From A Teaspoon To A Ton

If bringing a new pharmaceutical to market were a relay race, process chemists would be the ones running the oft-overlooked middle legs. There’s no roar from the crowd for them as they set off from the starting line, and they don’t get the glory of crossing the finish. But it’s those middle legs where races are won and lost, and if the process chemists can’t keep pace, there’s no way a drug will make it to the finish line.

“People don’t really know what it takes to do a synthesis scale-up and produce material on large scale,” said Ahmed Abdel-Magid, a longtime process chemist who is currently the chief scientific officer of the contract research organization Therachem Research Medilab. “We tackle everything.”

As process chemists tell it, if bringing a new pharmaceutical to market were really a relay race, not only would they run the race’s middle legs, they would also ensure the track was safe for the anchor runner; they would make certain that the race could be safely enjoyed by all; they would minimize the race’s environmental impact; and they’d be responsible for keeping the overall cost of race as low as possible.

“This is the ultimate in chemistry,” said Abdel-Magid, summing up his fondness for process research. “You’ve reached the stage where it’s becoming a project that people can spend millions of dollars on. The end result, if it succeeds and makes it all the way to the market, is that it will benefit people.”

To shed light on what precisely a process chemist does, Abdel-Magid started a series of process chemistry symposia at American Chemical Society national meetings in 2001. There have been five such symposia, the most recent of which was held during last month’s meeting in Boston, with one half-day session each in the Divisions of Medicinal Chemistry and of Organic Chemistry.

Medicinal chemists work to quickly develop a synthetic route that leads to multiple analogs, explained Jaan Pesti, a principal scientist in process chemistry at Bristol-Myers Squibb and one of the recent symposia’s organizers. Process chemists, on the other hand, “need to figure out how to make a single molecule very, very well,” he said.

“The fundamental thing is just getting your chemical transformations right, first and foremost, but after that, there are many aspects to the job that you have to consider,” added Robert Singer, a process chemist at Pfizer and another of the symposia’s organizers.

“We have to make sure that the product is safe for consumers. That’s obvious,” Singer said, noting that process chemists must also assure regulators that products are safe. Avoiding reagents or intermediates that might be potential mutagens or carcinogens is a priority, he said, but if potentially genotoxic materials must be used, they’re employed early in the route so they can be purged easily.

“A big part of process development from a business perspective is lowering costs,” Singer added. That includes optimizing yields, reducing excess amounts of reagents used, avoiding costly manufacturing processes whenever possible, and sidestepping reactions that generate expensive waste streams. “Process chemists need to use their knowledge of organic chemistry to solve this complex puzzle of meeting the needs of cost, safety, and regulatory challenges,” he said.

“The ability to take something that was only run on small scale and bring it to large scale in a reasonable amount of time, keeping in mind the factors of cost, safety, and intellectual property is quite a challenging proposition,” Pesti added. “We are always aware that even a difference of 10 days to approve a new drug could easily mean the difference in many lives.

“Without the ability to make a compound on a large scale, it’s essentially no more than a laboratory curiosity and it’s not going to be of any great benefit to the public,” Pesti continued. “The process chemist will take a synthesis that may have been done at 100-g scale and find out how to make it work on a metric ton scale.”

In mastering how to take a teaspoon of material and turn it into a ton, process chemists use the discovery group’s synthesis as a starting point. They learn its ins and outs and discover its drawbacks. The process chemist edits, revises, and often completely revamps the synthetic route altogether.

“Early in development, our job is to make material,” said Jade Nelson, an associate research fellow in process chemistry at Pfizer. “We start out enabling the manufacture of increasing quantities of supplies, which allows us to get the information we need to either terminate development if something is not going to be a product or to accelerate it to the market.”

At some point, Nelson said, process chemistry shifts to scaling the synthesis for manufacturing. “Our primary target is to accomplish something we can put into a manufacturing plant and potentially make tons of material.”

Nelson pointed to the process development of the β-lactam antibacterial candidate sulopenem as an example of how they retooled a synthesis for scale-up. Pfizer first explored sulopenem as an antibiotic in the 1980s. Its efficacy against gram-negative and anaerobic bacteria made it a promising candidate for treating pneumonia as well as complicated urinary tract and intra-abdominal infections.

