|Home | About | Journals | Submit | Contact Us | Français|
Platensimycin (1) is the flagship member of a new and growing class of antibiotics with promising antibacterial properties against drug resistant bacteria. The total syntheses of platensimycin and its congeners, platensimycins B1 (7) and B3 (9), platensic acid (2), methyl platensinoate (3), platensimide A (4), homoplatensimide A (5), and homoplatensimide A methyl ester (6) (Figure 1) are described. The convergent strategy developed toward these target molecules involved construction of their cage-like core followed by attachment of the various side chains through amide bond formation. In addition to a racemic synthesis, two asymmetric routes to the core structure are described: one exploiting a rhodium-catalyzed asymmetric cycloisomerization, and another employing a hypervalent iodine-mediated dearomatizing cyclization of an enantiopure substrate. The final two bonds of the core structure were forged through a samarium diiodide-mediated ketyl radical cyclization and an acid-catalyzed etherification. The rhodium-catalyzed asymmetric reaction involving a terminal acetylene was developed as a general method for the asymmetric cycloisomerization of terminal enynes.
The discovery of new antibiotics is an urgent medical priority due to the continuous emergence of drug resistant strains of pathogenic bacteria. As evolution drives the inevitable development of bacterial resistance to existing drugs, so the search for new treatments of bacteria-induced infections must continue.1,2 Natural antibiotics are produced by microorganisms as chemical defenses in their fight for scarce resources. Such molecules have been at the forefront of antibiotic research since the seminal discovery of penicillin.3,4 The screening of compounds for antibacterial activity is fraught with difficulty, however, due to two conflicting sets of problematic circumstances.5 Firstly, high throughput whole-cell assays of naturally occurring molecules struggle to distinguish specific antibiotic action from non-specific toxicity and can lead to the re-isolation of known compounds. Any hit must then be individually identified and its mechanism of action determined, requiring a significant investment of effort and resources. On the other hand, the use of target-based assays, while ensuring the hits are active against a known cellular target, has proven troublesome due to difficulties in realizing such activity in whole-cell settings. A team led by Wang and Singh at Merck recently addressed these problems by developing an antisense RNA assay for antibiotic activity.5 Briefly, they screened natural product isolates against a strain of Staphylococcus aureus in which the expression of the FabH and FabF condensing enzymes of the fatty acid biosynthesis pathway (FAS II) was inhibited by the presence of antisense RNA corresponding to the genes encoding both proteins. The strain is thus sensitized to fatty acid biosynthesis inhibitors. By screening samples against this strain and one expressing these enzymes normally in parallel, samples containing compounds that act through inhibition of FAS II could be identified by their differential effects on the two strains.5
Using this technique, Wang, Soisson, Singh, and coworkers identified several potent FAS II inhibitors, including platensimycin (1, Figure 1), which was isolated from a strain of Streptomyces platensis from a soil sample collected in South Africa.6 They determined the complex polycyclic structure of platensimycin and demonstrated that it acts through highly potent and selective inhibition of the elongation condensing enzyme FabF. As this mechanism of action is not shared by any clinically used antibiotic, platensimycin showed no cross-resistance and was found to be active against a broad range of pathogenic Gram-positive bacterial species. Platensimycin was also shown to be effective in vivo, clearing a model S. aureus infection in mice, although only at relatively high, continuously administered doses. Interestingly, the fatty acid inhibition strategy to combat bacteria has recently been challenged on the basis of experiments that showed their survival does not depend crucially on their own production of fatty acids since they can tap on the reserves of their hosts for such essential ingredients.7
Singh and coworkers also disclosed biosynthetic studies on platensimycin (1), indicating the terpenoid nature of the tetracyclic carboxylate “right-hand” portion,8 as well as preliminary chemical studies9 and the structures of several related natural products (Figure 1).10 Platensic acid (2) and its methyl ester (3) were isolated as natural products in their own right,10a as might be expected from the biosynthetic pathway, and were found to have virtually none of the antibiotic activity of platensimycin (1). The naturally occurring aliphatic amide platensimide A (4),10a which retains the right hand structural motif of platensimycin (1), was also found to be inactive, as was homoplatensimide A (5).10b The relative stereochemistry of the 2,4-diaminobutyrate unit of platensimide A (4) was verified by the semisynthesis of 4 from 2.10a Homoplatensimide A (5) and its methyl ester, homoplatensimide A methyl ester (6), contain a carboxylate moiety with the complete C20 skeleton of a diterpene, in keeping with the biosynthetic proposal for these compounds, bound through an amide linkage to glutamide.10b Platensimycins B1–B3 (7–9, Figure 1), three other natural congeners of platensimycin (1) with modest changes in the benzoic acid domain of the molecule, also show poor activity due to the lack of a free carboxylic acid moiety.10c Platencin (10, Figure 1) was also isolated from a strain of S. platensis.11 Although it shows an overall structural similarity with platensimycin (1), particularly in the common aromatic portion, the latter compound has a less-oxygenated ketolide domain. Platencin (10) possesses a similar biological profile to platensimycin (1), but, surprisingly, a slightly different mechanism of action.11
