When producing APIs or intermediates with chirality, ensuring that only the correct isomer is produced can be a problematic step. Dr Sarah Houlton highlights several new routes developed to achieve that goal
These days, if a new drug molecule contains a chiral centre, it is unlikely that the regulatory authorities will allow it onto the market in anything other than single isomer form. About the only exception is if two isomers rapidly inter-convert, rendering a chirally pure form impractical. It is thus essential that an effective, efficient method for synthesising the desired single isomer of a drug is developed before it reaches the market.
There are several strategies for introducing chirality into a molecule: employing stereo-selective chemical reactions; using chiral fragments from the "chiral pool" of common, cheap naturally chiral molecules; or making racemic or diastereomeric mixtures and then separating them. The bedrock of industrial chiral chemical synthesis is asymmetric hydrogenation, and numerous chiral catalysts and ligands are available, enabling many substrates to be selectively hydrogenated.
A recent example of this from Merck involves the hydrogenation of a pyridyl N-acyl enamide as the key chiral reaction in the synthesis of a potent bradykinin B1 antagonist.1 Bradykinin mediates physiological processes that accompany both acute and chronic pain and inflammation, and there are two classes of bradykinin receptor, B1 and B2, that are activated on kinin release. While B2 is present in many tissues and cells under normal conditions and is thought to be responsible for the acute pain sensation on injury, B1 is expressed in response to tissue damage, inflammation and bacterial infection and thus could be a target for drugs designed to treat chronic pain and inflammation. Merck scientists invented a compound that selectively blocks this receptor, and kilogram quantities were required for further development.
The convergent route they chose for the scale-up split the molecule into three fragments, with its single chiral centre in a bromofluoropyridine derivative. Several possible routes were tried, and they decided to use a route that started from the commercially available starting material 5-bromo-3-nitropyridin-2-ol. The hydroxide was replaced with a second bromine using POBr3, which was then substituted with a cyanide group using CuCN in EtCN, before converting it into an N-acyl enamide ready for chiral hydrogenation.
The catalyst search began with a comprehensive screen of numerous different chiral phosphines and rhodium sources at 10 mol%, using methanol as solvent at room temperature and 90psi of hydrogen. Several rhodium and phosphine combinations gave good results, with clean hydrogenation and high ee, and the best phosphines tried were BisP and Tangphos, both of which gave an ee of 99%. Tangphos was chosen because it was cheaper and easier to source. They found that using (S,S,R,R)-Tangphos along with Rh(COD)2[BF4] gave full conversion at catalyst loadings down to 0.1% with ee above 99%. The results were dependent on the quality of the enamide, and increasing the catalyst loading to 0.3% gave a more reliable process on scale-up (Scheme 1).
Another drug synthesis from the same process chemistry group at Merck uses a novel SN2 triflate displacement of a chiral a-trifluoro-methylbenzyl triflate with a leucine derivative.2 The drug, odanacatib, is a potent and selective inhibitor of cathepsin K, a cysteine protease that is extensively expressed in osteoclasts, the cells responsible for bone resorption. Because it is so well expressed in these cells, it provides a potential target for the treatment of osteoporosis, where too much bone is resorbed and insufficient new bone is made as part of the normal bone remodelling process.
The key step in the synthesis involved the stereospecific coupling of two fragments. Both of these contained a chiral centre, but the issue was preserving the chirality to give a single diastereomer and not a mixture. The first problem occurred with one of the chiral fragments - replacing a benzylic hydroxide with a leaving group without compromising the stereochemistry at the benzylic centre bearing a trifluoromethyl group. Mesylate and tosylate were unreactive under various different reaction conditions.
More success was achieved with tresylate, but it gave only 42% conversion and 5:1. Much higher conversions were seen with the triflate, but at the expense of much of the ee. By trying a variety of different solvents and bases, they found that using 2,6-lutidine as the base and cyclohexane as solvent gave the best results, with 95% conversion and 92% ee.