The project was shelved in the 1990s but revived in 2003, with two targets: sulopenem, which would be used as an intravenous antibiotic, and sulopenem etzadroxil, an ester prodrug that could be taken orally. The core molecule contains three basic units, which are assembled during the synthesis. First is the acetoxyazetidinone, a β-lactam building block that’s commercially available in bulk quantities. That’s a huge benefit, Nelson pointed out, because it supplies three of the compound’s five chiral centers.

The other two chiral centers can be found in the chiral sulfoxide, a moiety that differentiates sulopenem from other β-lactam or thiopenem derivatives. Those two chiral centers make the synthesis of the chiral sulfoxide costly. Oxalic acid derivatives make up the rest of the molecule’s core. And in the case of the prodrug, there’s also the ester moiety.

Although the process team used the discovery team’s route to make enough of the drug to carry out early clinical studies, Nelson said, “we knew as process chemists this was not something we would continue to use in a commercial setting. Almost every step has a real drawback associated with it.”

To begin with, the two most expensive building blocks are put together in the first step. By design that makes a synthesis more expensive because each subsequent step that doesn’t give 100% yield means a loss on the initial investment.

The next step, N-acylation of the β-lactam nitrogen with an oxalic acid derivative, had to be carried out at –50 °C. “Typical bulk reactors have a low-temperature limit of –20 °C,” Nelson explained. Go any lower and you’ll need specialized equipment. Also, to prevent acylation of the sulfoxide oxygen, the oxalic acid derivative that the discovery team used was an expensive acid fluoride that gave off hazardous HF.

The discovery team built the thiopenem ring using chloroform as a solvent. “That’s a standard no-no in process chemistry,” Nelson said. “Chloroform is toxic.” And a back-of-the-envelope calculation indicated that they’d need roughly 50,000 L of chloroform to get 150 kg of sulopenem. “So this was one of those reactions we knew we didn’t want to go forward with for long,” he noted.

The penultimate step, a palladium-catalyzed deprotection to give sulopenem, meant that the final product would need to be recrystallized to ensure that no palladium was getting into a drug that would be delivered intravenously. The complex recrystallization process took a big bite out of the yield, Nelson said.

Finally, the prodrug was prepared via an alkylation reaction. “The use of an alkylating agent in the last step of a synthesis is never a good idea because alkylating agents are by default compounds of concern. They can alkylate DNA, for instance, so the limits of acceptable levels are extremely low,” Nelson noted.

“Those were the things we needed to address if we were going to go forward in development and clearly before we were to put this in a commercial facility,” he continued. Ultimately, his team came up with a revamped synthesis that got around those obstacles.

To begin, they put off incorporating the pricey chiral sulfoxide until almost the end of the synthesis. By doing so, they were able to complete the N-acylation with an acid chloride instead of the acid fluoride. Not only was this reagent less expensive, but it also eliminated any hazardous HF by-products.

With the new intermediate, the team was then able to get the thiopenem ring to form in toluene, as opposed to chloroform, eliminating the major stumbling block of the discovery synthesis. They were then able to add the pricey chiral sulfoxide.

Finally, the group retooled its synthetic “endgame,” shuffling the deprotections and alkylation so that removal of the silyl protecting group was always the final step, rather than the Pd deprotection or the alkylation.

“In a lot of ways, process chemistry is common sense,” Nelson said. “It just takes a lot of work to get us to the end point.”

A process chemist approaches organic chemistry with an eye toward practicality, said David Mitchell, a process chemist with Eli Lilly & Co. “In discovery, you want to get to a target as quickly as you can so you can test it,” he said. Minimizing costs and maximizing yields aren’t necessarily a priority in those early stages of drug discovery, so chemists might take more steps en route to target simply because the steps may be well known.

Developing individual reactions at those early stages is pretty unusual, Mitchell explained. “It doesn’t make much sense to put a lot of effort into preparing a target when there’s a chance it might not work clinically,” he said. Once a molecule has shown enough promise to move on to the process stage, he continued, that’s when chemists revise the synthesis, keeping costs, regulations, and waste management in mind. “In process chemistry, where we’re playing for keeps, so to speak, you start taking a lot of the practical aspects into consideration.”

Mitchell’s colleagues recently handed him a target that featured a biaryl moiety in the structure. The original methodology used in the discovery of the compound took six steps to prepare the biaryl unit—a process that took seven days to complete.

