I'd like to know the meaning of these terms in a HCDC studies of a modified strain. Max specific yield (mg/g DCW) Max volumetric product yield (g/l) Specific product formation rate Qp (mg/g DCW/h)
Maximum specific yield = milligram product per gram dry cell weight (DCW). Maximum volumetric yield = gram product per liter culture. Specific product formation rate Qp = milligram product per gram dry cell weight per hour.
Fed-batch culture is, in the broadest sense, defined as an operational technique in biotechnological processes where one or more nutrients (substrates) are fed (supplied) to the bioreactor during cultivation and in which the product(s) remain in the bioreactor until the end of the run.  An alternative description of the method is that of a culture in which "a base medium supports initial cell culture and a feed medium is added to prevent nutrient depletion".  It is also a type of semi-batch culture. In some cases, all the nutrients are fed into the bioreactor. The advantage of the fed-batch culture is that one can control concentration of fed-substrate in the culture liquid at arbitrarily desired levels (in many cases, at low levels).
Generally speaking, fed-batch culture is superior to conventional batch culture when controlling concentrations of a nutrient (or nutrients) affects the yield or productivity of the desired metabolite.
Chemical pesticides are widely used. However, the extensive use of these pesticides also affects non-target insects and animals, thereby impairing ecosystem, potentially posing a risk to human health, and causing diseases, including cancer, reproductive disorders, neurological disorders, and allergies. Therefore, there is a consensus to reduce the use of pesticides . To better protect the ecological environment and human health, and to maintain sustainable agricultural development, it is urgent to develop “green” pesticides [2,3,4,5]. Fungi are the major source of biological pesticides [6,7,8]. Their metabolites, including toxic proteins and peptides, have the potential to be used in biological pest control [9,10,11].
Ribotoxins produced by some fungal species, such as Aspergillus, Hirsutella thompsonii and other entomopathogenic fungi, have the potential to be used as biological pesticides [10, 12, 13]. Specifically, the fungal ribotoxin hirsutellin A (HtA) produced by the invertebrate fungal pathogen H. thompsonii exhibits insect-specific cytotoxicity and strong insecticidal properties [14, 15]. Native HtA is a non-glycosylated monomeric protein comprising 130 amino acid residues and shows thermo stability and protease stability. As a comparison, HtA is 10–20 amino acids shorter than ribotoxins from Aspergillus, and has a low homology with them (25%), which is the main reason leading to significant differences in their biological functions [14, 16, 17]. Previous studies have reported that HtA is highly toxic to adult mites and aphids, lethal to moth and fly larvae, and shows oral toxicity to neonatal larvae of Aedes aegypti [14, 15]. However, the content of native HtA is low, with only 35 μg HtA isolated from the supernatant of 1 g of dried mycelium and less than 1 mg HtA purified from 1 l of fermentation broth . Although recombinant HtA (rHtA) has been successfully prepared using Escherichia coli (E. coli) expression system, less than 1 mg rHtA could be purified from 1 l of culture medium using a complex purification protocol . Moreover, E. coli-expressed rHtA is not the N-terminal native protein, resulting in potentially different biological activity. Additionally, E. coli produces large amounts of endotoxin, which needs to be removed before in vivo activity analyses. The failure to prepare large amounts of HtA has seriously limited the further development of its insecticidal potential.
Investigation of HtA bioactivity against insect pests requires large quantities of protein [14, 15]. In particular, the determination of the oral insecticidal activity of HtA against agricultural pests and its biological safety to mammals also requires a large amount of protein [14, 15]. Therefore, it is necessary to develop a heterologous protein expression system and efficient purification method to prepare a large amount of active rHtA. As a widely used high-level eukaryotic protein-expression system, Pichia pastoris (P. pastoris) has the ability to secrete recombinant proteins . Additionally, P. pastoris has the ability to produce gram-level amounts of secretory recombinant protein per litre of fermentation culture . Furthermore, P. pastoris does not produce endotoxin. Therefore, purified recombinant proteins can be directly used for in vivo experiments.
In this study, we reported a method for efficient expression and purification of rHtA from P. pastoris X33 by fed-batch fermentation. Also, we analysed the bioactivity of rHtA.
Arbutin is a hydroquinone glucoside that is widely distributed in various plants of Ericaceae such as Arctostaphylos uva-ursi and Vaccinium spp. It inhibits tyrosinase activity in melanin synthesis by slowly releasing the active component, hydroquinone, through hydrolysis of its glucoside group [ 5, 8, 16]. It is a safe and mild agent for treating hyperpigmentation disorders compared to hydroquinone and has immense market potential in cosmetics industry. Structurally, α-arbutin is an α-glucoside of hydroquinone and the α-glucosidic bond offers higher stability and efficacy on melanogensis than the β-form. [ 5, 24, 25]. Due to its better mode of inhibition of tyrosinase, production of α-arbutin has been investigated by several researchers. α-Arbutin is produced by the glucosylation of hydroquinone in the presence of purified enzymes [ 13, 18] or by whole microbial [ 14] or plant cell biocatalysis [ 9, 15, 27]. The use of highly purified enzymes in the production process will ultimately result in very expensive α-arbutin, and whole cells usage results in lower productivities due to mass transfer limitations imposed by the cellular membrane.
Development of whole-cell biocatalysts displaying the target enzyme directly on the cell surface is one of the methods of overcoming substrate permeation barrier and to improve product yields. For example, using Pseudomonas syringae ice nucleation protein (INP) as a carrier protein, enzymes such as levansucrase [ 10], carboxymethylcellulase [ 11], and organophosphorus hydrolase [ 23] have been successfully displayed on Escherichia coli surface for use as whole-cell biocatalysts in the related applications. In the previous study [ 26], we have successfully constructed engineered E. coli displaying transglucosidase on the cell surface using INP anchoring motif. The whole cells expressing INPNC-transglucosidase at surface with very high transglucosylating activity were applied as biocatalysts in the glucosylation of hydroquinone with maltose as the sugar donor, and a high yield of arbutin was obtained [ 26].
