Efficient synthesis

The CPhI conference held in Paris last October highlighted the need for the more efficient syntheses of intermediates. Dr Sarah Houlton explains

As ever more complex molecules emerge from the drug companies' research pipelines, and pressure to reduce manufacturing costs increases, so there is a growing need for faster, more efficient syntheses of pharmaceutical intermediates.Development chemists, particularly at fine chemical companies, are therefore searching for ever more effective processes, and even collaborating with academia to maximise the potential of new ideas. Several examples of this kind of work were given at the CPhI Conference held last October.

Carboxylic cyclopropane groups are common structural features in drug molecules. Examples include Pierre Fabre's milnacipran and Merck's cilastatin. As Sumitomo Chemical's Jim Birnie explained, two main techniques are used in the industrial synthesis of cyclopropane carboxylic acids: carbene addition to alkenes, and a-haloester addition to alkenes. The carbenes are generated from ethyl diazoacetate, with rhodium acetate catalysts commonly being used to trigger the decomposition. This, however, has limitations. Sumitomo Chemical has developed a new catalyst, Rh₂(OCOCPh₃)₄, which gives better yields than dirhodium tetraacetate, but exclusively racemic products; a chiral version is still needed.

Cyclopropane dicarboxylic acids are also common intermediates, and they are very flexible for subsequent functional group interconversion. Catalytic asymmetric cyclopropanation can be mediated by a chiral copper catalyst with salicylaldimine derivatives as chiral ligands. It is substrate specific, said Birnie, and the reaction with isobutene, for example, proceeds in 85% ee.

Sumitomo is also experienced in the resolution of racemic compounds. It has more than 5,000 biocatalysts - esterases, lipases and reductases - in its library for use in resolutions, which it can try out using a high throughput screening system. Genetic engineering techniques can then be used to create a practical industrial scale resolution process. It runs fermenters at a scale of up to 20m³.

The downside of resolution procedures is, of course, the low yield, as the theoretical maximum is only 50% for each enantiomer. This can be improved by employing an inversion technique to convert the undesired isomer into the correct one. An example from Sumitomo is shown in Scheme 1, where aluminium bromide plus hydrogen peroxide, or hydrogen bromide and oxygen, are used to invert both chiral centres simultaneously.

DSM's Rinus Broxterman explained some of his company's work on enantiopure amines and amino acids. Enantiopure amines are particularly powerful chiral building blocks because of their structural diversity. They range from simple non-functionalised alkyl amines, through more complicated compounds such as 2-aminoethan-1-ol, to α-amino acids and their derivatives. And then there are even more complex molecules, such as disubstituted amino acids, and β-amino acids.

Over 95% of all enantiopure amine derivatives can be made in four ways:

• Asymmetric reduction of carbonyl or imine substrates;

• Classical diastereomeric resolution;

• Stereoselective enzymatic resolution;

• Chirality transfer using a chiral auxiliary product

However, there is a bewildering array of structural variety between the molecules, alongside numerous technologically different production methods. So how should a chemical company decide which intermediates to offer? Broxterman explained that technology and process development should be driven by continuous evaluation of concrete market opportunities and important structural classes, while bearing in mind where the company's technological competencies lie.

In DSM's case, these are:

• Chiral technologies, notably crystallisation-induced resolution with in situ racemisation, and chirality transfer using D-PGA;

• Homogeneous catalysis: asymmetric homogeneous metal catalysed hydrogenation, using MonoPhos as ligand;

• Biocatalysis and biotransformations, notably biocatalytic resolution using aminopeptidases, proteases, lipases, penicillin G acylases and hydantoinases.

Crystallisation-induced asymmetric transformation involves the conversion of the more soluble isomer into the less soluble, using a racemisation aid, such as an aldehyde. Chirality transfer involves enantiopure amines being employed as chiral amino donor sources, here in the synthesis of amino acids and other amino functionalised compounds. The normal source is phenylglycinol, but this is overly expensive, and not simple to remove. DSM has developed the use of phenylglycyl-amine (PGA) instead, and PGA adducts can be valuable synthons. For example, treating them with ozone, gives different products, depending on the work up procedure employed.