The novel and intricate chemical structure and vitally important biological activity of platensimycin (1) made it an attractive target for total synthesis, and a number of groups have reported their synthetic studies toward this end. Following our initial reports on the total synthesis of (±)-platensimycin [(±)-1]12a and, later, of the natural (−)-enantiomer [(−)-1],12b several groups reported alternative elegant approaches.12c–k Additionally, our research group and those of others have prepared a number of biologically active analogues of platensimycin.13 We present here an improved synthesis of platensimycin (1) and a full account of our studies on the total synthesis of this fascinating target as well as the related compounds (2–7, 9) sharing its tetracyclic enone structural motif. Our studies on the total synthesis of platencin (10) are reported elsewhere.14
We began our studies by pondering a retrosynthetic analysis of the platensimycin molecule (1). The final route developed is depicted in retrosynthetic format in Figure 2. Our analysis began with the straightforward amide disconnection to generate aromatic amine 11 and platensic acid (2) as potential precursors. From here, a double disconnection of the propionate and methyl α-substituents revealed the tetracyclic ketolide 12 as the primary synthetic subtarget. The inherent structural bias of cage-like intermediate 12 was expected to impart stereocontrol in the construction of the C4 quaternary center. We considered a variety of disconnection strategies at this stage. We hypothesized that disconnection of the ether ring, as shown, was likely to be productive since it not only simplified the target significantly, but also offered the chance to control the C15 stereocenter as a result of geometric constraints, and complete the cage-like structure by way of well-understood chemistry. The most straightforward etherification substrate, tricyclic enone 13, contains a 4-hydroxyketone structural motif. Disconnection of the C9–C10 bond by way of a retro 1,4-addition was an intriguing prospect. This disconnection resulted in a spirocyclic 4,5-cyclohexadienone bearing a single stereogenic center (i.e. 14, Figure 2). Although the success of the 1,4-addition reaction was by no means assured and the preparation of a suitable substrate might prove challenging to prepare, we were attracted to this strategy by the possibility of exploiting the local symmetry of the dienone unit to set the all-carbon quaternary center at the heart of this subtarget. Inspection of the three-dimensional structure of substrates such as 14 indicated that if a successful 1,4-addition reaction could be achieved, the stereochemistry at C8 and C9 would follow as a matter of course. Additionally, several reactions could be used to effect the desired intramolecular 1,4-addition to the enone, including radical and Stetter-type processes, giving the approach some flexibility. Aldehyde 14 could act as a substrate (or suitable precursor for such a substrate) for several of these reactions. Setting the absolute stereochemistry of the single stereocenter of 14 (C12) would be the key to realizing an asymmetric synthesis of platensimycin, a challenge for which we envisioned two separate possible solutions. The first (a, Figure 2) involved forging the C12 stereocenter and the spirocyclic cyclopentyl ring in a single cycloisomerization step from a symmetrical, achiral 1,6-enyne precursor (15). This intermediate, in turn, could be traced back, through Stork–Danheiser alkylation chemistry, to enone 16. This route offered the advantages of a direct and potentially catalytic asymmetric synthesis, but faced the challenge of realizing an asymmetric cycloisomerization reaction of terminal alkyne substrates, which at the time was an unsolved problem. Our second strategy (b, Figure 2) involved separating the tasks of forming the spirocyclic system and the stereocenter. The former would result from a dearomatizing cyclization of a 4-substituted phenol such as 18. Our interest in the chemistry of hypervalent iodine reagents led us to propose an oxidative cyclization using an allylsilane as an internal nucleophile. The substrate for this reaction (18) was easily envisioned to derive from a ketone such as 19, which was expected to be available in enantiopure form via the alkylation of pseudoephedrine amide 20. In this case, the advantages lay in the likely ease of substrate preparation employing Myers’ alkylation technology using readily available starting materials. Within this route, a dearth of precedent for the dearomatization reaction gave us some concern. In the end, we would investigate both asymmetric approaches; initially, however, we focused on securing the relatively late key step, which was perceived to be the most challenging aspect of the proposed synthesis: the conversion of aldehyde 14 to cage structure 12, using racemic material.
Our urgent desire to investigate reactions of aldehyde 14 led us to prepare this compound in racemic form in the first instance. This endeavor also served as a preliminary investigation for the first asymmetric route to platensimycin [(−)-1]. Our optimized route to (±)-14 is shown in Scheme 1. Starting from ketone 17, we carried out sequential alkylation reactions using the established protocol of kinetic enolate generation with LDA. This allowed the installation of what would eventually become the quaternary center at C8 and all the carbon atoms needed for the lower region of the platensimycin enone domain to afford intermediate 16 in excellent overall yield. Reduction of enone 16 with DIBAL-H followed by treatment with aqueous HCl led to enone 22 in good yield.15 Short reaction time and maintaining the mixture at 0 °C prevented significant loss of the silyl ether. Treatment of enyne 22 with ruthenium catalyst ([CpRu(MeCN)3]PF6) in acetone resulted in a smooth and efficient cycloisomerization reaction,16 furnishing spirocycle 23 as an inconsequential 1:1 mixture of diastereoisomers. Formation of the TMS enol ether of enone 23 and oxidation gave dienone 24, which was converted into aldehyde (±)-14 by acidic hydrolysis of the TBS enol ether. The oxidation reaction could be carried out using stoichiometric amount of Pd(OAc)2, as reported previously,17a but was accomplished most efficiently using the mild catalytic system of Pd2(dba)3•CHCl3 in combination with diallyl carbonate.17b In either case, excellent selectivity was observed for oxidizing the TMS enol ether to afford 24.
