Here, several experts in the field discuss challenges and choices in microbial production, including type of organism, optimization of medium, and upscaling of production systems.
Yinjie Tang, PhD, Professor, McKelvey School of Engineering, Washington University, St Louis
Gergely Kosa, PhD, Senior Research Engineer, Bioprocess R&D, GE Healthcare
David Wood, PhD, Professor, Chemical & Biomolecular Engineering, The Ohio State University
GEN: In large-scale microbial production systems, glucose may represent the most convenient carbon source (Guo et al. Microb Cell Fact (2016) 15:24.) for the production of a variety of bulk chemicals. But its adoption may hamper efforts to avoid the environmental damage resulting from reliance on conventional industrial sources of raw materials. For large-scale synthetic programs, what are your recommendations for viable carbon sources, such as non-edible sugars?
Tang: Glucose is the optimal carbon source, but the cost limits its use for producing low-value chemicals or fuels. Lignocellulose is an appealing alternative, but there are at present a number of challenges. Biomass is bulky, and the physical/chemical treatments for biomass conversion into fermentable substrates (such as glucose, xylose, and lignin-derived products) is laborious. Such processing may also require a number of expensive enzyme additives.
When you also factor in transportation costs to the production site, lignocellulosic-based biomanufacturing may make the whole process non-competitive from an economic standpoint. In addition, there are frequently inhibitors present in crude biomass-based feedstock, which may have deleterious effects on fermentation performance. Microbiologists and metabolic engineers should merge their approaches to develop novel, non-model microbial platforms that have the innate capability to utilize non-edible sugars for bioproduction.
Kosa: Several non-food carbon sources can be used for industrial-scale fermentations to produce, among others, bulk chemicals (citric acid, lactic acid, 1,3-propanediol), biofuels (bioethanol, jet fuel), enzymes (rennet, lipase, invertase), and microbial biomass (single cell proteins).
Lignocellulosic molecules, the most abundant raw materials on Earth, can be converted into fermentable monosaccharides such as xylose, arabinose, mannose, and galactose. Crude glycerol (1kg/10kg biodiesel) is available in huge quantities. Food waste, such as whey (9 kg/kg cheese), and molasses (0.3kg/kg processed sugar) are also important substrates. Hydrophobic materials, such as animal fat and vegetable oils, or even gases, including syngas and CO2, can serve as viable carbon sources.
Wood: Many large-scale processes still rely on industrial-grade corn syrup due to its high glucose content and easily handled liquid form. As process scales become extremely large, as in the case of biofuels of bioplastics, there is a societal cost to effectively trading food for fuel or disposable plastic. In this case, the main alternate source under development would probably be waste cellulose or cellulose that can be rapidly and cheaply grown, as in the famous case of switchgrass. Although these sources have few societal disadvantages, there are a number of technical problems that need to be solved before cellulose can be easily incorporated into conventional processes. Primary among these are: the need to pretreat cellulose to break it into usable sugars, the lignin content of most cellulose, and the inability of some organisms to utilize the base sugars contained in cellulose. Work continues on solving these problems, which includes a range of approaches from re-engineering plants for decreased lignin content to identifying economical pretreatment strategies.
Another possible alternative is waste glycerol, an expected byproduct of the converting triglycerides into biodiesel. As the biodiesel industry expands, a tremendous amount of glycerol will be generated, which will make it more attractive as a simple carbon source. However, it remains to be seen if the currently used microbial strains will be able to utilize this source effectively.
GEN: In configuring a microbial bioprocessing program, it may be necessary to develop a rapid method for screening many samples to determine the optimal medium for the specific project at hand. Have you encountered this demand, and what do you recommend as the optimal solution?
Tang: We employ high-throughput methods to carry out multiple screenings involving micro-bioreactors. In this way we can simultaneously collect a range of data covering a variety of conditions. In advanced laboratories, such as the Joint Bioenergy Institutes, the Automated Robotic System is available to greatly speed up sample prep and analysis. From these collected results, researchers can further employ statistical analysis and machine learning to obtain optimal conditions.
