Diverse synthesis of C2-linked functionalized molecules via molecular glue strategy with acetylene

As the simplest alkyne and an abundant chemical feedstock, acetylene is an ideal two-carbon building block. However, in contrast to substituted alkynes, catalytic methods to incorporate acetylene into fine chemicals are quite limited. Herein, we developed a photoredox-catalyzed synthetic protocol for diverse C2-linked molecules via a molecular glue strategy using gaseous acetylene under mild conditions. Initiated by addition of an acyl radical to acetylene, two cascade transformations follow. One involves a double addition for the formation of 1,4-diketones and the other where the intermediate vinyl ketone is intercepted by a radical formed from a heterocycle. In addition to making two new C-C bonds, two C-H bonds are also created in two mechanistically distinct ways: one via a C-H abstraction and the other via protonation. This system offers a reliable and safe way to incorporate gaseous acetylene into fine chemicals and expands the utility of acetylene in organic synthesis.

A s the simplest alkyne, acetylene has traditionally been viewed more as a fuel than as an economical chemical feedstock, though it is widely used in chemical industry, with an estimated annual goal production of over one million tons 1 . The high-volume industrial use of acetylene includes the production of vinyl-containing monomers, such as vinyl amine, vinyl chloride, acrylic acid and its derivatives, used for polymeric materials, in other chemical commodities and as feedstocks [2][3][4] . The spatial orthogonality of the two independent π-systems in alkynes can be used for the discovery, design and control of the new cascade transformations 5 . However, in contrast to other substituted alkynes, a very limited number of catalytic protocols directly incorporate acetylene into fine chemicals. This is probably due to the greater inherent strength of the π-bonds 5 , higher activation energies for reactions 5 and, especially, apprehensions about handling an explosive 6 and flammable gaseous reagent. In fact, reactions exploiting acetylene at atmospheric pressure (1 atm) are uncommon [1][2][3][4][7][8][9][10][11][12][13][14] . Moreover, the instability of its possible intermediates, the terminal vinyl radical 15,16 and, especially, the cation 17 has traditionally restricted its conversion. Circumvention of these obstacles would undoubtedly boost the use of acetylene as a reagent in modern organic synthesis involving complex small molecules. Currently, the catalytic transformations of acetylene into fine chemicals are mainly focused on simple monofunctionalizations, such as vinylation 2,3,[7][8][9][10]13,14,18,19 , the Sonogashira coupling reaction 20,21 , and cyclization 2,22 . In addition, construction of other molecules with different levels of molecular complexity 11,12,[21][22][23] is of interest (Fig. 1a). To effectively introduce acetylene in synthesis, competing reactions, such as the generation of more reactive substituted alkynes or alkenes from acetylene, need to be avoided. Despite elegant synthetic developments, effective catalytic strategies to incorporate acetylene gas into fine chemicals remain scarce.
A variety of structurally complex and biologically active molecules or their precursors can be regarded as two components linked with a saturated two-carbon unit [24][25][26] (Fig. 1b). Ideally, it would be desirable to bridge two or more simple molecules simply through their exposure to a C2-synthon in the presence of a catalyst. This would provide an efficient platform to quickly construct the target molecules and offer molecular diversity via the molecular glue strategy 27 . As the simplest and abundant unsaturated two-carbon molecular building units, ethylene 12,28,29 and acetylene 2,3,7-12, [18][19][20][21][22][23] have been employed as C2 building blocks for the formation of fine chemicals. In fact, due to the presence of two addressable π-systems, acetylene can be used as a carbon glue 30 , a type of molecular glue, that connects molecules through sequential transformations of the two π-bonds, which mutually complements the ethylene transformation field 12,28,29,31,32 (Fig. 1c). Given 1,4-diketones have great importance as versatile intermediates for the synthesis of some bioactive molecules 33 , natural products and related compounds 34,35 , along with the ubiquity of bioactive molecules containing a carbonyl group linked to a heterocycle moiety 26 by a CH 2 -CH 2 bridge, developing a general methodology to quickly access these compounds via a molecular glue strategy from readily available molecules and abundant feedstock would be highly desirable, especially in drug discovery chemistry 36 .
