Modeling product formation in anaerobic mixed culture fermentations

Modeling Product Formation in Anaerobic Mixed Culture Fermentations Jorge Rodrı´guez,1,2 Robbert Kleerebezem,2 Juan M. Lema,1 Mark C.M. van Loosdrecht2 1Department of Chemical Engineering, Universidade de Santiago de Compostela, Schoolof Engineering, ru´a LopeGo´mezdeMarzoa s/n, 15782SantiagodeCompostela, Spain; e-mail: jorger@usc.es 2DepartmentofBiotechnology,DelftUniversityofTechnology, Julianalaan67, 2628BC Delft, The Netherlands Received 15 June 2005; accepted 14 September 2005 Published online 4 November 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/bit.20765 Abstract: The anaerobic conversion of organic matter to fermentation products is an important biotechnological process. The prediction of the fermentation products is until now a complicated issue for mixed cultures. A modeling approach is presented here as an effort to develop amethodology formodeling fermentativemixed culture systems. To illustrate this methodology, a steady- state metabolic model was developed for prediction of product formation in mixed culture fermentations as a function of the environmental conditions. The model predicts product formation from glucose as a function of the hydrogen partial pressure (PH2), reactor pH, and substrate concentration. The model treats the mixed culture as a single virtual microorganism catalyzing the most common fermentative pathways, producing etha- nol, acetate, propionate, butyrate, lactate, hydrogen, carbon dioxide, and biomass. The product spectrum is obtained by maximizing the biomass growth yield which is limited by catabolic energy production. The optimization is constrained by mass balances and ther- modynamics of the bioreactions involved. Energetic implications of concentration gradients across the cyto- plasmic membrane are considered and transport pro- cesses are associated with metabolic energy exchange to model the pH effect. Preliminary results confirmed quali- tatively the anticipated behavior of the system at variable pHandPH2 values. A shift fromacetate to butyrate asmain product when either PH2 increases and/or pH decreases is predicted aswell as ethanol formation at lower pH values. Futurework aims at extension of themodel and structural validation with experimental data. � 2005 Wiley Periodicals, Inc. Keywords: modeling; mixed culture; fermentation; car- bohydrate; anaerobic INTRODUCTION Industrial fermentation processes are typically performed using pure cultures and aimed at the production of high value products. For bulk-chemical production the pure culture process seems less attractive due to the high costs, like those associated with control of the culture performance and with working under strictly sterile conditions in order to prevent contaminations. Important equipment investments are neces- sary for these processes at industrial scale. In addition to this, pure culture fermentations require generally the use of pure and therefore more expensive substrates. The risk of contamination of the culture furthermore remains, since an unstable pure microbial culture is used. Mixed cultures consist of a stable mixed microbial population, as typically found in nature. The use of less pure substrates (even wastes or by-products) is possible with the subsequent costs implications. Mixed culture fermentations (MCF) did not find wide application at industrial scale because they present still important limitations. The products formed by MCF vary in amount and composition and the control of the optimumbalance among themicroorganisms is not straightforward and requires a better understanding of their behavior (Hesseltine, 1991). Production of Chemicals From Mixed Culture Fermentations During extraction of many agricultural products, large amounts of residues are produced. The use of this organic matter by the biotechnological industry is limited due to the large diversity of organic compounds present in these residues. Production of energy carriers (Claassen et al., 1999) or other valuable products by mixed culture fermenta- tions would bring utility to those useless wastes or by- products and also enable interesting downstream integra- tions. MCF is a potentially interesting technology for validation of these streams and generation of specific products. Products that can potentially be obtained by MCF include mixtures of volatile fatty acids, alcohols, lactate that may serve as building blocks in other processes. Several interesting applications exist for MCF processes. i) Production of biodegradable polymers such as 3-hydro- xyalkanoic acids (PHAs) and poly-3-hydroxybutyric �2005 Wiley Periodicals, Inc. Correspondence to: Jorge Rodrı´guez Contract grant sponsors: The Spanish Ministry of Education (FPU Pro- gramme AP2003-3164); The Dutch Technology Foundation (STW project DPC.5904) acid (PHB) has been extensively investigated (Lee, 1996). PHAs can effectively be produced by mixed cultures of bacteria by imposing a strong selection pressure on the mixed culture (Reis et al., 2003). ii) Biological hydrogen production by MCF has a large research interest (Benemann, 1996) due to its potential application as energy carrier. The hydrogen production yield depends stoichiometrically on the range of fermentation products formed. iii) Solvent fermentations for production of alcohols and acetone, butanol, or propanol by clostridial cultures has been of much interest in past and recent years (Du¨rre, 1998) due to their potential application as sustainable additives to gasoline. To which extent mixed cultures can be applied to the production of specific solvents, remains largely unclear. iv) The carbohydrate fermentation is furthermore a crucial step in the anaerobic digestion process for wastewater treatment and valorization of solid waste streams. In anaerobic digestion processes of wastewater, the initial fermentation of carbohydrates (acidogenesis) leads to a variety of products that are finally methanised by other microbial populations. The interest of carbohydrate fermentations in the frame- work of all these applications motivates