Cascade Metathesis Reactions for the Synthesis of Taxane and Isotaxane Derivatives

Abstract Tricyclic isotaxane and taxane derivatives have been synthesized by a very efficient cascade ring‐closing dienyne metathesis (RCDEYM) reaction, which formed the A and B rings in one operation. When the alkyne is present at C13 (with no neighboring gem‐dimethyl group), the RCEDYM reaction leads to 14,15‐isotaxanes 16 a,b and 18 b with the gem‐dimethyl group on the A ring. If the alkyne is at the C11 position (and thus flanked by a gem‐dimethyl group), RCEDYM reaction only proceeds in the presence of a trisubstituted olefin at C13, which disfavors the competing diene ring‐closing metathesis reaction, to give the tricyclic core of Taxol 44.


Introduction
Ta xol (paclitaxel), together with its derivatives Taxotere (docetaxel) and Jevtana (cabazitaxel) are the largest sellinganticancer drugs of all time, with sales of over three billion USD per year for Taxotere alone in 2010. [1] Originally indicated for the treatment of ovariana nd breast cancers, they are now widely prescribed to treat ab road range of malignancies. [2] The structureso ft hese three compounds only differ in terms of the functionalization of the amineo nt he side chain and the hydroxyl groups at C10 and C7 (Scheme1). Ta xol is currently being manufactured through plant-cell fermentation by Phyton Biotech,L LC, aD FB Pharmaceuticals Company for Bristol-MyersS quibb, whilstT axoterea nd Jevtana are produced by semisynthesis from 10-deacetylbaccatin III by Sanofi, which still requires an expensive extraction process of natural resources. There have been six total syntheses of Ta xol by the groups of Holton, [3] Nicolaou, [4] Danishefsky, [5] Wender, [6] Mukaiyama [7] and Kuwajima, [8] as well as three formal syntheses by the groups of Ta kahashi, [9] Nakada [10] as well as Sato and Chida, [11] but they all comprise at least 37 steps. [12] An efficient synthesis of active taxoid analoguesh as yet to be achieved, because of the sterically hindered, complex and highly functionalized structure of these compounds.
Ar apid synthesis of the tricyclic core of Ta xol where all of the functional groupsr equiredf or activity are present or in al atent form would facilitatea ccess to ad iverse array of novel taxoids with potentiala nticancer activity.W ereport here as ynthetics trategy featuring ac ascade ring-closing dienyne metathesis( RCDEYM) reactionthat allows access to the ABC tricyclic ring system of taxanes as well as taxane analoguest hat possess an ovel skeleton and cannot be prepared by semi-synthesis. [13]

Results and Discussion
Our initial retrosynthesis is outlinedi nS cheme 1. We aimed for af ormal synthesis of Taxol, so we chose the intermediate 4 described by Holton duringh is synthesis of this natural product as our primary target. [3] The Ar ing would be closed by ap inacol coupling between the ketones at C11a nd C12 in compound 5,a sp reviously described by Mukaiyama on as imilar substrate. [7] The ketone at C12 would be installed by hydration of alkyne 6.T he eight-membered Br ing would be formed by ar ing-closing metathesis (RCM) reaction [14] between the alkenes at C10a nd C11i nc ompound 7.T his key step was successfuli no ur synthesis of model BC bicyclic systems of Taxol (with no hydroxyl group at C7 and ab utyl side chain at C1). [15] Finally,t he metathesis precursor 7 would be assembled by aS hapiror eaction between hydrazone 8 and aldehyde 9.T his coupling reactionh as proved to be very diastereoselective on similar substrates duringo ur previous approaches to taxoids. [16] Our synthesis commenced with the preparation of aldehyde 9 (Scheme 2). Commercially available 3-pentyn-1-ol was oxidized with the Dess-Martin periodinane [17] (DMP) and the resulting aldehydew as subjected to aB arbier reaction with prenyl bromide under the Luche conditions [18] to furnish alcohol 10 in excellent yield. Oxidation of alcohol 10 gave the corresponding ketone 11,w hich was submitted to trimethylsilyl cyanidei nt he presence of at he tertiarya mine 1,4-diazabicyclo[2.2.2]octane( DABCO) as ac atalyst. The resulting cyanohydrin was reduced to the intermediate imine, which was hydrolyzed to give the racemic aldehyde( AE)-9 by exposure to silica gel. Optically active aldehyde 9 was also prepared in 99 % ee in as imilar fashion [19] using ac hiral amine base for the cyanation reaction, [20] but we chose to pursue the synthesis of the metathesis precursors with the racemic aldehydet ow iden the array of taxoidsg enerated,a nd to study the influence of the stereochemistry of the precursor on the RCM reaction outcome.