The next issue was how to couple this to (S)-g-fluoroleucine ethyl ester without losing the enantioselectivity in either compound. They had already established that the chirality in the benzyl fragment was compromised by the racemisation of the triflate as the triflation proceeded. Fortunately, this proved not to be an issue in the coupling reaction, as when the leucine fragment and potassium carbonate were added to the hexane solution of benzyl triflate that was isolated from aqueous work-up of the previous step, after heating to 65-70°C for 18- 24 hours, the coupled fragment was isolated in 95% yield and 84% diastereomeric excess. Further elaboration led to odanacatib without further loss of diastereoselectivity (Scheme 2).
A readily available molecule that contains chiral centres is often the cheapest way of introducing stereochemistry, as long as one exists that meets the requirements of the target molecule. Roche's oseltamivir (Tamiflu) has become extremely important in recent years, with governments and health authorities stockpiling large quantities of the drug, initially in case the H5N1 avian flu jumped species into humans, and more recently as a result of the H1N1 swine flu epidemic.
The synthesis that is used in the manufacture of oseltamivir uses shikimic acid as a starting material; this compound contains three chiral hydroxides. The process uses azide chemistry, which is potentially dangerous. This, along with a perception that shikimic acid was in short supply, led to several alternative syntheses being developed by third parties. However, according to process chemists at Roche in Basel, shikimic acid is now readily available both from the extraction of star anise, and from fermentation using a genetically modified strain of E. coli. They believe that their current synthetic process is better than all the non-shikimic acid alternatives, and have now published an improved synthesis that starts from the natural product.3
The acid group is first protected as its ethyl ester, and the hydroxyls as mesylates. Then comes the azide step, where sodium azide gives a regio- and stereoselective substitution of the allylic O-mesylate. There was a danger of aromatisation here - with consequent loss of all the stereocentres - which was prevented by adding the triethylamine base at 0-5°C. Treatment with triethylphosphite formed an aziridine, which was ring opened with 3-pentanol and boron trifluoride etherate as a Lewis acid catalyst. This was all achieved in 45% yield from shikimic acid, without the need for purification of any of the intermediates.
The nitrogen-phosphorus bond was then cleaved, followed by the addition of further sodium azide to replace the final O-mesylate with an azide group, which was then elaborated into the desired amine in the final step. The authors claim that this provides the most direct route thus far to oseltamivir, and while there are eight steps in the reaction sequence, the final product is made in an unoptimised 20% overall yield from shikimic acid, and with just three work ups and purifications being required. (Scheme 3).
Commercially available molecules that contain chiral centres but which are not themselves chiral as they contain a plane of symmetry also have potential, if they can be desymmetrised. An example comes from a programme at Bristol-Myers Squibb, where the medicinal chemists were looking for a drug candidate for the modulation of chemokine activity.4 Their target molecule requires a chiral intermediate that one might imagine coming from cis-1,2,3,6-tetrahydrophthalic anhydride. However, although techniques for opening up and desymmetrising this meso compound to give a monoester exist - one isomer comes from alcoholysis catalysed by cinchonine and quinine, and the other via cinchonidine and quinonidine - these reactions are not ideal.
The answer lay in an enzymatic desymmetrisation of a diester derivative of the parent diacid. They screened a collection of lipases, esterases and proteases, all of which are hydrolytic enzymes, for activity in hydrolysing the diester. Eleven of these gave at least 5% conversion to the correct monoester; none of the enzymes gave a diacid. Taking all factors into consideration, such as yield and ee, they proceeded with the immobilised lipase from Candida antarctica, Novozym 435. Further investigations into the reaction conditions led to the process being carried out at pH 8.5 and at 40°C, as this maximised both yield and ee. The reaction has been scaled up, and a total of 3.15kg of the monoester has been made from 3.42kg of the diester, in two batches with yields of 98.1 and 99.8%, and with an ee of 99.9% (Scheme 4).4
Resolution of a racemate remains a very common technique for producing chirally pure drugs, because synthesising mixtures is often the cheapest way. Although the theoretical maximum yield is 50%, this can be greatly improved if the undesired isomer can be recycled.
