Chiral centres in molecules


Dr Sarah Houlton outlines some of the methods that companies are using today to introduce chiral centres into molecules

Dr Sarah Houlton outlines some of the methods that companies are using today to introduce chiral centres into molecules

A large proportion of the drugs now reaching the market contain at least one chiral centre. The proportion in the pharma companies’ development pipelines is even higher. There are, fundamentally, three methods of introducing chiral centres into molecules. These are: using fragments from the chiral pool of readily available enantiopure molecules; using some form asymmetric reaction, which makes only one isomer, or by the resolution of a racemate.

Taking advantage of nature’s ability to synthesise chirally pure molecules can be the simplest and most efficient method, if a suitable intermediate is available. One such example is AstraZeneca’s (AZ) scale-up synthesis of the farnesyl transferase inhibitor AZD3409, which has potential in the treatment of breast cancer and other tumours.1

The medicinal chemistry route used trityl mercaptan to introduce a chiral sulphur atom. This compound is not readily available in bulk and is an extremely ‘atom inefficient’ method of adding a single atom in a synthetic route. Instead, AZ started with a readily available Boc-protected hydroxyproline derivative, which is converted to Boc-protected bicyclic thiolactone.

Initial scale-up attempts had limited success, with synthesis above a 20g scale having much lower yields, and a competitive polymerisation reaction found to be taking place. Reducing the proportion of sodium sulphide used in the reaction solved this issue. The reaction was quenched with aqueous citric acid solution. The process has been performed at a tonne scale.

Reducing the bicyclic thiolactone to a thiolactol with DIBAL gave a masked aldehyde, which – through a reductive amination process – could be reacted with the second fragment of the molecule. This molecule also contains an amino acid-derived chiral centre. The adduct was formed without any loss of stereochemistry, and could then be ring-opened to give the required thiol derivative, just a couple of steps away from the final product. This route has now been used to synthesise AZD3409 at a 170kg scale (see fig.1).

enantioselective reduction

The majority of enantioselective reactions used at an industrial scale are some form of asymmetric hydrogenation, such as the reaction used by scientists at Johnson & Johnson (J&J), in collaboration with Johnson Matthey, to make the integrin antagonist JNJ-26076713.2

The initial route was not ideal, as it included a low temperature lithium reaction. The reaction required an excess of reagents, which limited batch size and posed purification problems. It also included a reaction that generated a mixture of E and Z isomers of an unsaturated ester, making its enantio-selective reduction a challenge.

The hydrogenation created all four possible diastereomers, requiring two sequential chiral chromatography steps to separate them.

Their aim was to create a route that defined the geometry of the unsaturated ester. The ß-keto ester was converted to the enol triflate. Using Hünig’s base gave a mixture of E and Z isomers, while using sodium hydride resulted in substantial cleavage of the Boc protecting group. But when both were used together, the Z enol triflate was made selectively – with less than 3% of the product being the E isomer. This was then coupled with a quinolone to give the precursor for the asymmetric hydrogenation reaction. The process was carried out using 10% Pd/C in wet methanol, which gave a 51% yield of the expected mixture of two diastereomers. These were separated by chiral chromatography to give 20% of the desired product. Better results were obtained when the methyl ester was reduced to the acid before the asymmetric reduction reaction. This was achieved using lithium hydroxide, with a 68% yield.

The hydrogenation was then carried out with a ruthenium catalyst and (R)-XylPhanePhos as chiral ligand. The final product obtained was in 71% yield and contained an ee in excess of 99%. A small amount of undesired isomer was precipitated out and could be recycled. The methyl group could then be replaced to give the required intermediate for the synthesis (see figure 2).

asymmetric synthesis

The number of examples of other asymmetric synthesis reactions being run at a large scale is increasing, however. One example is chiral cyclopropanation, which was used by scientists at Bristol-Myers Squibb in the synthesis of a selective serotonin reuptake inhibitor.3 There are numerous different reactions that give the enantio-selective cyclopropanation of allylic alcohols, including the Simmons Smith reaction.

The BMS team reacted their indole allylic alcohol derivative with two equivalents of Zn(CH2I)2 in the presence of catalytic amounts of a chiral titanium complex. Although this gave the desired cyclopropane in 65% yield, the ee was unacceptably low at 17%. Better results were obtained with stoichiometric quantities of (R,R)-dioxaborolane – the desired adduct was prepared in 83% yield and 88% ee.