“As process chemists, we wondered, ‘Can we get this down to one step?’ ” Mitchell said. “When we had time to sit back and say, ‘Okay, how are we going to prepare this molecule on a larger scale?’ we looked at the literature and thought about developing a decarboxylative cross-coupling.”

Making biaryl moieties via cross-coupling reactions, such as the Suzuki-Miyaura cross-coupling, is popular in organic chemistry. But the starting materials, such as aryl halides or aryl boronic acids, aren’t always readily available, particularly on the large scales used in process work.

That’s why Mitchell said he turned to decarboxylative cross-coupling to make the biaryl. “It’s a methodology that was out there, but it hadn’t been developed or used that readily,” he explained. The major advantage with this route was that the aryl carboxylate starting materials could be purchased.

Mitchell and his coworkers took the known procedure and carefully varied its conditions to find the optimal palladium catalyst, solvent, amount of base, and reaction times and temperatures that gave the best yields.

Understanding the mechanism, Mitchell said, was also key. Although the reaction probably follows the same course as that of other palladium-catalyzed cross-coupling, there was, he noted, one critical difference for the decarboxylative reaction. “In our case the question is: How does this decarboxylation take place? In process chemistry you need to be mindful of the pathway of the reaction,” Mitchell explained. “You don’t want to have CO₂ gushing out, so you need to make sure it’s a controlled process.”

Along with his coworkers, Mitchell studied the evolution of CO₂ with respect to the formation of the coupled product. They noted that generation of the product and generation of CO₂ run parallel to one another, with the coupled product appearing before the CO₂ is released. “It’s a true catalytic process taking place,” Mitchell pointed out. “The decarboxylation step and product formation is a concomitant process.” Furthermore, CO₂ evolves gradually during the reaction, so the process is indeed controlled.

Mitchell rounded out the study by testing the cross-coupling with several different aryl carboxylic acids and found that many of them worked well. “This methodology is a nice supplement to the toolbox of organic chemistry that could enable, hopefully, a more practical approach to biaryl formation,” Mitchell said. He’s already used it to tackle two other molecules with biaryl functionality.

Susanne Kiau, a principal scientist in process research development at Bristol-Myers Squibb, was also recently looking to trim down a reaction scheme en route to the compound MIV-170/BMS-801585-01. The chief challenge in making this molecule, she said, is its 1,2,3-cis-substituted cyclopropyl moiety, which makes one side of the three-membered ring quite crowded.

“It is very difficult to build this molecule by traditional cyclopropanation methods,” Kiau said, because those methods all lead to trans substitution. “The only established way to build this kind of molecule in the literature is to tether the carbene to the alkene. Through strain in the transition state, you can guide the substituent to the cis position.”

That’s the route the discovery group took, and although that route was suitable for synthesizing as much as 5 kg of the material, it had some key problems when it came to scaling up, Kiau noted.

First, it was relatively lengthy, taking eight steps overall to reach the all-cis-substituted carboxylic acid intermediate. The route also required some sophisticated separation as well as cryogenic conditions for one step. “When you go to scale up, you’ll find that everything is much slower,” Kiau explained. Cooling and heating operations take much longer, she said. “There’s really no way around it.”

The final nail in the tethered route’s coffin was an unstable diazo intermediate. During preparation of small batches for preclinical and early clinical supplies of the compound, decomposition of this intermediate was only a minor drawback, but it would become a big problem during the longer reaction times needed for scale-up.

“We knew there was a potential that we’d need to deliver much more of the compound—on the 50-kg scale or higher—very soon,” Kiau said, so the process team was called in. “Our challenge was to find a route that was shorter, more efficient, and amenable to scale-up.”

Kiau and her colleagues decided to eliminate the tethered approach. Although they found very few examples of making all-cis-trisubstituted cyclopropanes by an intermolecular approach in the literature, they decided to see whether they could figure out a way to overcome the problem that most cyclopropanations would result in the less hindered trans isomer.

“In our process group, we’re always striving to understand our reactions and their mechanisms to be able to judge whether something is feasible and scalable based on first principles,” Kiau pointed out. With this approach, Kiau and her colleagues came up with three successful options, all of which use a chromene as starting material.

The first two—which Kiau refered to as the diastereoselective dehalogenation approach and the diastereoselective decarboxylation approach—use the same strategy: Instead of installing one substituent in the carbene addition, install two and then remove the less hindered one in the trans position. “It’s making lemons into lemonade,” she said. “Bring in two substituents and remove the one you don’t want.”

