Redesigning reactions is key to making the synthesis of drug molecules for late stage trials and commercial production cheaper, more efficient and safer. Dr Sarah Houlton looks at some recent successful examples that create better large-scale routes to chirally pure APIs.
With the importance of single enantiomer molecules in the pharmaceutical industry, it is perhaps unsurprising that the process chemistry groups within pharma companies are adept at solving chirality problems. Whether this involves applying reactions developed elsewhere or inventing their own to create better large-scale routes to chirally pure APIs, their efforts enable the practical synthesis of drug molecules for late stage trials and commercial production.
The chemical process r&d group at Amgen in Thousand Oaks, CA, US, for example, developed a catalytic asymmetric synthesis route to a tertiary benzylic carbon centre, using a phenol-directed alkene hydrogenation.1 These moieties are challenging for synthetic chemists to put together, and often functional ‘handles’ are left behind that must be removed. This is not very atom efficient. Amgen had identified a chiral phenol that was a key structural motif in the series of drug candidates, and could also be applied to the synthesis of the marketed overactive bladder treatment tolterodine (Detrol), which loses US patent protection later this year. Some form of asymmetric hydrogenation appeared to be the obvious synthetic choice.
Amgen had identified a chiral phenol that was a key structural motif in the series of drug candidates, and could also be applied to the synthesis of the marketed overactive bladder treatment tolterodine
The medicinal chemistry route used a chiral chromatographic separation to introduce the chirality, and they wanted to use an enantioselective route instead, and one that used a cheaper, readily available starting material, unlike the lab route. They took a cyclopentanone that could be made from cheap starting materials, and tried an aryl Grignard coupling in conjunction with dehydration to construct the cyclopentene derivative required for the hydrogenation reaction. However, they expected this Grignard reaction might prove challenging because of the potential for a competing ketone enolisation. Adding a lanthanide salt has been used in the past to improve the yields of such transformations, and this was successful using both lanthanum and cerium (III) chlorides. The required cyclopentenyl adduct was then simple to generate using an acid-mediated dehydration reaction.
With this adduct in hand, they turned to the asymmetric hydrogenation reaction, and screened a large panel of ligands for the hydrogenation reaction, using Rh(COD)2BF4 as the catalyst. Four of the ligands gave a conversion in excess of 95% and an ee above 50%; Josiphos SL-J-210-1 gave the best ee at 83%. Using this as the ligand, they then studied a range of solvents, additives, pressures and rhodium loadings to optimise the reaction. Solvent choice made a big difference; while reactions using methanol and ethyl acetate gave ees below 80%, THF at 94% and toluene at 96% were much better. Using the optimised conditions, with 0.1% catalyst loading, 5 mol% of triethylamine as base at 200 psi in THF as solvent, the reaction went to completion in an hour on a 98g scale, giving a crude ee of 98.7%, and 96% after filtration and trituration (Scheme 1).
Scientists in Takeda’s chemical technology department have also developed a new asymmetric hydrogenation to solve an API manufacturing problem, this time on a benzophenone-based compound as part of the large-scale synthesis of the cholesterol-lowering squalene synthase inhibitor TAK-475.2 Its benzoxazepine core had been created using an intramolecular Michael addition, and then condensation of the N-alkylated benzhydrol derivative with a fumaric acid. This was made in racemic form, and then an enzymatic resolution was used to create enantiopure material, but for the larger scale they wanted to develop an enantioselective synthesis.
Typically, chiral benzhydrols are made by the reduction of a prochiral benzophenone, or via the addition of air-sensitive phenyl nucleophiles to benzaldehydes. They thought that the way ahead might lie in an enantioselective borohydride reduction reaction, using lithal along with a bimetallic chiral Lewis acid. However, this gave only moderate enantioselectivity, and better results were achieved with (–)-DIP chloride.
An alternative strategy was to use a ruthenium-catalysed hydrogenation, which has literature precedent, and gave reasonable results in terms of enantioselectivity and catalyst efficiency, but further improvements were needed if it were to make a commercially viable route.
The initial route used the precatalyst [RuCl2{(S)-xylbinap}{(S)-daipen}] and it required high pressure. They believed this could be optimised by altering the BINAP analogue to give a system that ran at lower pressure while giving better enantioselectivity. To do this, they developed a range of new BINAP analogues to help them fine-tune the catalyst. They found that those with methyl and ethyl substituents at the 3,5 positions of the aryl groups on the BINAP gave much better enantioselectivities, typically giving ees of 94–95%. This was ascribed to steric and electronic effects of the ligand on the phosphorus atom. Meta substituents with greater steric bulk gave much lower enantioselectivities and conversions.