for modeling these processes from the perspective of control of the product formation. Control of a Mixed Culture Fermentation The control of the product spectrum obtained during fermentation of carbohydrates by mixed cultures has been studied in the past decades and some models have been proposed. Most of these models described the fermentation of carbohydrates considering a fixed stoichiometry (Batstone et al., 2002; Kalyuzhnyi, 1997; Vavilin et al., 1996; von Munch et al., 1999) mainly for integration in anaerobic digestion models. However, experimental investigation have repeatedly demonstrated the dependency of the products formed on the operational conditions (Horiuchi et al., 2002; Zoetemeyer et al., 1982a,b). To describe the variable stoichiometry of product formation during MCF some models have been proposed that predict the effect of hydrogen on the product scheme based on thermodynamic considerations (Costello et al., 1991; Mosey, 1983; Ruzicka, 1996). These authors however used thermodynamic or inhibitory hydrogen-dependent kinetic expressions for the fermentative reactions assuming a direct relation between thermodynamics and kinetics. These models do not incor- porate biochemical information and do not predict pH effects since they focus on the role of hydrogen. Successful application of metabolic models to anaerobic metabolism including the energetic effects of pH in the medium can be found in the literature (Beun et al., 2000; Kleerebezem and Stams, 2000). Herewe propose amodeling concept for prediction of the product formation in MCF. Metabolic network based assumptions are used leading to a simplified model of the MCF process. The model provides a mechanistic interpretation of the processes occurring and a structural identifiability of the most relevant parameters is possible. MODEL FOR ANAEROBIC MIXED CULTURE FERMENTATION The objective of the model is to predict the formation of products in anaerobic MCF under steady-state conditions. In this paper, we consider glucose as the only carbon and energy source while the main anaerobic fermentation products are included. The model proposed is based on the assumption that the fermentation products obtained will be those providing the maximum energy generation for growth processes to the mixed culture under the given conditions in themedium. This assumption is based on the hypothesis that natural selection, as imposed by evolution, has selected microbial populations able to maximize their growth efficiency under any given environmental conditions (Westerhoff, 1982). This results in the dominance of the microbial community with the highest biomass production rate under the prevailing environmental conditions. The model developed assumes a virtual organism capable of catalyzingmost typical pathways for glucose fermentation leading to the following range of products considered: hydrogen, carbon dioxide, lactate, ethanol, butyrate, propio- nate, acetate, and biomass. Besides the intracellular product formation pathways, transport of products across the cyto- plasmic membrane is considered. As opposed to thermo- dynamic black box approaches (Kohn and Boston, 2000), the gray box approach proposed here includes biological information and thereby reduces the number of degrees of freedom. The inclusion of common mechanisms and path- ways, like energy coupling to transport processes enables mechanistic interpretation of the results obtained. The virtual microorganism proposed here should be regarded as a representation of the different microbial strains involved in carbohydrate fermentation. Microbial diversity and dynamics of the process are neglected at this stage. The fact that bioreactions in anaerobic processes typically run very close to thermodynamic equilibrium (Mcinerney and Beaty, 1988), can justify for the assumption that the virtual microorganismwill behave similar to amicrobial consortium under steady-state conditions. Herewith it is assumed that metabolic efficiency dominates over phylogenetic diversity in the product formation and environment selects for the organism that is capable of catalyzing the thermodynami- cally most efficient set of reactions (Dollhopf et al., 2001). The core of the model developed consists of a bioreaction network where glucose and ammonium are the sole carbon and nitrogen sources respectively. The biomass growth is described as a general anabolic reaction fuelled by ATP. Finding the optimum fluxes vector through the biochemical network resulting in maximum biomass growth within the RODRI´GUEZ ET AL.: MODELLING PRODUCT FORMATION 593 thermodynamic feasible region, provides the net formation rate of products and biomass for the given environmental conditions. Bioreaction Network: Stoichiometry A metabolic network of the whole mixed culture is built up just by inventorying themost common catabolic bioreactions known for glucose fermentation to the products considered (Buckel, 1999) and posterior lumping of the most important pathways occurring in a MCF of that carbohydrate. Figure 1 presents the bioreaction network proposed including the transport of substrates and products. Pyruvate is the central branching metabolite in all pathways except for the biomass production pathway that is defined directly from glucose. Only the NADH/NAD redox-couple is explicitly considered for electron transfer. The potential role of the FAD/FADH2 electron carrier is identified and implicitly taken into account as will be outlined where required. Based on the inventory established by Buckel (1999) the following (lumped) reactions were considered (see Fig. 1 for the general scheme of reaction network and Fig. 2 for the algebraic stoichiometric representation). v1:Pyruvate formation fromglucose accompanied byATP and NADH generation via the Embden–Meyerhoff pathway. Only the stoichiometry of this conversion is considered, no kinetic description is established and consequently a pyruvate concentration needs to be assumed. v2: Butyrate production by reduction and decarboxylation of pyruvate consuming one acetate. One of the bioreactions of the lumped