The method worked well to prepare a geminal dibromocyclopropane wherein the trans bromine could be reduced. Although this method initially looked promising, the team would also have had to figure out a way to do the cyclopropanation reaction enantioselectively, so they turned to a slightly different approach.

With the diastereoselective decarboxylation approach, a geminal diester cyclopropane is prepared, and the ester in the trans position is diastereoselectively hydrolyzed and then decarboxylated. Kiau, however, noted one hang-up with this approach: Epimerization during decarboxylation destroyed the diastereoselectivity.

After tinkering with the decarboxylation conditions, Kiau and her colleagues decided to use Krapcho decarboxylation conditions employing a sterically hindered phenol as their hydrogen source. “As long as I can ensure that the hydrogen is coming in from the less hindered, trans side, I still get the cis isomer,” Kiau explained. Not only did this provide a scalable route to the trisubstituted cis isomer, she added, but it also expanded the chemistry of Krapcho decarboxylations.

The third route employed by Kiau’s team took advantage of recent reports of making cis-disubstitued cyclopropanes with the gold-catalyzed reaction of alkenes through vinylcarbenoids derived from propargyl esters. “Nobody had really made use of this approach to build 1,2,3-cis-trisubstituted cyclopropanes,” she said.

The chemistry worked, but it gave the desired material in only a meager 17% yield, and it required four equivalents of the chromene starting material. “Our chromene is very valuable. We don’t want to use it in excess. Our first step was to demonstrate that we could use it as the limiting reagent,” Kiau explained.

The researchers discovered that the propargylic ester reagent was decomposing in the course of the reaction. “If we constantly add propargylic ester, we can overcome this,” Kiau said. While screening conditions, they also found that concentration was important to the rate of reaction. “The faster the rate of reaction, the less competitive this decomposition is.” Armed with this information, they were able to boost their yield to 64%.

“This work was done within just four months,” Kiau pointed out. “And we not only solved our problem, but through our work we could contribute to the chemistry community.”

“In development these days there’s a lot of pressure to not make changes to routes,” said Karl Hansen, a scientific director in process research at Amgen. “But we’ve found when you take the time to do this and really explore the science, it pays back in dividends.”

Hansen’s team learned this lesson firsthand while developing the chemistry en route to AMG 221, a potential therapy for type 2 diabetes that targets the enzyme 11β-hydroxysteroid dehydrogenase type 1 (C&EN, April 27, 2009, page 31).

“It’s a pretty challenging structure,” Hansen pointed out. “It’s got two stereogenic centers, which are not trivial to prepare,” an exo-norbornylamine, and a tertiary chiral center in the thiazolone portion of the molecule. “The process group got involved pretty early on in trying to identify ways to prepare this compound in clinically enabling amounts,” he said.

The initial route started with commercially available norbornyl isothiocyanate, which was treated with ammonia to get racemic thiourea. Using chiral chromatography, the chemists got the desired configuration of the norbornyl moiety, to which they added the isopropyl thiazolone. Finally, they alkylated the compound with methyl iodide and again subjected their material to chiral chromatography to get AMG 221.

“We recognized that for the quantities we were going to need later on, we might want to come up with something other than chiral chromatography to power the synthesis,” Hansen said.

Their first revision to the synthesis was to make the thiourea building block in an enantioenriched fashion. “We employed an age-old technique of doing classical resolution of a compound,” Hansen explained. They took norbornene and did a Ritter reaction to get norbornyl amine, which gave them a compound they could easily resolve by making it into a diastereomeric salt that could be easily purified via recrystallization.

“Then the big question became, from a chemistry standpoint, how were we going to effect this chiral alkylation?” Hansen said. When looking at Koga chiral bases for the enantioselective alkylation, Hansen’s team observed that when two Koga bases were tethered together with a propyl group, the chemists saw remarkable diastereoselectivity.

The alkylation wasn’t without its drawbacks, however. The chiral base wasn’t commercially available in the large amounts they needed. Also, methyl iodide proved to be a fairly indiscriminate electrophile, often alkylating the nitrogens on both the base and the substrate.