In this study, the cost effective production of large amount of engineered E. coli cells anchoring surface-displayed transglucosidase for biotransformation by fed-batch fermentations was demonstrated. The cell concentration and the transglucosylation activity of the whole-cell biocatalyst for arbutin synthesis were used as the parameters for analyzing the results of this study.
2 MATERIALS AND METHODS
2.1 Algal strain and growth conditions
S. acuminatus GT-2 was isolated from South Lake of Guangzhou, China. Algal cells were maintained in a modified Endo growth medium, containing 30 g/L glucose, 3 g/L KNO3, 1.2 g/L KH2PO4, 1.2 g/L MgSO4•7H2O, 0.2 g/L trisodium citrate, 0.016 g/L FeSO4•7H2O, 2.1 mg/L EDTA-Na2, 0.03 g/L CaCl2•2H2O, 2.86 mg/L H3BO3, 0.222 mg/L ZnSO4•7H2O, 1.81 mg/L MnCl2•4H2O, 0.021 mg/L Na2MoO4, and 0.07 mg/L CuSO4•2H2O. To prepare the inoculants for fermentation, a single colony of S. acuminatus GT-2 was inoculated into 100 ml of modified Endo medium in a 250-ml Erlenmeyer flask and grown at 30°C for 5–6 days in a shaking incubator at the speed of 180 rpm, which was then used as inoculum for fermentation.
Bench-scale fermentation experiments were performed in a 7.5-L bioreactor (BioFlo & CelliGen 310 New Brunswick) with the initial working volume of 2.8 L. The pH was maintained automatically by the addition of 3 M NaOH or 1 M HCl solution. Aeration was maintained at 1 vvm with the airflow rate of 2.8 L/min. Dissolved oxygen (DO) was controlled automatically above 40% by coupling with the stirring speed. In fermentor batch medium, KNO3 was replaced by 0.84 g/L urea. Feeding medium used during fermentation process was the 25-fold concentrated batch medium, containing 750 g/L of glucose.
Pilot-scale fermentation was carried out in a 1,000 L stirred tank bioreactor (WKT 1000 L Yangzhong Weikete Biological Engineering Equipment Co., Ltd., China) containing 300 L medium. To shorten the culture period, the 1,000-L pilot-scale fermentation was inoculated with 40 L high-cell-density culture (80 g/L) from the 100-L bioreactor after 4 days’ fed-batch cultivation, which was inoculated from 7.5-L bench-scale fermentor. The initial biomass concentration in 1000-L fermentor cultivation was approximately 10 g/L. For 1000-L pilot-scale fermentation, the aeration rate and the agitation speed was initially set at 20 m 3 h −1 (1 vvm) and 80 rpm, respectively. The pressure of the inner bioreactor was kept at 0.035 Mpa.
For photoautotrophic culture, the growth medium BG-11 was used (Rippka, Deruelles, Waterbury, Herdman, & Stanier, 1979 ). Algal cells were grown in 750 ml BG-11 in an 800-ml column PBR (i.d. 5 cm) under continuous light (250 μmol·m −2 ·s −1 ) at 25 ± 2°C. Mixing and aeration were provided by bubbling air containing 2.0% (v/v) CO2. The cell culture was sequentially scaled-up to a 12-L panel PBR and a 380-L tubular PBR followed by a 1,300-L tubular PBR. The inoculum size during each step was 10% (v/v) of the total volume of culture media.
2.2 Induction of lipid production
Pilot-scale lipid production experiments were conducted in a 5,300-L tubular PBR (i.d. 5 cm) outdoors from June to September in 2016 (39° 97′ N 117° 06′ E, Sanhe, China). Algal cells grown in 7.5-L fermentor and 1,300-L PBR indoors were transferred to two parallel 5,300-L PBR and induced for lipid production in the N-limited BG-11 medium containing 1.1 mM nitrate for 13 days. For the outdoor experiments, CO2 was injected into the culture during daylight hours to maintain pH in the range of 6.5 to 6.8. The cooling system prevented the culture temperature from exceeding 35°C. During the night, the culture temperature was allowed to equilibrate to ambience.
2.3 Analytical procedures
Cell growth was monitored by measuring the dry biomass weight according to Chini Zittelli, Pastorelli, and Tredici ( 2000 ). The cell number was counted using a haemocytometer after appropriate dilution. The glucose concentration was determined with a Safe-Accu UG Blood Glucose Monitoring System (Model BGMS-1 Sinocare Inc., Changsha, China). The contents of total lipids were determined according to the method described in a previous study (Jia et al., 2015 ).
2.4 Technoeconomic analysis
The cost of heterotrophic cultivation for inoculum production was compared to that of conventional photoautotrophic culturing modes, including open-pond and PBR systems, both of which are widely used in algal industries and were thus used as references to evaluate the economic feasibility of heterotrophic cultivation here. The key input assumptions for TE analysis are summarized in Tables S1 and S2. To evaluate the effect of scale on production cost, it was assumed that two different inoculum production capacities are 1,000 and 10,000 tons per year on 300 operating days, respectively. A set of tubular photobioreactor for the imocula production is 98 m 3 , and the culture volume of the open pond was assumed to be 1,000 m 3 . According to our pilot-scale experimental results in the 1,000-L fermentor, we assumed the average harvest biomass concentration in a 120 m 3 inocula production fermentor is 200 g/L, achieved within 10 days in a batch of the production process. The initial biomass concentrations in open pond and tubular photobioreactor are 0.1 and 0.2 g/L, and their harvest biomass concentrations are 0.8 and 2 g/L, respectively. The collapse rate caused by biotic contamination in heterotrophic fermentation, open-pond, and tubular photobioreactor cultivations was assumed to be 10%, 15%, and 5%, respectively. For economic assumptions, all the capital and operating costs for both heterotrophic and photoautotrophic inocula production were estimated based on vendor quotes, previous studies, or standard engineering estimates. The biomass cost was calculated based on the model reported by Tapie and Bernard ( 1988 ). The cost structure includes the following two major parameters: capital investment costs and operating costs. The operating costs include fixed costs (e.g., labor, overhead, and maintenance) and variable costs (e.g., nutrients, power, CO2, and water). For heterotrophic and photoautotrophic culture, overhead is 60% and 2% of the installed equipment cost for labor and maintenance, respectively. The lifetime of open-pond and tubular PBR was assumed to be 10 and 15 years, respectively. For photoautotrophic culture, the CO2 was assumed to be supplied from a nearby power plant. Water was recycled in photoautotrophic culturing systems, and the evaporation rate of water in open pond was assumed to be 1 cm/day.
