Catalysts - whether chemical or enzymatic – are key to more profitable and sustainable manufacture of active pharmaceutical ingredients. They can also be the most effective way of introducing chiral centres into an API molecule by directing the synthesis to one enantiomer or the other.
Catalysts are key to more profitable and sustainable chemistry. Dr Sarah Houlton looks at some recent examples of smarter API synthesis aided by catalysts
Without catalysts – whether chemical or enzymatic – the processes used to make many APIs would be far more expensive inefficient. Whether the catalyst speeds up the reaction, increases yield, improves selectivity, allows reactions to run at a lower temperature, or reduces raw material use, it can make a huge contribution to the sustainability of chemical processes.
Catalysts can facilitate reactions that might otherwise not have been possible, which can result in a far more efficient overall synthetic route. They can also be the most effective way of introducing chiral centres into an API molecule by directing the synthesis to one enantiomer or the other.
Several of the statin drugs used to reduce cholesterol levels contain related chiral 3,5-hydroxy side-chains. A number of methods for the production of this side-chain have been developed, including the biocatalytic route developed by Codexis for Pfizer's atorvastatin (Lipitor) that won the EPA Presidential Green Chemistry award in 2006. Another enzymatic route has been reported by Slovenian generics company Lek Pharmaceuticals.1 The firm used pancreatin powder to facilitate the synthesis of a lactonised side-chain intermediate designed for use in the Wittig reaction that attaches the side-chain to the rest of the molecule.
The team had already developed and patented a method for the synthesis of the acetyl-protected version of this lactone from simple prochiral substrates, but there was a problem: the deacetylation to the bare lactone needed for the Wittig reaction involved an expensive and toxic tin-based reagent, and was low-yielding. Removing the acetyl group without destroying the lactone ring or removing the silicon-based protecting group on the hydroxyl proved challenging, and most other chemical reactions found in the literature gave low yields and poor selectivity, if they worked at all.
The answer lay in enzyme catalysis. Seven different enzymes were screened in water as solvent, six lipases from different sources, and pancreatin powder. By far the best results were seen with pancreatin; only this and the lipase from Thermocymes lanuginose gave 100% conversion from acetate to free alcohol, but the yield with the lipase was poor at 51%. The team speculated that this was a result of lactone ring opening and subsequent loss of material during work-up. The yield with pancreatin was much better (77%).
They then looked at improving the contact between substrate and catalyst to decrease catalyst load and increase yield; they suspected that an undesired second, ring-opening reaction was taking place that reduced the yield. Removing the cosolvent that had been used to improve the lactone’s miscibility with water caused the alcohol to precipitate out once it was formed, and stepwise addition of catalyst into the mixture of starting material and phosphate buffer gave almost complete conversion and a reduced catalyst load, with an optimal reaction time of 8 hours. Further optimisation of the extraction process led to the desired unprotected lactone being isolated in yields of up to 95% (Scheme 1).
Scheme 1: Production of chiral 3,5-hydroxy side-chains
Catalytic asymmetric hydrogenation is one of the most important techniques for introducing chirality into a molecule. A recent example of this is a key step in the synthesis of Merck and Actelion's investigational renin inhibitor MK-1597, with the hydrogenation of a tetra- substituted ene-ester.2
A substantial screening process was carried out to identify the best combination of metal catalyst precursor, ligand and solvent being tried. These test reactions were run using hydrogen at 500psi and a 1:1.05 ratio of metal precursor and ligand, at 50°C for 18hrs. Ruthenium, iridium and rhodium based metal precursors were all tested, with a total of 384 different combinations of metal, ligand and solvent. A handful of the rhodium options gave some chiral selectivity, but none of them with the common ligand families such as BINAP and DuPhos, where there was no conversion at all. Ruthenium performed better, with the combination of (COD)Ru(Me-allyl)2 and a Josiphos ligand giving an ee in excess of 90%, although the yield, at 30%, was low. Reducing the catalyst loading and lowering the reaction temperature improved the ee to 99%, although there was little impact on the yield.