In order to test the key metathesis reaction, we decided to use a7 -deoxy Cr ing as ac oupling partner in the Shapiro reaction. It is worth noting that removal of the functional group at C7 in Taxol did not result in as ignificant loss of bioactivity. [21] When hydrazone 12 (Scheme 3), prepared in 76 %y ield from the corresponding knownk etone, [22] was submitted to tBuLi for the Shapiro coupling, only degradation was observed. [23] We surmised that this was due to the deprotonation at the allylic position, and thus the alkene was maskeda saprotected primary alcohol. Enantiopureh ydrazone 13 [15b] was treated with aldehyde (AE)-9 using conditions we had developed previously. [15b] To our surprise, the reactiono nly proceeded in 20 % yield. Several additives were screened. Addition of MgBr 2 and ZnCl 2 did not lead to any of the desired product, but we observedadramatic increasei ny ield when dry CeCl 3 was stirred for 30 min with the vinyllithium reagent derived from hydrazone 13 before addition of aldehyde( AE)-9,a nd diols 14 a,b were obtained in 85 %c ombined yield after hydrolysiso ft he TMS ether.T he reason for this differencei nr eactivity between the model aldehyde (butyl side chain at C1) and (AE)-9(2-butynyl side chain at C1) is unclear. [15b] As had been observedp reviously for the model aldehydes, this reaction wash ighly diastereoselective, giving the trans diol compounds [24] 14 a and 14 b after hydrolysis of the trimethylsilyl ether.T he stereochemistry of 14 a and 14 b was assigned by comparing their proton NMR spectra with those of the corresponding model Shapiro adducts possessing ab utyl side chain at C1. [25] Diols 14 a and 14 b weret hen submitted separately to trityle ther hydrolysis, elimination of the resulting primary alcohol using the Grieco protocol [26] and protection of the C1-C2 diol as the cyclic carbonatee ster to furnish the metathesis precursors 15 a and 15 b in 75 %a nd 65 %o verall yields for the four steps, respectively.N oi ntermediate purification was required for these transformations.
We first tried out the key RCM reactiono nc arbonate 15 a, which possesses the opposite configuration at C1 and C2 compared to Taxol. Treatmento ft his compound with 10 mol %o f the second-generation Grubbs precatalyst in toluene at reflux for 24 hd id not lead to the desired cyclooctene, but gave tricyclic derivative 16 a instead( Scheme 4). This product resulted from an enyne metathesis reaction between the alkene at C10 and the alkyne at C13, furnishing the intermediate bicycle www.chemeurj.org 16 a',w hich further cyclized by ad iene metathesis to give 16 a in good yield. Thisi ntermediate 16 a' could be isolated as a1:1 mixture with 16 a if only 5mol %o ft he precatalyst was used for the reaction. The first enyne metathesis reactionw as not unexpected; [27] what was more surprising to us was the ease of formation of the strained tricyclic system in compound 16 a. This 14,15-isotaxane has an unprecedented skeleton, which is very similart ot hat of taxaned erivatives, except that the C14 and C15 carbonsh ave swapped positions, which places the gem-dimethyl group in the A-ring alone. In addition, the C2 stereogenic centerp ossesses the undesired configuration for Ta xol.
