Speed is of the essence when manufacturing chiral APIs for clinical trials, and separation is usually quickest. More efficient routes are essential for commercial production, but separation can still have a role to play
More than four-fifths of all drugs now in development are chirally pure. The majority of those that are not simply have no chiral centres. But for a drug that does have chiral centres to be accepted in racemic form by the regulators these days, there has to be an extremely good reason for it, such as the two forms interconverting spontaneously, whether on storage or in the body, rendering single enantiomer manufacture irrelevant.
Despite the huge - and growing - interest in chiral catalysts, enzymatic catalysis and asymmetric synthesis, the more than half of all chirally pure drugs are still made either starting from a commercially available simple chiral molecule (often a naturally occurring one such as an amino acid), or by using a resolution step where a racemate is synthesised and then the two enantiomers are separated. If the "wrong" isomer can then be reracemised to create more of the "right" one, this can actually be a very efficient process.
Another big advantage of using a separation process is that it is invariably much easier to develop a synthesis for a racemic mixture than a single enantiomer product. With time being of the essence in drug development - every day extra on a drug's development process is a day fewer the drug will be on the market with its patent intact - it is very common for a company to develop an initial manufacturing process that includes a separation. It will then often try and find an asymmetric synthesis route that would be more efficient as a final manufacturing process once the drug is on the market and the API is required in large quantities. Even then, however, it can be difficult to find a sufficiently selective route that is commercially viable.
A good example of this is the synthesis of Pfizer's development cholesterol regulating drug torcetrapib. The company already makes Lipitor (atorvastatin), the world's current biggest-selling medicine, which is designed to reduce levels of low density lipoproteins (LDL), as when these are elevated it is a major risk factor for raised blood pressure and heart disease. Raising levels of high density lipoproteins (HDL) has a similarly protective effect, but while there are numerous drugs - not just the statins - that can reduce LDL cholesterol, it is much more difficult to raise HDL cholesterol.
The observation that people who do not have a functioning copy of the cholesteryl ester transfer protein, or CETP, have high levels of HDL cholesterol led to the theory that inhibiting CETP could provide a HDL cholesterol raising therapy. As a result, Pfizer developed the CETP inhibitor torcetrapib, which is now in late stage trials for this indication. This meant that large quantities were required to supply materials for clinical trials.
As it is frequently quicker (and easier) to develop a synthesis for a racemate than single enantiomer and then separate the two isomers, this was the initial strategy the chemists used.1 The key step is the construction of a tetrahydroquinoline nucleus with the two substituents in a cis arrangement. They felt the obvious choice would be an aza Diels-Alder reaction of an N-arylamine with an N-vinyl compound, as this generally has a high degree of cis selectivity.
An imine was generated in situ from p-trifluoromethyl aniline and n-propanal, with titanium tetrachloride as a water scavenger. This was then reacted with a vinyl carbamate plus catalytic BF3OEt to give the desired tetrahydroquinolone. However, to start with this gave a low yield, ranging between 40 and 60%, although no trace of the trans isomer was observed. Trapping the imine as a benzotriazole adduct solved this problem, and again only the cis isomer was formed. The team felt that a total absence of trans isomer was surprising, and so used molecular modelling to try and establish what was happening. They found that the selectivity of the cyclisation was kinetically controlled, and the cis cyclisation had a lower activation energy.
The free amine was then acetylated and the benzylcarbamate protecting group removed, prior to separation of the two enantiomers. Although chiral chromatography can be used to achieve this, in practice at the scale required it is extremely inefficient in terms of cost and time so classical resolution via salt formation was by far the more sensible option. After screening several chiral acids, the hemi-tartrate salt was found to give the best results, with a 39% yield (which amounts to 78% if one takes into account the fact that the maximum yield is 50% because the rest is the wrong isomer) of product being obtained in 98% de. All that remained was to add the required substituents to the primary amine, which was straightforward. (Figure 1).
