Formal synthesis of (+)-lactacystin from l-serine

A formal, stereocontrolled synthesis of lactacystin has been completed from t-Bu-O-l-serine, providing the key intermediate 13, also useful for the generation of a range of C-9 analogues.


Introduction
The 20S proteasome is a large barrel-shaped protein comprised of 28 subunits. 1 The primary function of the proteasome involves the degradation of damaged proteins, a vital component of the ubiquitin proteasome pathway. Inhibition of the proteasome can lead to cell death. This property of the proteasome has made it a promising target for cancer therapeutics. 2 Microbial metabolites have provided a wealth of proteasome inhibitors (Fig. 1). Lactacystin 1 was discovered in 1991 bȳ Omura through extraction from the cultured broth of Streptomyces sp OM-6519, 3,4 aer observations that it induced differentiation of the mouse neuroblastoma cell line, a consequence of proteasome inhibition. Further studies, driven by several early efforts to prepare lactacystin and analogues, found that lactacystin undergoes cyclization to the b-lactone omuralide 2, which inhibits the proteasome 5 and can induce apoptosis. The potential for beta-lactone, gamma-lactam proteasome inhibitors was further highlighted by the discovery of the salinosporamides, e.g. 3, 6-8 and cinnabaramides, e.g. 4. 9 Similar lactam cores have also been discovered in the metabolite oxazolomycin 10 5, which possesses antibiotic activity. A number of strategies to access these cores to produce natural products and analogues of high therapeutic value have been reported. 11