2.5 Statistical analysis
The values are expressed as mean ± standard deviation. The statistical tests were performed by using one-way analysis of variance in SPSS (version 19.0). Statistically significant difference was considered at p < .05.
3 Results and discussion
3.1 Leucine auxotrophy and inhibitory effect
To set up a cultivation process for E. coli K12 ER2507 the effect of leucine on cell growth was estimated. Therefore, shaking flask cultivation of E. coli K12 ER2507 with various initial leucine concentrations (0.0–3.0 g/L) were performed and the cell density was measured within the first 22 h of growth (Fig. 1A). We found that the highest specific growth rate was obtained from the lowest initial leucine concentration of 0.1 g/L, however this concentration is not sufficient to achieve maximum cell density, which requires a minimum of 0.2 g/L. The experiment was finished after the given time interval with still growing cultures which may reach higher final cell densities from more leucine. Very little growth was observed if leucine concentration was above 3.0 g/L −1 (Fig. 1A, first three data points are hidden behind others). As expected no growth was determined if no leucine supplementation was applied to the medium (Fig. 1A, data shown for 16, 17, 18 h as well tested in more extended overnight culture). The strain used carries a leuB6 mutation, resulting in a lack of isopropylmalate dehydrogenase, an enzyme involved in the biosynthesis of leucine 9 . Leucine is required as a signal molecule for the leucine-responsive regulatory protein, which regulates the expression of genes involved in different cellular functions. Shimada et al. 12 as well as Calvo and Matthews 13 state that this protein has a strong impact on cell metabolism in E. coli.
Based on cell densities, maximum specific growth rates were calculated and shown in Fig. 1B. The highest value 0.43 h −1 corresponds to the lowest leucine concentration of 0.1 g/L −1 . A cultivation performed with 3.0 g leucine results in a specific growth rate of only 0.19 h −1 . Rothen et al. 4 report this inhibitory effect of leucine as well.
3.2 Determination of biomass yield from leucine
The results of the first experiments demonstrated the expected necessity of leucine for cell growth while even low concentrations reduce the specific growth rate of the cells. For the setup of a fed-batch process the yield coefficient of biomass from leucine is an essential parameter. An unbalanced leucine addition could result in growth limitation or inhibition due to leucine accumulation, as shown above.
To investigate these relationships a full factorial, three level design of experiments (DoE) approach was used. The factor settings and results are given in Table 1. Glucose concentration was chosen in a range of a low value (2 g/L) up to a typical initial value for the batch phase of a HCDC, while the leucine concentration range was estimated from the experiments shown in Fig. 1, indicating a strong inhibitory effect above 0.6 g/L −1 .
|Ni||cLeu (g/L)||cGlc (g/L)||μmax (h −1 )||(g/g)|
The contour plots in Fig. 2 depict the results of the variation of leucine and glucose concentration, respectively. Figure 2A shows the strong influence of leucine on specific growth rate, while a moderate effect of glucose is observed. Figure 2B displays a strong interactive effect of glucose and leucine on the yield coefficient yX/Leu. Rothen et al. 4 report for E. coli HB101, which is auxotrophic for thiamine, proline, and leucine, a similar reduction in growth rate and yield coefficient. They investigated the medium requirements in continuous cultivation and determined a biomass yield of (20 ± 1) g CDM (g leucine) −1 . Under the assumption of a low glucose concentration in the culture, this value is in excellent agreement with ours.
Using Eq. 1, a yield coefficient of 19.8 g CDM (g leucine) −1 can be obtained for the lower limits of the design space, 0.1 g/L leucine and 2 g/L glucose, respectively.
3.3 Leucine feeding strategy for fed-batch cultivation
Under consideration of the inhibitory effect of leucine, especially in the presence of a higher glucose level, a feeding strategy for achieving high cell density was developed. The DoE approach indicated an optimal growth rate and biomass yield coefficient in the presence of low leucine and glucose concentrations. However, a leucine limitation due to an undersupply must be avoided.
In a classical HCDC process, the concentrations of many substrates in the start medium are rather high to supply the cells with the required nutrients over the course of the batch phase or, in the case of some nutrients, throughout the entire process. A typical batch medium according to Korz et al. 3 contains about 25 g/L glucose, a concentration which is shown to significantly reduce cell growth rate as well as cell yield from leucine (Fig. 1). Thus instead of a classical batch/fed-batch process an improved high cell density cultivation (iHCDC) was necessary to overcome problems with high initial concentrations of glucose and leucine. We completely removed the batch phase by filling the reactor with a C source free medium, and began directly with substrate feeding. This strategy has been reported to increase process to process reproducibility highly 14 . The authors used an exponential feed rate corresponding to a specific growth rate, which was approximately three-fourth of the maximum specific growth rate to obtain a self-controlling growth resulting in a stable biomass concentration and specific growth rate at induction time.
