Gold-Catalyzed Cyclizations of cis-Endiynes: Insights into the Nature of Gold-Aryne Interaction

Ortho-Benzyne^[1]/aryne^[2] are reactive and versatile intermediates in organic synthesis.^[3] Their diverse chemistry has been further enriched via the formation of metal aryne complexes.^{[3b, 4]} They are mostly generated via elimination of two adjacent groups/atoms on an arene ring. The elimination strategy, however, can be synthetically limiting due to difficulties in access to suitably substituted arenes, demanding reaction conditions and/or operational hazard. A notable exception is a de novo aryne formation via intramolecular hexadehydro-Diels-Alder reactions.^[5]

Considering that the coordination of alkynes to gold complexes often serves as the point of entry into versatile gold catalysis,^[6] it is notable that aryne has seldom served as the substrate for gold catalysis. Although aryne is highly electrophilic due to its low lying LUMO, the HOMO of benzyne was computed to have an energy level similar to that of 2-butyne.^[7] Hence, it is not unreasonable to anticipate the interaction between electrophilic gold complexes and aryne would be plausible. In fact, the study by Zhang^[8] has invoked a benzyne gold complex albeit the interaction between in-situ generated benzyne and Ph₃PAuCl is not clearly defined, leaving doubt to its existence. In contrast, another study^[9] used in situgenerated benzyne to trap a gold intermediate en route to anthracene derivatives, where no interaction between gold and benzyne is proposed. We envisioned that gold aryne interaction, if understood, can serve to spur further advance in gold catalysis. Herein, we disclose a study inferring that the interaction between a cationic gold complex and an aryne (i.e., a gold aryne complex) is a transition state and leads to regioisomeric ortho-aurophenyl cation intermediates, which could either be trapped by weak nucleophiles or undergo intramolecular C-H insertions via α-carbene gold carbene mesoisomers.

In our attempt to generate gold aryne complexes, we reasoned conceptually that aryne could be formed upon the addition of a deprotonated terminal alkyne to the other C-C triple bond in a cis-enediyne substrate with the assistance of a metal catalyst (Scheme 1A). Notable in this design is that the strained C-C triple bond in aryne is originally the terminal alkyne in the substrate and, in contrast to the typical elimination approach, formally not the bond formed. To implement it, we envisioned that an additional metal could facilitate the pivotal cyclization,^[10] as shown in Scheme 1B, thereby potentially leading to the generation of a metal aryne complex. Moreover, we reasoned that a cationic Au(I) complex could serve as the metal catalyst for both M¹ and M² in Scheme 1B. It is well established that gold complexes can activate carbon-carbon triple bonds efficiently and, at the meantime, alkynylgold can be readily formed in situ upon reacting with terminal alkynes.^[11]

Scheme 1.

New approach to the aryne moiety and metal aryne

We^[11] have recently reported the generation of gold vinylidenes (i.e. A) from benzene-1,2-dialkyne via a highly regioselective, goldpromoted 5-endo-dig cyclization (Scheme 2a).^[12] DFT calculations on that energy surface revealed that this intermediate was formed by a bifurcation mechanism in competition with α-auronaphthyl cation B_ben from 6-endo-dig cyclization^[13] (see the SI of that work); moreover, B_ben rearranges to its more stable structural isomer C_ben following a low-barrier gold migration. The calculation also predicted the formation of a gold carbene intermediate E_ben from C_ben upon a carbene C(sp³)-H insertion in the following step, thereby indicating the contribution of a mesomeric α-carbene gold carbene (i.e., C’_ben). While the naphthalene-type products were not detected as bulky BrettPhos^[14] was used in place of PH₃ as the real ligand in our experiments, we were nevertheless intrigued by the 6-endo-dig reaction path.

Scheme 2.

a) With ring A = benzene and R = Me, previous DFT calculations predict a bifurcation to form two possible intermediates A_ben and B_ben; the structures in the computed mechanism are indicated by a subscript ben. b) With ring A = cyclohexene and R = Et, the computed mechanism for the formation of 2a; the structures in the computed mechanism are indicated by a subscript cy.

