Process performance models in the optimization of multiproduct protein production plants

In this work we propose a model that simultaneously optimizes the process variables and the structure of a multiproduct batch plant for the production of recombinant proteins. The complete model includes process performance models for the unit stages and a posynomial representation for the multiproduct batch plant. Although the constant time and size factor models are the most commonly used to model multiproduct batch processes, process performance models describe these time and size factors as functions of the process variables selected for optimization. These process performance models are expressed as algebraic equations obtained from the analytical integration of simplified mass balances and kinetic expressions that describe each unit operation. They are kept as simple as possible while retaining the influence of the process variables selected to optimize the plant. The resulting mixed-integer nonlinear program simultaneously calculates the plant structure (parallel units in or out of phase, and allocation of intermediate storage tanks), the batch plant decision variables (equipment sizes, batch sizes, and operating times of semicontinuous items), and the process decision variables (e.g., final concentration at selected stages, volumetric ratio of phases in the liquid-liquid extraction). A noteworthy feature of the proposed approach is that the mathematical model for the plant is the same as that used in the constant factor model. The process performance models are handled as extra constraints. A plant consisting of eight stages operating in the single product campaign mode (one fermentation, two microfiltrations, two ultrafiltrations, one homogenization, one liquid-liquid extraction, and one chromatography) for producing four different recombinant proteins by the genetically engineered yeast Saccharomyces cerevisiae was modeled and optimized. Using this example, it is shown that the presence of additional degrees of freedom introduced by the process performance models, with respect to a fixed size and time factor model, represents an important development in improving plant design.     

Techno-economic analysis of a transient plant-based platform for monoclonal antibody production

Plant-based biomanufacturing of therapeutic proteins is a relatively new platform with a small number of commercial-scale facilities, but offers advantages of linear scalability, reduced upstream complexity, reduced time to market, and potentially lower capital and operating costs. In this study we present a detailed process simulation model for a large-scale new “greenfield” biomanufacturing facility that uses transient agroinfiltration of Nicotiana benthamiana plants grown hydroponically indoors under light-emitting diode lighting for the production of a monoclonal antibody. The model was used to evaluate the total capital investment, annual operating cost, and cost of goods sold as a function of mAb expression level in the plant (g mAb/kg fresh weight of the plant) and production capacity (kg mAb/year). For the Base Case design scenario (300 kg mAb/year, 1 g mAb/kg fresh weight, and 65% recovery in downstream processing), the model predicts a total capital investment of $122 million dollars and cost of goods sold of $121/g including depreciation. Compared with traditional biomanufacturing platforms that use mammalian cells grown in bioreactors, the model predicts significant reductions in capital investment and >50% reduction in cost of goods compared with published values at similar production scales. The simulation model can be modified or adapted by others to assess the profitability of alternative designs, implement different process assumptions, and help guide process development and optimization.     

Technoeconomic analysis of semicontinuous bioreactor production of biopharmaceuticals in transgenic rice cell suspension cultures

Biopharmaceutical protein production using transgenic plant cell bioreactor processes offers advantages over microbial and mammalian cell culture platforms in its ability to produce complex biologics with simple chemically defined media and reduced biosafety concerns. A disadvantage of plant cells from a traditional batch bioprocessing perspective is their slow growth rate which has motivated us to develop semicontinuous and/or perfusion processes. Although the economic benefits of plant cell culture bioprocesses are often mentioned in the literature, to our knowledge no rigorous technoeconomic models or analyses have been published. Here we present technoeconomic models in SuperPro Designer® for the large-scale production of recombinant butyrylcholinesterase (BChE), a prophylactic/therapeutic bioscavenger against organophosphate nerve agent poisoning, in inducible transgenic rice cell suspension cultures. The base facility designed to produce 25 kg BChE per year utilizing two-stage semicontinuous bioreactor operation manufactures a single 400 mg dose of BChE for $263. Semicontinuous operation scenarios result in 4–11% reduction over traditional two-stage batch operation scenarios. In addition to providing a simulation tool that will be useful to the plant-made pharmaceutical community, the model also provides a computational framework that can be used for other semicontinuous or batch bioreactor-based processes.     