Our process starts with a glucose-free, low concentration leucine medium. Due to the reduced μset it is possible to keep both concentrations extremely low. μset of the initial exponentially increasing feed rate was chosen to result in one-third of the maximum cell specific growth rate. A second phase followed with reduced but still exponential feeding and therefore lower specific growth rate to avoid oxygen transfer limitation when reaching cell densities of more than 50 g CDM L −1 . Fast growing cells at high cell densities exhibit very high oxygen uptake rates, which must be met by the oxygen transfer capabilities of the reactor system. A third fed-batch phase, the production phase with for this situation optimized feeding data, was finished by depletion of the feed. All automated (MFCS) transition steps from one fed-batch phase to the next were triggered by reaching a predefined optical density (input of offline values) or calculated time to get this cell density during unwatched periods without sampling.
Leucine shows a low solubility at neutral pH, which increases with increasing pH 15 . Thus, the feeding of a sufficient amount of leucine was achieved by dissolving it in the base ammonia. Base addition is proportional to cell growth, due to its function as a nitrogen source. Reitzer 16 reported ammonia to be the N-source for highest growth rate and preferred nitrogen source for E. coli cultivation. From the biomass from leucine yield coefficient the required amount was calculated and dissolved in 25% NH3 , which was used as the base solution and dosed by the pH control unit during the cultivation.
Two cultivations with different amounts of leucine dissolved in ammonia are shown in Figs. 3 and 4. Both were carried out with the same parameters, μset,0 = 0.17 h −1 during first fed-batch phase and after 23 h with change to second fed-batch phase and μset,1 = 0.15 h −1 . The final growth rate was reduced to μset,2 = 0.12 h −1 after 31.4 h in cultivation A and 29.9 h in B.
The yield coefficient for the first cultivation (Figs. 3A and 4A) was derived from Eq. 1 for a nearly glucose-free medium (0.1 g glc L −1 ) and low concentration of 0.3 g leucine L −1 as yX/Leu = 19.3 g CDM (g leucine) −1 . It should be mentioned that this parameter set slightly extends the design space. From yX/Leu the amount of leucine required was calculated for 1.50 L feed solution, containing 750 g glucose L -1 , resulting with yX/Glc of 0.45 g CDM (g glc) −1 in 52.5 g leucine (L 25% NH3) −1 . Additionally, 0.3 g leucine L −1 was dissolved in 3 L start medium from which 0.5 L was used to prepare medium that was used for inoculum production in shaking flasks. This volume was later refilled as inoculum into the bioreactor to form 3 L inoculated medium with approximately 0.2 g/L leucine.
During fed-batch phase 1 after 28 h the acetate concentration increased drastically and growth dropped unexpectedly after 30.5 h, before initiation of the production phase. During the last phase of the cultivation the small increase in biomass together with increasing volume from feeding resulted in nearly constant cell concentration. The cultivation was finished after 33.5 h with a final CDM concentration of 74.9 g/L. The calculation of leucine concentration in 25% ammonia was based on the wrong assumption that the whole filling of the base bottle, 0.5 L, would be added during the cultivation. The used volume, dosed by the pH controller was much lower, resulting in a leucine undersupply with glucose accumulation (not analyzed) that may explain acetate formation, further growth reduction 17 , base overdosing caused by the acidification resulting in toxic ammonia concentration, and finally an increase of leucine concentration. The calculation of an actual yield coefficient for the leucine and glucose limited period until 30.5 h gave 18.7 g CDM (g leucine) −1 , which is below the value from the shaking flask cultivations and DoE evaluation. This further confirmed our conclusion of leucine underdosing during fed-batch phase 1.
The second cultivation (Figs. 3B and 4B) was carried out in the same manner but leucine concentration in the base was increased by 20% to 63.3 g leucine (L 25% NH3) −1 . The start medium concentration of 0.5 g leucine L −1 was considered and base consumption of 400 mL assumed. Finally, 1.5 L feed and 357 mL of the leucine/base solution were consumed forming 91.9 g CDM L −1 at a low final concentration of 0.68 g acetate L −1 and without any leucine accumulation.
This cultivation demonstrated a successful iHCDC strategy. HPLC analysis confirmed that accumulation of leucine and lower carbonic acids were prevented. The ratio of formed biomass to used leucine mass give an actual yield of 18.15 g CDM (g leucine) −1 for this cultivation strategy. This value is very close to the optimum yield from the DoE model and indicates that the DoE approach is an excellent tool to identify such parameters.
3.4 Demonstration of the usability of the iHCDC strategy by GFP production from recombinant E. coli
A proof of our new concept was the production of a gene product. Two cultivations were carried out under nearly identical conditions. Product concentration was evaluated from SDS-PAGE as shown in Fig. 5. Samples were prepared to ensure input of protein from an equal amount of biomass per lane to avoid overloading. Therefore, increasing density of the product band at 26.9 kDa is the result of increasing cell-specific product content. During the production phase cell density, CDM and cell-specific product load increased. Calculated from the latter, the product mass in the culture reached 6.05 g (mean from two cultivations) with a range of 1.10 g. With respect to the final product concentration or mass, the cultivations show a limited reproducibility, most likely due to product analytics and quantification by SDS-PAGE.
The final product mass was slightly below the maximum value that may indicate product degradation by proteases (Fig. 6). Unfortunately, the E. coli strain K12 ER2507 is not deficient in Lon and OmpT proteases, in contrast to strains such as BL21.
From averaging both cultivations, we derived a mean of 6.05 g GFP in 438 g CDM (range 3.9 g) corresponding to 1.30 g GFP L −1 (range 0.22 g/L). The productivity is in good agreement with the literature for other production strains 18 . With further effort to optimize the production phase with respect to induction cell density, inductor concentration, and cell-specific growth rate driven by the feeding rate, the product yield may increase significantly.
