Silver(I) Oxide‐/DBU‐Promoted Synthesis of Dihydrofuran Units through Allenyl Silver Formation

A formal [3+2] cyclization mediated by silver(I) oxide and 1,8‐diazabicyclo[5.4.0]undec‐7‐ene (DBU) is described herein. Through a broad variety of carbonyl compounds, this system can promote cyclization reactions with high yield (up to 85 %) and diastereoselectivity (up to 95:5) for a straightforward access to complex and congested dihydrofuran derivatives in one step under mild conditions. Based on DFT studies, the proposed mechanism would involve an allenyl silver intermediate.

Keywords: allenyl silver; dihydrofuran; intramolecular cyclization; isomerization

An easy way to complex: An unprecedented isomerization/allenyl silver formation/[3+2] cyclization promoted by a combination of silver salts and base is presented. The reaction, which is broad in scope, allows the straightforward and diastereoselective (up to 95:5 dr) synthesis of complex dihydrofuran frameworks under mild conditions from easily accessible substrates.

2,3‐Dihydrofuran subunits can be encountered in a wide range of natural and synthetic products exhibiting, in some cases, biologically interesting profiles.[1] Additionally, such scaffolds can be regarded as attractive precursors in an array of organic transformations.[2] Based on these considerations, considerable efforts have been devoted to the development of efficient accesses to 2,3‐dihydrofuran units by various approaches.[3–5] Among them, intra‐ or intermolecular [3+2] cycloadditions constitute a powerful tool to reach such heterocycles[4f, 6] and more generally, heterocyclic five‐membered rings. In this field, the vast majority of transformations encountered imply hetero 1,3‐dipolar reagents or their synthetic equivalent to reach a broad variety of heterocyclic molecules.[7] However, access to cyclic frameworks by means of cycloadditions is much more challenging when all‐carbon 1,3‐dipoles are involved due to the difficulty to generate such species.[8]

To overcome this issue, the activation of double or triple bonds by means of transition metals appears to be an attractive method. However, very specific substrates are required, and this constitutes a main drawback that narrows the field of application.[8] One area of improvement would be to take advantage of the strong alkynophilicity of "coinage metals" to carry out such cyclizations.

Indeed, in 2007 Zhang's group reported an elegant [3+2] cycloaddition for the diastereoselective synthesis of cyclopentanone enol ethers by using enones and allenyl MOM ethers mediated by gold catalysis.[9] One year later, the same group published the synthesis of different 2,5‐dihydrofuran motifs through Au‐containing all‐carbon 1,3‐dipoles from propargylic ketals, showing the ability of gold to promote such cyclizations.[10] In the course of their work on dual organocatalysis combining gold and amino catalysis, Kirsch et al. unexpectedly obtained a tricyclic cis‐fused dihydrofuran adduct by using a co‐catalytic amount of (Ph3P)AuOTf (10 mol %) and (c‐C6H11)(iPr)NH (20 mol %) in xylenes at 150 °C for 1 h (Figure 1).[11] Surprisingly, this transformation has only been underlined in a single case by the authors and, to the best of our knowledge, has not been reported again in literature. Interestingly, in 2010, Zhang et al. demonstrated that the combination of Ag2O and KOH was able to promote formal intramolecular [3+2] cycloaddition reactions, generating all‐carbon 1,3‐dipoles, with conventional heating[12] or under microwaves conditions.[13] However, the scope of this transformation remains restricted to N‐propargylamides bearing an electron‐withdrawing group on the nitrogen atom (Figure 1).

In the work developed here, we expect to reach bi‐ or tricyclic complex structures containing 2,3‐dihydrofuran cores in a single step from readily accessible compounds bearing an inactivated alkyne moiety. This reaction will proceed according to a formal AgI‐mediated [3+2] cyclization involving all‐carbon 1,3‐dipoles. This approach appears to be particularly challenging as we envisioned obtaining allenyl silver species through an unprecedented alkyne isomerization mediated by AgI (Figure 1). As far as we know, the formation of an allenyl silver has only been reported by the Vermeer group through a 1,4‐conjugated addition of an alkyl silver onto an ene‐yne system (path a) or by transmetallation of an allenyl lithium species (path b).[14] Moreover, the reactivity of such species has been little studied, and the intramolecular cyclization targeted herein is unprecedented in literature (Figure 1).

Silver was chosen as metal catalyst in this transformation as it is both significantly less expensive and more abundant than gold and less prone to undergo side‐reactions than copper (i.e. Glaser‐Hay coupling). Indeed, due to its [Kr] 4d10 5s1 configuration, silver possesses a strong alkynophilicity and its ability to π‐coordinate to unsaturated systems having low‐lying empty orbitals makes it extremely interesting especially in the field of alkyne‐based chemistry.[15] In a preliminary survey, we chose the combination of Ag2O/1,8‐diazabicyclo[5.4.0]undec‐7‐ene (DBU) in acetonitrile and 1,3‐keto ester 1 a as the substrate (Scheme 1).