Photocatalysis has recently emerged as a powerful platform for the direct functionalization and activation of organic compounds via open-shell pathways under mild conditions [37][38][39][40][41][42] . As one of the most widely exploited transformations within the realm of open-shell chemistry, the direct addition of carbon radicals to carbon-carbon π-bonds has been broadly leveraged to effect carbon-carbon bond formation with alkenes 43,44 . Encouraged by these developments, we wondered if it is feasible to utilize the highly reactive terminal vinyl radical, generated through the acyl radical addition to acetylene, to accomplish the C-H abstraction from suitable hydrogen donors, followed by a Giese radical addition [45][46][47][48][49] , resulting in the formation of 1,4-diketones or connecting the carbonyl group to heterocycles via a two-carbon unit (Fig. 1d). However, the direct addition of most common radicals to unactivated alkynes, especially acetylene, to obtain the terminal vinyl radical, is a really big challenge due to (i) the diminished rate of C-C bond formation owing to increased activation barriers 15,50,51 and (ii) the in situ generation of highenergy vinyl radical intermediates that are highly unstable, have a short lifetime, and readily participate in various undesirable open-shell pathways. To the best of our knowledge, direct addition to acetylene producing a terminal vinyl radical followed by functionalization is rarely reported 52 . In addition, the various radicals formed can react with the acetylene or the newly formed alkene generating undesired products. Importantly, how to selectively generate the desired product is also a big problem. Using the polarity matching effect, a subtle yet important element in radical addition process 53 , the generated product may be controlled by varying the electronic properties of the hydrogen donors. In addition, side reactions may be limited by quenching the radical intermediate formed by the C-H abstraction from the hydrogen donor.
In this work, we developed a photoredox-catalyzed synthetic protocol for diverse C2-linked molecules via a molecular glue strategy that employed gaseous acetylene under mild conditions (Fig. 1e). Formally, aryl ketones were linked to aryl ketones, or linked with heterocycles by CH 2 -CH 2 bridges, resulting in the formation of two C-C bonds and two C-H bonds. Mechanistic experiments demonstrate that the two C-H bonds are created in two mechanistically distinct ways, one via a C-H abstraction and the other via protonation.

Results and discussion
Reaction development. To start our investigation, the synthetic method for 1,4-diketones was explored with the commercially available α-oxocarboxylic acid 1a as the model substrate, acetylene gas as the C2-linker reagent, K 2 HPO 4 as the base, H 2 O as the hydrogen source and Ir[dF(CF 3 )ppy] 2 (dtbpy)PF 6 as the visible-light photocatalyst at room temperature under the irradiation of blue LEDs. From a mechanistic perspective, the highly reactive vinyl radical should abstract a hydrogen atom from a solvent Csp 3 -H bond because of the Csp 2 -H's higher bond dissociation energy (BDE) 54,55 . To our delight, the reaction in a solution of DCM/H 2 O (v/v) (3/2) afforded the desired 1,4-diketone 2a in 10% yield ( Table 1, entry 1). Examination of a range of photocatalysts revealed that Ir[dF(CF 3 )ppy] 2 (phen)PF 6 was superior with respect to reaction efficiency, yielding 2a in 16% yield (entries 2-4). Various solvents, such as acetone, N,Ndimethylformamide, acetonitrile, and tetrahydrofuran, were used instead of dichloromethane (entries 5-8) and the results indicated dichloromethane was more suitable for this transformation. The concentration of 1a and the amount of water may affect the reaction efficiency. When 1a was decreased to 0.05 M, 38% isolated yield of 2a was obtained (H 2 O (5 equiv.), 24 h) (entry 9). Prolonging the reaction time (36 h) did not improve the yield of 2a (entry 10). Further reducing 1a to 0.025 M improved the yield to 51% (entry 11). A slight decrease in the yield was observed when the concentration was further reduced (entry 12). Using other inorganic bases did not positively affect the reaction efficiency (entries [13][14][15]. Gratifyingly, the yield of 2a could be greatly improved to 79% by increasing H 2 O (20 equiv.) (entry 16). Further increasing the amount of water decreased the yield (entry 17), which means the amount of water has an important effect on the yield.