To tackle the first problem, the process team developed a route to make the large quantities of the base. To address the methyl iodide problem, they decided to swap intermediates: Rather than methylate an isopropyl thiazolone, they’d use isopropyl iodide to alkylate a methyl thiazolone. Surprisingly, the reaction worked with only a modest drop in yield. “It’s pretty rare that you make a change like this in organic chemistry and it works out this well,” Hansen pointed out.

With this first-pass revised synthesis, the process group was able to prepare 40 kg of AMG 221. But, Hansen said, “there were a number of things with our procedure that left room for improvement.”

Their principal problem was the large amount of chiral base needed for the alkylation. More than two equivalents were required to get high stereoselectivity, and the base actually weighed more than AMG 221. For every kilogram of active pharmaceutical ingredient (API) they made, they’d need between 5 and 6 kg of chiral base. “That’s a lot of material to have to move around,” Hansen noted.

In addition, getting high enantioselectivity of the norbornyl thiourea ended up being pretty expensive in the synthesis as designed, he said. “If we had a better way to make that, we could afford to get lower selectivity” in the final alkylation.

So the team took another look at the norbornyl thiourea, and it turns out that chiral chromatography resolves the enantiomers beautifully. Furthermore, they could prepare the compound in one step from norbornene using thiocyanate in a Ritter reaction. Realizing that this was the best way to make this chiral intermediate was a bit of a blow to his process chemist’s pride, Hansen joked. “They tell you, ‘Never do chromatography; that’s a sign of weakness.’ ”

With a cheaper route to the chiral thiourea, the process group reasoned that they could lose some of the diastereoselectivity in the alkylation step if it meant they could use a cheaper amine catalyst. After extensive screening, they discovered that ephredrine worked well in the reaction, providing a cheap catalyst that also gave them greater flexibility with the amount and type of solvent they used.

Hansen said the process group is currently looking into a fourth route to AMG 221 that involves preparing the two chiral building blocks separately and simply coupling them (J. Org. Chem. 2009, 74, 3833). “I actually think that it’s a better way to make the molecule overall,” he added, but noted that it hasn’t been scaled up yet.

“The chemistry is still really important, and we are continually going back and asking ourselves, is this where we need to be? Is this the best we can do?” Hansen said of his team’s philosophy of process development. “It’s rewarding for the scientists, but I think the programs benefit from it too. They benefit from holding ourselves to a high standard and trying to do the best chemistry we can.”

“The process chemist’s role is becoming extremely vital in the development of drug candidates,” noted Chris H. Senanayake, Boehringer Ingelheim’s vice president for chemical development. He pointed to the scaled-up synthesis of the cathepsin S inhibitor BI 9001, an API for treating rheumatoid arthritis. He and his colleagues in process research had to quickly develop a synthetic route to the compound, coming up with some new chemistry in the process.

The researchers had intended to build a phosphonate intermediate by using a Michaelis-Arbuzov reaction. Their plan was to convert an alcohol into a chloride and then react the chloro intermediate with triethyl phosphate, as is done in a traditional Michaelis-Arbuzov transformation. But the intermediate chloride proved to be too unstable, forming dimeric impurities with the starting materials. The process team discovered that it could modify the traditional Michaelis-Arbuzov phosphite into a chlorophosphite and couple the chlorophosphite directly with the alcohol. The reaction “may well be a novel transformation,” they note in a recent publication (J. Org. Chem. 2010, 75, 1155).

The team also had to look for novel reagents for an asymmetric hydrogenation in the reaction. Although several proprietary catalysts are available for this reaction, the cost to use them can dramatically drive up the overall cost of a process, according to Senanayake. “In chemical production, per-year costs can easily exceed $1 million for proprietary chiral catalysts,” he said. Because of this, Boeringer Ingelheim has developed a series of in-house ligands for asymmetric hydrogenation.

For the final peptide coupling in the synthesis, the process chemistry team had initially used a procedure that called for ethyl-(N´,N´- dimethylamino)-propylcarbodiimide hydrochloride and 1-hydroxybenzotriazole (HOBt). But HOBt and related benzotriazoles are classified as explosives in Europe, so the researchers had to find an alternative reagent. After extensive screening they discovered that 2-hydroxypyridine was a robust replacement that led to no racemization during the peptide coupling.

“Developing a cost-effective, safe, and environmentally sound synthesis is the hard part of making a target,” Senanayake said. “Anybody can make a ton of API. How you make that ton is the most important thing.” ■