Novartis-MIT Center for Continuous Manufacturing

Crystallization is crucial in the pharmaceutical industry as a separation process for intermediates and as the final step in the manufacture of active pharmaceutical ingredients (APIs). In this perspective article to celebrate 10 years of Crystal Growth & Design, we focus on three areas related to crystallization in the pharmaceutical industry: (1) advances in our understanding of the fundamentals of nucleation, (2) production and scale-up of novel solid forms, and (3) continuous processing. While the areas discussed are not new, they are areas, in our opinion, of significant current interest to the community engaged in crystallization in the pharmaceutical industry.

The pharmaceutical industry is investing in continuous flow and high-throughput experimentation as tools for rapid process development accelerated scale-up. Coupled with automation, these technologies offer the potential for comprehensive reaction characterization and optimization, but with the cost of conducting exhaustive multifactor screens. Automated feedback in flow offers researchers an alternative strategy for efficient characterization of reactions based on the use of continuous technology to control chemical reaction conditions and optimize in lieu of screening. Optimization with feedback allows experiments to be conducted where the most information can be gained from the chemistry, enabling product yields to be maximized and kinetic models to be generated while the total number of experiments is minimized. This Account opens by reviewing select examples of feedback optimization in flow and applications to chemical research. Systems in the literature are classified into (i) deterministic "black box" optimization systems that do not model the reaction system and are therefore limited in the utility of results for scale-up, (ii) deterministic model-based optimization systems from which reaction kinetics and/or mechanisms can be automatically evaluated, and (iii) stochastic systems. Though diverse in application, flow feedback systems have predominantly focused upon the optimization of continuous variables, i.e., variables such as time, temperature, and concentration that can be ramped from one experiment to the next. Unfortunately, this implies that the screening of discrete variables such as catalyst, ligand, or solvent generally does not factor into automated flow optimization, resulting in incomplete process knowledge. Herein, we present a system and strategy developed for optimizing discrete and continuous variables of a chemical reaction simultaneously. The approach couples automated feedback with high-throughput reaction screening in droplet flow microfluidics. This Account details the system configuration for on-demand creation of sub-20 μL droplets with interchangeable reagents and catalysts. These droplets are reacted in a fully automated microfluidic system and analyzed online by LC/MS. Feeding back from the online analytical results, a design of experiments (DoE)-based adaptive response surface algorithm is employed that deductively removes candidate reagents from the optimization as optimal reaction conditions are refined, leading to rapid convergence. Using the automated optimization platform, case studies are presented for solvent selection in a competitive alkylation chemistry and for catalyst-ligand selection in heteroaromatic Suzuki-Miyaura cross-coupling chemistries. For the monoalkylation of trans-1,2-diaminocyclohexane, polar aprotic solvents at moderate temperatures are shown to be favorable, with optimality accurately identified with dimethyl sulfoxide as the solvent in 67 experiments. For Suzuki-Miyaura cross-couplings, the optimality of precatalysts and continuous variable conditions are observed to change in accordance with the coupling reagents, providing insights into catalyst behavior in the context of the reaction mechanism. Future opportunities in automated reaction development include the incorporation of chemoinformatics for faster analysis and machine-learning algorithms to guide and optimize the synthesis. Adoption of this technology stands to reduce graduate student and postdoc time on routine tasks in the laboratory, while feeding back knowledge used to guide new research directions. Moreover, the application of this technology in industry promises to lessen the cost and time associated with advancing pharmaceutical molecules through development and scale-up.