Rhodium catalysed asymmetric hydrogenation is another good way of making chiral amines. In general, said Broxterman, both enantiomers are easily accessible. DSM has developed the proprietary MonoPhos class of ligands along with Ben Feringa at the University of Groningen, and has also licensed BICP from Penn State University, which can be successful with some substrates where MonoPhos is not ideal. Broxterman said that MonoPhos is very easy to use, giving good turnover numbers, yields and ees, and he believes it will become a very good 'work horse' process. As with other rhodium catalysts, it is essential to recycle the metal to make the process cost effective. DSM is running the synthesis of a-amino acids from substituted benzaldehydes at a pilot plant scale (Scheme 2), with a yield in excess of 80% and an ee of over 95% for both derivatives. The yield of the related hydrogenation to give amine products is not so good, and needs to be improved before this can make an impact, Broxterman added.

DSM is also working on using enzymes in resolution, for example the synthesis of enantiopure a-amino acids using L-amidase. A prochiral substrate is made using standard Strecker synthesis, and the enzyme is used to racemise the D-isomer. Penicillin acylase is another enzyme that is readily available because of its use in penicillin synthesis. But can it be applied to other fine chemicals? DSM has found that the penicillin acylase derived from E. coli is very selective for β-amino acids. 'It is selective, robust and you do not have to be very precise with the reaction time,' said Broxterman. The procedure shows promise for non-functionalised amines. In some cases, he added, a standard lipase can do the job too.

Another company that is working on new ligands for enantioselective catalysis is Solvias, as Marc Thommen explained at the conference. 'There are various factors that must be considered when selecting a catalyst, including cost, intellectual property issues, and the availability of large scale catalyst,' he said. If one is going to carry out a screening procedure to find the best ligand, it is essential to have a broad selection of ligands available, and ideally ones that can be tuned to tweak activity. The ligand system should be either available, or straightforward to make. And, of course, the relevant starting material should be available in bulk. He also added that access to screening capabilities is essential - relying on serendipity is not good enough. Various different high throughput systems are available on the market, but Solvias made its own. Once a 'hit' has been found, it is important to be able to scale up quickly to 50 litres, either directly or step-wise. And various different types of reactor should be available to cope with different substrates, particularly acidic ones.

Solvias is launching a ligand kit for screening purposes, which includes Josiphos, Walphos, Taniaphos (licensed from OMG), Mandyphos (again licensed from OMG), Rophos (from BASF) and Butiphane, which was originally developed by Ciba Specialty Chemicals and sold to Solvias. Thommen claimed that this is a pretty comprehensive ligand portfolio that is able to cover at least 90% of all prochiral substrates. The phosphines are added into the ligand at a late stage of their synthesis to allow for more flexibility.

Tuning the ligands to gain optimal performance requires expertise. Yield, ee, and turnover number can all be improved by subtle changes in the ligands. The ligand kit contains all six families of ligands with different phosphines and R groups, and is being sold through Strem Chemicals.

A good example of a customer process based on Solvias ligands is the route to dextromethorphane developed by Lonza (Scheme 3). By altering the R groups on Josiphos ligands, the ee was improved from 14% to 90%, and the turnover number raised from 200 to 1,500, with the overall activity improved by a factor of 200. This process is now running on a scale of 1.5 tonnes.

Thommen explained that it is essential to make the technology attractive to potential customers it is not merely sufficient to have a good ligand portfolio. He said that catalyst costs must be independent from catalyst performance; they should be sold at a kilogramme price, with IP included in the deal; they should be easy to license, and royalty-free; and no licence fees should be tied to added value generated in the catalytic step. 'All process improvements that lead to higher productivity or better selectivity should remain our customers' benefits they do not have to pay for improvements,' he explained.