A semicontinuous 3-column countercurrent solvent gradient purification (MCSGP) process

The recently developed continuous Multicolumn Countercurrent Solvent Gradient Purification (MCSGP) Process has been reduced to a fully equivalent semicontinuous setup with only three chromatographic columns and three gradient pump modules. Actually the 3-column MCSGP unit can even achieve better performance than the original 6-column process due to an additional degree of freedom, that is a different switching time for the "batch lane" and the "interconnected lane." Experimental results for the 3-column MCSGP unit of the purification of an industrial multicomponent peptide mixture containing 46% of Calcitonin on a reversed phase resin are compared with model simulations. It is concluded, that the model is well suited to predict the system behavior and therefore to design its optimal operating conditions.     

Bioprocess design and economic analysis for the commercial production of environmentally friendly bioinsecticides from Bacillus thuringiensis HD-1 kurstaki

A production process for B. thuringiensis (Bt) bioinsecticides was designed in detail, including alternative batch, low-density fed-batch (LDFB), and high-density fed-batch (HDFB) fermentation configurations. Capital and operating costs, as well as profitability based on simple rate of return, were performed using a purpose-written FORTRAN program, explicitly analyzing production of a water-based flowable product used in forestry applications.The total capital cost was 18 million dollars (Canadian dollars) for a stand-alone plant with base-scale capacity of 3 x 10(7) billion international units (BIU)/year. Raw material costs amounted to 1.5 million dollars yearly, of which approximately half was for formulation ingredients. Per-unit production cost rose sharply for scales of less than 1 x 10(7) BIU/year, but was little affected by scale above 3 x 10(7) BIU/year. Product cost was much lower at all scales for a LDFB as opposed to batch fermentation process, but HDFB gave relatively little additional cost benefit. Profitability analysis performed by co-varying scale and selling price showed that break-even occurred at a price of 0.45 dollars/BIU for a batch process at base scale, while with LDFB fermentation the same production volume sold at 0.35 dollars/BIU gave a 12% rate of return. Since the assumed base scale would represent 8-15% of current world Bt bioinsecticide production, based on value or volume, it was concluded that profitability would require some or all of the following elements: targeting higher-value markets such as disease vector control, in addition to forestry; a potentially lower plant capacity (although at least 1 x 10(7) BIU/year;) and coproduction of other large-volume microbial products to absorb capacity and match bioinsecticide output to market demand.     

Parametric study of a 6-column countercurrent solvent gradient purification (MCSGP) unit

The novel "multicolumn countercurrent solvent gradient purification" (MCSGP) process has been modeled for the purification of a polypeptide mixture characterized by a strong non-linear competitive adsorption isotherm. As a model system, the purification of an industrial polypeptide mixture containing 46% of the hormone calcitonin has been selected. The many impurities contained in the mixture have been lumped into three key impurities, which are selected as the ones eluting closer to the main component. The simulation model allows for a better understanding of the complex operating behavior of the multicolumn system, which has been experimentally investigated in a previous work. Through a systematic parametric analyses of the model behavior, the main operating parameters controlling the process performance in terms of purity and yield are investigated. The study of internal liquid and adsorbed phase concentration profiles along the unit for the different operating conditions allow elucidating the working principle of the new separation process. It is found that the MCSGP unit achieves much higher yields for a given product purity than the corresponding single-column batch units.     

Species classification from hyperspectral leaf information using machine learning approaches

Information on plant species is fundamental to forest ecosystems, in the context of biodiversity monitoring and forest management. Traditional methods for plant species inventories are generally inefficient, in terms of cost and performance, and there is a high demand for a quick and feasible approach to be developed. Of the various attempts, remote sensing has emerged as an active approach for plant species classification, but most studies have concentrated on image processing and only a few of them ever use hyperspectral information, despite the wealth of information it contains. In this study, plant species are classified from hyperspectral leaf information using different machine learning models, coupled with feature reduction and selection methods, and their performance is optimized through Bayesian optimization. The results show that including feature selection and Bayesian optimization increases the classification accuracy of machine learning models. Among these, the Bayesian optimization-based support vector machine (SVM) model, combined with the recursive feature elimination (RFE) feature selection method, yields the best output, with an overall accuracy of 86% and a kappa coefficient of 0.85. Furthermore, the confusion matrix revealed that the number of samples correlates with classification accuracy. The support vector machine with informative bands after Bayesian optimization outperformed in classing plant species. The results of this study facilitate a better understanding of spectral (phenotype) information with plant species (genotype) and help to bridge hyperspectral information with ecosystem functions.     