Another type of addition reaction was developed at scale by a group at Merck in New Jersey in the synthesis of a glucokinase activator.4 A big problem with the medicinal chemistry route was that it included an HPLC separation of enantiomers as the final stage, so an enantioselective route would represent a big improvement. However, the options for the chirally pure synthesis of α-arylpyrrolidones were limited, and not ideal for scale-up.

They decided to try the enantio-selective deprotonation of N-Boc-pyrrolidone; while this process is well known, it has not been successfully selectively arylated. The team started with a related organozinc reagent, where chirality is induced by chiral ion formation with lithium and a suitable chiral additive, and the in situ addition of zinc chloride to the resulting lithium species. There is precedent for the conformational integrity of secondary organo-zinc reagents, and for the addition of achiral alkylzinc compounds to aryl halides.

The alkylzinc der-ivative was formed using zinc chloride and (–)-sparteine, and then the aryl component added in situ. The first catalyst used, Pd2-dba3 with tBu3P-HBF4, led to competitive deprotonation of the aniline NH2 group, so a more active catalyst, Pd(OAc)2 with the same additive, was used instead, with 0.85 equivalents of zinc chloride giving the best results. Two six mole batches were prepared, giving a total of 2.13kg of the coupled product, with an ee of more than 90%.

Another chiral addition reaction was used by a team at Merck Sharp & Dohme in the UK, this time a Michael reaction in the synthesis of an oestrogen receptor beta selective agonist.5 The drug candidate is a tetracyclic molecule, with a chiral quaternary centre that poses synthetic challenges. The medicinal chemistry route was suitable for small quantities but not ideal for further scale-up, with problems including expensive reagents, an ozone depleting alkylating agent, and a chiral chromatography separation step to separate the isomers at the quaternary centre.

The chiral centre was set up using an asymmetric phase transfer organo-catalysis process developed at Merck. First, they constructed a chlorinated phenoxy substituted indanone, and then a base catalysed Michael addition of the enolate of this indanone to methylvinyl ketone, in the presence of a (+)-cinchonine derived quaternary ammonium phase transfer catalyst. They identified a number of factors that had an impact on the ee and purity of the addition product when the reaction was run at scale. Nine different cinchonine derivatives were tried, with ees varying from 20% to above 50%. The N-(2-naphthylmethyl) derivative was chosen as, although it did not quite give the best ee, the corresponding naphythylmethyl bromide was readily commercially available.

The best organic solvent was toluene, and during the reaction, sufficient agitation was essential because of the viscosity of the 50 wt% sodium hydroxide phase. They found that the optimum reaction protocol was to slurry the catalyst in a solution of indanone in toluene and the sodium hydroxide solution, and leave it for 14 to 16 hours, before adding the methyl vinyl ketone. The toluene layer was concentrated, and a Robinson annulation was then used to construct the third of the four rings in situ. The cyclised addition product was crystallised in an overall 85% yield, and with an ee of 52%.

After dissolving and crystallising the 52% ee material in isopropyl acetate, the mother liquors were found to contain material with more than 95% ee. On further concentration, the correct enantiomer could be crystallised and isolated in 47% yield (which equates to 94% of the theoretical yield) and in 97% ee. After further elaboration, the desired tetracyclic drug candidate was achieved in an overall 18% yield, with six intermediates having been isolated during the route. This method has been used to make more than 6kg of the molecule (see fig. 3).

A different stereoselective Michael reaction was used by a group at Novartis in New Jersey in the synthesis of the peptide deformylase inhibitor LCD320, one of a new class of antibacterial agents.6 Initially, they tried a route using a beta-lactam that they had used successfully for a previous, related compound, but leaving group issues limited its success, and they returned to the Michael addition they used initially for this previous compound.

The starting material for the Michael addition included a chiral oxazo-lidinone auxiliary to induce the stereochemistry. This was reacted with O-benzyl hydroxylamine; the reaction took three days in THF at 45°C, but increasing the temperature to speed up the reaction resulted in decreased stereoselectivity. Instead, they ran the reaction under more concentrated conditions, adding the solid starting material to a solution of two equivalents of the hydroxylamine in a minimal amount of toluene.