Reaction scope investigation. With the optimal reaction conditions in hand (Table 1, entry 16), we next investigated the scope of a variety of α-oxocarboxylic acids summarized in Fig. 2. The o-, mand p-methyl groups on phenyl α-oxocarboxylic acids were tolerated (2b-d). A number of phenyl α-oxocarboxylic acids bearing electron-donating groups, including not only 1°-alkyl groups but also more hindered 2°-and 3°-alkyl groups, were converted into the corresponding 1,4-diketones in good yields (2e-l). The 3,5-dimethyl substituted substrates was transformed into the desired compound in 68% yield (2m). Halogenated phenyl substrates were incorporated into the corresponding compounds with slightly lower yields (43-66%) (2n-q). The phenyl α-oxocarboxylic acid bearing a strong electronwithdrawing group underwent smoothly to afford 2r in 50% yield. Notably, 2-(2,3-dihydro-1H-inden-5-yl)-2-oxoacetic acid 1s and 2-oxo-2-(5,6,7,8-tetrahydronaphthalen-2-yl) acetic acid 1t provided the corresponding 2s and 2t in moderate yields. Importantly, the methoxy-substituted long chain substrate 1u was also suitable for this transformation giving 2u in moderate yield. Moreover, the 2-adamantanol-derived α-oxocarboxylic acid 1v and L-menthol-derived α-oxocarboxylic acid 1w reacted smoothly under the standard conditions to furnish the corresponding desired compounds 2v and 2w in moderate yields. These results show a great potential for the structural modification of an array of complex biological molecules in medicinal chemistry. Notably, various experiments with substrates containing redox non-innocent substituents, such as amino-, phenolic hydroxyl-, vinyl-, alkynyl-and hydroxyl-substituted, were also performed under the standard conditions. Unfortunately, these functional groups were not compatible with our system.
Considering these results, we turned our attention to test other hydrogen donors, specifically nucleophilic hydrogen donors. These donors would generate radical intermediates from C-H abstractions by the highly reactive vinyl radicals. The resulting C-H abstracted radical intermediate reacted well with the electron-deficient alkenes [44][45][46][47][48] to furnish the desired compounds containing the carbonyl group and nucleophilic component linked via a CH 2 -CH 2 bridge. In 2016, Knowles 56 reported an elegant intermolecular C-H functionalization to construct similar molecules in moderate yields using stoichiometric N-ethyl-4-methoxybenzamide as the abstractor. This abstractor has the potential to serve as a structurally modular catalyst for radical C-H functionalization. Though powerful, the preparation of the vinyl ketone and additional reagents to activate the substrate were required for the reported method. If the vinyl radical generated in our system could be used as the abstractor, it would be highly desirable. Due to the prevalence of furan-containing compounds 57,58 , THF was subsequently explored as a hydrogen donor and furan source, as well as the solvent. In fact, 13% yield of the tetrahydrofuran linked compound was isolated during the optimization ( Table 1, entry 8). A considerable increase in yield was obtained with a slight variation of the reaction conditions (Fig. 3). With respect to the 2-aryl-2-oxocarboxylic acid partner, we observed moderate to good yields of the desired products, which represent an important skeleton in a variety of bioactive molecules 11 . 2-Oxo-2phenyl-acetic acids bearing both electron-donating and electronwithdrawing substituents on the phenyl ring are suitable substrates (3a-3ab). Relatively lower yields were observed when an electronwithdrawing group was attached to the phenyl ring (3r-u), which could be ascribed to the reduced reductive quenching ability toward the photoexcited photocatalyst. Notably, 2-aryl-2-oxocarboxylic acids with synthetic handles, such as halides, were readily incorporated into the products (3t-u), which highlights the potential for the incorporation of these scaffolds into more complex targets. Evaluation of substrates containing reactive groups, such as the chemically and biologically abundant amide and ester, provided the corresponding products in moderate yield (3x-3z). The starting material, bearing an easily-oxidized thioether, is also tolerated in this system, furnishing the desired product 3aa in moderate yield. In addition, 2-naphthyl-and 1-fluorene-substituted glyoxylic acids are both suitable substrates, albeit in slightly lower yield (3ac-3ad). The reactive benzylic C-H bonds in the starting material, or in the corresponding product 3ad, remain intact under the reaction conditions. The simple benzoylformic acid was transformed to the corresponding product 3ae in 69% yield. Moreover, the mentholderived α-oxocarboxylic acid and 2-adamantanol-derived αoxocarboxylic acid reacted smoothly under the standard