(173f) Development of a Novel Continuous Enzymatic Reactive Crystallization Process for the Production of Amoxicillin Trihydrate on a Pilot Scale

Beta-lactam antibiotics are highly important drugs which have been used to treat numerous bacterial infections worldwide for eight decades. Their synthesis remains a focal point of many research groups due to their extraordinarily high global demand and subsequent shortages. Recently, the production of beta-lactam antibiotics, such as amoxicillin, has been adapted to a biocatalytic process to improve the sustainability and economic viability of the process. Amoxicillin, the largest antibiotic consumed worldwide (~30,000 tons/yr), is synthesized using immobilized penicillin G acylase (PGA) from E. coli in a batchwise process; afterwards, amoxicillin is crystallized and isolated from an aqueous solution using a pH swing crystallization. One disadvantage of the biocatalytic process is the product hydrolysis that PGA catalyzes in parallel to the synthesis reaction (Figure 1A). To counteract product hydrolysis, reactive crystallization may be used as a form of in situ product removal, protecting the product from enzymatic hydrolysis. While the reactive crystallization of amoxicillin by combining 6-aminopenicillinoic acid (6-APA) and D-4-hydroxyphenylglycine methyl ester (D-4-HPGME) has been demonstrated on a small batch scale, it has never been adapted to continuous processing, which is particularly advantageous due to the much higher volumetric productivity achieved. In this work, we demonstrate the development of a novel continuous process for the enzymatic reactive crystallization of amoxicillin trihydrate and the challenges met and addressed throughout that development.

The first challenge to address with adapting the system to a continuous mode of operation was the recycling of the biocatalyst. After deciding to move forward with an immobilized enzyme platform due to improved stability and ease in separating from the liquid phase, the challenge of a solid-solid separation was apparent. We designed a size-based separator which utilizes a stainless-steel mesh sieve designed to allow product crystals to pass through, but smaller than all biocatalyst support particles so that only product slurry would pass through the separator and into the product stream. Optimization of sieve size, product withdrawal rate, wet milling intensity, and support size were conducted to enhance separator and biocatalyst performance, with an emphasis placed on ensuring isokinetic product withdrawal and higher enzyme selectivity. It was determined that a 300 μm sieve size was large enough to ensure isokinetic product withdrawal and may be combined with 425-500 μm enzyme support particles which exhibit acceptable selectivity and activity. Additionally, the continuous process was modeled using a previously developed process model to identify potential operating conditions to avoid substrate or product precipitation while also yielding modest productivity and conversion. The continuous process was demonstrated on a pilot-scale (Figure 1B), where it was robustly operated for approximately 50 residence times without shutdown. HPLC was used for liquid-phase monitoring while focused-beam reflectance measurement (FBRM) and in situ microscopy were used for the solid phase. All product isolated was tested rigorously against USP standards and was of high quality. Turbidity, an output of the in situ microscope, appeared to be useful in detected disturbances such as clogs, while HPLC was useful in detecting changes in effective enzyme concentration and residence time. Additionally, disturbances to pH control and wet milling were intentionally generated during steady state operation to determine their impact on process stability and the rate of progress towards process upsets after introduction of the respective disturbances. It was determined that wet milling is critical to mitigate clogging and encrustation in tubing while a lack of pH control results in the accumulation of substrates, due to lowered enzyme activity, which may then precipitate and contaminate the product. Additionally, the selective recrystallization and recycle of unreacted substrates from mother liquor was investigated, which improved the conversion of both substrates to 85 and 95% for 6-APA and D-4-HPGME, respectively. Optimization of both recovery steps was performed to improve recycle yield and overall conversion. The overall schematic of the process can be seen in below in Figure 1C. Overall, we successfully demonstrated the development of a novel continuous pharmaceutical production process from concept to operation on a pilot scale.