Scheme 2: Asymmetric hydrogenation of a benzophenone by Takeda
The two best of these derivatives – both with two methyl groups and one with a 4-amino substituent – were then compared using mild hydrogenation conditions. The ee with the amino derivative actually improved at lower pressure, while it fell for that without. This suggested that an increased electron density on the phosphorus was beneficial.
Reaction temperature also had a big effect on enantioselectivity – although the reaction rate dropped along with a reduction in temperature, at 30–40°C a good balance of selectivity and productivity was achieved. At 60°C, the ee dropped to just 81%. Scaling up, adding THF to the isopropanol solvent increased the solubility of the substrate, and enabled the reaction to be run at a high concentration. These conditions enabled them to produce the desired inter-mediate on a large scale, with full conversion in 94% ee, and after recrystallisation, it was obtained with >99.9% ee in an 82% yield (Scheme 2).
A further application of asymmetric hydrogenation comes out of the chemical research and development group at Pfizer.3 The scale-up of imagabalin, a chiral β-amino acid that is being investigated as a voltage gated calcium channel ligand in generalised anxiety disorder, posed challenges in controlling the relative and absolute stereochemistry of two chiral centres within the molecule. Because the dose is projected to be fairly high, cost of goods is a significant issue, so they wanted to develop an asymmetric synthesis route rather than rely on some form of chiral separation where the overall yield would probably be lower.
The scale-up of imagabalin, a chiral β-amino acid that is being investigated as a voltage gated calcium channel ligand in generalised anxiety disorder, posed challenges in controlling the relative and absolute stereochemistry of two chiral centres within the molecule
The medicinal chemistry route was long – 10 linear steps. It started from a cheap chiral pool compound, (S)-(–)-(beta)-citronellol, with the second chiral centre introduced via a diastereoselective enolate alkylation, and enabled multi-kilogramme quantities to be made. However, it could not be scaled up further for safety reasons, as it used hazardous reagents such as chromium(VI) oxide. The Pfizer development team in the US came up with an alternative using two Michael additions, but this was likely to be too expensive, because two cryogenic reactions were involved, and the de was only 92%.
The team in Sandwich took on the project, and looked to create the chiral β-carbon via the asymmetric hydrogenation of an enamide precursor. This precursor was made as a mixture of E and Z forms in two steps from (R)-3-methylhexanoic acid, which is available in tonne quantities from commercial suppliers. For the asymmetric hydrogenation reaction on the E:Z mixture of enamides, a screen of chiral cationic rhodium complexes highlighted three possibilities, all of which gave the product with a de in excess of 90%. While the catalyst that included an R-binapine ligand gave the best stereoselectivity, with a de of 94–97%, it was less active and thus the catalyst loading had to be higher than the alternatives, at 200:1.
Scheme 3: Asymmetric synthesis used in scale-up of imagabalin by Pfizer
Although it gave a lower de at 92–94%, the catalyst with an (R)-trichicken-footphos ligand (originally developed by Pfizer scientists) gave a substrate to catalyst ratio of 950:1 under similar reaction conditions; improvements in the synthesis and enantiomeric purity of this commercially available ligand facilitated a reliable increase of the de to above 95%. Isolation and crystallisation raised this further, reaching the specification of >98% de. This catalyst was therefore chosen for scale-up. After optimisation, the reaction proceeded at a low hydrogen pressure of 0.5 barg at a temperature of 50–55°C in an alcoholic solvent – either methanol or ethanol. The reaction runs to completion in 12–24 h, giving a conversion in excess of 95%. Increasing hydrogen pressure to 3 barg led to a reduction in diastereoselectivity.
They have now made 30 batches in the pilot plant and the commercial facility, with a maximum batch size of 420kg. After the final elaboration to imigabalin hydrochloride, a total of more than 1.6 tonnes of the API have now been made in this way. They have since been working on an alternative route that avoids the protection-deprotection of the enamine intermediate, which would increase the efficiency of the process further (Scheme 3).
Chiral sulfinyl transfer agents can be used in asymmetric synthesis, and while the synthetic options for making chiral sulfinamides are limited, they have potential in the synthesis of chiral amines. As a result, a group at Boehringer Ingelheim in Ridgefield, CT looked to develop a more flexible way of making them to increase their synthetic usefulness in the creation of chiral amine pharmaceutical intermediates and ingredients.4 The result was a double displacement method that enabled them to prepare a variety of chiral alkene and arene sulfinamides.