We were pleased to find a complete and clean conversion of 1 a into the tricylic compound 2 a at 50 °C after 2 h. The cyclized product was obtained in 85 % yield with a total diastereoselectivity in favor of the cis‐adduct without detection of the homo‐coupling product.[16] The stereochemistry was determined by NOE experiments and further confirmed by a single‐crystal X‐ray diffraction analysis of p‐nitrobenzoyl derivative 3,[17] obtained after reduction of ester 2 a and subsequent esterification of the resulting alcohol with p‐nitrobenzoyl chloride (90 % yield over two steps, Scheme 1).

Encouraged by this preliminary result, we focused on the optimization the reaction conditions (Table 1). A short screening of silver salts highlighted the ability of inexpensive silver(I) oxide to cleanly promote the cyclization after 2 h with a satisfying yield (85 %, entry 1). Silver nitrate and silver antimonate afforded 2 a albeit in moderate yields, respectively 64 % and 63 % (entries 2 and 4) whereas silver sulfate furnished only 32 % of the desired product (entry 3).

1 TableReaction optimization. Full data optimization is available in the Supporting Information.

1 [a] Each attempt was conducted with 1 equiv. of base. [b] Isolated yield. [c] Only traces of expected product were detected by 1H NMR after 20 h. [d] Reaction carried out under ambient atmosphere. [e] Reaction carried out with O2 balloon. [f] Reaction performed at RT rather than 50 °C. [g] Reaction performed at 110 °C.

The reaction proceeded smoothly with K2CO3 as a base (80 %, entry 5) and pyrrolidine (80 %, entry 6) but only traces of the cyclized product were detected after 20 h by using NEt3 (entry 7). Interestingly, the reaction is able to proceed in a wide range of solvents, providing 2 a in satisfying yields (entries 8–10, see the Supporting Information). Finally, without DBU (entry 11) or silver(I) oxide (entry 12) no reaction was observed, establishing that the combination of silver reagent and base is mandatory to promote such cyclization (entries 11 and 12). Using 0.25 equiv. of Ag2O under air atmosphere afforded 2 a (78 %, entry 13) after a prolonged reaction time (48 h) which can be dramatically reduced to 24 h by conducting the reaction under O2 atmosphere (83 %, entry 14). Interestingly, reactions can also be run with a catalytic amount of DBU (0.1 equiv.) without significant lengthening of the reaction time (3 h instead of 2 h). However, for accuracy and reproducibility considerations, we chose to perform the reactions with 1 equivalent of DBU instead of 0.1. Optimized conditions do not provide the product when the reaction is conducted at room temperature (entry 15). Also, we performed the reaction under conditions close to those reported by Kirsch but no product was detected (entry 16).

Next, the scope of the reaction was examined: Our optimized conditions (Ag2O: 0.5 equiv. DBU: 1 equiv. in MeCN, [C]=0.24 m) were first applied to various cyclic ketones (Scheme 2).

The cyclization of five (1 a) or six‐membered ring substrates (1 b) occurred under mild conditions and good yields were obtained (up to 85 %) with a total diastereoselectivity in favor of the cis‐adduct. We assumed that the selectivity is imposed by the configuration at the spiro carbon. Hence, for geometry and distance considerations, the ketone can be only attacked in a syn‐periplanar fashion leading to the (±)‐cis isomer. Surprisingly, seven‐membered ring analog 1 c was reluctant to cyclize and two equivalents of silver were necessary to attain a good conversion. Additionally, the cyclized products 2 c/4 c were obtained as a mixture of two diastereoisomers (55:45 by 1H NMR) probably due to a higher conformational flexibility. Starting from the naphthalenone derivative 1 d, no cyclization occurred even under heating in a sealed tube at 100 °C.[18] Tricyclic ketone 2 e was obtained in 78 % yield from the corresponding symmetrical diketone 1 e. Noteworthy, these cyclization conditions were also suitable to heterocyclic compounds and piperidine derivative 2 f could be isolated in satisfying yield (80 %). To our delight, using conditions B (Scheme 3) allowed the extension to substrate with longer chain (1 g) in order to reach six‐membered‐ring analog with good yield and high selectivity for the cis‐isomer 2 g. The reactivity of linear ketones was next surveyed by varying the substitution of the malonate, the nature of the ketone and/or the influence of the functional group in beta position. We first studied the behavior of different linear compounds towards the optimized cyclization conditions by tuning the nature of the substituent in alpha position to the ketone (R2). Substitution by an allylic chain (compound 5 a) as well as neutral (5 c) or enriched (5 d) benzylic chains was not deleterious to the reaction and good reactivities were observed along with modest selectivities in favor of the anti‐adduct (63:37, 67:33 and 63:37, respectively). Astonishingly, the diastereomeric ratio raised to 83:17 when the aromatic moiety was substituted by a p‐nitro group (5 g).[17] A bulkier isopropyl chain 5 b delivered the corresponding bicyclic compound in 67 % yield with inversion of the diastereomeric ratio in favor of the syn‐adduct (7 b), evidencing that the steric hindrance at this position is not detrimental to the reaction but strongly influences the selectivity. Modification of the size of the ester provided the expected 2,3‐dihydrofurans in good yields, however without notable improvement of the diastereoselectivity. On the contrary, the anti‐isomer 6 h was exclusively obtained with 58 % yield when the lactone 5 h was used.[17] Unlike benzylic ketone 5 j, ketone 5 i was reluctant to cyclize under conditions A while applying conditions B afforded good conversion, providing bicyclic compounds 6 i/7 i and 6 j/7 j (in 42 and 57 % yield, respectively) with a similar selectivity towards the anti‐isomer, along with degradation products. Finally, the substitution pattern of the triple bond was examined. To this aim, the reaction was conducted onto different substituted alkynes bearing either methyl (1 h), phenyl (1 i) or TMS (1 j) groups (Scheme 4).