We also investigated an asymmetric version of the above route to aldehyde 14. No asymmetric variant of the ruthenium-catalyzed cycloisomerization reaction16 had been reported at the time; however, Zhang and coworkers had described an equivalent transformation catalyzed by chiral rhodium(I) diphosphine complexes.18 While, in general, excellent substrate scope and enantiomeric excesses were noted in these reports, examples incorporating terminal alkyne partners were conspicuously absent. In order to investigate the possibility of an asymmetric cycloisomerization, the synthesis had to be slightly modified to incorporate the second olefin in the enone ring so as to produce achiral substrate 15 (Scheme 2). We also prepared related substrates with the alkyne moiety capped with various functional groups as shown in Scheme 2. Thus, activation of the ketone group of 22 gave trimethylsilyl enol ether 25, which served as a common intermediate for all required substrates (Scheme 2). Oxidation of 25 with IBX•MPO19 led to dienone 26, with deprotection of the TBS ether giving parent prochiral substrate 15. Alternatively, the oxidation of cyclohexenone 22 could be achieved in a one-pot procedure through the introduction of a phenylselenide group followed by selenoxide formation and syn elimination as summarized in Scheme 2. Two further substrate analogues were prepared by taking advantage of the temporary masking of the ketone group as enol ether 25. Lithiation of the terminal alkyne followed by quenching with TMSCl or Mander’s reagent gave enol ethers 27 and 28, respectively (Scheme 2). In each case, oxidation with IBX•MPO (to afford 29 and 30, respectively) and acid-induced desilylation proceeded smoothly, giving prochiral substrates 31 and 32, respectively, as shown in Scheme 2.
As we expected from the lack of literature examples, our initial investigations into the asymmetric cycloisomerization of terminal enyne 15 proved disappointing. Exposure of this terminal alkyne to a mixture of [Rh(cod)Cl]2, (S)-BINAP, and AgSbF6, as reported by Zhang and coworkers,18c–g resulted in sluggish conversion and significant decomposition of the starting material upon prolonged reaction times, and, thus, only a very low yield of product aldehyde (−)-14 and poor recovery of substrate alcohol 15 (Scheme 3). Increasing the catalyst loading did not improve the yield of (−)-14, but accelerated the decomposition of 15. However, we were encouraged by the observation that aldehyde (−)-14, obtained in low yield (15%) under these conditions, was formed with excellent enantiomeric enrichment (95% ee). We made recourse to the ‘capped’ alkynes described above in an effort to improve the chemical yield of the reaction, and subjected alkynyl silane 31 to the same catalyst system (Scheme 3). Unfortunately, the expected spirocyclic aldehyde 33 was not detected in this case, presumably due to the steric hindrance of the TMS group.20
Fortunately, the ester-bearing substrate 32 proved amenable to cycloisomerization under asymmetric conditions, as shown in Scheme 4. Exposure of 32 to the standard rhodium catalyst system ([Rh(cod)Cl]2, (S)-BINAP, AgSbF6) rapidly gave spirocyclic ester (−)-34 in excellent yield. HPLC analysis using a chiral stationary phase indicated that (−)-34 was formed in greater than 95% ee.21 The sense of asymmetric induction using the rhodium(I)-BINAP system appeared to be uniformly consistent throughout the published examples,18 which span a variety of substrate types. We were, therefore, confident that (S)-BINAP would provide the correct absolute configuration required for our synthesis of (−)-platensimycin [(−)-1]. This expectation was borne out by comparison of optical rotations at a later stage (vide infra). The use of 32 in the successful asymmetric cycloisomerization reaction, however, came at the cost of removing the superfluous carboxylate moiety from product (−)-34. This was achieved by conversion of the potentially sensitive aldehyde group to the corresponding ethylene glycol acetal (−)-35, followed by saponification of the methyl ester to give carboxylic acid (+)-36, during which the exocyclic double bond had migrated to the endocyclic position, presumably to release some strain. Formation of the Barton ester (reaction with reagent 37 and EDC•HCl, Scheme 4) followed by photolysis in the presence of t-BuSH gave deacarboxylated compound (+)-38.22 Finally, hydrolysis of the acetal group within (+)-38 led to the corresponding aldehyde (+)-39 as shown in Scheme 4.
The asymmetric synthesis of aldehyde (+)-39 allowed us to demonstrate a formal asymmetric synthesis of (−)-platensimycin (−)-1;12b however, the requirement for a multi-step decarboxylation sequence amounted to an unsatisfactory overall efficiency. Obviously, the realization of an effective asymmetric cycloisomerization of enyne 15 would provide the most direct means to access the key intermediate aldehyde (−)-14.
At this point, we decided to embark on the development of an asymmetric enyne cycloisomerization reaction using terminal alkynes as substrates.23 Employing 1,6-enyne 40 as a substrate, we examined a number of Rh-based complexes in order to find the optimum catalytic system; our results are summarized in Table 1. Thus, while the performance of [Rh(cod)(MeCN)2]BF4 in the presence of (S)-BINAP was rather modest (36% yield of product 40a, after 12 h at 23 °C), the product was obtained with significant enantiomeric enrichment (90% ee, Table 1, entry 1). The complex generated in situ by mixing [Rh(cod)Cl]2 and AgOTf in the presence of (S)-BINAP led to a 60% yield and a comparable ee (91% ee, Table 1, entry 2), whereas the combination of [Rh(cod)Cl]2, AgSbF6, and (S)-BINAP employed by Zhang et al. with internal alkynes 18c–g gave 40a in 65% yield and 95% ee (Table 1, entry 3). Finally, it was found that the preformed rhodium catalyst [Rh((S)-BINAP)]SbF624 gave the best results, furnishing product 40a in 86% yield and > 99% ee (Table 1, entry 4). The absolute configuration of product 40a was determined by X-ray crystallographic analysis (see ORTEP drawing, Figure 3) of its p-bromophenyl carbamate derivative25 (40b, m.p. 99–101 °C, EtOAc/hexanes 1:1, prepared as shown in Scheme 5).