conditions to furnish the desired compounds 3af and 3ag in moderate yield. These results show the potential for the structural modification of an array of bioactive molecules. Considering the experimental results above and the requirements for construction of other heterocycle-containing compounds as well as the extension of this approach, some hydrocarbons, especially, heterocycles instead of tetrahydrofuran were subsequently tested (Fig. 4). It was found that this strategy could also be adapted for use with other five-membered heterocycle-based hydrogen donors (Fig. 4a). When substituted tetrahydrofuran was examined, the regioisomers 4 and 4′ were provided in moderate yield with a 4/1 ratio. N-Boc pyrrolidine reacted smoothly to furnish the corresponding compound 5 in 31% yield. Moreover, tetrahydrothiophene provided the desired compound 6 under similar conditions in 30% yield. These results suggest that the weak C-H bonds adjacent to heteroatoms in the five-membered heterocycles could be directly functionalized with acetylene using our system. However, the less reactive cyclopentane (a five-membered carbocycle) could not serve as the hydrogen donor to provide the desired product 7. In addition to five-membered heterocycles, we also examined the feasibility of other heterocycles with different ring sizes (Fig. 4b). The experiments showed that the three-membered cyclic ether, 1,2epoxypropane, could not be glued with acetylene to the carbonyl group for the formation of 8 or 8′, which might be attributed to the epoxides easily opened ring. To our delight, the fourmembered cyclic ether, oxetane, was successfully connected with acetylene, furnishing the desired ketone 9 in 26% yield, accompanied by 19% yield of 2a. As shown in Fig. 3, THF, a five-membered cyclic ether, gave a good result with 69% yield of 3ae. When the ring size was further increased from five to six, the corresponding ketone 10 tethering with tetrahydropyran (sixmembered cyclic ether) could be isolated as well, but with a lower yield (17%). However, no desired product 11 could be detected when the seven-membered cyclic ether, oxepane, was used as the substrate. It seems that the larger membered heterocycles (over six members) are not suitable partners. Linear ether (isopropyl ether) failed to provide the desired product 12, which is in line with the trend demonstrated in Fig. 4b. Interestingly, connecting 1,3-dioxolane with the carbonyl group was possible, giving regioisomers 13 and 13′ in a reasonable yield with a 5/1 ratio (Fig. 4c). The observed regioselectivity of H-abstraction in 1,3dioxolane may be attributed to the anomeric lowering of the C-H BDE 59   and 15′, which are an ideal precursor of unsymmetric 1,4diketones, were also obtained by switching from 1,3-dioxolane to 2-methyl-1,3-dioxolane. These outcomes illustrate that unsymmetric 1,4-dicarbonyl compounds could be accessed by using acetylene as a carbon glue, which greatly expands the synthetic scope of 1,4-dicarbonyl compounds. Compared to Knowles' strategy 56 , our protocol utilized the vinyl radical generated in situ, to activate the weak Csp 3 -H bond of heterocycles avoiding the need for the additional abstractor and the preparation of the corresponding vinyl ketones as substrates. Our strategy provides alternative access to this type of compound from abundant feedstock under mild conditions.
To illustrate the synthetic utility of our strategy, a series of experiments were conducted (Fig. 5). On a preparative scale, 1,4diketone 2q was isolated in 60% yield, which was slightly higher than that of the 0.3 mmol scale, suggesting that large-scale production might be feasible. Synthesis of enantiomerically enriched Ombitasvir 19 33 , an orally bioavailable and potent inhibitor, was achieved using chiral 18 which was prepared in 4 steps from 2q via sequential asymmetric reduction, methanesulfonation, nucleophilic cyclization 33 and coupling steps (73% yield for 4 steps, dr 5/1) (Fig. 5a). Moreover, simple transformations of the 1,4-dicarbonyl moiety afforded a range of synthetically useful scaffolds (Fig. 5b). For instance, heterocyclic molecules, such as substituted furan 20, substituted thiophene 21 and substituted pyrroles 22-23 were obtained via the cyclization of 2a under acidic condition or in the presence of Lawesson's reagent, ammonium acetate or benzylamine. The 1,4-dicarbonyl compound was easily transformed to naphthalene-2,3-diylbis(phenylmethanone) 24 in 98% yield under basic conditions with o-phthalaldehyde. Additionally, alkene 25 was accessed through a Wittig reaction with methyltriphenylphosphonium bromide. The derived 1,3-butadiene 26 was also obtained through a sequential reduction and elimination processes. Furthermore, the 1,4dicarbonyl compound 2a was transformed to cyclobutene 27 in good yield according to the known literature 61 .