The quantitative requirements of human diploid cells (strain MRC-5) for amino acids, vitamins and serum.

The uptakes of all essential amino acids, vitamins (except riboflavin), glucose and serum during growth of human diploid cells (MRC-5) were determined. The amino acid uptakes varied considerably with the conditions of culture. The glucose requirement is several times greater than that for mouse LS or human HeLa cells. These analytical results were used to modify the medium so as to ensure that an excess of all defined medium constituents was present and pH was not limiting during study of the serum requirements. It was then found that maximum cell populations were directly proportional to the serum concentration. Hence the growth was limited by the supply of an unknown growth factor in serum. The serum growth factor was not replaced by a mixture of over 60 vitamins, co-enzymes, hormones and other organic and inorganic compounds considered to be possible growth factors, although this mixture did not lower the growth rate and somewhat (22%) increased the yield from the serum growth factor. The unit of serum growth factor is precisely defined in terms of the amount in a standard batch of calf serum. This standard contains 10 units/ml whereas the other batch of serum used contained only 5 units/ml.     

Multi-objective optimization of glycopeptide antibiotic production in batch and fed batch processes

Fermentation optimization involves potentially conflicting multiple objectives such as product concentration and production media cost. Simultaneous optimization of these objectives would result in a multiobjective optimization problem, which is characterized by a set of multiple solutions, knows as pareto optimal solutions. These solutions gives flexibility in evaluating the trade-offs and selecting the most suitable operating policy. Here, ε-constraint approach was used to generate the pareto solutions for two objectives: product concentration and product per unit cost of media, for batch and fed batch operations using process model for Amycolatopsis balhimycina, a glycopeptide antibiotic producer. This resulted in a set of several pareto optimal solutions with the two objectives ranging from (0.75 g l⁻¹, 3.97 g $^-1) to (0.44 g l⁻¹, 5.19 g $^-1) for batch and from (1.5 g l⁻¹, 5.46 g $^-1) to (1.1 g l⁻¹, 6.34 g $^-1) for fed batch operations. One pareto solution each for batch and for fed batch mode was experimentally validated.     

Advanced model‐based control strategies for the intensification of upstream and downstream processing in mAb production

Current industrial trends encourage the development of sustainable, environmentally friendly processes with minimal energy and material consumption. In particular, the increasing market demand in biopharmaceutical industry and the tight regulations in product quality necessitate efficient operating procedures that guarantee products of high purity. In this direction, process intensification via continuous operation paves the way for the development of novel, eco‐friendly processes, characterized by higher productivity and lower production costs. This work focuses on the development of advanced control strategies for (i) a cell culture system in a bioreactor and (ii) a semicontinuous purification process. More specifically, we consider a fed‐batch culture of GS‐NS0 cells and the semicontinuous Multicolumn Countercurrent Solvent Gradient Purification (MCSGP) for the purification process. The controllers are designed following the PAROC framework/software platform and their capabilities are assessed in silico, against the process models. It is demonstrated that the proposed controllers efficiently manage to increase the system productivity, returning strategies that can lead to continuous, stable process operation. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:966–988, 2017     