The generality and scope of this new asymmetric reaction was examined and the results are summarized in Tables 2 and and3.3. As seen from entries 2–10 (Table 2), the process tolerates well all cis allylic alcohols tested as substrates (42–50, Table 2), leading to good to excellent yields and enantiomeric excesses of the cycloisomerized products (42a–50a, Table 2). The reaction, however, performed only modestly in the case of trans allylic alcohol 41 (Table 2, entry 1), giving aldehyde 40a (the same product obtained from the cis allylic alcohol 40, Table 1) in 45% yield and 29% ee. The reaction performed well with substrates containing a wide range of structural motifs, linking the two reactive ends of the molecule, namely sulfonamides (Table 2, entry 2), amides (Table 2, entries 3 and 4), 1,1-diesters (Table 2, entries 5 and 6), 1,1-sulfones (Table 2, entry 7), bis-aromatic moieties (Table 2, entry 8), and ether linkages (Table 2. entry 9). It was also demonstrated that amide carbonyl groups in conjugation with either the olefin or alkyne reactive moieties are tolerated in this process (Table 2, entries 3 and 4, respectively). Most notable is the success of this asymmetric reaction with the bromoacetylenic substrate 50 (Table 2, entry 10), leading to vinyl bromide 50a, a product that could be utilized in a variety of further transformations, thus expanding considerably the scope of the method.
Table 3 depicts a number of experiments undertaken in order to explore further aspects of the asymmetric cycloisomerization. Thus, TBS ether 51 and allylic acetate 52 (Table 3, entries 1 and 2, respectively) entered the reaction successfully to afford enol ether 51a and vinyl acetate 52a, respectively, in high yield and ee, whereas entries 3 (substrate 53) and 4 (substrate 54) demonstrate that the reaction tolerates sulfur and nitrogen substituents, respectively, on the ene side chain. Finally, it was shown that a heteroatom is not necessary in the ene side chain for the success of the reaction as demonstrated with substrate 55 (Table 3, entry 5).
Much to our delight, substrate 15, the precursor to platensimycin [(−)-1], entered the asymmetric cycloisomerization with 5 mol % catalyst ([Rh((S)-BINAP)]SbF6) to afford the desired spiro dienone aldehyde (−)-14 in 86% yield and > 99% ee, as shown in Scheme 6. As mentioned above, application of Zhang’s protocol18c–g to substrate 15 led to inferior yield (15%) and lower ee (95%) (see Scheme 3). Although we cannot pinpoint precisely the reasons for the observed difference between the two catalyst system at the present time, we can, however, speculate that the observed results may have their origins in the fact that our catalyst24 is more well-defined than the various species that must be present in the previously employed catalyst system.18c–g
Alongside our efforts to develop an asymmetric cycloisomerization route to aldehyde (−)-14, we also investigated the alternative approach outlined retrosynthetically in Figure 2 (route b). In this approach, we planned to forge the spirocyclic ring system through a dearomatizing cyclization. Such processes commonly employ the combination of a phenol with a strong electrophile, such as a protonated diazoketone or allyl halide.12f,g,26 The oxidation of phenols in the presence of a suitable nucleophile has been used to achieve a similar transformation, particularly using heteroatom or aromatic nucleophiles,27 although examples with carbon-based nucleophiles, aside from aromatic systems, are rare.28,29 We favored the use of an allyl silane nucleophile under oxidative conditions due to an interest in exploring the feasibility of such a transformation and the potential ease with which a suitable substrate might be prepared, although misgivings remained as to the success of the key step. Allylsilane nucleophiles had been employed only rarely in oxidative additions to phenols. Quideau and coworkers have reported the intermolecular addition of allyl trimethylsilane to naphthols mediated by iodine(III) reagents; however, they observed that good yields were obtained only when the phenolic partner was pre-treated with the oxidant followed by addition of the silane, suggesting an incompatibility between allylsilanes and iodine(III) oxidants.28b
As indicated in Figure 2, we adopted a plan which called for the installation of the stereogenic center found within key aldehyde 14 prior to the dearomatizing cyclization. As expected, the substrate for the cyclization was prepared in a straightforward manner using Myers’ asymmetric alkylation protocol (Scheme 7). Thus, known carboxylic acid 5630 was coupled with (S,S)-pseudoephedrine 57 to give amide (−)-20 in quantitive yield. Alkylation according to Myers’ procedure31 with benzylic bromide 5832 gave amide (−)-59 in excellent yield, albeit with rather low selectivity (crude de ~85%) when compared with literature examples. However, a single recrystallization of the crude product provided (−)-59 as a single diastereoisomer in 75% yield. Cleavage of the auxiliary group from (−)-59 with methyllithium then provided methyl ketone (−)-19 in excellent yield with enantiomeric excess of > 98%.33 Treatment of ketone (−)-19 with KHMDS and Comins reagent (60) gave enol triflate (−)-61 in excellent yield, although the efficiency of this reaction was highly dependent on the quality of the Comins’ reagent used. Kumada coupling of (−)-61 with (trimethylsilyl)methyl Grignard reagent [2.5 mol % Pd(PPh3)4, LiCl] and selective desilylation of the phenol under basic conditions (NaOH) furnished dearomatization substrate (−)-18 in excellent overall yield.