Kosa: Screening of various media traditionally has been performed in shake flasks. Due to advances in metabolic engineering and process automation, high-throughput screening platforms such as parallel microbioreactors (100 mL) or microtiter plates (mL scale in 24-,48-, 96-well plates or L range up to 1536 well plates) are now preferred. State-of-the-art systems with monitoring and control options of process parameters (pH, DO, feeding, etc.) are available on the market. The optimal platform for microbial medium screening will depend on the required throughput, the need for process monitoring and control, and, of course, the available budget. The medium screening plan is based on the Design of Experiment (DoE) approach.
Wood: In general, our initial decisions about growth media are driven by the cost and estimated yields of the product we are trying to produce. In the case of biopharmaceuticals, for example, the selling price is sufficiently high that the cost of raw materials is generally not a major consideration (although the rise of biosimilars and the drive to make drugs available to larger populations may change this). In the case of commodity chemicals, however, we look at the overall costs of each type of medium and then decide which ones are likely to make sense. Once this determination is made, we acquire samples from the various vendors and carry out small-scale fermentations, usually in shake-flasks.
As an example, a difficulty we encounter with many inexpensive nitrogen sources is that they can contain chemicals that are toxic to some microbes and in many cases have large amounts of solids in the raw feeds. These difficulties can be evaluated in a small fermentation protocol, and once the final sources are identified, we would go forward with conventional DoE approaches to optimize various nutrient levels and feeding strategies. Having a good statistician to help design the experiments and analyze the data can greatly expedite this process.
GEN: In mammalian cell-based bioprocessing protocols, the choice of cell lines is limited, usually to CHO and several other possibilities. In the case of microbial-based bioprocessing choices, however, the opportunities are much wider, encompassing many different species. Could you discuss your system of choice (bacterial, yeast, fungal) and what special advantages it possesses for your program?
Tang: Mammalian-cell cultivation is so extremely expensive that it can only be employed for producing high-value molecules, such as antibodies. In microbial engineering, there are two philosophies for strain development: 1) engineer well-established model hosts as broad manufacturing platforms, including Saccharomyces cerevisiae, Bacillus subtilis, and Escherichia coli, and 2) develop novel microbial hosts and optimize their endogenous pathways for synthesizing specific products.
In my laboratories, we help a local biotech company, Arch Innotek, with the production of nutraceuticals using the non-model yeast Yarrowia, since metabolically yeast behave comparably to a plant cell and Yarrowia is GRAS (generally regarded as safe). In contrast, bacteria frequently generate toxins, which make them less than optimal for producing pharmaceuticals. However, certain bacterial species can degrade lignin and are also amenable to genetic engineering. We have employed some species (for example, Rhodococcus) for lignin valorization projects funded by Department of Energy. Therefore, different hosts offer different advantages. We should employ them on a case-by-case basis.
Kosa: The choice of microorganism will depend on the existing metabolic pathways leading to the desired product or on the possibility to efficiently engineer the microorganism. Growth and production rates, scalability, and regulatory concerns are also important factors when deciding upon the cell factory. Yeasts have been used for thousands of years for beer, wine, and bread production. S. cerevisiae’s genome has been sequenced and can be easily manipulated to produce biopharmaceuticals and other fine chemicals.
Filamentous fungi, especially recombinant ones, are the primary choice for the production of industrial enzymes (lipases, proteases, etc.). Certain antibiotics and polyunsaturated fatty acids (PUFA) are also produced by filamentous fungi. Bacteria are naturally good acid producers (lactic, acetic acid), while Actinomycetes are important for the manufacturing of antibiotics. E. coli, a model organism, can be easily modified by molecular tools to produce heterologous proteins for biopharmaceutical applications (insulin for example).
Wood: Our laboratory is currently limited almost entirely to E. coli and one of a number of HEK293 strains or CHO cells. Since these are well-developed and have strong regulatory records, they tend to be the safest bet for now. It has been recognized that expression hosts will have a major impact on biopharmaceutical costs going forward however, and this has led to active research in identifying new strains for extremely high production of recombinant protein therapeutics.
Although a number of these have been reported, two companies that are generating some interest are Pfenex (based on Pseudomonas fluorescens), which was spun out of Dow Chemical, and Dyadic International (based on the Myceliophthora thermophila fungus). Both of these companies have developed novel expression hosts and are moving quickly to demonstrate them with biosimilars and other potential therapeutics. Pfenex has a number of products in clinical trials now.
These hosts have the ability to produce products using simple growth media and, in the case of Dyadic, have the potential to produce complex glycoproteins with extraordinarily high yields. As with any new host, the ability to produce glycosylation that is compatible with the human immune system is a major consideration, and the FDA will need to approve any resulting products.
