Formulating chiral routes

Hydroformylation has been used industrially for years, and yet in the world of chiral synthesis it has not achieved its full potential. But, as Dr Sarah Houlton shows, companies are looking at the chemistry with new interest

The perfect industrial asymmetric reaction is selective and high yielding, and minimises reagents, solvents and byproducts. Many reactions run well in the lab, but are difficult to scale up. Equally, established industrial scale processes are all too often unselective when chiral variants are attempted. Achieving selectivity at scale is essential - but rarely straightforward.

Hydroformylation is a reaction that has been run for many years at large scale yet, despite its great potential in chiral synthesis, a production scale enantioselective version has so far been elusive. But recent developments, particularly at Dowpharma, could be about to make this technology available for the chiral synthesis of pharmaceutical intermediates.

The hydroformylation reaction adds a single carbon unit, and is an obvious disconnection for a synthetic chemist to make in a retrosynthetic analysis. Essentially, it adds an aldehyde unit, which can then be elaborated by oxidation to an acid or by reduction to an alcohol. An amine could even be added, amounting to aminomethylation. If the process could be carried out chirally, then it is potentially a very powerful synthetic technique.

history notes

Achiral hydroformylation has been in industrial use since the 1940s with a great deal of success. Indeed, the Oxo process where, for example, propylene is hydroformylated to butanal and then reduced to butanol, is carried out as a continuous process at a worldwide scale of up to 6m tpa of Oxo products. The reaction was initially patented in 1943 by Eastman, using a homogeneous cobalt catalyst to initiate the reaction with carbon monoxide and hydrogen. This process was put into large scale production by Ruhrchemie.

Later, rhodium catalysts were substituted for the cobalt ones, and Union Carbide commercialised a rhodium triphenylphosphine catalyst for the butanol process in the 1970s. This is believed to be the largest volume reaction using a homogeneous transition metal catalyst currently in production.

However, the reaction has yet to be used industrially in its asymmetric form. The first attempt to extend hydroformylation to chiral products was made in the early 1990s, with styrene as a substrate. The hydroformylation of vinyl arenes would give the nucleus of the profen class of drugs, but these are largely sold in racemic form, and those that have been switched to their single enantiomer form have low annual sales. Naproxen is sold as the S enantiomer, as the off isomer is extremely toxic.

However with desired enantiomer selling for around only US$50/kg, the low potential returns mean that development of an asymmetric chemocatalytic route would be hard to justify.

Limited success with asymmetric hydroformylation has been achieved on a narrow substrate range with the use of rhodium catalysts based on Takasago's Binaphos ligand.¹

ligand systems

However, it is not ideal as a catalyst, as its functional group tolerance is limited, making it difficult to apply in the real world to pharmaceutical intermediates, which often have complex substitution patterns. Several other catalyst ligand systems have been explored, particularly for the hydroformylation of vinyl arenes. Others that have had some success include Chiraphite from Union Carbide² and Claver's sugar based systems.³ These are all used with a rhodium precursor, creating the catalyst in-situ. Binaphos gives an ee of 97%, but a branched:linear ratio of only 7:1 for the asymmetric hydroformylation of styrene. While Chiraphite gives a lower ee of 88%, the branched:linear ratio is much better at 50:1.

current work

At the beginning of this year, a team at DSM published a paper detailing their results using Binaphos as a ligand in the asymmetric hydroformylation of allyl cyanide.4 Dowpharma has also been working on this reaction, using a completely different asymmetric ligand that has been developed in house. Chirotech had been actively involved in asymmetric catalysis for many years, but recognised the potential to apply its catalytic skills to hydroformylation only after the acquisition by Dow.

As scientist and technology area leader Ian Lennon explains, the acquisition meant the Chirotech team instantly had a whole range of new technologies at its disposal. 'We saw the synergies between the achiral hydroformylation that Dow had acquired along with Union Carbide, and our expertise in creating new catalysts,' he says. 'And when you add the high throughput screening capabilities that Dow has in Midland, MI, US, there was the potential for tying all three together to create a powerful team to investigate asymmetric hydroformylation.'

Regioselectivity, enantioselectivity and the catalyst itself all needed addressing. Regioselectivity is a particular problem as, depending on which end of the double bond the hydroformylation occurs at, the product is either linear or branched. For the industrial, achiral process the linear isomer is typically required, whereas asymmetric reactions are much more likely to demand branched products. The catalyst is instrumental in directing the regioselectivity of the reaction. If the process is extended to 1,1'-disubstituted olefins, then both of the possible regioisomers will be chiral, with one creating a tertiary chiral carbon, and the other a quaternary centre.

modular approach

As with any chiral reaction, good enantioselectivity is key. And, as well as controlling both regio- and enantioselectivity, ideally the catalyst needs to be stable, easy to handle, and cheap to make. A further potential problem is chemo-selectivity ensuring that hydroformylation occurs, rather than hydrogenation.