Set it and forget it: The combination of feedback control and continuous-flow operations in microreactors (see picture) enables online and fully automated reaction optimization. A Heck reaction demonstrates the potential for rapid multivariable reaction optimization while requiring a minimal amount of material. Optimal conditions are quickly scaled-up by a factor of 50. Detailed facts of importance to specialist readers are published as ”Supporting Information”. Such documents are peer-reviewed, but not copy-edited or typeset. They are made available as submitted by the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

An automated, droplet-flow microfluidic system explores and optimizes Pd-catalyzed Suzuki-Miyaura cross-coupling reactions. A smart optimal DoE-based algorithm is implemented to increase the turnover number and yield of the catalytic system considering both discrete variables-palladacycle and ligand-and continuous variables-temperature, time, and loading-simultaneously. The use of feedback allows for experiments to be run with catalysts and under conditions more likely to produce an optimum; consequently complex reaction optimizations are completed within 96 experiments. Response surfaces predicting reaction performance near the optima are generated and validated. From the screening results, shared attributes of successful precatalysts are identified, leading to improved understanding of the influence of ligand selection upon transmetalation and oxidative addition in the reaction mechanism. Dialkylbiarylphosphine, trialkylphosphine, and bidentate ligands are assessed.

Kinetic information is used to determine the optimal reaction conditions, to successfully scale up a reaction from the laboratory to the pilot plant, and to improve process control. Obtaining accurate kinetics using conventional benchtop equipment and techniques, however, requires numerous experiments and can be complicated by sluggish mixing and heat-transfer rates. To improve the speed and efficiency in gathering reaction kinetics, we present an automated, silicon microreactor system that uses a sequential experimentation framework driven by model-based optimization feedback for online reaction rate parameter determination. The method, based on Information Theory and Bayesian Statistics, first selects the appropriate global reaction rate expression. After determining the correct rate law, a D-optimal strategy precisely estimates the pre-exponential and activation energy of the rate constant. The approaches are validated experimentally with a model system, the Diels−Alder reaction of isoprene and maleic anhydride in DMF. The benefits of quickly obtaining this information with an automated microreactor system are further demonstrated by successfully scaling the Diels−Alder reaction by a factor of 500 from a microreactor to a Corning flow reactor.

An automated, continuous flow system for the online, multivariable optimization of a chemical reaction is presented. Time and material required for an optimization trial are minimized by performing reactions in an integrated silicon microreactor and incorporating an HPLC for inline monitoring of the reaction performance. We use the system to optimize two different reactions to describe the potential impact of this system for reaction development. First, we demonstrate the broad operation capabilities by incorporating several feedback algorithms to optimize a weighted objective function involving the yield and the throughput of a Knoevenagel condensation reaction. After illustrating how system operations can be adapted for individual reactions, we perform a multiparameter optimization to maximize the yield of benzaldehyde in the oxidation pathway of benzyl alcohol to benzaldehyde to benzoic acid. A significant feature of the automated system is the ability to perform “black-box” optimization where no a priori information of the reaction parameters is required.

Continuous flow chemistry is being used increasingly; however, without detailed knowledge of reaction engineering, it can be difficult to judge whether dispersion and mixing are important factors on reaction outcome. Understanding these effects can result in improved choices of reactor dimensions and give insight for reactor scale-up. We provide an overview of both dispersive and mixing effects in flow systems and present simple relationships for determining whether mixing or dispersion is important for a given flow system. These results are summarized in convenient charts to enable the experimentalist to identify conditions with potential mixing or dispersion problems. The information also expedites design changes, such as inclusion or changes of mixers and changes in reaction tube diameters. As a case study, application of the principles to a glycosylation reaction results in increased throughput and cleaner product profiles compared to previously reported results.

Although nanoporous materials have been explored for controlling crystallization of polymorphs in recent years, polymorphism in confined environments is still poorly understood, particularly from a kinetic perspective, and the role of the local structure of the substrate has largely been neglected. Herein, we report the use of a novel material, polymer microgels with tunable microstructure, for controlling polymorph crystallization from solution and for investigating systematically the effects of nanoconfinement and interfacial interactions on polymorphic outcomes. We show that the polymer microgels can improve polymorph selectivity significantly. The polymorphic outcomes correlate strongly with the gel-induced nucleation kinetics and are very sensitive to both the polymer microstructure and the chemical composition. Further mechanistic investigations suggest that the nucleation-templating effect and the spatial confinement imposed by the polymer network may be central to achieving polymorph selectivity. We demonstrate polymer microgels as promising materials for controlling crystal polymorphism. Moreover, our results help advance the fundamental understanding of polymorph crystallization at complex interfaces, particularly in confined environments.