A wide range of conditions were screened for the proposed oxidative cyclization reaction of (−)-18, some of which are summarized in Table 4. Iodine(III) reagents were quickly determined to be the most effective promoters, with PhI(OAc)2, PhI(O2CCF3)2, and PhIO all giving the desired spirocyclic product (−)-62 in low to moderate yield under a variety of solvent and temperature conditions. Non-nucleophilic polar solvents proved the most effective media for this reaction, as might be expected for this type of transformation.34 The combinations of the milder reagent PhI(OAc)2 and either 2,2,2-trifluoroethanol (TFE) or 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) were equally effective, with the former solvent favored on the basis of cost (entry 4, Table 4). Under the optimized conditions, phenolic allylsilane (−)-18 was converted to (−)-62 in 68% yield as shown in Scheme 7. Removal of the acetal protecting group from this intermediate revealed aldehyde (−)-14 in high yield.
The dearomatization reaction is presumed to proceed through an initial interaction of the iodine reagent with the phenol, followed by oxidation of the aromatic ring and either concomitant or step-wise addition of the nucleophile. Such a mechanism is in accordance with accepted reactivity of phenols with iodine(III) reagents27 and with the results of Quideau mentioned above.28b,c Although we have not identified and characterized any of the by-products formed from the reactions of (−)-18, we observed that when the TBS ether of (−)-18 was treated with PhI(OAc)2 in TFE it underwent rapid proto-desilylation. This result, while not involving an oxidation reaction, indicates that the allylsilane group is indeed reactive under the reaction conditions. Based on this observation, we assume that at least some of the by-products generated under the dearomatization conditions employed may come through this parallel pathway.
The dearomatization reaction of phenolic allylsilanes was investigated further using several simpler substrates (Scheme 8). In each case, treatment of the phenol with the iodine(III) reagent led to cyclization to give the products shown, suggesting considerable generality of this process for the formation of spirocyclic systems. As shown, both branched and linear allylsilanes participated in this process satisfactorily. Canesi and coworkers have recently reported similar reactivity involving both allylsilane and enol ether nucleophiles in iodine(III)-mediated reactions of phenols to generate a variety of products, including (±)-aspidospermidine.35
Having secured access to aldehyde 14, both in its racemic and enantioenriched forms, we began exploring the possibility of converting it to the next intermediate, such as tricyclic species 13, through formation of the C9–C10 bond (see Scheme 9). Our initial investigations focused on ionic processes such as the nucleophilic carbene-catalyzed Stetter reaction. However, treatment of (−)-14 with several hetereocyclic carbene species reported to promote the Stetter reaction36 gave no tricyclic products. We attributed this failure to the strain involved in forming the norbornane motif of the targeted tricyclic structure. In an alternative approach from these laboratories,12d it proved possible to forge the C9–C10 bond using a Stetter reaction36 to form a decalin system lacking the bridging ring, which was installed later. In an effort to increase the reactivity of substrates along similar lines, we prepared the TMS cyanohydrin of (−)-14 (structure not shown) and attempted to induce cyclization through treatment with a strong base. Again, the results were disappointing and recourse was made to the ethoxyethyl (EE) protected cyanohydrin (−)-66, prepared in excellent yield from (−)-14 through standard chemistry, as shown in Scheme 9. In this case, treatment of (−)-66 with KHMDS at 0 °C led to rapid cyclization to form tricyclic product (−)-67 in high yield (86%). The stereocenters at C9 and C10 were both formed as shown with complete control, although the product remained a mixture of two diastereoisomers due to the stereocenter within the EE group. The cyanohydrin moiety could be removed in one step by sequential treatment with aqueous acid and base to give diketone (+)-68 in 89% yield. The stereochemistry at C9, predicted based on geometric considerations in the 1,4-addition reaction, was confirmed at this stage by nOe studies. Reduction of (+)-68 with DIBAL-H gave a diol, which could be oxidized selectively with MnO2 to furnish alcohols 13a and 13b as a mixture of inseparable diastereoisomers (13a:13b ca. 1:5). Unfortunately, the major isomer (13b) was found to be the undesired compound, as determined by the ability of only the minor isomer (13a) to engage its exocyclic olefin in a ring forming reaction (vide infra). Despite screening a range of reductants, we were not able to improve the ratio of these products. Additionally, alcohol 13b was found to be resistant to inversion, with Mitsonobu conditions giving no reaction, and displacement of the corresponding mesylate or triflate proving fruitless.
In parallel with the studies outlined above, we also investigated the potential of a samarium diiodide-mediated ketyl radical cyclization as a means to convert (±)-14 to (±)-13a/13b. Such a process has the advantage of giving (±)-13a/13b directly in a single step.37 Additionally, the inherent stereochemical bias of these processes would be expected to favor the selective formation of the desired stereoisomer (±)-13a. Initial attempts to cyclize (±)-14 were uniformly unsuccessful, despite a range of conditions being applied (Table 5). Optimization efforts were hampered by a lack of identifiable by-products. Small quantities of (±)-13a/13b (ca. 2:1 ratio) were isolated from a reaction employing HMPA and MeOH at low temperature (entry 4). This reaction was further optimized to provide a 2.2:1 mixture of (±)-13a and (±)-13b in up to 51% yield (entry 5). In this case, the major component in the reaction mixture was the desired isomer (±)-13a. Notably, the best results were obtained when the samarium reagent was added very rapidly, suggesting that the mixing rate is important. The mixing rate was especially crucial in preparative scale experiments. This result, although to some extent disappointing, allowed for the preparation of sufficient quantities of 13a for further studies [the samarium diiodide-induced ring closure (entry 5, Table 5) was also carried out with (−)-14 to afford (−)-13a/(+)-13b38]. Other reagents for the preparation of ketyl radical intermediates,39 for example, titanium(III) species, AIBN/n-Bu3SnH and Et3B/O2/n-Bu3SnH, were investigated for this transformation (14 → 13a/13b) without success.
