Removing Impurities

"Nothing comes without a hidden cost or a flip side," says Christopher J. Welch, head of the analysis and preparative separations group within process research at Merck. "Asymmetric catalysis is a great boon but carries with it the problem of having to deal with residual metals." As the use of organometallic catalysts in synthesizing chiral and nonchiral compounds has been increasing, so too has the need for chemists to find ways to remove metal-related impurities.

Contributing to the problem are ever more stringent regulatory restrictions on the level of metals allowed in pharmaceutical products. Routine testing with a validated method is required for specific metal residues from catalysts, explains David Pears, chief scientific officer of Reaxa. This testing is in addition to regulatory requirements that may stipulate general heavy-metal tests relating to overall production quality, including metal from all contamination sources.

"Catalysis is now such a major feature in pharmaceutical production that trace-metal reduction to low parts-per-million levels is an increasingly important challenge for the industry," Pears says. "Patient safety is clearly the dominant consideration and the key driver that dictates critical high performance from metal-scavenging products and processes."

Specifically, transition-metal catalysts are being explored in a variety of reactions and chemistries (see page 40). On the plus side, as catalyst use is developed and optimized, catalyst performance improves and required loadings decrease. Nevertheless, chemists still may find dealing with residual metals to be a stumbling block in developing new synthetic routes.

"Five or 10 years ago, it would have been very difficult to obtain the required levels," says Kapa Prasad, distinguished fellow in process R&D at Novartis. "With the combination of methods available and the knowledge that has developed, we can reach those levels routinely." Although not difficult to carry out, finding the right method is still largely trial and error, Prasad points out.

Prasad and coworker Christine E. Garrett have reviewed what they call “the art of meeting palladium specifications” in active pharmaceutical ingredients (APIs), describing four possible attacks: distillation, adsorption, extraction, and crystallization (Adv. Synth. Catal. 2004, 346, 889). Key to choosing a method, Prasad tells C&EN, “is not adding another operation to a process.” For example, using an adsorbent when there's already a filtration step may be fine, but needing to switch solvents to make it work is to be avoided.

A common problem with widely used palladium catalysts, Prasad points out, is that the state of the metal at the end of a reaction is not easily predicted. "In a given reaction all forms could coexist, in which case a single method may not be able to remove everything,"; he says. The metal also is expensive, and many researchers want to recover, not just remove, it.

Before taking on potential metal problems, it's worth considering how much value a catalytic step offers. At the Chiral USA 2005 meeting in Princeton, N.J., in July, Yongkui Sun, director of Merck's Catalysis & Reaction Discovery & Development Laboratory, described using a critical rhodium-catalyzed asymmetric hydrogenation late in the synthesis of MK-0431 (C&EN, Sept. 13, 2004, page 28). "You can strategically place the high-value asymmetric hydrogenation step at the end of the synthesis so you don't lose ‘value' as you go along the synthesis," he said.

Some attendees questioned having the metal-catalyzed transformation occur so late in making the API, as putting it earlier would afford the opportunity to lose residual metal in subsequent steps. "What we found with the metal issues is that you can always find a solution," Sun responded. "Sometimes very simple things just work out really well." In this case, the Merck researchers used an Ecosorb adsorbent found through a screening process.

"These days, there is some metal removal issue in most of our drug compounds coming forward," Welch says, “and that would not have been the case five or 10 years ago."To address the problem, his lab created a 96-well microplate assay to screen adsorbents (J. Sep. Sci. 2002, 25, 847). Recently, they've been using a more convenient microtube-based screen of 12 to 30 adsorbents that were frequent "hits" in the older method (Org. Process Res. Dev. 2005, 9, 198).

Economics was one factor behind the change, Welch explains, because the larger screen included some exotic and expensive materials, costing thousands of dollars per pound, as well as very inexpensive ones, such as activated carbon at just a few dollars per pound. "Although we are looking down the line at production applications, even the amount we'd use in a development campaign can be prohibitive for some of these materials," he says.

Adsorbent performance varies with reaction conditions, the nature of the metal complex, and, for residual catalysts, the type and amount of ligands. Because it's nearly impossible to predict which adsorbent will work, Welch recommends screening actual reaction streams. "We also look through the reaction sequence to see where we can best nail the metal problem," he adds.

Several companies, including Reaxa, Engelhard, and Johnson Matthey, have developed new specialty materials with strong affinities for metals. Sigma-Aldrich is now distributing Reaxa's QuadraPure scavenging resins. The lead product, QuadraPure TU, is a macroporous bead that has thiourea functionalities and is designed to result in low levels of extractable impurities when used in pharmaceutical processing. Other products have carboxylic acid, phosphonic acid, diacetate, amino, or other functionalities to bind up to 15 metals under different reaction conditions.

One advantage of metal-specific products versus traditional activated carbon, Reaxa says, is having lower losses in yield. High-surface-area carbons with nonspecific adsorption mechanisms tend to trap both impurities and desired products. The company is working with major pharmaceutical firms to generate data on trace-metal removal to meet drug production standards. Reaxa plans to launch additional QuadraPure resins.

Early this year, Engelhard introduced its metal-scavenging agents. They are free-flowing powders made from inorganic substrates, including silica, alumina, and activated carbon. Effective on different metals, they are designed for fixed-bed or slurry applications using aqueous and organic solvents.

Johnson Matthey has three of its Smopex metal scavengers on file with the Food & Drug Administration for use in product streams containing APIs. Smopex-111 is a styryl thiol-grafted polyolefin fiber that can remove palladium, platinum, rhodium, and copper. Similarly, Smopex-105 is a vinyl pyridine-grafted polyolefin fiber that can pick up anionic platinum group metals and complexes, and Smopex-102 is an acrylic acid-grafted fiber for cation scavenging of nickel, iron, or chromium. The 0.3-mm-long Smopex fibers are mechanically and chemically very stable, the company says.