An automated multitrajectory optimization platform with continuous online infrared (IR) monitoring is presented. The production rate of a Paal–Knorr reaction is maximized within a constrained temperature and residence time design space. The automated platform utilizes a microreactor system to carry out optimizations with low material requirements and implements a micro IR flow cell for continuous online monitoring of reaction conversion. The approach to steady state at each set of reaction conditions is assessed continuously before the objective function is evaluated and reactor conditions move to the next set point. Several optimization algorithms are tested for their performance on a complex objective terrain. Each function comes to agreement on the optimal conditions but requires a significantly different number of experiments to reach the final conditions. Additionally, multiple objective functions are compared to analyze the trade-off between production rate and conversion.

Crystallization processes can be batch or continuous. Potential advantages such as operating at steady state, small equipment size (relative to batch), and ability to recycle are encouraging the pharmaceutical industry to investigate continuous processes. In this work, a continuous cooling crystallization process for the immunosuppressant drug cyclosporine was developed. A multistage mixed suspension mixed product removal (MSMPR) crystallizer was employed which allowed simple analysis of kinetic parameters employing the population balance. Experimentally, the continuous crystallization system was able to operate without any clogging issues for more than four residence times. The experimental yield and purity of the crystals was determined as 71% and 96%, respectively (without recycle) and 87% and 94%, respectively (with recycle). In a batch cooling crystallization experiment, carried out under conditions similar to those of the continuous experiment without recycle, the experimental yield and purity of the crystals were 74% and 95%, respectively. The equilibrium distribution coefficients of cyclosporine impurities were measured experimentally as a function of impurity % of the starting solution. The distribution coefficients increase with a decrease in the purity of the starting solution, indicating a decrease in purification. The MSMPR model was used to estimate the nucleation and crystal growth rate kinetic parameters for cyclosporine crystallization and to evaluate the effect of process conditions on the purity of the crystals and the process yield. Results showed that the temperature of the third stage has a large impact on the final purity of the crystals. As the temperature of the third stage increases, the purity of the crystals also increases while the yield of the process decreases. The effect of recycle ratio on both crystal purity and process yield was also evaluated. A 93% process yield was obtained with a recycle ratio of 0.9. The yield of the process can be significantly improved by increasing the recycle ratio while the crystal purity decreases.

Active ingredients in most pharmaceutical products are complex organic molecules that require crystallization as a purification and isolation step that results in a pure product at a high process yield. Knowledge of the operating conditions required to obtain crystals with the desired crystal shape, polymorph, and morphology is critical during process development. This paper describes a two-stage mixed suspension mixed product removal (MSMPR) continuous reactive crystallization procedure developed for Aliskiren hemifumarate. This process was able to crystallize Aliskiren hemifumarate at both high purity (>99%) and high yield (>92%). A model of the crystallization was developed through the simultaneous solution of a population balance equation, kinetic expression for crystal growth and nucleation, and a mass balance. Experimental data were fit to the model to obtain kinetic parameters for crystal growth and nucleation. After including equilibrium distribution coefficient data, the model was used to optimize crystal purity and yield of the product by adjusting the operating temperature and residence time. This process has been integrated into an end-to-end continuous manufacturing system developed at MIT.