A good example of this is a process recently published by Piramal Healthcare in Huddersfield for sertraline (Pfizer's Zoloft).5 It involves the resolution of a chiral amine. Many different strategies for this molecule that include a racemisation process already exist, but this new strategy integrates diastereomeric crystal resolution and catalysed amine racemisation that could, in theory, give quantitative yields and very low waste.
Diastereomeric resolution is widespread in pharmaceutical manufacture because it is simple to operate. A second chiral centre is introduced temporarily into the molecule to create diastereomers, which are easier to separate than enantiomers because their shapes are different, rather than merely being mirror images of each other. However, the drawbacks include low yields, low productivity, long cycle times and large amounts of waste, which increase if multiple recrystallisations or recovery of the chiral acid are necessary.
The process the Piramal scientists developed has the potential to be applied to a wide variety of chiral amines because, unlike many other processes, the racemisation step does not rely on a strongly acidic proton at the chiral centre. In the sertraline example, the current commercial synthesis uses a simulated moving bed chromatographic separation of a racemic tetralone, followed by a diastereoselective reductive amination, which gives at least 95% of the desired cis isomer of sertraline. The "wrong" isomer of the tetralone can be racemised using an alkoxide base. But it does not alter the fact that there are four possible diastereomers of sertraline, leaving the maximum theoretical yield of the correct product without using any racemisation at 25%.
The new resolution process uses an iridium based catalyst, [IrCp*Cl2]2 (Cp* = pentamethyl cyclo penta diene), with potassium iodide as a stoichiometric additive to speed the process up. They have termed this system SCRAM, and it is air stable, tolerates both weak acids and bases, and works for many different primary and secondary amines, although it is less effective if there is a strongly electron withdrawing group on the aromatic ring. Chiral alcohols are also potential substrates.
With sertraline, both chiral centres need to be racemised, and as the precursor sertralone is very cheap, to be cost-effective the recycling process needs to be efficient if recovering it is going to be worthwhile. After a mandelic acid mediated crystallisation to separate the desired isomer, the SCRAM catalyst was used to set up a novel simulated dynamic thermodynamic resolution. First, the amine is racemised at 80°C using 0.1% of the catalyst. The process takes just 30 minutes, and then a base is used to racemise the other chiral centre. The whole process is carried out in one pot, with one solvent, and the correct isomer is then separated by crystallisation, giving sertraline that is more than 95% pure and contains levels of iridium below that which is allowed by the regulators. (Scheme 5).
Enzymes are also commonly used to carry out resolution processes. In an example from Bristol-Myers Squibb, a developmental oral TACE inhibitor, BMS-561392, was scaled up with an enzymatic resolution as the key chiral step.6 TACE, or TNF-a convertase, is implicated in numerous different inflammatory conditions, including rheumatoid arthritis, and so a compound that selectively blocks it has therapeutic potential. BMS-561392 was selected for development as it exhibited good pharmacokinetics and efficacy in preclinical models.
The discovery route included numerous chromatographic steps, including a diastereomeric separation and the protection and deprotection of a phenol, that all added to the length of the stepwise synthesis and material loss. If the separation could be carried out selectively on the unprotected phenol, then that would give a much more efficient process.
Literature precedent suggested that pig liver esterase could be used to hydrolyse a-allylphenyl glycine ethyl esters in a stereoselective manner, and the required ester gave the crude resolved ester in 88% yield, and 96% once it had been crystallised. Interestingly, when this process was attempted on the benzyl protected phenol, the results were significantly inferior, with the ee dropping to just 70%. This proved to be the key step in the new synthesis, as the phenol could be alkylated with 4-chloromethyl-2-methyl quinoline after the resolution (Scheme 6). In total, more than 150kg of this developmental drug were made in this way over a two-year period. mc