Although this route provided a fast route to sufficiently large quantities of torcetrapib to progress the drug through clinical trials, and several multi-kilo batches were prepared, a more effective synthesis would be infinitely preferable - and much cheaper - for large scale production. In the light of the fact that they had seen not even the slightest trace of trans isomer in the racemic synthesis, the team wondered whether using a cyclisation where the precursor incorporated one of the chiral centres would also result in a cis only product. As one of the chiral centres would already have been set, this would by default lead to a single enantiomer product.2
Retrosynthetic analysis led to a commercially available starting material, (R)-2-amino-1-butanol, to introduce the first chiral centre. In addition to the chiral amine functionality, this also includes the required ethyl group and an alcohol "handle" for further elaboration. Protection of the amine with a Boc group followed by cyanation of the alcohol via the mesylate and amine deprotection gave an intermediate that could be coupled with chloro-4-(trifluoromethyl)benzene. The cyanide function in the resulting adduct was converted to an amide, and then an imide by adding benzylchloroformate, setting up the desired intermediate for a reductive cyclisation. Ultimately, conditions were established using sodium borohydride that gave an 82% yield of the cyclised intermediate - the single enantiomer version of the same intermediate in the racemic synthesis, and hence just four steps away from chirally pure torcetrapib.
Despite this success, the team still felt the synthesis could be improved further. The tetrahydroquinoline still included the benzylcarbamate protection group, which was actually not necessary for the synthesis. The extra steps required to remove it and replace it with one of the nitrogen substituents required in the final product adds inefficiency to the process. Instead of benzylchloroformate, the acylation reaction was carried out using methylchloroformate, which gave the desired adduct in 94% yield, at a 7kg scale. The reduction and cyclisation process was carried out in exactly the same way as for the benzylchloroformate adduct, with the sodium borohydride used in pellet form to improve process safety. Again, no trans cyclised product at all could be detected by NMR analysis, and 5.3kg were produced, representing an 80% yield.
This intermediate could then be elaborated further into torcetrapib itself, by a different method to the racemic version because it includes a methylchloroformate moiety and not the original benzylchloroformate. This ultimately led to a multikilogram synthesis of the drug candidate, in six steps and a 37% overall yield from the deprotected cyanide intermediate. (Figure 2).
Merck Sharp & Dohme is investigating an inhibitor of [gamma]-secretase as a potential treatment for Alzheimer's disease. This protease enzyme plays a critical role in the production of amyloid-[beta] peptide, which is one of the main pathological characteristics of the disease. As a result, it represents a potential target for drug intervention in Alzheimer's.
A team at Merck has developed a practical and efficient synthesis of a chirally pure form of the inhibitor which can be run at a multikilo scale.3 It is a challenging structure to synthesize, with a trisubstituted octahydro-1H-2,1-benzothiazine-2,2-dioxide core. One of its four chiral centres is an unusual tertiary sulfone bearing carbon. The team's approach was to use this tertiary sulfone stereocentre to control all the relative stereochemistry in the molecule. The key intermediate is a [gamma]-amino alcohol cyclohexane derivative, which also includes this tertiary sulfone. This was created using an intramolecular [3+2] nitrile oxide - olefin cycloaddition, which could be carried out in a one pot process giving an 84% yield and a ds of 96%.
However, after several unsuccessful attempts to effect a chiral synthesis of this intermediate, with particular problems with racemisation of its precursors being experienced, it was ultimately made in racemic form and separated by classical resolution. A survey of chiral acids led to the dibenzoyl-D-hemitartrate salt being identified as the best option. They found that it was possible to resolve it using just 0.25 equivalents of dibenzoyl-D-tartaric acid, but this gave much poorer yields. Ultimately, despite the separation, they managed to synthesise the desired drug candidate in a 13% overall yield, with 10 intermediates being isolated in the process. The three diastereoselective transformations each proceeded with a ds of at least 96%, all of them being created by substrate based induction. (Figure 3).