A process model to estimate biodiesel production costs

'Biodiesel' is the name given to a renewable diesel fuel that is produced from fats and oils. It consists of the simple alkyl esters of fatty acids, most typically the methyl esters. We have developed a computer model to estimate the capital and operating costs of a moderately-sized industrial biodiesel production facility. The major process operations in the plant were continuous-process vegetable oil transesterification, and ester and glycerol recovery. The model was designed using contemporary process simulation software, and current reagent, equipment and supply costs, following current production practices. Crude, degummed soybean oil was specified as the feedstock. Annual production capacity of the plant was set at 37,854,118 l (10 x 10(6)gal). Facility construction costs were calculated to be US dollar 11.3 million. The largest contributors to the equipment cost, accounting for nearly one third of expenditures, were storage tanks to contain a 25 day capacity of feedstock and product. At a value of US dollar 0.52/kg (dollar 0.236/lb) for feedstock soybean oil, a biodiesel production cost of US dollar 0.53/l (dollar 2.00/gal) was predicted. The single greatest contributor to this value was the cost of the oil feedstock, which accounted for 88% of total estimated production costs. An analysis of the dependence of production costs on the cost of the feedstock indicated a direct linear relationship between the two, with a change of US dollar 0.020/l (dollar 0.075/gal) in product cost per US dollar 0.022/kg (dollar 0.01/lb) change in oil cost. Process economics included the recovery of coproduct glycerol generated during biodiesel production, and its sale into the commercial glycerol market as an 80% w/w aqueous solution, which reduced production costs by approximately 6%. The production cost of biodiesel was found to vary inversely and linearly with variations in the market value of glycerol, increasing by US dollar 0.0022/l (dollar 0.0085/gal) for every US dollar 0.022/kg (dollar 0.01/lb) reduction in glycerol value. The model is flexible in that it can be modified to calculate the effects on capital and production costs of changes in feedstock cost, changes in the type of feedstock employed, changes in the value of the glycerol coproduct, and changes in process chemistry and technology.     

Economic simulation of batch and continuous aqueous two-phase purification for viral products

Vaccine manufacturing strategies that lower capital and production costs could improve vaccine access by reducing the cost per dose and encouraging localized manufacturing. Continuous processing is increasingly utilized to drive lower costs in biological manufacturing by requiring fewer capital and operating resources. Aqueous two-phase systems (ATPS) are a liquid–liquid extraction technique that enables continuous processing for viral vectors. To date, no economic comparison between viral vector purifications using traditional methods and ATPS has been published. In this work, economic simulations of traditional chromatography-based virus purification were compared to ATPS-based virus purification for the same product output in both batch and continuous modes. First, the modeling strategy was validated by re-creating a viral subunit manufacturing economic simulation. Then, ATPS capital and operating costs were compared to that of a traditional chromatography purification at multiple scales. At all scales, ATPS purification required less than 10% of the capital expenditure compared to chromatography-based purification. At an 11 kg per year production scale, the ATPS production costs were 50% less than purification with chromatography. Other chromatography configurations were explored, and may provide a production cost benefit to ATPS, but the purity and recovery were not experimentally verified. Batch and continuous ATPS were similar in capital and production costs. However, manual price adjustments suggest that continuous ATPS plant-building costs could be less than half that of batch ATPS at the 11 kg per year production scale. These simulations show the significant reduction in manufacturing costs that ATPS-based purification could deliver to the vaccine industry.     

Money matters (II): costs of maize inbred line conversion schemes at CIMMYT using conventional and marker-assisted selection

This article presents selected results of a study carried out in Mexico at the International Maize and Wheat Improvement Center (CIMMYT) to compare the cost-effectiveness of conventional and biotechnology-assisted maize breeding. Costs associated with the use of conventional and marker-assisted selection (MAS) methods at CIMMYT were estimated using a spreadsheet-based budgeting approach. This information was used to compare the costs of conventional and MAS methods for a particular breeding application: introgressing an elite allele at a single dominant gene into an elite maize line (line conversion). At CIMMYT, neither method shows clear superiority in terms of both cost and speed: conventional breeding schemes are less expensive, but MAS-based breeding schemes can be completed in less time. For applications involving tradeoffs between time and money, relative profitability can be evaluated using conventional investment theory. Using a simple model of a plant breeding program, we show that the optimal choice of a breeding technology depends on the availability of operating capital. If operating capital is abundantly available, the "best" breeding method will be the one that maximizes the net present value (i.e., MAS), but if operating capital is constrained, the "best" breeding method will be the one that maximizes the internal rate of return (i.e., conventional selection). This insight may help to explain why private firms tend to invest more aggressively in biotechnology than public breeding programs, which are more likely to face budgetary constraints.     

Column flow reactor using acetohydroxyacid synthase I from Escherichia coli as catalyst in continuous synthesis of R-phenylacetyl carbinol

We tested the possibility of utilizing acetohydroxyacid synthase I (AHAS I) from Escherichia coli in a continuous flow reactor for production of R-phenylacetyl carbinol (R-PAC). We constructed a fusion of the large, catalytic subunit of AHAS I with a cellulose binding domain (CBD). This allowed purification of the enzyme and its immobilization on cellulose in a single step. After immobilization, AHAS I is fully active and can be used as a catalyst in an R-PAC production unit, operating either in batch or continuous mode. We propose a simplified mechanistic model that can predict the product output of the AHAS I-catalyzed reaction. This model should be useful for optimization and scaling up of a R-PAC production unit, as demonstrated by a column flow reactor.     