Yet another approach is to avoid or reduce metal contamination from the start. In theory, heterogeneous or immobilized catalysts bound to supports are more easily separated than homogeneous catalysts and less likely to introduce metal into the reaction. This is true only if the metal remains bound and is removed during isolation of the product. In reality, leaching of the metal can occur, and various approaches, each with their proponents, target improved catalyst stability and performance.

Reaxa's EnCat catalysts use a highly cross-linked microporous polyurea matrix that retains the activity of the catalyst and allows for it to be reclaimed and reused. According to Reaxa, residual metal levels for its palladium EnCat products are typically less than 10 ppm. The company plans on expanding the EnCat family to include ligand-coencapsulated products and other transition-metal catalysts.

Engelhard has a catalyst immobilization technology in which [Rh(COD)2 ]+BF₄^- is fixed to a support, which then allows introduction of different ligands through simple ligand-exchange reactions. For example, in asymmetric hydrogenations using the (R,R)-MeDuPhos ligand, the company says, the activity and selectivity of the immobilized catalysts are comparable with those of their homogeneous analog.

Johnson Matthey has taken a few approaches to immobilizing catalysts, explains Christopher F. J. Barnard, scientific consultant at the company's technical center in Reading, England. FibreCat uses a functional group on a polymer backbone in place of one of the ligands to bind the catalyst. Its Cataxa technology attaches the metal atom itself to an alumina support through a heteropolyacid. Unmodified chiral ligands can be substituted freely onto the supported metal precursor complex.

Methods vary for removing or possibly reducing metals

Heterogeneous catalysis
Immobilized homogeneous catalysts
Biphasic catalysis
Reorganized synthetic routes
Organocatalysis
Purification

• Adsorption on solid or bound trimercaptotriazine, bound ethylenediamine, activated carbon, glass-bead sponges, polymeric fibers, or silica-bound scavengers
• Crystallization using N-acetylcysteine, thiourea, hemi-maleate salt, or tributylphosphine
• Distillation
• Extraction using N-acetylcysteine, l-cysteine, or tributylphosphine and lactic acid

NOTE: Examples of some of the above methods used in synthesizing pharmaceutically relevant compounds appear in Adv. Synth. Catal. 2004, 346, 889.

Recently, Barnard and coworkers looked at an easy and scalable method for anchoring chiral rhodium diphosphine catalysts to unmodified carbon supports through simple sorption (Org. Process Res. Dev. 2005, 9, 164). Although the binding efficiency changed with the support, certain basic carbons were particularly successful and resulted in no leaching. "It's not a panacea, but under the right conditions, it is a very cost effective way to go," he says.

To test the system, they studied the hydrogenation of a prochiral C5C bond. The reactions with anchored catalysts showed similar activity to the homogeneous reactions and a marked increase in enantiomeric excess. The improvement, which has been reported with other immobilized systems, is believed to arise from more steric control over the substrate's interaction with the catalyst, Barnard says.

"Immobilized catalysts have to have enantioselectivity and activity at least the same or better than the homogeneous system, or there is just no point," says Graham J. Hutchings, professor of physical chemistry at Cardiff University, in Wales. "From what I can see, many in industry are quite happy with nonimmobilized systems and cleaning up afterwards."

Hutchings believes the stereoselective containment effects with immobilized systems are not being explored or exploited enough to improve enantioselectivity and yield recoverable and reusable catalysts. "We often find that chiral ligands that are not effective in homogeneous reactions suddenly become much more effective when immobilized," he explains. "Setting out to immobilize the best homogeneous catalyst is not the best way forward."

Instead, Hutchings advocates looking at “less successful” homogeneous catalysts and devising strategies to improve selectivity (Chem. Soc. Rev. 2004, 33, 108). He considers encapsulation methods elegant, but complex, whereas adsorption methods, while easy, tend to create nonstable catalysts. Immobilization by covalent tethering produces stable catalysts, he adds, and is a good option for catalysts that can't be tied down through electrostatic interactions.

Hutchings and collaborators at Johnson Matthey have reported good catalytic performance, ready recoverability, and no leaching when using simple ion-exchange methods to immobilize rhodium diphosphine complexes in a microporous material (Org. Biomol. Chem. 2005, 3, 1547). Catalyst complexes can be immobilized directly or formed by exchanging chiral ligands onto anchored [Rh(COD)2 ]+BF₄^-.

Other catalyst producers, such as Degussa and Solvias, cover the bases with homogeneous and immobilized or chirally modified heterogeneous catalysts, as well as metal scavengers or adsorbers. Together the companies developed catASium F214, a platinum-on-alumina catalyst in which the catalytic active surface is modified with cinchona alkaloids.

Homogeneous reactions may still be the best option. Immobilized catalysts can be very good, says Hans-Ulrich Blaser, chief technology officer of Solvias, but few are available commercially, and producing new high-performing ones is time-consuming. "When you have a new chemical entity, where you need a new process and the time window is quite small, you can't spend time screening homogeneous catalysts to hopefully find one and then immobilize it and not be guaranteed that you'll get the same performance."

For industrial applications, selectivity, activity, productivity, and cost considerations are all important factors. Despite the cost, Blaser says, pharmaceutical users especially do not want to reuse catalysts because even slight changes in performance can change the profile or stability of a process. "We try to optimize for once-through catalyst use," he explains, and removing trace metals will probably be an issue no matter what catalyst is used.

Thus, catalyst separation is seldom a reason for not using a catalytic process, he says. "It's the price you have to pay."