Reactions onto 1 h or 1 i were ineffective under both conditions (A or B). On the contrary, when trimethylsilylalkyne 1 j was submitted to conditions A, it was smoothly converted into the cyclized product 2 a (R=H). This result suggests a preliminary insertion of silver into the C−Si bond, which is consistent with the well‐known propensity of terminal alkynes and TMS alkynes to undergo a metalation step under similar conditions.[19, 20]

In order to provide insights into the mechanism and to provide a rationale for the diastereoselectivity, we studied the first five steps of the AgI‐mediated cycloaddition reaction mechanism by means of density functional theory (DFT) approaches.[21] The computed reaction path together is presented in Figure 3, while computational details together with the labelling scheme are reported in Supporting Information. Two most probable isomers of BH+ and the two most probable approaches of B to Ag have been computed (see the Supporting Information for the related structures). We would like to underline that other trials starting from other B to Ag geometry always lead to one of the two species here reported. The data discussed here correspond to the arrangements of the B/BH+ and reactants providing the lowest energies.

A multistep pathway is proposed to explain the reaction mechanism up to the first cyclization process involving the carbonyl group and carbons of alkynyl residue (Figure 1). The analyzed reaction mechanism proceeds through five steps. Starting from the very stable silver alkynyl intermediate (I), the initial step corresponds to the proton transfer from conjugated acid of DBU (BH+ in Figure 2) to I leading to the formation of intermediate II. This step proceeds with a barrier of 14.40 kcal mol−1 (Table S1 in the Supporting Information) and leads to the formation of stable intermediate II. A subsequent proton transfer from II to DBU (B in Figure 2) generates the silver allenyl intermediate (III). The energetic barrier to overcome for its formation is found to be 12.07 kcal mol−1 (Figure 2). Of note the back reaction from II to I is characterized by a very small barrier so that we can assume that I is easily reformed. Following the energetic span model,[22] we can therefore assume that the energy span is related to the difference between the energy of I and TS2, leading to an effective barrier of 22.45 kcal mol−1 determining the reaction turnover. The III allenyl intermediate is characterized by practically equivalent C1−C2 and C2−C3 bonds (1.30 Å). From this intermediate, a first cyclization occurs via the TS3 transition state: the C3−C4 bond decrease achieving a value of 2.20 Å and suggesting the closure of the cycle. TS3 lies at 10.61 kcal mol−1 higher in energy with respect to the III and leads to the formation of intermediate IV, characterized by a C3−C4 bond of 1.57 Å and thus by a fully formed five‐carbon cycle. Of note in IV the silver atom approaches the oxygen (Ag−O bond of 2.15 Å), while C1−Ag and C2−Ag are elongating up to 2.70 Å. Comparing the atomic (Mulliken) charges computed for the TS3 and the IV, it is found that C4 becomes more electrophilic (from −1.91 |e−| to −0.22 |e−|), while the oxygen becomes more nucleophilic (from −0.34 |e−| and −0.49 |e−|). This makes possible to envisage a proton transfer from the conjugated acid BH+ of DBU to the oxygen atom of IV. The very low energy barrier (2.23 kcal mol−1) computed for this reaction confirmed this hypothesis.