Also in an early stage of development, with a synthesis that relies on a separation, is a hepatitis C viral polymerase inhibitor at Pfizer. The medicinal chemistry approach, as is frequently the case, involved a racemic synthesis followed by chiral HPLC separation. While this is perfectly adequate on a small scale, it is far too expensive to be practical on a larger scale. In addition, the route was not amenable to the development of a future asymmetric synthesis, and some of the intermediates provided some serious stability concerns for large-scale manufacture. As a result, the development group created a completely new route, initially as a pilot process involving enantiomer separations for the sake of speed, but which has the potential for asymmetric synthesis to be applied in future.4
The medicinal chemistry route involved alkyne intermediates, which were eliminated in the development process, and a retrosynthesis was designed that centred on the preparation of a chiral tertiary alcohol. A key step was the synthesis of a prochiral ketone, which was converted into a [beta]-hydroxy acid using a two-step, one pot process. The resulting product was separated by resolution. Promising initial results were obtained using (S)-methylbenzylamine, which gave a 95% ee, but a yield of just 17%. Screening led to the identification of (1R,2S)-(+)-cis-1-amino-2-indanol as a resolving agent, with 0.5 equivalents giving the desired single isomer in 95% ee. This could be increased to above 99% by recrystallisation.
Palladium catalysed carbon-carbon coupling reactions have become particularly important tools in organic synthesis in recent years, especially in the preparation of pharmaceutical ingredients and intermediates. A group at the Universidade Estadual de Campinas in Sao Paulo, Brazil, has used one of these, the Heck reaction, to make paroxetine, Lundbeck's big-selling selective serotonin reuptake inhibitor antidepressant Citalopram.5
The Brazilian team claim that Heck reactions using arenediazonium salts have been underutilised, despite the fact that they offer several advantages over more traditional electrophiles. For a start, phosphine ligands are not necessary with the catalysts, which reduces costs and simplifies the process, not least by eliminating the time-consuming search for the best ligand. They are also usually faster, and can be carried out under aerobic conditions, making them much easier to handle because there is no need to exclude air from the reaction mixture.
As part of an investigation into the versatility of the Heck reaction of arenediazonium salts with complex unsubstituted acrylates, they looked at the possibility of making paroxetine in this way. Although numerous syntheses of the drug have been reported, none have used Heck chemistry to create the arylpiperidine moiety it contains. By starting with 4-fluorobenzene diazonium tetrafluoroborate and reacting it with an aza-endocyclic acrylate derivative, good yields of the desired adduct were obtained. This was then straightforward to elaborate into the drug itself, both by established routes and by a new procedure that gave a total synthesis of paroxetine in seven steps and a 20% overall yield (Figure 4). The group adds that a number of the Heck adducts they created during the investigation have great potential as synthetic intermediates for more complex molecules, and they are continuing their research in this area.
Another example of a new way of making an old drug chirally comes from a group at Dr Reddy's Laboratories in India.6 Esomeprazole is the chiral switch version of the successful proton pump inhibitor omeprazole (Losec), which is marketed by AstraZeneca as Nexium and which it introduced as a patent defence strategy when the older drug was reaching the end of its patent life. Like its racemic predecessor, the new version decreases the amount of acid in the stomach and is used for the treatment of ulcers, gastro-oesophageal reflux disease, erosive oesophagitis and other conditions that result from an excessive production of acid in the stomach.
Unusually, its chiral centre is sulfur-based rather than carbon, and the available methods for synthesising chiral sulfoxides include optical resolution or asymmetric oxidation of the corresponding sulfide. The former have proved more effective, as the yields for the asymmetric oxidation of the sulfide were poor. Similarly poor results are achieved with the oxidation of sulfides in the presence of a chiral catalysts, such as (-)-menthol or a substituted chiral succinate, or an optically active substituted N-halocaprolactam. Another less than successful method uses an inclusion complex with cyclodextrin. While purities in excess of 90% could be achieved, the complex is unstable and the process needs to be carefully monitored throughout.
A more successful approach uses a titanium mediated oxidation of a prochiral sulfide with cumene hydroperoxide in the presence of (S,S)-diethyl tartrate and a base. Titanium can also be used as a catalyst together with a chiral auxiliary such as a tartaric acid ester. The group at Dr Reddy's has found that, by taking racemic omeprazole, it can be converted into esomeprazole. They first formed a sodium salt and then transformed it into a titanium complex in the presence of diethyl-D-tartrate. It is further reacted with L-(+)-mandelic acid to form diastereomeric salts. The S-isomer was then simple to separate out, and then turned into free esomeprazole with a reported chiral purity of 99.97%, as determined by chiral HPLC.