The isomerization from B_ben to C_ben was calculated to go through a transition state along the reaction coordinate that is an arynecoordinated gold complex (i.e., D_ben). The phenyl cationic structures B_ben and C_ben can be viewed as resulting from gold atom slippage to the ends of the coordinating triple bond in D_ben. As a result, B_ben/C_ben appear to be the appropriate structures from the interaction between cationic gold(I) and aryne; moreover, they should be more electrophilic than typical arynes due to their aryl cation nature, which would be consistent with known silver catalysis[15] where insitu generated benzyne in the presence of Ag⁺ reacts with significantly enhanced electrophilicity.

With no success with benzene-1,2-dialkyne substrates in achieving the desired 6-endo-dig cyclization, we switched to the cisendiyne 1a, and the reaction discovery and optimizations are shown in Table 1. Much to our delight, the C-H insertion product 2a was indeed formed when the substrate was treated with BrettPhosAuNTf₂ (5 mol %) and 2,6-lutidine N-oxide (5a, 0.5 equiv) in 1,2-dichloroethane (DCE) (entry 1). The formation of 2a can be readily rationalized by invoking a mechanism shown in Scheme 2b, which is further supported by further DFT calculations (see SI). The additive 5a, in line with our previous study, acted as a mild base to facilitate the formation of the alkynylgold entry species in Scheme 2. This mechanism can also explain the formation of the phenol 4a, which is attributed to the addition of adventitious H₂O selectively to the ortho-aurophenyl cation C. The side product 3a, though in trace amount, is likely formed following a mechanism similar to that reported by Liu using PtCl₂ as the catalyst.^[16]

Table 1.

Initial reaction discovery and optimization.^[a]

en.	catalyst	base	temp/time	2a/3a/4a[b]
1	BrettPhosAuNTf2	5a	rt/17 h	61%/trace/9%
2	BrettPhosAuNTf2	-	rt/24 h	24%/trace/4%
3	BrettPhosAuNTf2	5b	rt/24 h	28%/trace/4%
4	BrettPhosAuNTf2	TsNa	rt/24 h	30%/trace/5%
5	BrettPhosAuNTf2	lutidine	60 °C /24 h	61%/trace/4%
6	tBuBrettPhosAuNTf2	5a	60 °C/24 h	45%/trace/10%
7	JackiePhosAuNTf2	5b[c]	rt/20 h	50%/ trace/trace
8	Mor-DalPhosAuNTf2	5a	60 °C/11 h	83%[d]/trace/trace
9	Mor-DalPhosAuCl/NaBArF 4	5a	60 °C/20 h	76%/ trace/trace
10	Ph3PAuNTf2	5a	60 °C/3.5 h	8%/54%/trace

^[a]Reaction run in flame-dried vial using dry DCE as solvent. [1a] = 0.05 M.

^[b]Estimated by ¹H NMR using diethyl phthalate as the internal reference.

^[c]1.1 Equiv.

^[d]Isolated yield.

Other base additives and gold complexes were then examined to improve the intended chemistry. In the absence of a base, only 24% of 2a was obtained (entry 2). Other bases such as 2,6-dibromopyridine N-oxide (5b, entry 3) and TsNa (conjugate acid pKa = 1.99^[17], entry 4) could not improve the reaction, either. With more basic lutidine (entry 5), the reaction was slower but resulted in an identical yield to that of entry 1. With regard to the gold catalyst, ^tBuBrettPhosAuNTf₂ (entry 6), a bulkier catalyst than BrettPhosAuNTf₂, and JackiePhosAuNTf₂ (entry 7), a more Lewis acidic catalyst, were less efficient. However, Mor-DalPhos^[18] turned out to be the most efficient ligand, affording 2a in 83% yield (entry 8). The use of less coordinating BAr^F₄^-,^[19] however, did not lead to further improvement (entry 9). While in all the cases discussed above 3a was formed in trace amounts, much to our surprise the Gagosz catalyst, Ph₃PAuNTf₂,^[20] promoted the formation of 3a at the expense of 2a (entry 10). It appears that the smaller steric size of the ligand Ph₃P is the cause for the switch of selectivity.^[21]