Their starting pot was the amino acid phenyl glycine, which is readily available in both R and S forms. These were first tosylated, and the adduct reacted with methylmagnesium bromide in THF to give the amino alcohol; the tosylation activation could not be avoided as the direct reaction of the Grignard with the amino acid gave a very poor yield of amino alcohol. The next step was to make a 1,2,3-oxathiazolidine-2-oxide derivative in a high diastereoselective excess. This was achieved via slowly adding a base to a mixture of the tosylated amino alcohol and thionyl chloride in THF at –40°C, giving a 90% yield and a de in excess of 90%, at a 200g scale. The de was improved to 99% by recrystallisation.
Scheme 4: Route to chiral amines from Boehringer Ingelheim
They then looked to convert this to a range of sulfinamides with potential as chiral transfer agents, testing a range of Grignard reagents in conjunction with the base LHDMS. First, the Grignard was added slowly at –78°C to the heterocyclic compound in THF to give a sulfinate intermediate. The base was then added, the reaction mixture warmed to room temperature, and the result was a good yield and ee of sulfinamide. Importantly, the chiral auxiliary could be recovered for reuse. They have further extended the utility of the oxathiazolidine-2-oxide by using it to make chiral sulfoxides and ferrocene-sulfinyl derivatives; the latter of these also have potential in asymmetric synthesis (Scheme 4).
Sometimes, the biggest problem with creating a chiral compound is actually making the starting material for the process
Sometimes, the biggest problem with creating a chiral compound is actually making the starting material for the process. Merck, for example, developed a route to a single isomer chiral cyclopropanone from a readily available vinyl boronate in six steps.5 Commonly made via the Simmons-Smith cyclopropanation of vinyl ethers or the Kulinkovic cyclo- propanation, neither of these processes is particularly effective at making chiral cyclopropanols, even with the use of chiral catalysts or auxiliaries, because of low yields and problems in removing chiral auxiliaries, and the substrate scope is limited.
The Merck group developed a six-step synthesis that relies on an enzymatic reso-lution to enrich the enantiomeric excess of the product. While the precursor racemic alkynyl cyclopropanol could be made from the cyclopropanol chloride on a lab scale, there was an issue with the large-scale use of the hazardous lithium acetylide ethylene di-amine complex which introduces the alkynyl unit.
The advantage of this approach was that it eliminated two steps from the process – a protection and subsequent deprotection.
The complex itself is a commercially available solid that is fairly stable in the air, but requires a highly polar solvent such as DMSO to react with an alkyl chloride, and 2.1 equivalents of the complex were required. This is a problem – it generates one equivalent of acetylene gas as a byproduct.
Scheme 5: Merck route to an isomer of cyclopropanone
A further problem is the incompatibility of strongly basic reagents like this with DMSO, and the elevated temperature that was required to make the reaction proceed at an acceptable rate. So careful safety investigations were required ahead of performing the reaction on a large scale.
Tests were carried out in a closed system accelerating rate calorimeter, corrected for thermal inertia, to mimic the situation in a 500-litre vessel. Sure enough, the DMSO was unstable once the temperature rose above 50°C, raising the potential of a dangerous thermal runaway. Indeed, the runaway accelerated with an approximate adiabatic temperature increase of 170 K, heating the reaction to 210°C. At this point, the DMSO decomposed further, giving a very violent, uncontrollable reaction, with 4000psi being generated, and the calorimeter failed. ‘At this point,’ they say, ‘it was obvious that the process was not safe to scale up.’
A screen of other polar, aprotic solvents suggested that DMPU was a much safer choice; although adiabatic heating still occurred, it was much more thermally stable as the second decomposition did not take place. The acetylene problem was solved by introducing a sacrificial base, with n-hexyl lithium in THF being added to the chloride, before adding this mixture to the complex in DMPU.
Finally, they were in a position to separate the enantiomers, once an acetyl group had been installed on the hydroxyl unit. The separation was achieved using the Candida lipase B enzyme Novozym 435 from Novozymes, following a screen of lipases and amidases in different formats. By shaking the racemic compound with the enzyme in MTBE as solvent and a 0.1M aqueous solution of potassium phosphate dibasic, there was no deacetylation, and the reaction proceeded, without overhydrolysis that would reduce the enantioselectivity, to give the desired chiral cyclopropyl derivative in 96% ee. It has now been scaled up to produce 2.81kg of the desired enantiomer.
references
1. S. Caille et al. J. Org. Chem. 2011, 76, 5198
2. M. Goto et al. Org. Process. Res. Dev. 2011, 15, 1178
3. M. Birch et al. Org. Process. Res. Dev. 2011, 15, 1172
4. Z.S. Han et. al. J. Org. Chem. 2011, 76, 5480
5. E.M. Bassan et al. Org. Process. Res. Dev. 2012, 16, 87