Mechanistic studies. To further gain mechanistic insights, control experiments were performed (Fig. 6). In the presence of the radical trap TEMPO, the reaction completely shut down (Fig. 6a, top), indicating that a radical intermediate might be involved in this transformation. The reaction of 1a was performed for 2 h under standard conditions, affording 2a in 24% yield (Fig. 6a, bottom). Additionally, no change in the yield of 2a was observed when the same reaction was conducted for 2 h and then for an additional 22 h without light (Fig. 6a, bottom). These experiments indicate that the reaction is a visible-light photocatalysis process. Considering benzaldehyde was detected as a by-product during the reaction condition optimizations, we explored the reaction of 1d and benzaldehyde under the standard conditions (Fig. 6b, top) to verify whether benzaldehyde was the reaction intermediate. No cross-over product 28 was observed, strongly suggesting that benzaldehyde is not likely an intermediate of this transformation. According to our initial assumption and the inherent reactivity of the vinyl radical, direct abstraction of a hydrogen atom from hydrogen donors to produce the vinyl phenyl ketone is reasonable. As a result, the reaction of stoichiometric vinyl phenyl ketone 29 and 2-(4methylphenyl)-2-oxoacetic acid 1d was performed under the standard conditions without acetylene, affording the expected product 28 in 26% yield (Fig. 6b, bottom). The result showed that the corresponding vinyl phenyl ketone is the key intermediate in our system. To get further insight into the hydrogen source, chloroform-d was used as the solvent instead of dichloromethane in the reaction of vinyl phenyl ketone 29 and 2-(4-methylphenyl)-2-oxoacetic acid 1d (Fig. 6c, top) and furnished 28 in 64% yield with 0% D-incorporation in the α-position of the carbonyl group according to 1 H NMR analysis. In sharp contrast to this phenomenon, > 90% D was incorporated into the molecule 30a when dichloromethane-d2 was used in the model reaction (Fig. 6c, entry 1), indicating one of the hydrogens next to the carbonyl group comes from the solvent. When the corresponding potassium salt of benzoylformic acid 1aa was tested under the standard conditions using dichloromethane-d2 as the solvent and D 2 O as an additive, the corresponding product 31a bearing > 90% D-incorporation in each methylene site was isolated without 30a being detected (Fig. 6c, entry 2). Additionally, > 90% D-incorporation was verified in one of the methylene sites using D 2 O as the additive (Fig. 6c, entry 3), meaning another hydrogen adjacent to the carbonyl group originates from water. To further exclude that the deuterated product was generated from 2a under standard conditions through H/D exchange with D 2 O, 2a was  (Figure 6c, entry 4). Interestingly, no deuterated products 30a or 31a were found with the 96% recovery of 2a. This experiment suggested that the CH 2 of 2a cannot be deuterated through H/D exchange with D 2 O under standard conditions. Taken together, these results demonstrated the water and solvent are both hydrogen sources but used in the different stages of the reaction process. This is further supported by the mechanistic experiments for 3ae (see Supplementary Fig. 6 and Supplementary Fig. 7).
