Raman spectroscopic based chemometric models to support a dynamic capacitance based cell culture feeding strategy.

Multiple process analytical technology (PAT) tools are now being applied in tandem for cell culture. Research presented used two in-line probes, capacitance for a dynamic feeding strategy and Raman spectroscopy for real-time monitoring. Data collected from eight batches at the 15,000 L scale were used to develop process models. Raman spectroscopic data were modelled using Partial Least Squares (PLS) by two methods-(1) use of the full dataset and (2) split the dataset based on the capacitance feeding strategy. Root mean square error of prediction (RMSEP) for the first model method of capacitance was 1.54 pf/cm and the second modelling method was 1.40 pf/cm. The second Raman method demonstrated results within expected process limits for capacitance and a 0.01% difference in total nutrient feed compared to the capacitance probe. Additional variables modelled using Raman spectroscopy were viable cell density (VCD), viability, average cell diameter, and viable cell volume (VCV).

Industrial application of heat- and mass balance model for fluid-bed granulation for technology transfer and design space exploration.

This work demonstrates the application of state-of-the-art modeling techniques in pharmaceutical manufacturing for fluid bed granulation at varying scales to successfully predict process conditions and ultimately replace experiments during a technology transfer of five products. We describe a mathematical model able to simulate the time-dependent moisture profile in a fluid bed granulation process. The applicability of this model is then demonstrated by calibrating and validating it over a range of operating conditions, manufacturing scales, and formulations. The inherent capability of the moisture profile to serve as a simple, scale-independent surrogate is shown by the large number of successful scale-ups and transfers. A methodology to use this 'digital twin' to systematically explore the effects of uncertainty inherent in the process input and model parameter spaces and their impact on the process outputs is described. Two case studies exemplifying the utilization of the model in industrial practice to assess process robustness are provided. Lastly, a pathway to leverage model results to establish proven acceptable ranges for individual parameters is outlined.

Biotechnology Based Process for Production of a Disulfide-Bridged Peptide.

A disulfide-bridged peptide drug development candidate contained two oligopeptide chains with 11 and 12 natural amino acids joined by a disulfide bond at the N-terminal end. An efficient biotechnology based process for the production of the disulfide-bridged peptide was developed. Initially, the two individual oligopeptide chains were prepared separately by designing different fusion proteins and expressing them in recombinant E. coli. Enzymatic or chemical cleavage of the two fusion proteins provided the two individual oligopeptide chains which could be conjugated via disulfide bond by conventional chemical reaction to the disulfide-bridged peptide. A novel heterodimeric system to bring the two oligopeptide chains closer and induce disulfide bond formation was designed by taking advantage of the self-assembly of a leucine zipper system. The heterodimeric approach involved designing fusion proteins with the acidic and basic components of the leucine zipper, additional amino acids to optimize interaction between the individual chains, specific cleavage sites, specific tag to ensure separation, and two individual oligopeptide chains. Computer modeling was used to identify the nature and number of amino acid residue to be inserted between the leucine zipper and oligopeptides for optimum interaction. Cloning and expression in rec E. coli, fermentation, followed by cell disruption resulted in the formation of heterodimeric protein with the interchain disulfide bond. Separation of the desired heterodimeric protein, followed by specific cleavage at methionine by cyanogen bromide provided the disulfide-bridged peptide.

Modeling the oxidation of methionine residues by peroxides in proteins.

We report the use of molecular modeling to predict the oxidation propensity of methionine residues in proteins. Oxidation of methionine to the sulfoxide form is one of the major degradation pathways for therapeutic proteins. Oxidation can occur during production, formulation, or storage of pharmaceuticals and it often reduces or eliminates biological activity. We use a molecular model based on atomistic simulations called 2-shell water coordination number to predict the oxidation rates for several model proteins and therapeutic candidates. In addition, we implement models that are based on static and simulation average of the solvent-accessible area (SAA) for either the side chain or the sulfur atom in the methionine residue. We then compare the results from the different models against the experimentally measured relative rates of methionine oxidation. We find that both the 2-shell model and the simulation-averaged SAA models are accurate in predicting the oxidation propensity of methionine residues for the proteins tested. We also find the appropriate parameter ranges where the models are most accurate. These models have significant predictive power and can be used to enable further protein engineering or to guide formulation approaches in stabilizing the unstable methionine residues.

Modeling of pan coating processes: Prediction of tablet content uniformity and determination of critical process parameters.