Effect of production method and gene amplification on the glycosylation pattern of a secreted reporter protein in CHO cells

We have investigated the independent effects of selective gene amplification (using the dhfr amplifiable selection marker) and culture operating strategy (batch vs repeated fed-batch vs semicontinuous perfusion) on the glycosylation of a recombinant reporter protein (secreted alkaline phosphatase, SEAP) produced in transfected Chinese hamster ovary (CHO) cells. HPLC analyses coupled with susceptibility to various exoglycosidases were used to determine the N-glycosylation profile of SEAP samples. The dhfr amplified cell line yielded an almost 10-fold increase in specific productivity as compared to that of the unamplified cell line. The glycosylation pattern of the reporter protein produced in batch bioreactor cultures of the amplified cell line showed only slight differences as compared to the glycosylation pattern of the protein from batch bioreactor cultures of the unamplified cell line. In contrast, analysis of SEAP glycosylation structures from the protein isolated from semicontinuous perfusion cultures indicated that both relative glycan content and extent of sialylation were increased as compared to samples isolated from repeated fed-batch cultures. These results suggest that the slow growing perfusion cultures produce more completely glycosylated proteins than the faster growing repeated fed-batch cultures.     

Histochemical studies of pollen: storage pockets in the endoplasmic reticulum (ER)

Summary The pollen grain of cotton (Gossypium hirsutum) was examined histochemically at the light and electron microscope level. The cytoplasm of the pollen contains an unusual storage unit which consists of a pocket of endoplasmic reticulum (ER) containing lipid droplets and dictyosome vesicles. The ER pockets are large enough to be seen with the light microscope if thin enough sections are used (0.3–1.5). The results of the histochemical analyses show that the dictysome vesicles are rich in carbohydrate and contain protein and lipid as well. The ER contains large amounts of protein which may be arginine rich. Some carbohydrates may also be present in the ER. The ER is covered with ribosomes so that the pockets are unusually rich storage units containing abundant protein, carbohydrate, lipid and RNA. The light microscope localization of carbohydrates was confirmed by the periodic acid-silver method. Other storage units in the cytoplasm were also studied. A new method for the embedding of plant tissue for thin sectioning for light microscopy is presented.This work was supported by a Public Health Service fellowship 5-F2-GM-22, 031-02 from the National Institute of General Medical Sciences, by NSF grant GB 3460, by NIH grant 5-RO1-CA 0356-10 and by the Miller Institute for Basic Science.     

Restructuring upstream bioprocessing: technological and economical aspects for production of a generic microbial feedstock from wheat

Restructuring and optimization of the conventional fermentation industry for fuel and chemical production is necessary to replace petrochemical production routes. Guided by this concept, a novel biorefinery process has been developed as an alternative to conventional upstream processing routes, leading to the production of a generic fermentation feedstock from wheat. The robustness of Aspergillus awamori as enzyme producer is exploited in a continuous fungal fermentation on whole wheat flour. Vital gluten is extracted as an added-value byproduct by the conventional Martin process from a fraction of the overall wheat used. Enzymatic hydrolysis of gluten-free flour by the enzyme complex produced by A. awamori during fermentation produces a liquid stream rich in glucose (320 g/L). Autolysis of fungal cells produces a micronutrient-rich solution similar to yeast extract (1.6 g/L nitrogen, 0.5 g/L phosphorus). The case-specific combination of these two liquid streams can provide a nutrient-complete fermentation medium for a spectrum of microbial bioconversions for the production of such chemicals as organic acids, amino acids, bioethanol, glycerol, solvents, and microbial biodegradable plastics. Preliminary economic analysis has shown that the operating cost required to produce the feedstock is dependent on the plant capacity, cereal market price, presence and market value of added-value byproducts, labor costs, and mode of processing (batch or continuous). Integration of this process in an existing fermentation plant could lead to the production of a generic feedstock at an operating cost lower than the market price of glucose syrup (90% to 99% glucose) in the EU, provided that the plant capacity exceeds 410 m(3)/day. Further process improvements are also suggested.     