In an effort to assess the influence of the nature of the diol protecting group on the outcomeo ft he metathesis reaction, which wass hown to be crucial for model compounds, [15b] the benzoate 17 a was prepared by addition of phenyllithium to the carbonate 15 a (Scheme 4). Unfortunately,b enzoate 17 a did not undergo metathesis when treated with the Grubbs 2 precatalyst, but slowly decomposed.
As and C2 on the outcomeo ft he metathesis reactions. [15b] These isotaxanes possess the undesired configurationa tC 1. Isotaxane 16 b was crystalline, and its X-ray crystallographic analysis [28] (Figure 1) established its tricyclic structure andc onfirmed the configuration of the carbonate-bearing stereocentersa tC 1 and C2.
The isotaxanes 16 a, 16 b and 18 b represent an ovel class of Ta xol analogues, andc ould be transformed into potentially active compounds. Indeed, taxanes such as tasumatrols E, F and G ( Figure 2), isolated from Taxus sumatrana,d on ot possess the classical ABC 6,8,6-tricyclic system of Taxol;h owever, they exhibit more potent activity than Ta xol in vitro against four human cancercell lines. [29] An easy way to circumventt he unwantedd ienynem etathesis cascade is to perform the alkyne hydration before the RCM step, andt his has been achieved in excellent yield (Scheme 5). Diol 19 was prepared in three steps from the Shapiro adduct 14 b in 68 %o verall yield. Treatment of alkyne 19 with the Gagosz catalyst [30] in the presence of water did not give the corresponding ketone but hemiketal 20. [31] Fortunately,c ompound 20 underwent ring-closing metathesis in 98 %y ield to form the BC ring system of Taxol 21.W ork is in progress for the completion of the synthesis of the tricyclic core of Ta xol according to the retrosynthesis shown in Scheme 1.
On the other hand, we also wanted to take advantage of this very efficient metathesis cascadet os ynthesize the ABCt ricycle of Taxol, and our revisedr etrosynthesis is shown in Scheme 6. The 2-ene-1,4-diol unit of compound 4 would be installed by aT i III radical-mediated opening of the corresponding 1,3-diepoxide, which can be generated from the 1,3-diene moietya tC 10-C13 of compound 22.   www.chemeurj.org dation sequence of the C3-C4 olefin would lead to the ketone at C4. [5] Tricycle 22 would be formed by am etathesis cascade reactionf rom dienyne 23,w here the alkyne at C11a nd the alkene at C13 have swappedp ositions compared to those in compound 15 b.I no rder to direct the metathesis cascader eaction so it starts with the olefin at C10 and not with the one at C13, we elected to have ad isubstituted olefin at C13, which would react more slowly with the metathesis precatalysts. It is important to note that this extra methyl group will not be present in the metathesis product 22,b ut will be part of the propene released after the diene metathesis reaction. Disconnectiono fd ienyne 23 through the C2ÀC3 bond reveals the two precursors aldehyde 24 and hydrazone 25.
The synthesis of aldehyde 24 in its racemic form was not as straightforwarda st he synthesis of the corresponding aldehyde 9.I ts tarted with ester 26, [32] obtained by propargylationo f ethyl isobutyrate (Scheme 7). Attempts to isomerize the terminal alkyne of 26 into the internal one with potassium tert-butoxide only resultedi nd egradation products.F ortunately, this isomerization reaction was successful on the corresponding acid 27,a nd acid 28 was obtained in 94 %y ield. Addition of crotylmagnesium chloride to the corresponding Weinreb amide (compound 41,s ee Scheme 9f or structure) furnished acomplexmixture of products, so we next turned to the crotylation of aldehyde 29.T reatment of this aldehyde with crotyl magnesium chloride in the presence of aluminumt richloride [33] led to a1 :1.5 mixture of a and g crotylation products. Fortunately,a llyl transfer from 2,3-dimethyl-4-penten-2-ol catalyzed by tin(II) triflate [34] gave the desired alcohol 30 (as an inconse-quential3:1 mixture of E/Z isomers) in 76 %y ield after 2d ays. Oxidation of 30 with 2-iodoxybenzoic acid (IBX) followed by homologation of the resultingk etone 31 furnished aldehyde (AE)-24 [35] in good overall yield.