In this work, we genetically engineered our E. coli MBE producer strain making an improvement of 30.6 % in the final titer in batch cultures. After that, a fed-batch microbial fermentation process was designed and optimized, achieving a maximum yield of 790.2 ± 6.9-mg MBE L −1 and a volumetric productivity of 15.8 ± 1.1-mg MBE (L h) −1 . Under these cultivation conditions, addition of propionate, n-octanol, and oleic acid to growth medium was carried out. Although addition of exogenous substrates is not convenient from an economically point of view, this was done to obtain maximum yield of MBE. Finally, we scaled-up the production of bacterial MBE, developed a purification method, and measured physicochemical properties related to their thermal behavior.
DSC TO and TCOM values, as well as PP, revealed that these molecules exhibit a more extended range of work at low temperatures than jojoba oil, that is, a better fluidity and enhanced cold‐temperatures performance. At the same time, MBE present TOX and TPK values close to those corresponding to jojoba oil, indicating that the thermo-oxidative properties of both oils are comparable.
These results encourage us to further characterize MBE oil by extensive physicochemical properties determinations, such as measurement of viscosity index and refraction coefficient. This will give us a more precise idea about concrete MBE applications, as well as generating a more clear comprehension of the principles that underlie the relationship between chemical structure and possible industrial uses for oleochemicals. Finally, given the remarkable diversity of MBE compounds that could be synthesized in vivo, it is likely that a similar microbial engineering approach can yield high quality or “selected” esters of designed structures suitable for defined or specific applications.
Schirmer, A., Rude, M.A., Li, X., Popova, E. & del Cardayre, S.B. Microbial biosynthesis of alkanes. Science 329, 559–562 (2010).
Gronenberg, L.S., Marcheschi, R.J. & Liao, J.C. Next generation biofuel engineering in prokaryotes. Curr. Opin. Chem. Biol. 17, 462–471 (2013).
Woolston, B.M., Edgar, S. & Stephanopoulos, G. Metabolic engineering: past and future. Annu. Rev. Chem. Biomol. Eng. 4, 259–288 (2013).
Paddon, C.J. & Keasling, J.D. Semi-synthetic artemisinin: a model for the use of synthetic biology in pharmaceutical development. Nat. Rev. Microbiol. 12, 355–367 (2014).
Kim, E., Moore, B.S. & Yoon, Y.J. Reinvigorating natural product combinatorial biosynthesis with synthetic biology. Nat. Chem. Biol. 11, 649–659 (2015).
Nielsen, J. et al. Engineering synergy in biotechnology. Nat. Chem. Biol. 10, 319–322 (2014).
Na, D. et al. Metabolic engineering of Escherichia coli using synthetic small regulatory RNAs. Nat. Biotechnol. 31, 170–174 (2013).
Lidstrom, M.E. & Konopka, M.C. The role of physiological heterogeneity in microbial population behavior. Nat. Chem. Biol. 6, 705–712 (2010).
Müller, S., Harms, H. & Bley, T. Origin and analysis of microbial population heterogeneity in bioprocesses. Curr. Opin. Biotechnol. 21, 100–113 (2010).
Taniguchi, Y. et al. Quantifying E. coli proteome and transcriptome with single-molecule sensitivity in single cells. Science 329, 533–538 (2010).
Li, G.W. & Xie, X.S. Central dogma at the single-molecule level in living cells. Nature 475, 308–315 (2011).
Guimaraes, J.C., Rocha, M. & Arkin, A.P. Transcript level and sequence determinants of protein abundance and noise in Escherichia coli. Nucleic Acids Res. 42, 4791–4799 (2014).
Zenobi, R. Single-cell metabolomics: analytical and biological perspectives. Science 342, 1243259 (2013).
Paige, J.S., Nguyen-Duc, T., Song, W. & Jaffrey, S.R. Fluorescence imaging of cellular metabolites with RNA. Science 335, 1194 (2012).
Love, K.R., Panagiotou, V., Jiang, B., Stadheim, T.A. & Love, J.C. Integrated single-cell analysis shows Pichia pastoris secretes protein stochastically. Biotechnol. Bioeng. 106, 319–325 (2010).
Mustafi, N., Grünberger, A., Kohlheyer, D., Bott, M. & Frunzke, J. The development and application of a single-cell biosensor for the detection of L -methionine and branched-chain amino acids. Metab. Eng. 14, 449–457 (2012).
Labhsetwar, P., Cole, J.A., Roberts, E., Price, N.D. & Luthey-Schulten, Z.A. Heterogeneity in protein expression induces metabolic variability in a modeled Escherichia coli population. Proc. Natl. Acad. Sci. USA 110, 14006–14011 (2013).
Delvigne, F., Zune, Q., Lara, A.R., Al-Soud, W. & Sørensen, S.J. Metabolic variability in bioprocessing: implications of microbial phenotypic heterogeneity. Trends Biotechnol. 32, 608–616 (2014).
Lu, X., Vora, H. & Khosla, C. Overproduction of free fatty acids in E. coli: implications for biodiesel production. Metab. Eng. 10, 333–339 (2008).
Xu, P. et al. Modular optimization of multi-gene pathways for fatty acids production in E. coli. Nat. Commun. 4, 1409 (2013).
Blazeck, J. et al. Harnessing Yarrowia lipolytica lipogenesis to create a platform for lipid and biofuel production. Nat. Commun. 5, 3131 (2014).
Zhang, F., Carothers, J.M. & Keasling, J.D. Design of a dynamic sensor-regulator system for production of chemicals and fuels derived from fatty acids. Nat. Biotechnol. 30, 354–359 (2012).
Lawrence, M.S., Phillips, K.J. & Liu, D.R. Supercharging proteins can impart unusual resilience. J. Am. Chem. Soc. 129, 10110–10112 (2007).
Lütke-Eversloh, T., Santos, C.N. & Stephanopoulos, G. Perspectives of biotechnological production of L-tyrosine and its applications. Appl. Microbiol. Biotechnol. 77, 751–762 (2007).