With the optimized conditions in hand, we then studied the scope of the reaction featuring C-H insertions, and the results are shown in Table 2, entries 1-8. In these reactions, 4Å molecular sieve was added to minimize the formation of phenol derivatives. Various types of aliphatic C-H bonds, including a primary C-H (entry 1), a tertiary C-H (entry 2), a benzylic C-H (entry 3) and a C-H α to a silyoxy group (entry 4) were readily inserted, leading to indane derivatives in mostly excellent yields. Changing the cyclohexane ring in the endiyne substrates discussed so far revealed that a smaller cyclopentene ring hindered the reaction as the corresponding substrate mostly remained unchanged; however, a bigger cycloheptene ring was inconsequential, and the indane 2f was isolated in 79% yield (entry 5). Moreover, the ring could be completely removed (entry 6) or replaced with n-butyl at either end of the C-C double bond (entries 7 and 8) without significantly impacting the reaction yield. Notably, in several cases the further optimized conditions deviated from the optimized ones shown in Table 1, entry 8 by the catalyst (e.g., BrettPhosAuNTf₂ in entries 2, 4 and 6) and/or the base (e.g., 5b in entry 6).

Table 2.

Reaction Scope of aryne gold Chemistry^[a]

^[a][1] = 0.05 M. Reactions were run in flame-dried vials in the presence of 4 Å MS except entry 13. Isolated yield were reported.

^[b]MorDalPhosAuNTf₂ was used.

^[c]BrettPhosAuNTf₂ was used.

^[d]Wet

Besides the C-H insertions by the carbene-type reactivity from the mesoisomer C’_cy (Scheme 2), the phenyl cation character from the mesoisomer of C_cy could also be harvested. For example, a tethered phenyl group could be the reacting nucleophile partner, and the reaction led to the formation of 9,10-dihydrophenanthrene (i.e, 2j, entry 9) in a good 80% yield. When a phenanthrene was the tethered nucleophile, the reaction resulted in a direct transformation of an endiyne into pentacyclic benzo[e]acephenanthrylene in a serviceable yield (entry 10). We were pleased to find out that a tethered HO group worked well as the nucleophile, and the tetrahydrooxepin 2l was formed in 90% yield (entry 11). This highly efficient formation of a 7-membered O-heterocycle and the lack of competing C-H insertion are particularly noteworthy, suggesting that C_cy/C’_cy prefers to react as a phenyl cation (i.e., in the form of C_cy). This observation is consistent with cationic gold complexes as strong soft Lewis acids^[6] but weak π-back donors.^[22] Similarly, a tethered sulfonamide was able to trap the cationic intermediate to deliver the seven-membered N-heterocycle 2m in a moderate yield (entry 12). Again, no competing intramolecular C-H insertion was detected.

During the scope study of C-H insertion reactions, we found that a 4-chlorobutyl group at the alkyne terminus did not undergo C-H insertion into the side chain. While this result again suggests that the carbene reactivity derived from C’_cy might be limited, the chloride substrate was used to examine intermolecular trapping of C_cy. With wet DCE as solvent, the phenol 2n was indeed formed in a very good yield (entry 13); similarly, ethanol and p-toluic acid, also used in excess, were suitable trapping nucleophiles as well, and the phenyl ether 2o and the phenyl ester 2p were isolated in 79% and 68% yields, respectively (entries 14 and 15). Importantly, these intermolecular trapping reactions were highly regioselective as the isomer anticipated in each case from trapping of the phenyl cation of type B_cy (see Scheme 2) was not detected.

(1)

Attempts to switch the regioselectivity were unsuccessful. For example, subjecting the endiyne 1q to the typical reaction conditions in DCE led to the formation of the phenol 5 in 75% yield in 42 h (Eq. 1). Neither the regioisomeric phenol 5’ nor the C-H insertion product 2q by an α-carbene gold carbene regioisomeric with C’ was not detected, which is consistent with the DFT calculations (Figure 1A), where B_cy-Me has a very low barrier to rearrangement to C_cy-Me and is 6.5 kcal mol^-1 less stable. It is expected that B_cy-Me would not be trappable, due to its predicted very short lifetime in solution. The reason for the enhanced stability for C_cy-Me could be attributed to the contribution of its mesomeric bent allene form, C’’_cy-Me (Figure 1B),^[23] consistent with the DFT optimized structure, in which the bond length of C2-C3 is 1.33 Å, significantly shorter than 1.40 Å in naphthalene. In contrast, the corresponding bond length for B_cy-Me is 1.45 Å.

Figure 1.

A) The rearrangement of B_cy-Me to C_cy-Me, energy diagram with electronic energies and free energies in methylene chloride in parentheses, and structures optimized at the M06/6- 31+G(d,p)/SDD(Au) level. B) The resonance structures of C_cy-Me.