The dienynes 32 a,b and 34 a,b were prepared using asimilar reactions equence to the one used for compounds 15 a,b and 17 a,b,a sd escribed in the preliminarya ccounto fo ur work. [13] Metathesis reactions of carbonates 32 a,b andb enzoates 34 a,b with Grubbs 2p recatalyst did not produce tricyclicc ompounds, but led to the bicycles 33 a,b and 35 a,b,r espectively, resultingf rom as imple diene RCM between the olefins at C10 and C13 (Scheme8). [13] Compound 33 a was crystalline, and its X-ray crystallographic analysis [36] (Figure 3) confirmedt he configurationa tC 1a nd C2 of the metathesis precursors 32 a and 34 a.
We had assumed that in the case of compounds 15 and 17 (Scheme 4), the initial enyne methathesis (between C10 and C13) would be favored compared to the alternative diene metathesis (between C10 and C11) because it would lead to am ore stable tricyclicp roduct after subsequent diene metathesis, but it seems that in all cases the first RCM takes place with the less hindered unsaturated functional group having no neighboring gem-dimethyl group. Since this gem-dimethyl group is part of the Taxol skeleton, it is not possible to relieve the steric hindrance at the propargylic positioni nc ompounds 32 and 34,b ut another option is to increase the steric hindrance of the alkene at C13, so the undesired diene RCM is disfavored. We thusembarked on the synthesis of metathesis precursors bearing at risubstituted olefin at C13. The synthesis of the aldehyde (AE)-36 required for their preparation is shown in Scheme 9. Prenyl transfer to aldehyde 29 from 2,3,3-trimethyl-4-penten-2-ol was unreliable, with yields of 37 rangingf rom 15 to 61 %. Ketone 38 wast hen obtained by IBX oxidation. An umpolung synthesis of 38 was also achieved. Prenylation of dithiane 39,p repared from aldehyde 29,f urnished 40 in excellent yield. Hydrolysis of the dithiane moiety gave ketone 38. However,t his route was not very convenient on large scale, so as al ast resort prenylation of the Weinreb amide 41 derived from acid 28 (Scheme7)w as also attempted. To our surprise, this reaction was very clean and afforded ketone 38 in 95 % yield. In this fashion,a ldehyde (AE)-36 [37] was obtained after homologation of 38 in 7s teps and 66 %o verall yield from ethyl isobutyrate.
Compounds 42 a,b and 43 a,b bearing at risubstituted olefin at C13 were synthesized in the same way as compounds 32 a,b and 34 a,b as previously described, then subjected to the Grubbs2precatalyst in toluene at reflux (Scheme 10). [13] We were disappointed to find out that carbonate 42 a and benzoate 43 a,p ossessing the undesired configurations at C1 and C2, furnished the bicyclic compounds 33 a and 35 a that we had already observed for the metathesis reactions of 32 a and 34 a (Scheme 8). Taxol-like benzoate 43 b also underwentd iene RCM to produce 35 b.H owever,T axol-like carbonate 42 b furnished compound 44 after RCDEYM, which corresponds to the tricyclicc ore of Taxol, along with the undesired bicyclic compound 33 b. [13] In order to confirmt he structure of the highly strained tricyclic product 44,w ec onverted it to the crystalline p-nitrobenzoate derivative 45 by hydrolysis of the carbonate and acylation of the resulting secondary alcohol (Scheme 11). X-ray crystallographic analysiso f45 [38] not only confirmed the tricyclic structure, but also established withouta mbiguityt he configurations at C1 and C2 for the Taxol-like series of compounds.