In the corresponding transition state (TS4), the Ag atom is found to bridge the oxygen and the alkynyl lateral group (C1−Ag, C2−Ag, Ag−O bonds in the 2.30–2.40 Å range), whereas the C4−O bond is elongated up to 1.41 Å and the O−H bond is forming (1.21 Å). Indeed, the V intermediate shows a proton bonded to the oxygen (O−H bond 1.02 Å) and the silver atom coordinated to the carbons of alkynyl residue (C1‐Ag, C2‐Ag around 2.30–2.35 Å).

At this stage, the reaction proceeds via a second cyclization process involving the alkynyl lateral group and the oxygen to form the dihydrofuran cycle[23] which actually we computed to be extremely thermodynamically favored (ΔE=−119.16 kcal mol−1, see the Supporting Information).

In order to disclose the origin of the diastereoselectivity, we computed the first three steps of the reaction mechanism following two approaches leading to anti or syn diastereoisomers (Figure 3; all computational details are reported in the Supporting Information). All computed energy barriers are similar for both diastereoisomers paths nonetheless we observe that IIa is effectively more stable than IIb of 4.19 kcal mol−1. Indeed, the destabilizing interaction between the oxygen lone pairs of both carbonyls (O−O bond 2.80 Å) disfavors the syn adduct. With the aim of further validating the hypothesis that regioselectivity may be related to the presence (or absence) steric interactions for one of the adducts, we extended the study of reaction step II to compound 5 e. (Figure S3).

The reaction then proceeds with similar barriers both along the a and b pathway, suggesting that the final product depends on the most favorable initial intermediate. Indeed, with the aim of further validating our hypothesis, we extended the study of the reaction steps of compound 5 d. In this case, comparing the two optimized diastereoisomers, it is worth to notice that while the energy difference between the anti/syn adducts for the first step is similar to the value observed for compound 5 e (3.43 kcal mol−1), it decreases down to 1.39 kcal mol−1 at step II. Because such energy difference is very low, we can assume that this is the origin of the absence of diastereoselectivity. Indeed, for this compound, the destabilizing interaction is substantially decreased upon substitution with less sterically hindered methyl group, since the oxygen lone pairs of both carbonyls are located further apart in this case (O−O bond 3.00 Å, Figure S3).

Experimentally, we ran the reaction on bulkier amide 8 moiety leading to a drastic enhancement of the diastereomeric ratio (anti/syn: >95:5; Scheme 5).

However, we were unable to confirm the stereochemistry at this stage. In order to determine the relative configuration of the cyclized product, the amide 9 was subsequently reduced into the corresponding amine 10 in 73 % yield over two steps. A NOE experiment unambiguously evidenced anti adduct 10 as the major compound. In parallel, performing the reaction on ethyl ester 5 e under conditions B did not allow improvement of the diastereomeric ratio (50:50), underlining the primary importance of the nature of the electron withdrawing group on the diastereoselectivity. It is noteworthy that conditions B allowed a better conversion without influencing the diastereoselectivity.

This unprecedented formal silver‐mediated [3+2] cyclization appears as a valuable tool for the synthesis of complex and/or congested polycyclic dihydrofurans. As an example, the approach developed herein would permit straightforward access to precursor natural compounds that possess an angularly fused tricyclic lactone backbone like alliacol or teucrolivin, under mild conditions (Figure 4).

In summary, we have developed a new formal [3+2] cyclization promoted by silver(I) oxide and DBU from readily available starting β‐dicarbonyl compounds. The reaction is broad in scope, proceeds under mild conditions, and allows an access to complex and congested molecules in a single step. This transformation is highly tolerant towards a wide range of solvents and bases (in stoichiometric or in catalytic amounts), thus facilitating its adaptation to structurally diverse and potentially sensitive substrates. DFT calculations have demonstrated the formation of an allenyl silver intermediate by isomerization of the corresponding alkyne. Although descriptions of this species remain particularly scarce in the literature, its formation is a prerequisite in the promotion of the reaction. Moreover, DFT calculations have also allowed us to determine the thermodynamic origin of the syn/anti diastereoselectivity for β‐dicarbonyl compounds bearing an ester group. More generally, through using this methodology, we have highlighted a new reactivity of silver salts towards terminal alkynes. Thanks to this unprecedented isomerization, we have opened the way to a new field of transformations. The peculiar reactivity of allenyl silver species is currently under investigation in our laboratory.

B.Y. thanks the China Scholarship Council for a Ph.D grant. I.C. and A.P thank the European Research Council (ERC) for funding under the European Union's Horizon 2020 research and innovation program (grant agreement no. 648558, STRIGES CoG grant). We would like to thank Prof. P. Belmont for fruitful discussions and single‐crystal X‐ray diffraction service from ICMMO‐UMR8182 Orsay.

The authors declare no conflict of interests.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

GRAPH: Supplementary