Pittard, J., Camakaris, H. & Yang, J. The TyrR regulon. Mol. Microbiol. 55, 16–26 (2005).
Liu, D., Xiao, Y., Evans, B.S. & Zhang, F. Negative feedback regulation of fatty acid production based on a malonyl-CoA sensor-actuator. ACS Synth. Biol. 4, 132–140 (2015).
Doroshenko, V. et al. YddG from Escherichia coli promotes export of aromatic amino acids. FEMS Microbiol. Lett. 275, 312–318 (2007).
Chou, H.H. & Keasling, J.D. Programming adaptive control to evolve increased metabolite production. Nat. Commun. 4, 2595 (2013).
Conrad, T.M. et al. RNA polymerase mutants found through adaptive evolution reprogram Escherichia coli for optimal growth in minimal media. Proc. Natl. Acad. Sci. USA 107, 20500–20505 (2010).
Nakata, K., Koh, M.M., Tsuchido, T. & Matsumura, Y. All genomic mutations in the antimicrobial surfactant-resistant mutant, Escherichia coli OW66, are involved in cell resistance to surfactant. Appl. Microbiol. Biotechnol. 87, 1895–1905 (2010).
Foster, P.L. Stress-induced mutagenesis in bacteria. Crit. Rev. Biochem. Mol. Biol. 42, 373–397 (2007).
Dietrich, J.A., Shis, D.L., Alikhani, A. & Keasling, J.D. Transcription factor-based screens and synthetic selections for microbial small-molecule biosynthesis. ACS Synth. Biol. 2, 47–58 (2013).
Raman, S., Rogers, J.K., Taylor, N.D. & Church, G.M. Evolution-guided optimization of biosynthetic pathways. Proc. Natl. Acad. Sci. USA 111, 17803–17808 (2014).
Veening, J.W., Smits, W.K. & Kuipers, O.P. Bistability, epigenetics, and bet-hedging in bacteria. Annu. Rev. Microbiol. 62, 193–210 (2008).
Jablonka, E. & Raz, G. Transgenerational epigenetic inheritance: prevalence, mechanisms, and implications for the study of heredity and evolution. Q. Rev. Biol. 84, 131–176 (2009).
Kiviet, D.J. et al. Stochasticity of metabolism and growth at the single-cell level. Nature 514, 376–379 (2014).
Keasling, J.D. Manufacturing molecules through metabolic engineering. Science 330, 1355–1358 (2010).
Tanaka, A. & Nakajima, H. Application of immobilized growing cells. Adv. Biochem. Eng. Biotechnol. 42, 97–131 (1990).
Barber, W.P. & Stuckey, D.C. The use of the anaerobic baffled reactor (ABR) for wastewater treatment: a review. Water Res. 33, 1559–1578 (1999).
Dahl, R.H. et al. Engineering dynamic pathway regulation using stress-response promoters. Nat. Biotechnol. 31, 1039–1046 (2013).
Zhang, F. & Keasling, J. Biosensors and their applications in microbial metabolic engineering. Trends Microbiol. 19, 323–329 (2011).
Fernandes, R.L. et al. Experimental methods and modeling techniques for description of cell population heterogeneity. Biotechnol. Adv. 29, 575–599 (2011).
van Heerden, J.H. et al. Lost in transition: start-up of glycolysis yields subpopulations of nongrowing cells. Science 343, 1245114 (2014).
Wang, B.L. et al. Microfluidic high-throughput culturing of single cells for selection based on extracellular metabolite production or consumption. Nat. Biotechnol. 32, 473–478 (2014).
Levine, E. & Hwa, T. Stochastic fluctuations in metabolic pathways. Proc. Natl. Acad. Sci. USA 104, 9224–9229 (2007).
Oyarzún, D.A., Lugagne, J.B. & Stan, G.B. Noise propagation in synthetic gene circuits for metabolic control. ACS Synth. Biol. 4, 116–125 (2015).
Lee, T.S. et al. BglBrick vectors and datasheets: a synthetic biology platform for gene expression. J. Biol. Eng. 5, 12 (2011).
Engler, C., Kandzia, R. & Marillonnet, S. A one pot, one step, precision cloning method with high throughput capability. PLoS One 3, e3647 (2008).
Kempe, K., Hsu, F.F., Bohrer, A. & Turk, J. Isotope dilution mass spectrometric measurements indicate that arachidonylethanolamide, the proposed endogenous ligand of the cannabinoid receptor, accumulates in rat brain tissue post mortem but is contained at low levels in or is absent from fresh tissue. J. Biol. Chem. 271, 17287–17295 (1996).
Juminaga, D. et al. Modular engineering of L -tyrosine production in Escherichia coli. Appl. Environ. Microbiol. 78, 89–98 (2012).
Solving the Puzzle of Cell Culture Optimization
Manufacturing engineers formerly used complex media from animal-derived products, but the burgeoning, modern-day industry is quickly adopting chemically defined media wherever possible. [JVisentin/Getty Images]
Biologics have transformed the therapeutic landscape. Twenty years ago, the pharmaceutical industry could hardly have envisioned the growth and broad impact that biologics would have today, according to Kendra Hightower, Ph.D., senior study director at Metabolon: “Now, they’re not just part of the therapeutic landscape—they’ve become key defining components of most pharmaceutical pipelines.”
While reducing costs, increasing production, and meeting regulatory requirements have always driven biopharma to greater heights, the complexity of manufacturing biologically produced molecules has made these demands more mountainous than ever. “A bioprocess is a puzzle,” explained Dr. Hightower, “and this puzzle has thousands and thousands of pieces, and those pieces need to fit together correctly to make the final picture. If you change those pieces, either intentionally or unintentionally, that could alter the final picture.”