The resulting slurry was heated to 48°C for three hours, and the resulting clear solution was part distilled to speed the reaction up further; it ran to completion in about 40 hours. The product was precipitated out, and although the diastereoselectivity in the reaction mixture was only three or four to one, the unwanted diastereomers did not precipitate out. The correct compound was obtained in 60 to 64% yield, and contained less than 1% of the wrong diastereomer.

However, when this was scaled up to the pilot plant, the diastereo-selectivity dropped from 3:1 to 2:3, even though they could not detect any difference by HPLC or NMR between the lab scale and pilot scale batches of starting material. Their first thought was that the problem could have been water in the starting material, but this was not the case.

The alternative – that it was a very minor impurity instead – was the cause. A small amount of residual lithium chloride remained from the formation of the starting material, which acted as a Lewis acid in the Michael reaction and affected the diastereoselectivity. This was proved by running the reaction at lab scale using a batch of starting material that had already given good diastereoselectivity and deliberately adding lithium chloride to the reaction; the diastereo-selectivity reversed. By modifying the pilot plant work-up conditions for the synthesis of the starting material, the problem was solved.

The current commercial synthesis of the influenza neuramidinase in-hibitor oseltamivir (Tamiflu), discovered at Gilead Sciences and commercialised by Roche, is expensive as it relies on shikimic acid as a starting material. Roche chemists in Basel have developed an alternative route using a desymmetrisation process to avoid the use of expensive chiral pool molecules.7

The new starting point is the cheap, commercially available compound 2,6-dimeth oxyphenol. This was used to make a diethyl isophthalate derivative, which was hydrogenated to give the all-cis meso diester. Several catalysts were tried, and the best results were obtained with 5% Ru–Al2O3 in ethyl acetate at 60°C; only small amounts of side products were obtained, which represents very good selectivity bearing in mind that 10 possible diastereomers (six racemates and four meso compounds) could have been formed.

This compound was then ready for the crucial desymmetrisation reaction, once the two methyl ether groups had been cleaved. After a comprehensive screening of potential enzymes, the most selective was found to be pig liver esterase, which gave the desired S-monoacid in 96 to 98% ee, and almost qualitative yield. This proceeded happily at a high substrate concentration and 35°C, which they speculate is a result of the substrate’s insolubility and the product’s hydrophilicity.

They also found that the dimethyl ether compound could also be desymmetrised with several commercially available lipase enzymes, but this gave the wrong enantiomer, the R-monoacid. Further elaboration of the S-monoacid gave the required oseltamivir as the phosphate salt in an overall yield of about 30%, in 10 steps from 2,6-dimeth oxyphenol (see fig. 4).

cost reduction

The intermediates N-Boc-2-hydroxy methyl morpholine and the related carboxylic acid are used in the synthesis of a variety of different drugs, particularly those that act on the central nervous system such as reboxetine. However, they are expensive, whether as racemates or single enantiomers, because starting materials are costly, routes are long and inefficient, and there are process issues. A chemist at Pfizer’s Groton site has developed a synthesis for both these intermediates from the readily available chiral substance epichlorohydrin.8

The epichlorohydrin is reacted with N-benzylethanolamine to give a chiral chlorohydrin adduct, which was not isolated because it is relatively unstable. This was then converted to a chiral epoxide which again was not isolated for stability reasons. The next product – the result of a hydroxide mediated cyclisation – was not isolated either, and the ratio of desired six-membered ring morpholine to undesired seven-membered ring oxazepine was 7:3. The morpholine was separated by selective succination, followed by isolation and cleavage of the succinate ester.

Finally, the benzyl protecting group was replaced by a Boc group using Pd-C catalysed hydrogenation followed by treatment with Boc2O. Alternatively, the benzyl groups in morpholine– oxazepine mixture could be replaced with Boc groups under similar conditions, and the resulting mixture subjected to Tempo oxidation conditions. This selectively oxidised the primary alcohol of the morpholine to the acid and the secondary alcohol of the oxazepine to the ketone. These were easily separated, giving the second useful intermediate from essentially the same process – and the chirality remained intact throughout the synthetic route, as shown in figure 5.