Cultivation and Bioprocessing Techniques & Design of Experiments (DoE) Dr. Simon K.-M. R. Rittmann 1. Closed batch (serum bottles, usually anaerobic) 2. Batch (e.g. Erlenmeyer flask, uncontrolled conditions) 3. Batch (bioreactor, (un)controlled conditions) 4. Fed-batch (bioreactor, controlled conditions) 5. Continuous culture (bioreactor, controlled conditions) Rittmann et al. 2012, Rittmann et al. 2015 Cultivation & Bioprocessing Techniques Closed batch cultivation of Methanothermobacter marburgensis, Methanothermococcus okinawensis, Methanocaldococcus villosus and Methanosarcina soligelidi in 120 mL serum bottles. Closed batch Mauerhofer et al. 2018 Closed batch flush and purge weigh determine pressure weigh incubate 0 1 2 3 4 5 6 pressure mass 4H2 + CO2  CH4 + 2H2O Taubner & Rittmann, 2016 Closed batch Cummulative H2O production of M. marburgensis, M. villosus and M. okinawensis in 120 mL serum bottles. Growth conditions: T = 65, 80 and 60°C, respectively, V = 50 mL, n = 9, 3 and 3, respectively. Taubner & Rittmann, 2016 Closed batch MER = methane evolution rate OD = optical density Taubner & Rittmann, 2016 Closed batch General mass balance Bioprocessing techniques Satzverfahren (batch) Kontinuierliche Kultur (continuous culture) Spektrum Akademischer Verlag, 2006 Batch & Fed-batch Mass balance for batch and fed-batch Batch & Fed-batch Mauerhofer et al. 2018 Batch Biological CH4 production – fed-batch © Simon K.-M. R. Rittmann Fed-batch Fed-batch fermentation of Methanothermobacter marburgensis in the Eppendorf bioreactor system. Fed-batch Mauerhofer et al. 2018 Fed-batch Growth (OD578nm) of M. okinawensis (left) and M. marburgensis (right) in fed-batch cultivation mode. Run 1 and run 2 are replicates. Abdel Azim et al., 2017 Fed-batch M. okinawensis M. marburgensis Methane evolution rate (MER) of M. okinawensis (left) and M. marburgensis (right) in fed-batch cultivation mode. Run 1 and run 2 are replicates. Abdel Azim et al., 2017 M. okinawensis M. marburgensis Fed-batch Abdel Azim et al., 2017 Specific methane evolution rate (qCH4) of M. okinawensis (left) and M. marburgensis (right) in fed-batch cultivation mode. Run 1 and run 2 are replicates. M. okinawensis M. marburgensis Fed-batch Results from the exponential fed-batch cultivation using M. marburgensis. For each run (colour legend) are presented the values of X, x, µ on the xaxis. Run 3 (orange bar) had the highest biomass (X [g]) and biomass concentration (x [g L-1]). Abdel Azim et al., 2017 Exponential fed-batch M. marburgensis Results from the exponential fed-batch cultivation using M. marburgensis. For each run the values MER, qCH4, CH4 offgas are presented on the xaxis. Run 2 (red bar) showed the highest MER and qCH4. During run 6 the highest CH4 off-gas concentration was obtained. Abdel Azim et al., 2017 M. marburgensis Exponential fed-batch Model for biomass concentration (x) calculated from the exponential fed-batch data of M. marburgensis. Abdel Azim et al., 2017 Exponential fed-batch Mauerhofer et al. 2018 Continuous culture Mass balance for continuous culture Continuous culture Spektrum Akademischer Verlag, 2006 Liquid limitation Continuous culture Principles of continuous culture bioprocessing  Liquid limitation: thin line  Gas limitation: bold line Continuous culture Seifert et al., 2014 Continuous culture Continuous culture Rittmann et al. 2012 vvm [L L-1 min-1] D [h-1] qCH4[mmolg-1h-1] Continuous culture Rittmann et al. 2018 Rittmann et al. 2018 Continuous culture Figure 1. An overview of dynamic process conditions which can be used in bioprocess development.  