The final bond construction required to complete the platensimycin cage-like structural motif was etherification between the alcohol group at C10 (i.e. O16) and C15 (see Scheme 10). Treatment of a mixture of (−)-13a and (+)-13b [obtained from samarium diiodide-mediated cyclization with (−)-14] with PhSeCl in the presence of K2CO340 at low temperature resulted in the smooth and rapid etherification of the major component (−)-13a, confirming its stereochemistry and providing C17-selenated platensimycin cage structure (+)-69, which underwent AIBN-induced radical reduction to give deselenated compound (−)-12. We were encouraged by the ease with which the proposed etherification reaction occurred, and subsequently discovered that exposure of the mixture of (−)-13a and (+)-13b to either TFA (excess) at 0 °C, or triflic acid (1.5 eq.) at −78 °C was sufficient to induce cyclization, forming cage structure (−)-12 in excellent yield (based on the proportion of (−)-13a in the mixture of diastereomers, Scheme 10). This finding made our two-step selenation-based route to (−)-12 unnecessary. Interestingly, alcohol (+)-13b was recovered unchanged from the reactions with PhSeCl and TFA, while upon the treatment with triflic acid its exocyclic double bond partially migrated to the endocyclic position (structure not shown). A sample of (−)-12 crystallized from CDCl3/CH2Cl2 to give colorless needles (m.p. 65–67 °C) suitable for X-ray crystallographic analysis.41 The X-ray derived ORTEP drawing of (−)-12 is shown in Figure 4.
Endocyclic aldehyde (+)-39 was also investigated as a substrate for the samarium(II)-mediated cyclization/acid induced-etherification sequence. The successful cyclization of this substrate was essential to our early asymmetric synthesis efforts, to ensure that the cycloisomerization of ester-capped alkyne 32 would lead to a total synthesis of platensimycin [(−)-1]. As shown in Scheme 11, aldehyde (+)-39 proved to be a viable substrate for the ketyl olefin cyclization reaction, leading to alcohol 70 as a single diastereoisomer (1H NMR spectroscopic analysis). Treatment of 70 with TFA resulted in clean conversion (36% yield for two steps) to cage-structure (−)-12, confirming the stereochemistry of 70 and completing the formal asymmetric total synthesis of platensimycin [(−)-1]. The dependency of the stereochemical outcome of the samarium(II)-mediated cyclizations on subtle differences in substrate structure is intriguing, and points to the sensitivity of the reaction to the precise structure and conformation of the spirocyclic substrate.
The configuration at C10 of cyanohydrin cyclization product (−)-67 is correct with respect to the orientation of the masked hydroxyl group, and we sought to exploit this in an alternative route to the platensimycin cage structure [(−)-12]. Thus, hydrolysis of the ethoxyethyl ethyl (EE) group with aqueous acid followed by etherification of the resulting product with TFA (0 °C) gave C10-cyano cage compound (+)-71 in high yield (81% for two steps) as shown in Scheme 12. Attempts to cleave the undesired carbon-carbon bond by treatment of the lithium anion of (+)-71 with LDBB42 failed to provide (−)-12 (Scheme 12). Although the cyanohydrin addition route did not provide an efficient means to access the cage structure of platensimycin, it did provide access to the bioactive analogue carbaplatensimycin,13b,c and may potentially lead to other analogues with chemical modification on the cage such as C10-cyano platensimycin.
With a viable route to the core structure (12) of platensimycin in hand, we turned our attention to the installation of the side chain and completion of the total synthesis. Enone 12 could be methylated in excellent yield by exposure to KHMDS followed by addition of MeI in a THF-HMPA mixture (Scheme 13). The methyl group was introduced with almost complete stereoselectivity, giving 72 as essentially a single diastereoisomer, as judged by 1H NMR spectroscopic analysis. A second alkylation was then investigated using various electrophiles. As might be expected, the enolate of 72 was found to be rather unreactive, and all attempts to install a side chain using 3-oxygenated propyl iodides (e.g. 73 or 74) failed. Treatment of the potassium enolate of 72 with bromide 75 rapidly transferred the ester group to the enolate oxygen of the substrate to form the corresponding enol carbonate. However, this enolate did react with allyl bromide to provide the desired allylated compound 76 along with significant quantities of the O-allylated product. Recourse to allyl iodide overcame this complication, giving 76 in good yield (83%) and, again, essentially as a single diastereoisomer, verifying our retrosynthetic hypothesis. The conversion of 76 into platensic acid (2) was also problematic at first. The allyl side chain was not reactive toward bulky hydroboration reagents; treatment with BH3•THF led to hydroboration of the allyl group with concomitant reduction of the enone carbonyl moiety, allowing isolation of diol 77 in low yield, along with several by-products. Oxidation of 77 with DMP then furnished aldehyde 78 in low overall yield from 76 (steps a and b were carried out with both racemic and enantiopure materials; steps c and d were carried out only with racemic compounds).