Assessing the activity of a living system using metabolomics requires robust, high-throughput analytical systems that can accurately identify metabolites, a comprehensive understanding of biochemical pathways, and expertise in analytics and biochemistry to successfully interpret and apply the results. [ReptileB488/Getty Images] In late August 2017, leaders in the bioprocessing industry gathered in Boston, MA, at Cambridge HealthTech Institute’s Bioprocessing Summit conference to discuss cost-effective strategies to ensure a picture-perfect product every time.S
Searching for Hidden Elements with Metabolomics
At the conference, Dr. Hightower illustrated how biopharmaceutical manufacturers can use a metabolomics approach to gain insight into the “active biology” of a bioprocess that can help minimize process variability. Dr. Hightower’s company, Metabolon, has specialized in metabolic profiling since 2000, and both Metabolon’s vast chemical reference library and their multiple points of matching strategy for metabolite identification enable highly accurate and comprehensive metabolite measurements.
Metabolomics expands on traditional methods that use 10 to 20 indicative biochemical molecules, like oxygen, carbon dioxide, and glucose, to track cell metabolism, growth, and productivity. While these metabolites “happen to be really useful,” said Kirk Beebe, Ph.D., senior director at Metabolon, “there are hidden elements of active biology where a precision metabolomics approach could add to the lactates, ammonias, and amino acid profiling that people currently do.”
These “hidden elements” can provide valuable knowledge about cell health and protein production. For example, glutathione can relay information about oxidative stress levels and the redox capacity of the cells—an important metric for recombinant protein production, which relies on redox potential for disulfide bond formation, proper protein folding, and protein secretion.
Assessing the activity of a living system using metabolomics requires robust, high-throughput analytical systems that can accurately identify metabolites, a comprehensive understanding of biochemical pathways, and expertise in analytics and biochemistry to successfully interpret and apply the results. However, the reduction in cost and time-to-consumer achieved by applying an extensive metabolomics approach to bioprocess development and monitoring makes these challenges worth overcoming.
While Metabolon builds understanding around the integral role different pieces play in creating a reliable process, companies such as Merck, Valitacell, Cell Culture Company (C3), and Roche have focused their efforts on the pieces themselves, with their respective work on media development, cell-line selection, bioreactor design, and continuous process monitoring.
Metabolomics can assess factors that influence the active biology of the cells in the bioprocess such as genetics, operational environments, and nutritional requirements. Metabolon’s technology provides a broad, informative assessment of biochemical space (i.e., active biology), expanding the avenues for optimizing production systems beyond conventional technologies.
Finding the Perfect (Cell Culture Media) Recipe through Systems Biology
Wai Lam Ling, Ph.D., Merck’s senior principal scientist and group leader of biologics upstream process and media development, provided GEN with an analogy that compared motorcycles to small-molecule drugs and jet planes to biologic therapies: like biologics, jet planes are complex to design and manufacture. Thus, to get their products off the ground, biomanufacturers need to select the right components—starting with cell culture media—to create a robust, reliable process from production cell lines.
Manufacturing engineers formerly used complex media from animal-derived products, but the burgeoning, modern-day industry is quickly adopting chemically defined media wherever possible. In contrast to complex media, chemically-defined media consists of individual, known components and is devoid of animal-derived materials and hydrolysates, which removes much of the mystery and variability inherent in complex media. However, with anywhere from 50 to 100 different components to balance, finding the perfect recipe for chemically defined media is still an industry headache.
Even well-planned design-of-experiment methodologies for media development can be both time- and resource-consuming. By using a systems biology approach that leverages the tools used in drug discovery, such as next-generation sequencing, shotgun proteomics, and metabolite profiling, Dr. Ling and her team can visualize how changing the selection, concentration, and chronology of different components affects cell productivity and product quality. Integrating the data with other important product attribute measurements, like glycosylation, allows the team to generate predictive models for media optimization and reduce the time and money spent finding the perfect formulation to complement their bioprocesses.
Identifying Clonal Instability Culprits through ChemStress Fingerprinting
While tailoring the media to the process can mitigate cellular stress caused by a bioreactor environment, starting out with a robust, stable cell line is essential to a successful process. Nutrient starvation, oxidative stress, toxicity, unfriendly pH or osmolarity levels, and other stressors can cause clonal instability, where productive cell lines become unproductive or produce errant proteins that require extensive purification.
To suss out instability, biopharmaceutical manufacturers often use subculturing studies that assess cell productivity over multiple (60 to 160) generations. However, these studies are both time- and resource-intensive, and, according to Terry McWade, CEO at Valitacell, they’re not always effective: “High numbers of cells are being brought to the last stage [of development], and then companies are finding that many of those cells are actually not stable.”
In 2014, Valitacell founders McWade and Jerry Clifford asked themselves if they could recreate the bioreactor environment by coating a 96-well plate with small-molecule chemicals that mimic different stressors, prompting the development of Valitacell’s latest product: ChemStress Fingerprinting.
Combined with a software platform that plots growth and product titer for each chemical challenge, ChemStress produces a fingerprint unique to the cell clone and culture media combination tested. Scientists can use changes in the fingerprint to detect clonal instability (Valita™STABILITY), as well as batch-to-batch variation in culture media (Valita™QC).
“The key exciting part of the [ChemStress] technology,” said McWade, “is the ability to determine much, much more quickly which cells are likely to be unstable.” While conventional subculturing studies can take 60 to 160 days, ChemStress Fingerprinting can detect clonal instability in 24 days.
Decreasing the time and money spent on cell line development can relieve some of the economic pressure biopharmaceutical companies face as newer, more targeted therapies emerge, requiring more projects, faster development, and more efficient processes. Not only has this pressure prompted innovative new solutions, but it has also provided incentive for manufacturers to re-evaluate established technologies. Some companies are even designing media specifically for continuous biomanufacturing operations.