An ideal pharmaceutical crystallization process produces a pure product at a high yield while minimizing energy input, the process equipment footprint, and its complexity. A good candidate for such a process is a single-stage mixed-suspension, mixed-product removal (MSMPR) crystallizer with recycle (SMR) system, where the characteristics of the refined crystal are controlled by the crystallization conditions of the MSMPR and the yield is manipulated by the recycle ratio. In this study, two continuous SMR systems, for the cooling crystallization of cyclosporine and the antisolvent-cooling crystallization of deferasirox, were developed. Both systems were designed to maintain the desired operating conditions inside the MSMPR crystallizer. For cooling crystallization, the recycle stream was concentrated via vacuum evaporation. For antisolvent-cooling crystallization, the desired solvent to antisolvent ratio was maintained by controlling the flow rates of feed, antisolvent, and recycle streams. The maximum experimental yield and purity of the crystals were determined as 91.8% and 94.3%, respectively (for cyclosporine) and 89.1% with 0.2 ppm impurity A, respectively (for deferasirox). For cyclosporine, this yield is 5.5% higher than that of a multistage MSMPR with a recycle system. Additionally, the SMR system is relatively simple, having a lower operational demand, in terms of space and number of unit operations required.

An automated, continuous flow droplet screening system is presented, enabling real-time simultaneous solvent and continuous variable optimization. An optimal design of experiments strategy is applied to the alkylation of 1,2-diaminocyclohexane in 16 μL droplets, with scale-up demonstrated. Analysis of segmented flow results suggests correlation of yield with solvent hydrogen bond basicity.

Automated continuous flow systems coupled with online analysis and feedback have been previously demonstrated to model and optimize chemical syntheses with little a priori reaction information. However, these methods have yet to address the challenge of modeling and optimizing for product yield or selectivity in a multistep reaction network, where low selectivity toward desired product formation can be encountered. Here we demonstrate an automated system capable of rapidly estimating accurate kinetic parameters for a given reaction network using maximum likelihood estimation and a D-optimal design of experiments. The network studied is the series–parallel nucleophilic aromatic substitution of morpholine onto 2,4-dichloropyrimidine. To improve the precision of the estimated parameters, we demonstrate the use of the automated platform first in optimization of the yield of the less kinetically favorable 2-substituted product. Then, upon isolation of the intermediates, we use the automated system with maximum a posteriori estimation to minimize uncertainties in the network parameters. From considering the steps of the reaction network in isolation, the kinetic parameter uncertainties are reduced by 50%, with less than 5 g of the dichloropyrimidine substrate consumed over all experiments. We conclude that isolating pathways in the multistep reaction network is important to minimizing uncertainty for low sensitivity rate parameters.

If continuous processing is to be employed in pharmaceutical production, it is essential that continuous crystallization techniques can meet the purity and yield achievable in current batch crystallization processes. Recycling of mother liquor in steady state MSMPR crystallizations allows the yield in the equivalent equilibrium batch process to be met or exceeded. However, the extent to which yield can be increased is limited by the buildup of impurities within the system. In this study, an organic solvent nanofiltration membrane was used to preferentially concentrate an API (deferasirox, M.W. = 373 Da) and purge the limiting impurity 4-hydrazinobenzoic acid (MW = 152 Da) from the mother liquor recycle stream in a mixed solvent (THF:ethanol) antisolvent (water) system. Incorporation of the membrane recycle allowed yields of 98.0% and 98.7% to be achieved. This compares to the following: a control MSMPR run without a membrane (70.3%), an equivalent batch process (89.2%), and the current commercial batch process (92%). Comparable product impurity levels were measured for the following: the MSMPR membrane recycle experiments (0.15 ppm and 0.22 ppm), the MSMPR control (0.13 ppm), and batch (0.32 ppm) control experiments. All processes met the regulatory specifications of a maximum of 3 ppm of the impurity 4-hydrainobenzoic acid.

Ionic liquids have been proposed as functional replacements for harmful and hazardous volatile organic solvents. However, limiting their use in this way does not fully explore the potential chemical benefits of their solvating properties, which stem from the inherent differences between ionic liquids and single molecule solvents. These differences can be used to facilitate alternative and improved reaction outcomes. This review will highlight a range of examples, involving materials preparation and organic synthesis, in which substantial progress towards the understanding and targeted application of ionic liquids is demonstrated. In addition, a number of studies will be cited where unanticipated outcomes have been observed and the relationships between these outcomes and ionic liquid structural effects will be analysed, casting new light onto these studies.