A: shift-up  B: shift-down  C: ramp-up  D: ramp-down  E: pulse  F: oscillation Spadiut et al. 2013 Dynamic process conditions Dynamic condition Modus operandi Changed condition Identification/Optimisation of Shift rapid change of parameter(s) followed by stable condition(s) physical parameters (D, T, rpm, vvm, light intensity, feed profile) chemical parameters (pH, nutrient concentrations, osmolality) physiological parameters (µ, qi) maximum physiological capacity (µmax, qs,max) productivity (qp,max) maintenance energy stress response metabolism morphology changes limitations Ramp continuous and slow change of parameter(s) productivity yields growth and production kinetic morphology viability limitations physiological capacity Pulse sudden change of parameter(s) productivity uptake rates yields growth kinetics short time cellular response product and metabolite release unscramble physiological and metabolic changes strain characteristic parameters (qs, qp) Oscillation controlled short up and down ramp(s) in a defined or changing frequency and/or amplitude productivity growth kinetics metabolic and physiological optimisation heat and stress response quality improvement metabolite formation Dynamic process conditions Spadiut et al. 2013 Rittmann et al. 2012 Identification of the maximum specific CH4 evolution rate (qCH4,max). Dynamic process conditions Bernacchi et al. 2014 Dynamic process conditions Mauerhofer et al. 2018 Anaerobic cultivation techniques Mauerhofer et al. 2018 Anaerobic cultivation techniques Mauerhofer et al. 2018 Anaerobic cultivation techniques Mauerhofer et al. 2018 Anaerobic cultivation techniques • DoE was fouded by Ronald Fisher (UK), who basically developed factoral experiments as well as ANalysis Of VAriance • George Box developed basis for optimisation of DoE designs (Response Surface Modeling (RSM) • „To find out what happens if you change something, is necessary to change it.“ • „Esentially all models are wrong, but some are useful“ • Within the DoE concept Gen‘ichi Taguchi (Japan) developed a qualitative approach (Taguchi-Methodology) Gen‘ichi Taguchi George Box Ronald Fisher Introduction – DoE Why do we need DoE? • Which (process/cultivation/environmental) parameters have which influence on which variables (response of an organism)? • How can we determine with a minimum of experiments which parameters and interactions of parameters are beneficial/detrimental for the cultivation of an organism in an experimental design space? Introduction – DoE Why do we need DoE? • Classical way to perform an experiment is to vary one parameter (factor) at a time  OVAT (one-variable-at-a-time) Drawbacks • Time consuming • Interactions • Maybe the optimum will not be identified Introduction – DoE Why do we need DoE? • Determine parameters (independent variables), which influence responses (dependent variables). • Optimise cultivation (process) • Improve growth, product quality, quantity... DoE requires • Planning of randomised experiments • Dicipline • Application of statistics http://www.umetrics.com Introduction – DoE Source: http://www.gmpua.com/World/GMPManual/daten/autorenteil/kapitel_07/07_i.htm (face centered) DoE designs 1.) Screening • Full or fractional factorial designs • Resolution V designs are best to be used, but also • Resolution IV designs are possible 2.) Optimization or modelling • Response Surface Model (RSM) • Cetral composite, Box-Behnken or Taguchi-design 3.) Verification Types of DoE experiments Screening • Good for first experiment(s) • Can consider lots of variables • Usually only two levels of each variable • Relatively few runs • Limited if any ability to identify interactions • (depending on the design) • Risky? P.G Mathews, 2012 DoE – Screening designs P.G Mathews, 2012 Screening • Useful for estimating main effects and interactions • Fractional factorial design can be used for screenign many factors to find the significant few DoE – Screening designs Color coding represents the design resolution: green = resolution V design or higher, yellow = resolution IV design and red = resolution III design Design Expert (Stat-Ease Inc., USA) DoE – Screening designs Factorial Effects Aliases [Est. Terms] Aliased Terms [Intercept] = Intercept [A] = A + BCE + DEF [B] = B + ACE + CDF [C] = C + ABE + BDF [D] = D + AEF + BCF [E] = E + ABC + ADF [F] = F + ADE + BCD [AB] = AB + CE [AC] = AC + BE [AD] = AD + EF [AE] = AE + BC + DF [AF] = AF + DE [BD] = BD + CF [BF] = BF + CD [ABD] = ABD + ACF + BEF + CDE [ABF] = ABF + ACD + BDE + CEF Factor Generator E = ABC F = BCD Factorial Effects Defining Contrast I = ABCE = ADEF = BCDF Factorial Effects Aliases [Est. Terms] Aliased Terms [Intercept] = Intercept [A] = A [B] = B [C] = C [D] = D [E] = E [AB] = AB + CDE [AC] = AC + BDE [AD] = AD + BCE [AE] = AE + BCD [BC] = BC + ADE [BD] = BD + ACE [BE] = BE + ACD [CD] = CD + ABE [CE] = CE + ABD [DE] = DE + ABC Factor Generator E = ABCD Factorial Effects Defining Contrast I = ABCDE 25-1 26-2 Resolution 5 design Resolution 4 design DoE – Screening designs Optimization • Good follow-up experiment to a screening experiment • Fewer variables - generally the most important ones • Often three or more levels of each variable • Provide a more complex model for the process DoE – Optimisation designs P.G Mathews, 2012 Central composite • Each numeric factor is varied over 5 levels • plus and minus α (axial points) • plus and minus 1 (factorial points) • usually three to six center points If factorial factors have to be added the central composite design will be doublicated for every combination of the categorial factor levels DoE – Optimisation designs DoE – Examples closed batch Taubner et al. 2018 CH2O [µL L-1] NH4Cl [g L-1] turnoverrate[h-1] Turnover rate in [h-1] as function of CH2O and NH4Cl concentrations. The turnover rate reached its maximum value at low CH2O concentration. At high CH2O concentration the turnover rate is higher for low NH4Cl concentrations. This study was based on a DoE approach. Taubner et al. 2018 DoE – Examples closed batch Organism: Nitrososphera viennensis Factors: • Ammonia concentration 1, 2.5 and 4 mM • Pyruvate concentration 0.1, 0.8 and 1.5mM • Temperature 37, 42 and 47 °C Calculation of: • NH4 uptake rates [mmol L-1 h-1] • NO2- production rates [mmol L-1 h-1] • Cell counts • Specific growth rate [h-1] from NO2- production rate Stieglmeier et al., 2014 DoE – Examples batch µ [h-1] Temperature [°C] c(pyr) [mM] The graph illustrates the effect of pyruvate concentration (c(pyr)) [mM] and temperature [°C] on the growth rate (µ) [h-1] of EN76T at a fixed ammonium concentration (c(NH4 +)) of 2.5 mM. Based on the results of the closed batch cultivation and the subsequent generated response surface model (RSM), the optimal conditions for the cultivation of EN76T within this three-factorial design space could be retrieved. The optimal cultivation conditions using µ as target value for maximization were calculated as follows: c(pyr) = 1.15 mM, c(NH4 +) = 2.03 mM, temperature = 42.02 °C, with a desirability of 0.854. Data points of the individual experiments are presented in red or rose colour. DoE – Examples batch Mauerhofer et al. 2018 M. thermaggregans DoE – Examples fed-batch M. thermaggregans Mauerhofer et al. 2018 DoE – Examples fed-batch Run DM [h-1] pH DS [L L-1 d-1] T [°C] rpm vvm [L L-1 min-1] ratio (H2/CO2) DN [L L-1 d-1] CNH4+ [mmol L-1] x [g L-1] MER [mmol L-1 h-1] qCH4 [mmol g-1 h-1] CH4 offgas [Vol.%] Y(X/CH4) [C- mol/mo l] rx [C- mmol L-1 h-1] DoR-balance C- balance N- balance 1 0.043 6.16 0.012 60 654 0.16 3.0 0.014 54.2 0.92 16.4 17.8 4.5 0.07 1.21 96.8% 75.3% 84.1% 2 0.055 7.83 0.013 60 1243 0.19 4.9 0.051 111.9 1.35 45.1 33.4 13.6 0.05 2.25 95.7% 101.8% 142.3% 3 0.211 7.84 0.010 60 1242 0.49 3.0 0.019 26.6 1.14 105.0 91.9 11.6 0.07 7.35 98.5% 80.7% 84.5% 4 0.057 6.16 0.049 61 1243 0.50 5.0 0.016 37.4 3.80 114.0 30.0 12.9 0.06 6.73 94.6% 99.5% 95.0% 5 0.21 6.16 0.053 61 1245 0.20 3.0 0.042 58.5 0.88 71.4 80.9 28.4 0.08 5.72 101.9% 95.6% 103.5% 6 0.059 6.15 0.011 70 1242 0.50 3.0 0.062 128.7 2.76 99.2 36.0 10.5 0.05 4.96 97.2% 124.0% 108.3% 7 0.202 6.15 0.012 70 1245 0.20 5.1 0.016 38.2 0.79 65.2 82.8 23.8 0.08 4.89 103.3% 93.7% 82.8% 8 0.161 7.85 0.009 69 650 0.16 3.1 0.044 64.3 0.27 17.8 66.0 5.0 0.08 1.33 101.6% 96.6% 100.7% 9 0.056 7.85 0.044 70 1245 0.19 2.9 0.014 44.4 1.75 65.2 37.2 26.0 0.05 3.00 97.0% 95.1% 110.4% 10 0.207 7.84 0.051 69 1245 0.49 5.0 0.048 44.7 1.29 111.3 86.7 12.7 0.07 8.13 98.2% 110.0% 101.9% 11 0.110 6.98 0.033 63 951 0.28 4.0 0.034 47.8 1.30 60.7 46.7 11.7 0.07 4.37 101.1% 82.4% 92.3% 12 0.111 6.99 0.025 64 947 0.29 3.9 0.018 37.8 1.26 61.5 48.7 11.5 0.07 4.31 103.7% 82.1% 94.0% 13 0.114 6.99 0.026 65 950 0.30 4.0 0.017 49.0 1.16 54.6 47.0 9.4 0.07 4.04 96.7% 129.8% 109.1% 14 0.107 6.99 0.020 65 946 0.29 3.9 0.026 45.6 1.05 59.5 56.7 11.0 0.06 3.39 95.1% 106.0% 105.2% 15 0.061 6.14 0.055 70 1245 0.20 5.1 0.045 100.3 1.77 67.8 38.2 25.8 0.05 3.32 101.1% 90.9% 109.3% 16 0.065 5.53 0.010 67 1242 0.49 2.9 0.055 96.6 1.98 99.2 50.0 10.7 0.04 3.97 95.4% 71.2% 112.5% 17 0.064 8.41 0.053 68 1242 0.19 2.9 0.004 18.2 2.06 66.4 32.2 27.1 0.06 4.01 106.0% 76.0% 93.0% 18 0.212 5.6 0.059 59 1242 0.19 2.9 0.047 39.7 0.84 71.2 85.0 31.5 0.08 5.43 103.5% 85.4% 84.3% Multivariate model generation from a nona-factorial DoE DoE – Examples conti culture Bernacchi et al. 2014 Plot of nitrogen dilution rate (DN) [d-1] versus medium dilution rate (DM) [h-1] in order to analyze growth to product yield (Y(x/CH4)). Individual levels of Y(x/CH4) are indicated through lines and boxes within the graph. The analysis shows that Y(x/CH4) varies by adjusting DN, DM, or both. An increase of DN reduces Y(x/CH4). However, an increase of DM increases Y(x/CH4). Plot of the nitrogen dilution rate (DN) versus the agitation speed in order to analyze Y(X/CH4) [C-mol/mol]. Individual levels of Y(x/CH4) are indicated through lines and boxes within the graph. Y(x/CH4) is highest at the lowest investigated range of both factors. DoE – Examples conti culture Bernacchi et al. 2014 • Modde (Umetrics, Sweden) • Design Expert (Stat-Ease Inc., USA) • Statistica (StatSoft, USA) • R Commander DoE – Software