DSM's Johannes de Vries explained the pros and cons of aromatic substitution reactions such as Heck and Suzuki couplings, probably the most important class of homogeneous transition metal catalysed reactions in fine chemicals production. On the pro side, they are highly selective, avoid the use of strong bases like BuLi for the carbon-carbon bond formation, they are compatible with many functional groups and can be used at a late stage of a total synthesis without protection-deprotection strategies having to be employed.

difficult to obtain

However, the palladium used as a catalyst is very expensive, and aromatic halides can be expensive and difficult to obtain. Potential solutions, he said, would be an easy on-site palladium recycling process, a cheaper palladium catalyst, and if non-halide aromatic substrates could be used, then so much the better.

De Vries believes that palladium is much the best catalyst for Heck reactions, used with phosphine ligands. High selectivity to α or β substituted double bonds is possible, and it is frequently used for large scale production. But it can require as much as 5mol% of palladium, and the phosphine ligands are hard to remove. DSM has discovered that it is possible to carry out Heck couplings without chiral ligands on iodinated aromatics, where the palladium(II) acetate is reduced in situ to an unknown Pd(0) compound (Scheme 4).

It works very well at moderate temperatures of 50-80°C, and triethylamine is used as a base alongside a polar aprotic solvent, usually NMP. One problem is the precipitation of palladium black but, said de Vries, could this actually provide a way of recycling the catalyst?

Over 99% of the palladium precipitates, but when it has been reused it has always had very low reactivity, and even the addition of solid carriers such a silica or Celite leaves the palladium much less reactive than the original catalyst. However, they discovered that adding 1-2eq of iodine served to reoxidise the metal back to Pd(II) and restore its original activity.

Amino acids are particularly important as pharmaceutical intermediates. '18% of the top 500 drugs include an amino acid in their synthesis,' explained Wacker's Thomas Maier. 'This is only likely to increase in future.' They are versatile building blocks, with a chiral centre and reactive groups that can be used for further elaboration into both peptide and non peptide compounds. Natural amino acids are generally produced by fermentation or extraction, as this is much the cheapest route. However, chemical methods predominate in the production of unnatural amino acids such as D, β and γ analogues. These chemical syntheses often require multiple steps, and frequently require a resolution to be performed at some point, with the inevitable deleterious effect on the yield.

However, Wacker has been using biotechnological procedures to make unnatural amino acids, and as Maier explained at the conference, they can even use fermentation. The company has extensive experience in carrying out biotransformations using enzymes and micro-organisms, and it runs a 3,000 tonne biotransformation plant for the manufacture of cyclodextrins in the US. It has now applied metabolic engineering to the production of amino acids.

Metabolic engineering is the redirection of enzyme catalysed reactions that take place in a living cell to improve the production of a compound that is synthesised by the cell. A microbial cell is, essentially, a chemical reactor. Many organisms can be used to make chemicals, for example E. coli has around 4,300 genes that code enzymes to catalyse reactions.

genetic engineering

The activity of certain enzymes can be improved by altering the genes, and genetic engineering can also be used to block enzymes, or redirect the cell's efforts to produce what would normally be a minor byproduct. Wacker has also found that the same bacteria engineering can be used to make unnatural amino acids. In the biosynthetic process for making cystine, an enzyme replaces an acetate unit with a thio group.

If certain other nucleophilic molecules are added to the reaction broth instead of a sulphur source, they are recognised by the enzyme and incorporated into the amino acid, leaving an unnatural side chain on the molecule.

Many are sulphur-based, but there are nitrogen and carbon compounds, and even an example that incorporates selenium. This biosynthetic route is more efficient, and it automatically gives extremely high ees. 'Fermentative processes give good economics,' said Maier. 'We are already running the cysteine process at large scale, and it will be easy to scale up the syntheses of unnatural compounds because the process engineering is already established for the natural compound.'