Mechanistic proposal. Based on the previously reported literature 54,55,[62][63][64][65][66] and the aforementioned mechanistic studies, a plausible mechanism for this transformation is proposed in Fig. 7. Upon irradiation with visible light, the photocatalyst Ir[dF(CF 3 ) ppy] 2 (phen)PF 6 I is known to access the highly oxidizing excited state (ES) species II ([Ir 3+ ]*), Which could be reductively quenched by the anion of α-oxocarboxylic acid 1 to generate an acyl radical 32 62,65 through a decarboxylative pathway, and the reduced species III. The generated 32 could be captured by acetylene gas to afford vinyl radical species 33 which is highly unstable and quickly abstracts a hydrogen from dichloromethane to provide the electron-deficient vinyl ketone 34 accompanied by the generation of 35. This proposed process is in line with the fact that the BDE of the Csp 2 -H bond (~110 kcal/mol) 54,55 exceeds that of Csp 3 -H bond of dichloromethane (~95 Kcal/mol) 66 . The reduced species III could be oxidized to regenerate the photocatalyst I by 35, furnishing the carbon anion 36 which could be protonated to regenerate dichloromethane. Subsequently a Giese radical addition reaction occurs with another molecule of acyl radical 32, which adds quickly to alkene 34 to produce a new carbon radical 37. Then a single-electron reduction of radical 37 by III affords the carbon anion 38 and regenerates the photocatalyst I. The desired product is obtained after protonation of 38. The phenomenon that nearly no electron-deficient alkene 34 was observed could be attributed to the fact that 34 is more highly reactive than the parent acetylene. Similarly, vinyl radical species 33 could also abstract hydrogen from heterocycles bearing weaker C-H bonds, affording alkene 34 and nucleophilic radical intermediate 39 which is rapidly captured by 34 to produce 40. Then the sequential single electron transfer (SET)/protonation process occurs to provide the final product 3. It is noteworthy that dichloromethane functions as solvent, as hydrogen donor, and as oxidant precursor to facilitate the regeneration of the reduced photocatalyst. In addition to serving as hydrogen donors and solvents (special conditions), heterocycles also work as the nucleophilic components due to the higher nucleophilicity of their corresponding radicals over that of 35. In summary, a diverse synthesis protocol for C2-linked functionalized molecules with gaseous acetylene was developed, which provided an efficient method to quickly access a variety of compounds through connecting two components together via a molecular glue strategy with readily available substrates and abundant acetylene gas under mild conditions. A series of 1,4diketones, which are important precursors of an array of heterocycles, were quickly constructed. Importantly, the reaction system was expanded to construct heterocycle-containing compounds. These compounds form when the intermediate vinyl ketone is intercepted by a radical intermediate formed from a heterocycle C-H abstraction. In both of these transformations, acetylene is incorporated in the final product as a CH 2 -CH 2 bridge. Additionally, mechanistic studies demonstrated that one of the formed C-H bonds is created via a C-H abstraction from the hydrogen donor, such as dichloromethane or a heterocycle, and the other via protonation from water or methanol. Moreover, this approach provides ready access to interesting, functionalized molecules and expands the utility of acetylene in organic synthesis, which will inspire new perspectives for value-added chemical synthesis using acetylene and promote the renaissance of catalytic transformations for acetylene.

Methods
Materials. Unless otherwise noted, all the materials were obtained commercially and used without further purification. All the solvents were treated according to general methods. Flash column chromatography was performed over silica gel (300-400 mesh). See Supplementary Methods for experimental details.
General procedure A for the synthesis of 1,4-dicarbonyl compounds. To an oven-dried 25 mL flask, Ir[dF(CF 3 )ppy] 2 (phen)PF 6 (0.003 mmol), 2-aryl-2oxocarboxylic acid (0.3 mmol), K 2 HPO 4 (0.36 mmol), H 2 O (20 equiv.) and DCM (12 mL) were added sequentially under N 2 . The flask was degassed through three freeze-pump-thaw cycles under acetylene and then an acetylene gas balloon was attached through a long syringe needle. The reaction mixture was irradiated by 12 W blue LEDs at a distance of 5 cm for 24 h at room temperature. The reaction mixture was filtered through a short pad of silica using ethyl acetate. The filtrate was concentrated in vacuo before it was purified by flash chromatography on silica gel to afford the desired product.
General procedure B for the synthesis of heterocycle-containing compounds.
To an oven-dried 25 mL flask, Ir[dF(CF 3 )ppy] 2 (phen)PF 6 (0.003 mmol), 2-aryl-2oxocarboxylic acid (0.3 mmol), K 2 HPO 4 (0.36 mmol), H 2 O/MeOH (1/1, 40 equiv.) and THF (12 mL) were added sequentially under N 2 . The flask was degassed through three freeze-pump-thaw cycles under acetylene and then an acetylene gas balloon was attached through a long syringe needle. The reaction mixture was irradiated by 12 W blue LEDs at a distance of 5 cm for 24 h at room temperature. The reaction mixture was filtered through a short pad of silica using ethyl acetate. The filtrate was concentrated in vacuo before it was purified by flash chromatography on silica gel to afford the desired product.