Organoytterbium Ate Complexes Extend the Value of Cyclobutenediones as Isoprene Equivalents

Changing course: While organolithium and Grignard reagents favor addition to C1 of A (R=Me), the corresponding organoytterbium reagents add to C2 (R=tBu). Computational studies provide insights into the nature of organoytterbium species and their reactivity, and a total synthesis of (-)-mansonone B illustrates the utility of the method in terpenoid synthesis. Tf=trifluoromethanesulfonyl.


In memory of Margaret Harrowven
3-Methyl-4-methoxycyclobuten-1,2-dione (1 a) has long been recognized as a valuable isoprene equivalent in natural products total synthesis, particularly for the preparation of hydroquinones, quinones, and their benzannulated analogues (e.g. Scheme 1). [1, 2a] It is readily introduced to a substrate as the electrophilic component in organolithium or Grignard addition reactions, for example, 1 a!2. Here, the differential reactivity of the C1 and C2 carbonyl groups in 1 a adds to its value as a synthon by providing a reliable and predictable means of achieving the ubiquitous head to tail connectivity of isoprene units. [1, 2a] Though regiocontrol in the addition of Grignard and organolithium reagents to 1 a is of critical importance, it also imposes a severe limitation. As evidenced by the example in Scheme 1, while the method is convenient for the synthesis of benzohydroquinones such as 3, it proves cumbersome when targeting the regioisomeric series 7. In this case addition of a carbon nucleophile to the vinylogous ester carbonyl (C2) of 1 a is required to achieve the desired outcome. Consequently, a protecting group strategy must be invoked to mask the more reactive C1 carbonyl (e.g. 1 a!4). [2] Addition of the Grignard or organolithium reagent to C1 is then followed by deprotection (e.g. 5!6), which can be difficult to achieve efficiently because of the presence of the acid sensitive tertiary alcohol and enol ether functions.
Herein we report an expedient solution to that longstanding problem and reveal some hitherto unknown facets of organoytterbium reactivity. In essence, while organolithium and Grignard reagents favor addition to the C1 carbonyl of cyclobutenedione (1 a), the corresponding organoytterbium reagents give exclusive addition to the C2 carbonyl of 1 b (Scheme 2). [3] The discovery was made during an optimization study aimed at reducing side reactions resulting from deprotonation of the C3 methyl substituent in 1 a. In the addition of PhLi, for example (Scheme 3), these appeared to limit the yield of the Scheme 1. Use of 1 a as an isoprene equivalent in the synthesis of benzohydroquinones.
Scheme 3. Regioselectivity in the additions of various phenyl organometallics to 1 a and 1 b. THF = tetrahydrofuran. adduct 2 a to approximately 70-80 % when employing standard protocols (e.g. THF at À78 8C). [4] Inspired by reports from the groups of Molander and Procter on the efficient addition of organoytterbium reagents to carbonyl compounds, [5] we decided to examine the addition of phenylytterbium triflate to 1 a. [2] The result we attained was quite unexpected, as the reaction led to the formation of both C1 and C2 adducts, 2 a and 11 a, respectively, in the corresponding yields of 50 % and 41 % (Scheme 3). Having noted similar anomalies in the additions of PhLi and PhMgBr to 1 b, we were delighted to find that the addition of phenylytterbium triflate to 1 b resulted in a complete reversal of the normal regiochemical course, thus giving the C2 adduct 12 a in a remarkable 92 % yield upon isolation. [5,6] To demonstrate the value of the method, a series of arylytterbium reagents were prepared and reacted with 1 b. These included ortho-, [7] meta-, [7] and para-substituted aromatics with a variety of substituents (+ I, ÀI, and + M), and a heteroaromatic example [7] (see Scheme 4 and the Supporting Information). In each case the C2 adducts 12 were formed in excellent yield. This result was in stark contrast to the corresponding organolithium additions to 1 a where the major product 2 resulted from addition to the C1 carbonyl. [7] For completeness, all adducts were subjected to thermolysis and aerial oxidation, thus leading to the complementary series of benzoquinones 15 and 16.
The complementary reactivity extended to vinyllithium and vinylytterbium triflate additions to 1 a and 1 b as demonstrated by the syntheses of the hydroquinones 23-25 and 27 (Scheme 5). [1] Notably, vinylytterbium triflate derived from vinylmagnesium bromide displayed the same regiochemical preference as those derived from vinyllithium intermediates. Interestingly, the addition of the vinyllithium 18 to 1 a proved unusual in that it gave 27 directly after a bicarbonate quench at À78 8C and warming to ambient temperature. The minor product of the reaction was the C2 adduct 21 b, [7] rather than the corresponding hydroquinone.