GEN: In microbial bioprocessing, E. coli is the most widely used host, due to its ease of cultivation, the wide variety of strains developed for specific tasks, and its ability to proliferate in simple media. However, it does possess some disadvantages, including amino acid misincorporation, a perennial problem. Although in some cases amino acid substitutions do not have serious consequences for the quality of the synthesized protein, in other instances protein performance may by compromised. In your experience, have you encountered this issue, and how have you dealt with it?
Tang: In our laboratory, a more common problem has been the deterioration of highly engineered producing strains during process scaleup (including E. coli and yeasts). Adding a number of beneficial mutations or heterologous genes (from the point of view of the genetic engineer) can place significant metabolic burdens and cause metabolic shifts, which unavoidably cause host genetic stability to decline over time. Essentially, nature is telling the cell to do what is in its own best interest rather than what we re-program them to do. To stabilize strain performance for industrial applications, we have to take advantage of the cell’s native biosynthesis capability and strengths and avoid over-engineering our strain.
Wood: We have not had major problems with misincorporation of amino acids in E. coli or other hosts. Most product-related impurities, of which this would be one, are difficult to separate at large scale and generally become part of the therapeutic that is eventually given to patients. The logic is that if the product demonstrates safety and efficacy in clinical trials, then attempting to remove trace amounts of these impurities might not be worth the effort. This would apply to charge variants as well, which often appear in small amounts as product-related impurities and are similar in nature to amino acid misincorporations.
In most cases, these impurities become part of the product “fingerprint” and, in some cases, are considered a product attribute when releasing lots of a drug. Their significance is becoming somewhat greater when comparing innovator (reference) molecules to biosimilars, where the specific makeup of each drug can be used to demonstrate similarity. Although initially ignored during approvals, these small variations in protein composition can also now be used to build defensive patents for innovator companies.
GEN: Bioethanol has potential as a substitute vehicle fuel, but in order to overcome the requirements for a non-industrial carbon source, alternatives to glucose from corn or other crops plants is a necessity. Can you discuss what has been your experience with alternative sugars produced by plant materials other than corn?
Tang: One of my former PhD students (Dr. Ni Wan) is now working as a data scientist for Monsanto. She explains that big data approaches and machine-learning tools are used effectively to optimize corn yields and reduce the cost of sugar production. On the other hand, fast-growing algae (especially cyanobacterium) have strong potential for converting CO2 into bioproducts (such as butanol) or renewable feedstock (such as lactate and sucrose). Photobiorefinery offers another promising route for the production of a variety of bulk chemicals.
Kosa: First generation bioethanol is produced from food crops, such as corn, sugarcane, and sugar beet. The food/feed versus fuel debate is limiting the expansion of first-generation bioethanol. Second-generation bioethanol is produced from non-edible lignocellulosic materials, such as residues from forestry (soft and hardwood) and agriculture (corn stover, sugarcane bagasse etc.) The rigid structure of these materials must be broken down by pretreatment through mechanical and/or chemical processes, followed by enzymatic hydrolysis of sugar polymers into monosaccharides.
Microalgae, due to their high oil content, have been mainly considered as a promising source for third-generation biodiesel production. Nonetheless, certain algal species contain high amounts of starch that can be a substrate for the yeast to produce third-generation bioethanol.
Wood: As I mentioned earlier, the conventional wisdom is that waste cellulose or highly optimized cellulosic crops will eventually be used to generate biofuels. Notably, however, there has been a push to move away from bioethanol and toward biobutanol more recently. This is mainly driven by the ability of conventional car engines to run on butanol with little or no modification, which is not true with ethanol. Butanol also has a similar vapor pressure and energy density to conventional gasoline, which suggests it might be a more easily adopted alternative.
An interesting side note is that butanol has a distinctive odor, which has driven some developers away from it as a fuel source. Regardless, one can imagine a time when biodiesel is generated in large algae farms where the conversion of the algal triglycerides to diesel generates large amounts of glycerol. If the glycerol can then be used to generate microbial biobutanol, then the potential is there for a fully carbon-neutral means to generate diesel fuel as a substitute for gasoline in conventional engines. As with all alternate fuels, however, the cost of producing and refining these products currently prevents them from being competitive to fossil fuels.