Despite its widespread occurrence in nature and broad application in industrial practice, nucleation of crystalline materials remains largely unpredictable and therefore difficult to control. In this work, we demonstrate a new method to control nucleation with polymer microgels by tuning their microstructure to vary systematically the degree of nanoscopic confinement and its effects on nucleation. We find that the polymer microstructure has a significant impact on nucleation kinetics. Moreover, there exists an optimum polymer mesh size at which the rate of nucleation is dramatically enhanced, the degree to which depends on the extent of polymer-solute interactions. With easily tunable microstructure and chemistry, polymer microgels offer a promising approach for the rational design of materials for controlling nucleation from solution.

Continuous crystallization process has potential advantages such as lower cost and improved flexibility in pharmaceutical production when compared to batch crystallization. A good continuous crystallization process should achieve a high product yield and purity comparable to current batch crystallization processes. For compounds that have low growth rates, a high yield is difficult to achieve without long residence times. Solids recycle is a potential solution for this problem as it can increase the surface area of crystals in the crystallizer thus increasing the mass deposition rate. In this study, solids recycle was used in a two-stage continuous mixed-suspension, mixed-product removal (MSMPR) cooling crystallization. Manual solids recycle and the use of a designed column for automatic slurry concentration were employed. The crystallization of cyclosporine, which has very low growth rate (about 0.1 μm/min) at low temperatures in acetone, showed only 65.0% yield in a two-stage MSMPR without solids recycle. With solids recycle to the second stage and both stages, 75.3% and 79.8% in yield were achieved, respectively. The product purity remained the same, while the yield was enhanced. A population balance model was developed to estimate the final yield of continuous process with solids recycle. The simulation results showed that optimization in stage number, stage temperatures, and solids recycle ratios could improve the yield to 83.9% in four-stage MSMPR crystallization with solids recycle. This yield was close to the batch yield at equilibrium, i.e., 86.0%.

Biocompatible materials that can control crystallization while carrying large amounts of active pharmaceutical ingredients (APIs) with diverse chemical properties are in demand in industrial practice. In this study, we investigate the utility of biocompatible alginate (ALG) hydrogels as a rational material for crystallizing and encapsulating model APIs that present drastically different solubilities in water. Acetaminophen (ACM) and fenofibrate (FEN) are utilized as the model hydrophilic and hydrophobic moieties, respectively. ALG hydrogels with different ALG concentrations (hence different mesh sizes) are utilized as heteronucleants to control the nucleation kinetics of ACM from solution. ALG hydrogels with smaller mesh sizes showed faster nucleation kinetics. We hypothesize that this behavior is due to the interplay between the polymer–solute interactions and the mesh-induced confinement effects. The loading of ACM into hydrogels by equilibrium partitioning is quantified and found to be inversely proportional to ALG concentration. For hydrophobic model APIs, loading via equilibrium partitioning is inefficient; hence, we suggest emulsion-laden hydrogels where emulsion droplets are encapsulated inside the hydrogel matrix. The incorporation of emulsion droplets inside hydrogels enables the high loading of the hydrophobic API leveraging the high solubility of the hydrophobic API in the dispersed emulsion droplets. By carefully choosing the emulsification method and the dispersed phase, we demonstrate significant loading (up to ∼80% w/w) and crystallization of the stable form of FEN. Our results provide new insights for designing biocompatible nucleation-active materials capable of carrying industrially significant amounts of water-soluble and insoluble APIs in the crystalline form.

Small droplets of supersaturated hen egg white lysozyme (HEWL) solution were exposed to intense linearly polarized laser pulses with different wavelengths, intensities, and pulse durations. Laser irradiation under some conditions significantly increased the number of droplets in which crystals were observed in a given time period, compared with nonirradiated samples. As a general rule, nonphotochemical laser induced nucleation (NPLIN) in lysozyme solutions was more effective with shorter aging time, 532-nm wavelength, higher peak intensity, and shorter pulse duration. Bovine pancreatic trypsin (BPT) was also examined using NPLIN, showing the potential application of NPLIN to other proteins.