Biomass transportation model and optimum plant size for the production of ethanol

A detailed model based on a non-dimensional transportation factor is developed to assess the economics of biomass collection, transportation, and storage. The optimum plant size for bio-refineries is investigated; ethanol production from corn stover via dilute acid hydrolysis is presented as a case study. The conversion of straight-line, farm-to-plant distances to road distances via a winding factor leads to a shift in the distribution of transportation distances towards shorter hauls. The capital investment scaling exponent was calculated using the model developed at the National Renewable Energy Laboratory (Aden et al., NREL/TP-510-32438, 2002) and found to be 0.7. The cost of the delivered corn stover is proportional to the square root of the inverse of the farmer participation; as a consequence, bio-fuel producers intending to use agricultural residues as feedstock should work towards a farmer participation of fifty percent. Costs associated with storage represent a significant portion of the production cost.     

Genetic programming assisted stochastic optimization strategies for optimization of glucose to gluconic acid fermentation

This article presents two hybrid strategies for the modeling and optimization of the glucose to gluconic acid batch bioprocess. In the hybrid approaches, first a novel artificial intelligence formalism, namely, genetic programming (GP), is used to develop a process model solely from the historic process input-output data. In the next step, the input space of the GP-based model, representing process operating conditions, is optimized using two stochastic optimization (SO) formalisms, viz., genetic algorithms (GAs) and simultaneous perturbation stochastic approximation (SPSA). These SO formalisms possess certain unique advantages over the commonly used gradient-based optimization techniques. The principal advantage of the GP-GA and GP-SPSA hybrid techniques is that process modeling and optimization can be performed exclusively from the process input-output data without invoking the detailed knowledge of the process phenomenology. The GP-GA and GP-SPSA techniques have been employed for modeling and optimization of the glucose to gluconic acid bioprocess, and the optimized process operating conditions obtained thereby have been compared with those obtained using two other hybrid modeling-optimization paradigms integrating artificial neural networks (ANNs) and GA/SPSA formalisms. Finally, the overall optimized operating conditions given by the GP-GA method, when verified experimentally resulted in a significant improvement in the gluconic acid yield. The hybrid strategies presented here are generic in nature and can be employed for modeling and optimization of a wide variety of batch and continuous bioprocesses.     

Trypanosoma cruzi 5S rRNA arrays define five groups and indicate the geographic origins of an ancestor of the heterozygous hybrids

Isolates of the etiological agent of Chagas disease, Trypanosoma cruzi, have been subdivided into six subgroups referred to as discrete typing units. The subgroups are related through two distinct hybridisation events: representatives of homozygous discrete typing units I and IIb fused to form discrete typing units IIa and IIc, whose homozygous genotypes have features of both ancestral types; a second fusion between strains of homozygous discrete typing units IIb and IIc created the heterozygous hybrid strains discrete typing units IId and IIe. The intergenic region of the tandemly repeated 5S rRNA array displays four variant sequence classes, allowing the discrimination of five discrete typing units. The genome project reference strain, CL Brener, is a hybrid discrete typing unit IIe strain that contains both discrete typing unit IIb and IIc classes of 5S rRNA repeats in distinct arrays present on different chromosomes. The CL Brener discrete typing unit IIb-type array contains approximately 193 repeated units, of which about one-third contain a 129 bp sequence that replaces a majority of the 5S rRNA sequence. The 129 bp 'invader' sequence was detected within the arrays of all hybrid discrete typing unit IId and IIe strains and in a subset of discrete typing unit IIb strains. This array invader replaces the internal promoter elements conserved in 5S rRNA. The discrete typing unit IIb Esmeraldo strain contains approximately 135 repeats and shows a region of homology to the array invader in the 5' flank of the array, but no evidence of the invading sequence element within the array. A survey of additional discrete typing unit IIb strains revealed a split within the subgroup, in which some strains contained invaded arrays and others were homogeneous for the 5S rRNA. The putative discrete typing unit IIb ancestor of the hybrid discrete typing units IId and IIe more closely resembles the extant Bolivian/Chilean IIb isolates than the Brazilian IIb isolates based on the correlation with the array invader.