They thought the yield might be low because the pyridine moiety in the substrate might be inhibiting the catalyst, and adding a Brønsted acid might improve it, although they were concerned that the Boc protecting group might be removed under these conditions. Despite this concern, they found that HBAF4.OEt2 bumped the conversion up to 96%, a significant improvement, and the ee remained high. The reaction proceeded well on scale-up, with 2.3kg of the desired chiral intermediate being produced in 84% isolated yield, and 99% ee (Scheme 2).
Scheme 2: Synthesis of MK-1597
A metal catalyst was also used in the synthesis of GSK1360707F, GlaxoSmithKline's serotonin, noradrenaline and dopamine re-uptake inhibitor designed by the chemical development group at the company's site in Research Triangle Park, US.3 Here, the key step was a metal catalysed enyne cyclo-isomerisation.
The first generation route to this compound had used a Suzuki coupling to make the biaryl and then used a double alkylation with a dihalomethane to create the fused cyclopropane ring. Although this was scaled up to 15kg, the route was far from ideal as it created polychlorinated biphenyl byproducts, the atom efficiency was poor, and the cost of goods was high.
Rather than coupling two pre-formed aryl groups together, they thought it might be possible to use a cyclisation reaction to create the carbon skeleton instead, and synthesised a 1,6-enyne with the correct functionality installed for the final product. They then set out to investigate its enantioselective cyclo-isomerisation.
Related substrates had given very slow reaction rates with the usual platinum catalysts, so they looked to use gold(I) instead. They ran 32 separate trial reactions using AuCl(SMe)2 as the metal source, and variously toluene or dichloro-methane as solvent, one of 10 different phosphine ligands, and the silver salt being silver antimony fluoride, silver triflate or silver tetrafluoroborate.
The gold:silver ratio was 1:1, 1:2 or 2:3. The only one of these factors that proved essential for good conversion and enantio-selectivity was the ligand, with R-tol-BINAP giving by far the best results. Of the other possibilities tried, dichloromethane and silver tetrafluoroborate were the best options. These conditions, with a 1:1 ratio of gold to silver, gave 100% conversion, and 59% ee of the correct isomer. Although this ee is low, it still represents an improved process over the original, because of the absence of problematic PCB by-products (Scheme 3).
Scheme 3: Metal catalysed enyne cyclo-isomerisation
simplicity for success
Sometimes in a catalytic reaction, less can be more, as a team from GlaxoSmithKline in Stevenage, UK found out when they were scaling up the synthesis of the spirocyclic glycine transporter inhibitor GSK2137305: the answer to improving their process lay in removing one of the components of the reaction altogether.4
In the discovery synthesis, HPLC purification was required to remove a regio-isomeric impurity from the copper catalysed cross coupling of an aryl bromide with 4-methyl imidazole. This is clearly impractical on a commercial scale. This Ullman-type arylation of nucleophiles has garnered a lot of interest recently, notably for the coupling of azoles, which is difficult to achieve using standard palladium-catalysed procedures.
The discovery route used copper(I) iodide and L-histidine, but this required 0.6eq of the copper salt and 1.2eq of L-histidine, and there were several competing side reactions, notably the histidine reacting with the aryl bromide substrate, which led to the difficulty they had encountered in isolating the product cleanly.
The obvious first step was to screen various different solvents for the reaction, but this proved unsuccessful. They then thought that perhaps the presence of 4-methyl imidazole as the coupling partner might actually render the L-histidine ligand – which was the source of many of the problems – superfluous. So they tried the reaction without the ligand and, sure enough, the reaction worked cleanly, with DMSO as solvent at 130°C. The reaction time was slightly longer at 36 hours, but the product was isolated as a 4:1 mixture of regioisomers, which could be purified by crystallisation. They had found milder conditions that worked, but this proved the cheapest overall, with the added advantage of being straightforward to operate on a multikilo scale (Scheme 4).
Scheme 4: Cross coupling of an aryl bromide
Sometimes, the unexpected can actually help rather than hinder a reaction. This was the case in the development of a preparative scale synthesis of the CETP inhibitor anacetrapib by the process research group at Merck Frosst in Canada.5 The most likely culprit for low-level impurities is usually the catalyst: in this case, however, the contaminant was actually in the solvent used for a metal catalysed direct arylation reaction used to make a biaryl alcohol from a bromoanisole and a 2-aryloxazoline.