(2)

It is well known that arynes can undergo [4+2] cycloaddition with furan;^[3] however, our attempts to perform the reaction with the endiyne substrates in the presence of gold catalysts led to no desired product. On the other hand, the reaction of 1n with trans-p-methoxycinnamaldehyde did lead to the formation of the [4+2] adduct 6. The formation of the [2+2+2] adduct 7 as the major product suggests that 6 is most likely generated via a step-wise process involving an ortho-aurophenyl cation of type C_cy.

(3)

(4)

(5)

Deuterium labeling studies were performed to gain further experimental support for the reaction mechanism. First, the alkyne terminus of 1c was labeled by a deuterium. With 10 equivalents of H₂O added, the reaction resulted in nearly complete loss of the isotope (Eq. 3). On the other hand, with a similar amount of D₂O as the additive, the non-labeled 1c proceeded to give 2c with significant amount of deuterium at the C4 but little at the C9 (Eq. 4). These two experiments confirm that the alkyne terminal hydrogen is removed as a proton during the reaction and the C4 hydrogen is installed via protonation. The origin of the C9 hydrogen was revealed experimentally by deuterium-labeling the side chain in a related substrate (Eq. 5). All the isotope-labeling results are consistent with the mechanism shown in Scheme 2b.

The remaining question with this chemistry is how the lack of the fused benzene ring in the substrates led to a complete switch of the initial cyclization from 5-endo-dig^[11] to 6-endo-dig (Scheme 2). It can be explained by considering the difference in gained aromatic stabilization: when ring A is a benzene ring, the additional resonance energy gained upon the 6-endo-dig cyclization is 28 kcal mol^-1 (64 kcal mol^-1 for naphthalene and 36 kcal mol^-1 for benzene); however, when ring A is not a benzene or nonexistent, the aromaticity gained by the newly formed benzene ring would be 8 kcal mol^-1 larger at 36 kcal mol^-1. Our DFT calculations in methylene chloride solution at 298 K confirm this idea. We find that the free energies for the formation of A and B (Scheme 2) favor B by 12.0 kcal mol^-1 when the fused ring is cyclohexene and by 15.6 kcal mol^-1 when there is no fused ring, which are much larger than 2.2 kcal mol^-1 difference with the fused ring as a benzene, where the 5-endo-dig product is favored.^[11] It is not unreasonable to imagine in the bifurcation mechanism that making the 6-endo-dig product much more stable relative to the 5-endo-dig product could skew the energy surface in a fashion that the 6-endo-dig path is favored.

In summary, we have disclosed a dual gold-catalyzed cyclization^[24] of cis-endiynes, which offers a de novo approach to various substituted indanes, heterocycle-fused benzenes and phenol derivatives. This represents a dramatic switch of the initial cyclization from the previously observed 5-endo-dig^[11] to 6-endodig in the bifurcation mechanism. Besides its synthetic utility, this chemistry offers for the first time insights on the interaction between cationic gold(I) and aryne. This interaction can be viewed as leading to a gold-aryne transition state that results in formation of two regioisomeric ortho-auroaryl cation intermediates, which can be considered as the results of the extreme slippage of the metal along the highly strained C-C triple bond. The more stable of these aryl cation forms reacts very regiospecifically with a range of nucleophiles in a manner similar to arynes, though the cation is more electrophilic. While cycloaddition reactions observed with arynes could not be realized, new reactivities with these polarized forms such as facile intramolecular insertions into C(sp³)-H bonds via mesomeric α-carbene gold carbenes are realized.

Experimental Section

General procedure for the dual gold catalyzed cyclization of cis-endiynes

2,6-dibromopyridine N-oxideor lutidineN-oxide(0.15 mmol), BrettPhosAuNTf₂(15.3 mg, 0.015 mmol) or Mor-DalPhosAuNTf₂ (14.4 mg, 0.015 mmol), and 4 Å Molecular Sieves (75 mg) were added in this order to a solution of cis-enediyne 1 (0.30 mmol) in DCE (6.0 mL) at room temperature. The reaction mixture was stirred at 60 °C and the progress of the reaction was monitored by TLC. Upon completion, the mixture was concentrated and the residue was purified by chromatography on silica gel (eluent: hexanes/ethyl acetate) to afford the desired product 2.