Results & discussion
Our synthetic work has focused on the development of a functionalized lactam core from glycine, 12,13 and recently using L-leucine 6 as a starting material to produce a formal synthesis of the C9-deoxyomuralide analogue, using the natural chirality of the amino acid to direct the synthesis. 14 Formation of the carbon skeleton of omuralide was achieved in 4 steps (Scheme 1). Peptide coupling of PMB-protected leucine 7 to the malonic acid benzyl ester 8 provided the precursor 9 to the key Dieckmann cyclization/alkylation step. Cyclization was induced with TBAF, and subsequent addition of methyl iodide provided two diastereoisomers. The major, 10b, was isolated and used in an acylation using Mander's reagent to give 11, thus completing the carbon skeleton. Six further steps produced pyroglutamate 12, which can be cyclized to give 9-deoxyomuralide in one step.
With this methodology in place, we turned our attention to a serine-derived route to lactacystin. This would provide a hydroxy group in the C9 position, which previous SAR studies have shown to be key for effective proteasome inhibition.
We envisaged that intermediate 13 could be synthesized from a suitably protected tetramic acid-like core 14 (Scheme 2). A removable benzyl ester would be used to help direct acylation of lactam 15 using Mander's reagent; 15 could in turn be formed using the cyclization/alkylation procedure previously developed. The Dieckmann cyclization precursor 16 could be synthesized from peptide coupling of a suitably protected O-t-Bu-L-serine 17. This starting material was used due to the size of the t-butyl group, its tolerance towards a wide range of conditions, and because both enantiomers are commercially available.
PMB protection of the serine derivative 17 was carried out using a modication of a procedure by Vázquez using the PMB sulphite adduct 18. 15 We have previously observed 14 epimerization during imine formation, and so a one-pot procedure was developed. Aer work up, the protected serine 19 could be used without further purication. Peptide coupling to benzyl malonic ester 8 provided the Dieckmann cyclization precursor in 49% yield over the two steps. Analysis using chiral stationary phase HPLC showed that only negligible epimerization had occurred at this point, providing 16 in 97% ee (Scheme 3).
With the Dieckmann precursor 16 in hand, we turned our attention to the tandem cyclization/alkylation step (Scheme 4).
When subjected to the conditions used previously (TBAF in THF, then MeI addition), the cyclization/alkylation proceeded with good yield and provided a 10 : 1 ratio of diastereoisomers 15a and 15b, in favour of 15a, where the newly added methyl group is situated on the same face of the molecule as the t-BuO group.
Our previous work on the leucine analogue 14 showed that addition of the methyl iodide is in that case preferentially introduced opposite the isobutyl moiety (Scheme 5). This outcome was expected due to the bulky nature of the amino acid group and the assumed planarity of the intermediate. The diastereoisomers were isolated in a 1 : 2 mixture and the observation was conrmed by single crystal X-ray analysis of 20, derived from the minor diastereoisomer 10a by PMB removal.
We were therefore surprised to nd that our serine derivative afforded the diastereoisomers in a 10 : 1 mixture favouring the diastereoisomer 15a with the methyl group cis to the t-BuO group, presumably because in forming 15b the benzyl ester unit is forced nearer to the t-BuO, so raising the transition state energy of that pathway.
On analysis by chiral HPLC the major diastereoisomer was found to have a disappointing ee of 44%. Running the reaction at decreased temperatures provided higher ee at the expense of yield. Further optimization (Table 1), including increased reaction times, lower temperatures and the separation of the cyclization and alkylation steps, achieved ees of up to 79% and yields of 66%. The ee could be improved further through recrystallization from isopropanol, which produced 97% ee. We suspect that the decrease in ee is primarily the result of epimerization at the C5 position in the mixture of diastereoisomers 15a/b following alkylation, leading to the (presumed) more stable diastereoisomer 15a. Lower temperatures reduce the degree of this epimerization, leading to the decreased diastereoselectivities (3 : 1) but higher ees.
During our work with the L-leucine derived analogue 9, we were able to isolate a 1 : 2 ratio of diastereoisomers 10a and 10b in 9% ee and 79% ee respectively (Scheme 5). 14 Our results with leucine strongly indicate that the partial racemization occurs aer the alkylation. If racemization occurred solely before alkylation, the ratio of diastereoisomers in the racemic material should match the ratio of diastereoisomers in the enantiopure material. As we observed the diastereoisomers in a 1 : 2 ratio at 9% and 79% ee respectively, the vast majority of racemization must occur aer alkylation. Due to the structural similarity between our L-leucine-derived lactams and our serine-derived lactams, it seems likely that the mode of racemization is analogous.
To investigate the stereoselectivity further, the benzyl ester was replaced by a methyl counterpart (Scheme 6). Coupling of 19 to the half malonic methyl ester potassium salt in an analogous procedure to that of 16 proceeded in good yield, to produce 21. Once treated with our cyclization/alkylation procedure, NMR analysis showed that the cyclization occurred efficiently, but we found that 22a/b decomposed if le in contact with silica gel for extended periods of time. Partial purication was therefore completed with a silica plug to produce a 3 : 1 mixture of diastereoisomers; the mixture was treated with CAN removing the PMB group (Scheme 6). The puried major isomer 23a was obtained as a colourless crystalline solid. Analysis by single crystal X-ray diffraction showed that the methyl group was still preferentially added to the same face as the tert-butoxy group, allowing us to conclude that the benzyl ester was not the primary inuence on diastereoselectivity in our system.
The system was further investigated by replacing the PMB group with a phenyl (Scheme 6). Chan-Lam coupling of the serine derivative with phenyl boronic acid and subsequent peptide coupling to the benzyl malonic half ester 8 produced the Dieckmann cyclization precursor 24. Compound 24 was subjected to our cyclization conditions. The methylated product 25 was produced in a 5 : 1 ratio of diastereoisomers according to the 1 H NMR spectrum, albeit in a low yield, perhaps due to the conformation required for cyclization being more unfavourable. A crystal of the major diastereoisomer suitable for X-ray analysis was obtained. Once again, in the major product the methyl group had been introduced to the same face as the tert-butoxy.
With 15a in hand, we sought to install the methyl ester moiety that would eventually form the beta lactone found in omuralide. Acylation using Mander's reagent at À40 C was found to produce compound 26 in good yield with no observed O-acetylation. In addition, only one diastereoisomer could be observed (Scheme 7). A small drop in ee was observed during the acylation, to 69%, perhaps due to the presence of small quantities of the minor diastereomer 15b. Using recrystallized 15a, the ee of 26 was 84%. Single crystal X-ray analysis of 26 was carried out on both the racemic and enantiomerically pure (obtained by recrystallization) forms, the latter conrming that the absolute stereochemistry of our C9 centre is as found in (+)-lactacystin.
Removal of the benzyl ester from 26 by hydrogenolysis led to tetramic acid derivative 14 as a mixture of diastereoisomers. The mixture proved unstable towards silica gel and so was used in the next step without purication. On treatment with Smethyl p-toluenethiosulfonate 27, in a similar manner to Pattenden, 16,17 a 4 : 1 mixture of inseparable diastereoisomers 28a/ b was formed. NOESY experiments conrmed that the major observed diastereoisomer was the desired one, with the methyl thioether trans to the t-BuO group (Scheme 8).
Removal of the t-butyl group from 28a would allow completion of the formal synthesis to form the Corey intermediate 29a. 18 Treatment of 28a/b with a 1 : 1 TFA/DCM mixture led to an inseparable mixture of the diastereoisomers 29a/b in a 2 : 1 mixture aer purication by column chromatography. To our surprise, analysis by chiral HPLC showed a drop in ee from 84% to 58% and 41% respectively for 29a and 29b (Scheme 9). A mechanism for this epimerization did not seem obvious, and a simpler substrate was synthesized to probe the reaction further. Use of methyl iodide as alkylating agent in place of 27 added a second methyl group to the C7 position, giving 30 (Scheme 10). This gem-methylated analogue of our substrate was chosen for two reasons, to eliminate a stereogenic centre, which could conrm C5 as the epimerizing centre, and to reduce the number of potentially reactive functional groups. Removal of the t-butyl group with TFA under the conditions used for 28a/b gave 31 in racemic form. Further analysis revealed that the stereocentre remained unchanged upon treatment of 30 with TFA. The racemization occurred during preparation for the purication by column chromatography, where crude 29a/ b was dissolved in DCM and adsorbed onto silica gel, perhaps a result of a retroaldol process. We were pleased to nd that changing to a wet loading method, where crude 31 was loaded onto the chromatography column in the eluting solvent using petroleum ether/ethyl acetate, resulted in complete preservation of the ee.
Returning to our original system, the 4 : 1 mixture of 28a/ b was treated with TFA/DCM to produce 29a/b. Analysis by 1 H NMR spectroscopy showed retention of the 4 : 1 ratio in the crude mixture. Silica gel was added, and the NMR spectrum obtained again, now showing a ratio of 3 : 1.3. Aer leaving 29a/b in contact with the silica gel for 16 h, the 1 H NMR spectroscopy was repeated, the spectrum this time showing a 3 : 2 ratio. Despite largely preventing the epimerization, we were unable to separate the diastereoisomers 29a/b efficiently. Because of this problem we changed the order of deprotection and reduction, aiming to produce the later Sodium borohydride reduction of the inseparable 4 : 1 diastereoisomeric mixture 28a/b led to two separable isomers 32a and 32b in a 4 : 1 ratio. Analysis of the products by HPLC using a chiral stationary phase indicated ees of 67% and 66% for 32a and 32b respectively. The desired rst eluting diastereoisomer 32a was recrystallized from IPA, to our delight providing 32a with 99% ee. Treatment of the ltrate with TFA/ DCM resulted in the desired lactam 13 without further epimerization, thus completing our formal synthesis of (+)-lactacystin in six steps from 16.