Their first attempt used 2.5mol% of the ruthenium(II) catalyst and 10mol% triphenyl phosphine as ligand, in the presence of 2eq K2PO4 as base in NMP as solvent, with the reaction run at 120°C for 20hrs. The conversion was good, and the reaction profile was generally clean.
They then set about optimising the process. Using electron rich, electron deficient and bidentate phosphine ligands all gave lower yields, while changing the ruthenium complex gave no improvement in either cost or efficiency. Potassium and caesium carbonates were equally effective as the base component, but organic bases such as triethylamine or Hünig's base were unsuccessful. Less polar solvents such as toluene and xylene gave lower yields than the polar, aprotic solvents NMP, DMA and DMF.
The reaction proved reproducible at a catalyst loading of 5mol%, but reducing this to 0.5–1mol% led to significant variations in conversion, with conversions as low as 30% and as high as 99% being observed; the variability was even worse at larger scale. Their first thoughts were that the degree of agitation might be having an effect, but this was discounted when side-by-side reactions using mechanical and magnetic stirring gave similar results. No conclusive results were seen by changing various other process conditions, such as adding the catalyst and ligand to a warm reaction mixture.
The answer lay in the solvent – the biggest variation occurred when different batches of NMP were used. Careful analysis of the different batches showed that in some batches, low levels of γ-butyrolactone was present. It turned out that the reproducibility and efficiency of the catalytic reaction was increased when there was more γ-butyrolactone in the mixture. If the NMP contained none of the contaminant at all, the results were poor. They also muse that most ruthenium catalysed direct arylation reactions are reported using NMP as the solvent, and this might be a result of this contamination, especially as differences in reactivity between other polar, aprotic solvents such as DMF and DMA were often marked (Scheme 5).
Scheme 5: Synthesis of anacetrapib
Catalysts used to improve processes don’t have to be metal complexes or enzymes – sometimes the addition of a tiny amount of a simple organic molecule can have a dramatic effect on the outcome of a reaction. A good example of this is in the synthesis of the corticotropin releasing factor 1 receptor antagonist pexacerfont developed by the process group at Bristol-Myers Squibb in New Brunswick, NJ, US.6 The final steps of the synthesis involved the substitution of a hydroxyl group by a chlorine, which gave the substrate for the addition of a chiral amine side-chain by nucleophilic displacement with the amine. The chlorination is typically carried out using neat phosphorus oxychloride at reflux, but this provides both safety and environmental challenges on scaling up, so milder conditions for this reaction would be extremely desirable.
First, they tried heating the substrate with diisopropyl ethylamine and POCl3 in toluene, with 1eq benzyl- tributylammonium chloride (BTAC) as a chloride ion source, which has in the past been shown to increase the reaction rate. However, the work up gave a red, oily solution that had to be filtered through silica gel to remove the colour, and the product is too soluble in toluene, which made crystallisation a challenge. Reducing the amount of BTAC to 0.6eq and adding acetonitrile as a cosolvent gave a homogeneous solution, which made work-up simpler and gave a better yield. They found it also worked without any BTAC at all, and in acetonitrile alone, both of which made product isolation simpler.
Even better results were achieved when a catalytic amount (0.01eq) of the amine base DABCO was added, as it accelerated the chlorination reaction. This gave complete conversion to the chloride in acetonitrile as solvent, in just one hour at 20°C. When the reaction was attempted under the same conditions without the DABCO, it took 48hrs to reach 90% conversion. This gave the desired intermediate, which was converted to the API without isolation. This two-step process has been carried out at a scale in excess of 40kg, giving a yield of 82% (Scheme 6).
Scheme 6: Final step in synthesis of pexacerfont
1. V. Troiani, J. Cluzeau and Z. Casar, Org. Proc. Res. Dev. 2011, 15, 622
2. C. Molinaro et al. J. Org. Chem. 2011, 76, 1062
3. N.M. Deschamps et al. J. Org. Chem. 2011, 76, 712
4. J.P. Graham et al. Org. Proc. Res. Dev. 2011, 15, 44
5. S. Ouellet et al. J. Org. Chem. 2011, 76, 1436
6. S. Broxer et al. Org. Proc. Res. Dev. 2011, 15, 343