'We approached the design of the ligands in a modular fashion,' says Chris Cobley, a senior research chemist at Dowpharma. 'This way, it was possible to build a library of more than 200 ligands quickly, and then screen them rapidly against a chosen substrate. Having identified a promising hit, the system could then be optimised using factorial design methods.'

different ligands

The library was created using parallel synthesis techniques, and contains a mixture of biphosphite, biphosphoramidite and bisphosphorodiamidate ligands. 'These ligands are all C2 symmetric, which makes them simpler and cheaper to make,' says Cobley. 'In contrast, Binaphos does not have this symmetry, as it contains both phosphine and phosphite moieties, rather than two identical phosphorus-containing units. The published synthesis takes six steps, and even if unreported optimisation has been achieved, it is still a lengthy synthesis by comparison.' Vinyl acetate and allyl cyanide were used as test substrates to explore the properties of the different ligands in the library. Both of these model compounds give products with potential pharmaceutical relevance, as shown in Scheme 1. Vinyl acetate leads to a-acetoxy aldehyde which can be elaborated into a chiral 1,2-diol or amino alcohol. And if allyl cyanide is hydroformylated, the potential downstream products include 4-amino-2-methylbutanol and 4-hydroxy-3-methylbutyronitrile.

investigative assays

Screening was simplified by making use of the Symyx PPR 48 well parallel reactor at Dow in Midland. The parallel reactor contains six eight-cell modules, which are loaded by a three-axis liquid handling robot, and all operations are carried out in a triple glove box. Pressure and temperature can be controlled independently, and a wide range of homogeneous catalytic reactions can be carried out. Along with rapid GC analytical assays to investigate the products, all 200 catalysts could be tested with both substrates in a few days.

This process led to a ligand dubbed Kelliphite after one of the Dow chemists who discovered it. Like the rest of the ligands in the library, it takes two steps to assemble: a chiral diol is reacted with PCl3, and then added to a biphenyl unit as shown in Scheme 2.

The first run with this new ligand on allyl cyanide gave an ee of 65%, full conversion and a selectivity of 19:1.

'This compares well with DSM's published results,' Cobley says, 'which had a 3:1 branched:linear ratio, with 73% conversion and 66% ee, using 0.2 mol% of catalyst, and four equivalents of ligand to each rhodium atom.

'By using the factorial design process, which took a week, we got the selectivity up to 23:1, an ee of 81%, and a substrate:catalyst ratio of 3000:1. And, of course, the catalyst is much less expensive to make.' Initial results with vinyl acetate were even better, with 90% ee and a 208:1 branched:linear ratio.

'Another advantage is that the reaction runs neat, with no solvent,' Cobley adds. 'This gives much improved reaction rates. And because we are running the reaction neat, the expensive rhodium can be recovered by partitioning into hexane.'

Once a good catalyst system had been identified, the team set about trying to apply it to a real world synthetic example. R-4-Amino-2-methylbutanol was identified as a potential intermediate for TAK-637, which is being developed by Japanese company Takeda Chemical Industries as a treatment for urinary incontinence.

Allyl cyanide was first hydroformylated, and then selective hydrogenation of the aldehyde gave 4-hydroxy-3-methylbutyronitrile, itself a potentially useful intermediate for pharmaceutical actives. If the nitrile is hydrogenated in the presence of an acid such as tartaric acid, the salt of the amino alcohol is formed, which can be crystallised to provide, ultimately, the free amine in high enantiomeric excess (Scheme 3).

future hopes

While asymmetric hydroformylation has yet to be put into production, the work that has already been carried out clearly proves its potential. Obviously, a large scale trial will not be carried out until there is a potential commercial application, but the Dow team is certain it can be applied to processing at any scale required. 'I'm very confident that when a customer wants it, we will be able to scale the reaction up using the equipment we already have,' explains Greg Whiteker, research specialist at Dow in South Charleston, WV, US.

'We have a big technology advantage from the large scale hydroformylations we already do, thanks to our legacy from Union Carbide.'

Lennon agrees. 'While we're never going to need to run it at multi-tonne scale, given a good intermediate there should certainly be a demand for hundreds of kilos to low tonne amounts. We're now also looking at the hydroformylation of other novel substrates. The work on allyl cyanide is very encouraging, and we believe the reaction has great potential.'