The addition of methylytterbium triflate to 1 b also proved facile (Scheme 6). In contrast, the analogous sequence with nBuLi required a 2:1 equivalence with ytterbium(III) triflate to achieve a high yield of the C2 adduct 28 b. [8] The photochemical rearrangement of the adduct 28 a into the furanone 29, [9] and the facile generation of the diorganylcyclobutendiones 30 by exposure of these adducts to acid, provide further illustrations of the methods utility.
For completeness, the addition of phenylalkynylytterbium triflate to 1 b was examined (Scheme 7). The reaction proved sluggish (9 h at À78 8C) in comparison with other additions and regioselectivity was attenuated, with the formation of cyclobutenones 34 and 33 in 69 % and 15 % yield, respectively. In contrast, the addition of the organolithium 31 to 1 a gave the cyclobutenone 32 as the sole isolated product.
Thermolyses of the adduct 32 and 34 exposed a further anomaly, with the former giving a 5:4 mixture of the quinone 35 and cyclopentenedione (E)-36, while the a latter gave only the cyclopentenedione (Z)-37.
To gain insights into the mechanistic course of organoytterbium addition reactions we decided to initiate a computational study of the reaction of PhYb(OTf) 2 with 1 b. However, these investigations proved fruitless, with DFT calculations predicting a preference for reaction at the ketonic C1 carbonyl (i.e. DG°A < DG°D, Figure 1). [10] The results led us to question the nature of the organometallic reagent. Based on limited literature precedent, [11] we had assumed that the reaction of PhLi with Yb(OTf) 3 would first give the ate complex PhYb(OTf) 3 Li, which in turn would collapse to PhYb(OTf) 2 and LiOTf. When this premise was tested computationally, our calculations indicated that loss of a triflate ligand from the ate complex would cost approximately 25 kcal mol À1 at À78 8C! Moreover, when the reactivity of [PhYb(OTf) 3 ] À towards 1 b was examined, the predicted outcome mirrored our experimental findings by revealing a kinetic preference for phenyl addition to the C2 carbonyl (i.e. DG°H < DG°E, Figure 1).
The modeling also exposed some seminal facets of organoytterbium reactivity (Figure 1). In particular, the prereaction complexes of [PhYb(OTf) 3 ] À ·1 b (F and G) each showed single ytterbium-carbonyl interactions leading to trapezoidal transition states. The strength of those interactions was a key factor in determining the regiochemical course of addition, with DG G less than DG F by about 2.5 kcal mol À1 . Additionally, steric interactions in the addition phase play an important role, as evidenced by 1) the need to rotate the tBu ether during C1 addition (F!E, Figure 1) at a cost of about 2.6 kcal mol À1 [12] and 2) the low selectivity observed for addition of [PhYb(OTf) 3 ] À to 1 a compared to that for 1 b (Scheme 3).
Finally, to demonstrate the methods potential in terpenoid synthesis we decided to target mansonone B, a natural product whose identity has yet to be determined with rigor but seemed likely to be as depicted in 43. [13] Our synthesis began with (À)-menthone (39; Scheme 8), which was readily transformed into the required vinyllithium reagent 41 by means of a Shapiro reaction. [14] Transmetallation with ytterbium triflate, and subsequent addition of the resulting vinylytterbium intermediate to 1 b, gave a 7:3 diastereomeric mixture of the C2 adduct 42 in 94 % yield. Thermolysis of that mixture under continuous flow at 110 8C with concomitant aerial oxidation, gave the quinone 44, from which our target 43 was readily derived. Pleasingly, the physical and spectral characteristics displayed by our synthetic sample (À)-43 matched those reported for mansonone B in its isolation paper, [13a] thus confirming the regiochemical and relative stereochemical identity of the natural product. Its absolute configuration remains uncertain because of a lack of available optical rotation data for comparison.
In conclusion, we have shown that organoytterbium additions to 1 b are easy to effect, proceed in excellent yield, and follow a complementary regiochemical course to related organolithium additions to 1 a. Alkyl, aryl, hetaryl, vinyl, and alkynyl ytterbium reagents participate in the reaction, greatly extending the utility of cyclobutenone rearrangements by providing a predictable means of introducing this isoprene unit in a regioselective manner. Highlevel computational studies have provided seminal insights into the nature of organoytterbium intermediates and their reactivity. Of particular note is the realization that reactions proceed via ytterbium ate complexes. . Keywords: density functional calculations · flow chemistry · natural products · rearrangement · ytterbium