Fortunately, the allyl side chain of 76 was amenable to selective cross metathesis, and this provided a mild method for the required oxidation (Scheme 14). Thus, cross metathesis of 76 with vinyl pinacol boronate (79)43 using Grubbs’ second generation initiator 8044 (25 mol %) gave olefinic boronate 82 in good yield (81%) as an inconsequential mixture of isomers (E:Z ca. 6:1). The use of the Hoveyda–Grubbs second generation initiator 8145 gave similar results (81% yield) with much lower catalyst loading (10 mol %). Oxidation of the olefinic boronate 82 was accomplished by heating with trimethylamine-N-oxide,46 to give aldehyde 78 in much higher overall yield than the hydroboration route described above. This two-step conversion of allyl groups to aldehydes was first employed by Danishefsky et al. in connection with their work on epothilone analogues,47 and has subsequently proved applicable to a range of platensimycin analogues.13a–c Further oxidation of 78 gave platensic acid [(±)-2 and (−)-2] in high yield,48 treatment of which with TMS-diazomethane gave methyl platensinoate [(±)-3 and (−)-3] in quantitative yield as shown in Scheme 14. The latter compound is also naturally occurring.10a
Recently, Corey and Yeung reported the synthesis of a platensimycin analogue in which the propionate side chain was installed using the more direct method of 1,4-addition to methyl acrylate.13d Although we had investigated such a reaction initially with little success, these recent results prompted us to reinvestigate this transformation. Under Corey’s conditions (methyl acrylate, t-BuOK, THF/t-BuOH) we found that methylated cage intermediate (−)-72 was transformed into methyl platensinoate (3) directly; however, the yield of this transformation was modest and the desired compound was isolated, along with several by-products. This situation was improved by switching to t-butyl acrylate (Scheme 15), which allowed direct formation of t-butyl platensinoate (−)-83 in high yield (84%). t-Butyl ester (−)-83 could be cleaved quantitatively by treatment with TFA to give platensic acid [(−)-2]. This is now the preferred method for the conversion of (−)-72 to (−)-2.
For the completion of platensimycin (1) from platensic acid (2), we required a suitably substituted aniline coupling partner. We chose to prepare aniline derivative 84 in the first instance, in which the carboxylic acid is protected as the methyl ester and the phenols are masked as MOM ethers. The choice of protecting groups was influenced by their ease of removal and by synthetic considerations. We prepared 84 from 2-nitroresorcinol (85, Scheme 16), a convenient and inexpensive starting material that incorporates three of the necessary substituents. MOM ether formation (MOMCl, i-Pr2NEt) followed by hydrogenation of the nitro group (Pd/C, H2) gave aniline derivative 86 in excellent yield, which later served as a coupling partner for the synthesis of platensimycin B3 [(−)-9]. Boc protection of 86 (Boc2O, neat) gave the corresponding carbamate in quantitative yield. Generation of the dilithio species by treatment with 2.2 equiv of n-BuLi, followed by addition of Mander’s reagent [MeOC(O)CN], did give the desired product (87), however, yields were low due in part to incomplete ortho-lithiation even at 5 °C, resulting in the generation of N-acylated by-products. This problem was overcome by in situ protection of the carbamate anion with one equivalent of TMSCl, followed by addition of a second equivalent of n-BuLi to effect ortho-lithiation. Addition of Mander’s reagent followed by mildly acidic workup then gave the desired ester 87 in good yield. Removal of the Boc protecting group was also problematic, with even 0.1% TFA in CH2Cl2 causing rapid decomposition of 87. Fortunately, thermolysis of the Boc group was clean; heating 87 in 1,2-dichlorobenzene at 205 °C for 7 min gave aniline 84 in 83% yield. Shortly after us, Giannis et al. reported another efficient synthesis of 84, employing ester 88 as a key intermediate.49 Taking advantage of 88, we readily prepared aniline 89, a coupling partner required for the platensimycin B1 synthesis, through ammonolysis with aqueous ammonium hydroxide, followed by hydrogenation of the nitro group as described above (see Scheme 16).