We developed an engineering model for predicting the active pharmaceutical ingredient (API) content uniformity (CU) for a drug product in which the active is coated onto a core. The model is based on a two-zone mechanistic description of the spray coating process in a perforated coating pan. The relative standard deviation (RSD) of the API CU of the coated tablets was found to be inversely proportional to the square root of the total number of cycles between the spray zone and drying zone that the tablets undergo. The total number of cycles is a function of the number of tablets in the drying zone, the spray zone width, the tablet velocity, the tablet number density, and the total coating time. The sensitivity of the RSD to various critical coating process parameters, such as pan speed, pan load, spray zone width, as well as tablet size and shape was evaluated. Consequently, the critical coating process parameters needed to achieve the desired API CU were determined. Several active film coating experiments at 50, 200, and 400 kg using various pan coaters demonstrated that good correlation between the model predictions and the experimental results for the API CU was achieved.

Experimental and model-based approaches to studying mixing in coating pans.

The focus of this study was the determination of mixing patterns and rates inside a cylindrical coating pan. The research for this study was divided into two parts. The first part examined the mixing pattern and the movement of tablets inside of a coating pan experimentally. The second part consisted of using a DEM (Discrete Element Model) simulation to evaluate mixing in the coating pan in silico. Mixing was investigated as a function of the rate of rotation of the pan and the number of revolutions. Mixing rates were measured in two directions--axial--from the front of the unit to the back of the unit along its axis and radial/angular--in the plane orthogonal to its axis. Radial/angular mixing was faster than axial mixing--the coating pan was found to be well-mixed across the axis within 2-8 revolutions as compared to 16-32 revolutions needed for the pan to be well-mixed along the axis. The DEM simulation used for this study is capable of predicting how fast the tablets mix in the coating pan. It does so by explicitly modeling the motion of individual tablets in the unit. Model predictions were verified by comparing the simulated mixing in the coating pan to the experiments. The simulated mixing process is found to be slightly slower than the experimentally observed mixing, which means that the simulations give a conservative estimate of mixing rates. The model can also be used to calculate the residence time distribution of the tablets in a spray zone of a given area.

Permeability and reactivity of Thermotoga maritima in latex bimodal blend coatings at 80 degrees C: a model high temperature biocatalytic coating.

Thermostable polymers cast as thin, porous coatings or membranes may be useful for concentrating and stabilizing hyperthermophilic microorganisms as biocatalysts. Hydrogel matrices can be unstable above 65 degrees C. Therefore a 55-microm thick, two layer (cell coat + polymer top coat) bimodal, adhesive latex coating of partially coalesced polystyrene particles was investigated at 80 degrees C using Thermotoga maritima as a model hyperthermophile. Coating permeability (pore structure) was critical for maintaining T. maritima viability. The permeability of bimodal coatings generated from 0.8 v/v of a suspension of non-film-forming 800 nm polystyrene particles with high glass transition temperature (T(g) = 94 degrees C, 26.9% total solids) blended with 0.2 v/v of a suspension of film-forming 158 nm polyacrylate/styrene particles (T(g) approximately -5 degrees C, 40.9% total solids) with 0.3 g sucrose/g latex was measured in a KNO3 diffusion cell. Diffusivity ratio remained above 0.04 (D(eff)/D) when incubated at 80 degrees C in artificial seawater (ASW) for 5 days. KNO3 permeability was corroborated by cryogenic-SEM images of the pore structure. In contrast, the permeability of a mono-dispersed acrylate/vinyl acetate latex Rovace SF091 (T(g) approximately 10 degrees C) rapidly decreased and became impermeable after 2 days incubation in ASW at 80 degrees C. Thermotoga maritima were entrapped in these coatings at a cell density of 49 g cell wet weight/liter of coating volume, 25-fold higher than the density in liquid culture. Viable T. maritima were released from single-layer coatings at 80 degrees C but accurate measurement of the percentage of viable entrapped cells by plate counting was not successful. Metabolic activity could be measured in bilayer coatings by utilization of glucose and maltose, which was identical for latex-entrapped and suspended cells. Starch was hydrolyzed for 200 h by latex-entrapped cells due to the slow diffusion of starch through the polymer top coat compared to only 24 h by suspended T. maritima. The observed reactivity and stability of these coatings was surprising since cryo-SEM images suggested that the smaller low T(g) polyacrylate/styrene particles preferentially bound to the T. maritima toga-sheath during coat formation. This model system may be useful for concentrating, entrapment and stabilization of metabolically active hyperthermophiles at 80 degrees C.