We then set to optimize the yield of the desired tricyclic compound 44.T he 44/33 b ratio was the same under different concentrations ranging from 310 À3 to 15 10 À3 m, [13] so all metathesis reactions were performed at 510 À3 m.T oluene at reflux provedt ob eab etter choice than 1,2-dichloroethane at reflux (80 8C) or xylene at reflux (140 8C). [13] Various precatalysts were then screened( Ta ble 1). No reactionw as observed with the less reactive Grubbs 1p recatalyst, so we tested secondgeneration ruthenium complexes.T he Hoveyda-Grubbs precatalyst HG2 gave an improved yield of the desired compound 44 compared to the Grubbs 2p recatalyst (69 vs. 45 %), and so did the Grela complex, which possesses an itro substituent on the benzylidene ligand. Pleasingly,t he HG2 derivativeZ han-1B, which possesses a N,N-dimethylsulfonamido group gave the  best yield (70 %) of compound 44.I ts eems that RCDYEM is favored with precatalysts possessing high initiation rates.
The different ratios observed with the Hoveyda-Grubbs-type precatalysts cannotb ee asily rationalized. Indeed, metathesis of substrate 42 b with any precatalyst will result in the same carbene (Scheme 12). This intermediate should then lead to the same ratios of 44 and 33 b after cyclization, releasing the same isopropylidene catalyst.T he only difference between the reactions is the ligand released after the first catalytic cycle, which could recombine with the isopropylidene catalyst to reform the precatalyst.T op robe the influence of the ligand, am etathesis experiment was run with 10 mol %o ft he Hoveyda-Grubbs2precatalyst and 300 mol %o ft he corresponding ligand,b ut the observed ratio of 44 and 33 b was very similar to the one observed withoutany added ligand.
Attempts to convert bicyclic product 33 b to the desired tricycle 44 by heating it in toluenea tr eflux in the presence of the Grubbs 2o rZ han-1B precatalyst were unsuccessful, even Scheme11. Synthesis and ORTEP(50 %probability) representation of p-nitrobenzoate 45.a)2Na q. NaOH, 1,4-dioxane, 80 %; b) p-nitrobenzoyl chloride, DMAP,Et 3 N, CH 2 Cl 2 ,80%. under microwave conditions. When 10 mol %o fp recatalyst was used, only 33 b was recovered, and with 100 mol %o fp recatalystd ecomposition occurred. Ring opening of 33 b in the presenceo fe thylene was not considered, because it would lead to ac arbene unsubstituted at C13, carbene' (Scheme 13), which would undergo diene metathesis preferentially.I na n effort to regenerate the carbene with at risubstituted olefin at C13, bicycle 33 b was submitted to the above conditions in the presenceo f2 -methyl-2-butene, the reagent whichG rubbs and co-workersh ave employed for the synthesis of trisubstituted olefins from their terminal homologuesb yc ross metathesis, [39] but to no avail (Scheme 13). These resultss eem to indicate that the formation of compound 33 b is not reversible, and that the metathesis reactions leading to 33 b and 44 are under kinetic control.

Conclusions
In summary,w eh ave synthesized Ta xol analogues with an unprecedented skeleton as well as the tricyclic core of Taxol in avery efficient fashion.The key step in these syntheses is acascade ring-closing dienyne metathesis (RCDEYM) reaction, leading to either 14,15-isotaxane tricyclic ring systems or the tricylic ring system of Ta xol in one operation from simple precursors, by judicious choice of the position of the alkyne (C13 for isotaxanes or C11f or taxanes). Furthermore, in the case of the taxane synthesis, we have shown that we can direct the course of the crucial metathesis reaction by adding at emporary methyl substituent to the olefin at C13, whichd oes not appear in the structure of the tricycle. Calculations rationalizing the different outcomeso ft he metathesis reactions of compounds 42 a,b and 43 a,b,w hich strongly dependo nt he stereochemistry and the protecting group of the diol at the C1 and C2 positions in the metathesis precursors, will be reported in due course.

Experimental Section
All experimental details can be found in the Supporting Information. The material includes compound characterization, crystal structures of 16 b and 33 a,a nd copies of spectra for all new compounds.