The coupling of platensic acid (2) with aniline derivative 84 was carried out by treatment with HATU and Et3N in DMF, providing amide 90 in 86% yield (Scheme 17). Deprotection of 90 was achieved through a one-pot procedure involving methyl ester hydrolysis (aq. LiOH in THF, 45 °C) followed by MOM cleavage (addition of aq. HCl and continued heating at 45 °C), providing platensimycin (1) in high overall yield. Both racemic and enantioenriched platensimycins were prepared starting with (±)-14 and enantioenriched (−)-14. Enantioenriched platensimycin [(−)-1] was identical in every respect to a reference sample of the natural material.6
Furthermore, platensimycins B1 [(−)-7] and B3 [(−)-9] were efficiently synthesized in a similar manner from platensic acid [(−)-2] and anilines 89 and 86, respectively, as shown in Scheme 18. Notably, aniline 89 with two free hydroxyl groups showed good reactivity and stability under these amide forming coupling conditions, ensuring that platensimycin B1 [(−)-7] could be formed directly in one step. However, the protecting group-free version of aniline 86 was not stable enough for handling, presumably due to the lack of an electron-withdrawing functional group on the aromatic ring. A two-step sequence was, therefore, necessary to synthesize platensimycin B3 [(−)-9] as shown in Scheme 18. Thus, coupling of bis-MOM protected aniline 86 with platensic acid [(−)-2] under the HATU-Et3N conditions, followed by removal of the MOM groups from the resulting product under acidic conditions, led to platensimycin B3 [(−)-9] in 85% overall yield. The physical properties of both synthetic platensimycins B1 [(−)-7] and B3 [(−)-9] matched those reported in the literature for the naturally derived materials.10c
Naturally occurring platensimide A [(−)-4] consists of platensic acid [(−)-2] coupled to (S)-N4-acetyl-2,4-diaminobutyrate in place of the aniline component of platensimycin. Singh et al. accomplished a semisynthesis of platensimide A [(−)-4] from naturally occurring platensic acid [(−)-2] as part of their isolation studies.10a They coupled platensic acid with mono-Boc-protected methyl diaminobutyrate, and subsequently installed the acetamide group. Within the context of a total synthesis (i.e. using synthetic platensic acid), we hoped to complete a more convergent route in which the acetamide was already in place. To this end, we prepared amine (+)-91 from Cbz-glutamine t-butyl ester 92 (Scheme 19). Hoffmann degradation of 92 was carried out according to the literature procedure [PhI(O2CCF3)2],50 however, rather than isolating the amine product, the latter was converted in situ to its acetamide derivative (−)-93 with Ac2O (70% overall yield), which served the dual role of saving one step and facilitating isolation. Hydrogenolysis of the Cbz group from (−)-93 then gave amine (+)-91 in 92% yield (Scheme 19). Coupling of (+)-91 with the activated ester generated from platensic acid [(−)-2] and HOBt under the conditions of EDC•HCl furnished platensimide A t-butyl ester (−)-94 in 94% yield. Hydrolysis of the latter with TFA then yielded synthetic platensimide A [(−)-4] in high yield (99%), whose physical properties were consistent with those of the naturally occurring product.10a
A similar strategy was employed for the total synthesis of the also naturally occurring homoplatensimide A [(−)-5, Scheme 20]. In this case, extension of the side chain of the platensimycin core was required. This was achieved through the highly selective Wittig reaction between stabilized ylide 95 and core aldehyde (−)-78, which gave unsaturated ester (−)-96 in excellent yield (88%) as a single geometrical isomer (Scheme 20). Hydrolysis to the corresponding carboxylic acid (97, 94% yield) and coupling of the crude product with glutamine t-butyl ester [(+)-98, obtained by hydrogenolysis of its Cbz derivative 92] under the HOBt/EDC•HCl conditions gave homoplatensimide A t-butyl ester [(−)-99] in 82% yield. Ester hydrolysis of the latter compound with TFA was again successful in revealing homoplatensimide A [(−)-5], which displayed physical properties in accord with those reported for the natural material.10b Homoplatensimide A methyl ester [(−)-6] was also prepared and fully characterized by treatment of (−)-5 with TMSCHN2 (98% yield) as shown in Scheme 20. Its physical data matched those reported for the natural product.10b
In conclusion, we developed several effective synthetic strategies for the total synthesis of platensimycin (1), a newly discovered antibiotic possessing a novel mechanism of action. In addition to a racemic synthesis, two complementary asymmetric approaches to a key synthetic intermediate [i.e. (−)-14] were devised. The first strategy was based on well known asymmetric alkylation technologies and a novel dearomatizing cyclization reaction of a phenolic allylsilane. The reaction was developed as a viable and general method for the oxidative cyclization of phenolic allyl silanes to spiro hemiquinones. The second strategy featured a rhodium-catalyzed asymmetric cycloisomerization reaction especially developed for this total synthesis, which, however, turned out to be of general applicability and scope, accommodating a wide range of terminal enyne substrates. The developed convergent total synthesis of platensimycin proceeded through enone 9, representing the cage core of the molecule, and which has been employed by all other syntheses reported to date12a–l of this natural product. Efficient strategies to install the side chain were also developed to ensure reliable syntheses of platensimycin and related compounds. In addition to providing access to synthetic platensimycin, this work has facilitated the total syntheses of a number of designed platensimycin analogues13a–c and its naturally occurring congeners, platensimide A [(−)-4], homoplatensimide A [(−)-5], homoplatensimide A methyl ester [(−)-6], and platensimycins B1 [(−)-7] and B3 [(−)-9]. The latter accomplishments represent the first total syntheses of these natural products. The reported synthetic strategies and methods are expected to facilitate the construction of further designed analogues of platensimycin and to find further applications in chemical synthesis.
We thank Drs. D. H. Huang, R. Chadha, and G. Siuzdak for NMR spectroscopic, X-ray crystallographic, and mass spectrometric assistance, respectively. We also gratefully acknowledge A. Nold and Dr. H. Zhang for assistance with chiral HPLC, and Profs. X. Zhang, A. Lei, D. Crich, Drs. P. G. Bulger, L. A. McAllister, and J. S. Chen for helpful discussions. We are especially grateful to Dr. S. B. Singh for providing an authentic sample of natural platensimycin and NMR spectra of homoplatensimide A, and verification of the optical rotation of platensimide A. Financial support was provided by grants from the National Institutes of Health (USA), National Science Foundation (CHE-0603217) and Skaggs Institute for Research, graduate fellowships from Bristol-Myers Squibb and Eli Lilly (both to A.L.) and Novartis (to S.P.E.), and a postdoctoral fellowship from Merck Sharp & Dohme (to D.J.E).