Valitacell’s ChemStress technology can recreate the stressors experienced by cells in a bioreactor that can lead to clonal instability. After a three-day culture on the plate, cell growth and product titer are measured and plotted against the chemical challenge to produce a fingerprint unique to the cell clone and culture media combination. Changes in the fingerprint over time can be used to predict clonal instability.
Perfusion Hollow-Fiber Bioreactors: An Old Trick for New Processes
Patented in 1974 by Kruznak et al., perfusion hollow-fiber bioreactors (PHFB) consist of bundles of thousands of hollow fibers constructed from semipermeable membranes and enclosed in a cylindrical cartridge. Similar to capillaries within the body, the porous membrane structure of the fibers allows oxygen and nutrient transport to cells immobilized on the exterior of the fibers and removal of waste products. The fiber bundles also increase the surface area available for cell growth considerably, which promotes a higher, more tissue-like density (
108–9 cells/mL) than fed-batch bioreactors (
One of the greatest advantages of PHFBs, however, is the ability to maintain cells in culture for several months at a time. Scott Waniger, vice president of bioprocessing at C3, remarked, “biotechnology has kind of been lagging behind the rest of the industrial world when it comes to continuous manufacturing.” PHFBs could make a significant contribution to closing this gap, especially compared to the 14 – 21-day runs typical of fed-batch reactors. Continuous manufacturing strategies can significantly reduce costs, enabling the production of “difficult-to-express” proteins that fall below the standard production rate of 1 to 10 g/L, making them too costly to produce using conventional methods.
C3, a contract development and manufacturing organization (CDMO), has used perfusion technology since 1981. Their largest bioreactor, the Acusyst (automated cell culture system) Xcellerator, runs 20 single-use hollow-fiber cartridges in parallel to accommodate a 1,600-L volume. The refrigerator-sized bioreactor is modular and can be scaled out by adding more units to increase the volume for large-scale production. In addition to providing a linearly scalable system, the design also eliminates the need for seed trains, which can increase time to production and requires additional bioreactors to gradually grow the culture before inoculating the final full-capacity production tank.
Continuous Process Monitoring with a Virtual Sensor
Designing a successful process to produce a particular biologic depends not only on fitting together the right cell culture media, cell line, and bioreactor, but also on careful control and monitoring of the system.
One of the most relevant parameters used for bioprocess control and monitoring is cell growth. These measurements are typically obtained through manual sampling and off-line cell count or optical density measurements during cultivation. However, Wolfgang Paul, Ph.D., senior scientist and group leader at Roche Innovation Center Munich, Roche Diagnostics, is developing a new approach to allow continuous monitoring of cell growth without manual sampling or the addition of bulky hardware.
The method involves a soft sensor, also referred to as a virtual sensor, that can calculate cell growth based on data already collected: oxygen input, carbon dioxide output, heating and cooling units, pH value, base addition, and agitation speed. The soft sensor uses multiple linear regression and artificial neural network models to relate these data to cell growth.
Loosely modeled after neural networks in the brain, artificial neural networks use interconnected nodes that imitate neurons to pass along information. According to Dr. Paul, the various nonlinear fitting abilities make this machine-learning approach a promising model for generating the type of accurate estimations needed for their application.
“We initiated the development of the soft sensor on small-scale bioreactors, because there are many vessels (mostly single-use bioreactors) in use with no space for additional hardware sensors,” said Dr. Paul. While their initial studies focused on smaller 250-L bioreactors, the soft sensor should, in theory, function independently of bioreactor size.
According to Dr. Paul the benefits of a non-invasive, continuous monitoring system that can reduce the need for manual sampling are “very prominent for the small-scale bioreactors,” which play an increasingly important role in process development and characterization for manufacturing large-molecule therapeutics.
Biologics have transformed the therapeutic landscape, and they continue to do so as the emergence of targeted and personalized therapies require even more innovation from the biomanufacturing industry. Manufacturers will need more flexible processes that can deliver consistent, high-quality products in less time and at lower cost than ever before to accommodate the growing needs of the industry and ensure process reliability.
This work was partially carried out in the Biotechnology Core Laboratory, NIH/NIDDK. We would like to thank Prof. Dr. Joseph Shiloach for his kind assistance. We also would like to thank the NIH Fellows Editorial Board (FEB) for editorial assistance.
AOX, alcohol oxidase DO, dissolved oxygen FDH, formate dehydrogenase FLD, formaldehyde dehydrogenase OD600, optical cell density at 600 nm pNPG, p-nitrophenyl -β-D-glucopyranoside YPD, yeast extract peptone dextrose.
High cell density cultivation of the chemolithoautotrophic bacterium Nitrosomonas europaea
Nitrosomonas europaea is a chemolithoautotrophic nitrifier, a gram-negative bacterium that can obtain all energy required for growth from the oxidation of ammonia to nitrite, and this may be beneficial for various biotechnological and environmental applications. However, compared to other bacteria, growth of ammonia oxidizing bacteria is very slow. A prerequisite to produce high cell density N. europaea cultures is to minimize the concentrations of inhibitory metabolic by-products. During growth on ammonia nitrite accumulates, as a consequence, N. europaea cannot grow to high cell concentrations under conventional batch conditions. Here, we show that single-vessel dialysis membrane bioreactors can be used to obtain substantially increased N. europaea biomasses and substantially reduced nitrite levels in media initially containing high amounts of the substrate. Dialysis membrane bioreactor fermentations were run in batch as well as in continuous mode. Growth was monitored with cell concentration determinations, by assessing dry cell mass and by monitoring ammonium consumption as well as nitrite formation. In addition, metabolic activity was probed with in vivo acridine orange staining. Under continuous substrate feed, the maximal cell concentration (2.79 × 10 12 /L) and maximal dry cell mass (0.895 g/L) achieved more than doubled the highest values reported for N. europaea cultivations to date.
This is a preview of subscription content, access via your institution.