using concentrations as high as 0.1 M (12% increase; Scheme 6A, right), which may indicate the formation of a donor−acceptor complex (EDA). 32, 33 Thus, we set out to study the relevance of this possible EDA interaction in the photoactivation of this reaction. Stern−Volmer studies evidenced a linear quenching of the steady-state fluorescence of BuNAH (10; Scheme 6B; blue) with increasing concentrations of the redox-active ester 2a. Nevertheless, the linear decrease in luminescence intensity is not a definitive proof of the mechanism by which this phenomenon occurs. 34 Therefore, the fluorescence lifetime of the excited state 10* (τ 0 (10*) = 1.08 ns) was measured using Time-Correlated Single Photon Counting (TCSPC). This study revealed a decrease in the lifetime of excited BuNAH (10*, Scheme 6B; purple) upon increase of the concentration of redox-active ester 2a. However, the significantly different slopes of the steady-state and lifetime Stern−Volmer plots were not consistent with a conventional dynamic quenching scenario. 34 Instead, the data supports the formation of a nonemissive EDA complex 10·2 in equilibrium with the free 10 (Scheme 6B; right). The corresponding equilibrium constant could be estimated through fitting of the steadystate and lifetime data (K eq ∼ 7; see SI). 34 Consistently, no additional luminescence bands corresponding to the EDA complex 10·2 could be observed in either excitation or emission spectra (see SI). At this point, it is unclear which of these coexisting dynamic and static interactions between BuNAH (10) and the redox-active ester 2 are most important for the reactivity. However, it is known that the formation of EDA complexes is affected by changes in the substrate, solvent, concentration, and/or temperature. 32 The fact that the reaction is not inhibited in dilute conditions disfavors the EDA complex to be critical in the photoactivation of this system. 24 In this sense, the direct reduction by photoexcited dihydronicotinamides without engagement in donor−acceptor complexes 32,33 has been documented but only in the context of more activated alkyl halide substrates. 17b The expected intermediacy of free-diffusing alkyl radicals was demonstrated by the different ratios of the products 4ab,ab' that were obtained using the 5-hexenyl radical clock precursor 2ab at different initial concentrations (Scheme 6C). To discern the fate of the radical intermediate that would result from the addition of the alkyl radical into the electron-deficient olefin, we conducted a series of experiments with the dideuterated BuNAH derivative 10-d 2 (Scheme 6D). These experiments revealed that hydrogen atom transfer (HAT) from BuNAH (10) is the main process to quench the putative radical addition product. 8g,h,19c−e Further control experiments confirmed that the solvents (DMSO and H 2 O) do not exchange with 10-d 2 under the reaction conditions and do not have any relevant role in the HAT process (see SI). The involvement of a radical chain mechanism was studied measuring the average quantum yield. This was determined in triplicate at 20−25% conversion of 2a, obtaining a value of 2.9 ± 0.5, which points to the propagation of a radical chain. 31 The mechanistic proposal in Scheme 7 comprises the electron−proton−electron transfer manifold that is typical in radical reductions mediated by dihydronicotinamides 14,17,35 and our own experiments (Schemes 5 and 6). Photoinduced electron and proton transfer from dihydronicotinamide 10 to

Scheme 6. Mechanistic Studies
Journal of the American Chemical Society pubs.acs.org/JACS Article the redox-active ester 2 through the dynamic and/or static mechanisms discussed above (Scheme 6B) produces the carbon centered radical 14, a nicotinyl radical 15, phthalimide (16), and CO 2 . The radical 14 adds to the olefin 3 to produce the radical 17, which after concerted 8g or stepwise 35 HAT yields the coupling product 4 and the nicotinyl radical 15. 8g The latter could reduce the redox-active ester 2 to produce the pyridinium salt 18, CO 2 , and the alkyl radical 14 that propagates the chain reaction (see Scheme 6C,D). 8g The formation of the pyridinium salts 18 derived from BuNAH (10) and NADH (11) and their kinetic correlation with the formation of the product 4 has also been evidenced by in situ NMR monitoring (see SI).

■ CONCLUSIONS
Herein, we report that the dihydronicotinamides BuNAH (10) and NADH (11) promote the photocoupling of redox-active esters and Michael acceptors upon illumination with blue light. These reactions do not require external photocatalysts or additives, have no detectable background reactivity, can run in water, and have an unusually high rate even at low concentration. This system has demonstrated its utility in the functionalization of DNA macromolecules in extremely dilute conditions. The mechanistic experiments demonstrate the multifaceted role of these dihydropyridines as photoinitiators, reductants, and hydrogen-atom donors to drive this fast photocoupling using a minimal homogeneous system. This work introduces NADH (11) as an autonomous photoreductant and opens prospects for new artificial coupling reactions that our group is currently investigating.