pathway corresponds to the reduction of crotonyl-CoA to butyryl-CoAwith FADH2 as electron donor under formation of FAD. Subsequent reduction of FAD is assumed to be accompanied by NADH oxidation by translocation of one proton across the cytoplasmic mem- brane. Assuming a Hþ/ATP yield of 3 this implies that crotonate reduction is accompanied by NADþ production and the formation of 1/3 ATP. Figure 1. Bioreaction network proposed for themixed culture fermentation of glucose. [Color figure can be seen in the online version of this article, available at www.interscience.wiley.com.] 594 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 93, NO. 3, FEBRUARY 20, 2006 v3: Propionate production by reduction of pyruvate through the succinate–fumarate pathway lumped. Propio- nate formation is assumed to be accompanied by the pro- duction 1/3 ATP during FADH2 dependent fumarate reduction, analogue as in butyrate production. v4: Acetate production by pyruvate oxidation and decarboxylation. One ATP is obtained by substrate level phosphorylation. v5: Lactate production by reduction of pyruvate. v6: Ethanol production by reduction and decarboxylation of pyruvate. v7: Hydrogen production by oxidation of excess NADH. This reaction is assumed to proceed close to thermodynamic equilibrium. Herewith the hydrogen partial pressure has a direct impact on the oxidation state of the NADH/NAD couple and the thermodynamic state of the different product formation pathways (Mosey, 1983). v8: Hydrolysis of ATP for maintenance and ATP driven transport processes that will be explained below. v9: Biomass growth is set up as a lumped anabolic reaction (Heijnen, 2001) producing biomass from glucose and ammonium and using ATP as energy source, the detailed stoichiometry is shown in Figure 2. The bioreaction network proposed (Fig. 1) considers 18 intracellular species from which 12 can be exchanged with the environment and 6 are present only inside the cell, namely ADP, ATP, phosphate, pyruvate, NADþ, and NADH. These are assumed not to be transported over the cellmembrane and therefore considered as conserved moieties under steady- state conditions. Figure 2 presents the proposed network set-up in an algebraic form where the 18 first rows correspond to the intracellular species and the last 12 to the extracellular species. Columns correspond either to bioreactions or to transport fluxes, thus the first nine columns are the bioreactions considered and the last eleven columns correspond to the transport of substrates and products defined for the network from Figure 1. Multiplication of the resultant 30� 20matrix by a vector of 20 fluxes (9 for bioreactions and 11 for transport) leads to a vector of net formation of all species. Application of mass balances to the intracellular species, whose net production must be zero in steady state, is performed by a calculation using the null space of the 18 first rows of the stoichiometry matrix, a base of six vectors is obtained. All feasible vectors of fluxes in terms of mass balances must be a linear combination of the six base vectors of the null space obtained. The network has therefore six degrees of freedom after application of mass balances to the internal conserved moieties. Thermodynamic Constraints Besides the mass balances, additional constraints to the biochemical reactions are provided by thermodynamic considerations. The Gibbs free energy change of a certain bioreaction or transport process must be negative to be feasible with a positive flux. According to Equation 1 the activities of substrates and products in a certain process affect the Gibbs free energy change of the reaction. �G ¼ �Go þ R � T � ln Y j a gj j ! ð1Þ where aj is the activity (typically concentration in diluted solutions) of the species Sj for a certain process/reaction with the stoichiometry g1 � S1þ g2 � S2þ . . .þ gj � Sjþ . . .¼ 0. This implies that the accumulation of fermentation products can lead to thermodynamic limitation of a certain bioreaction. Figure 2. Algebraic representation of the network proposed for the mixed culture fermentation of glucose. RODRI´GUEZ ET AL.: MODELLING PRODUCT FORMATION 595 The concentrations and partial pressures of products in MCFs are dependent on liquid and gas handling in the reactor. The operational conditions including the substrate (glucose) concentration degraded will define the product spectrum obtained. The type of reactor (e.g., a chemostat), reactor pH (leading to higher or lower concentration of the undissociated forms of the acidic products), partial pressures of gases (particularly hydrogen and CO2 that play an important role) are the variables to be considered in this work. Transport Processes: Bioenergetics The number of ATP-molecules consumed and produced in the intracellular reactions v1 to v9 as shown in Figure 1 have all been identified from generalized metabolic schemes. Besides this direct metabolic energy, the bioenergetic impli- cations of the transport of acidic species (t3 to t6) is important in modeling the overall catabolic energy production, mainly when considering the effect of the medium pH. The transport of chemical species over the cellular membranes can lead to generation or consumption of proton motive energy depending on the environmental conditions (Konings et al., 1995; van Maris et al., 2004) given that the membrane potential can be interconverted into ATP equivalents (Mitchell, 1979). Energy-mediated transport is considered for the excretion of the undissociated acidic products.When the productsmust be transported against a concentration gradient active trans- port is considered with energy consumption as ATP. This implies that energy investment is required at high extra- cellular product concentrations. The opposite situation happens at low extracellular con- centrations when transporting an organic acid (acetate, propionate, butyrate, or lactate) across the membrane down a concentration gradient and potential energy can be con- served by symport with a variable number of protons. This symport transport has been reported for lactate (Konings, 1985; Konings et al., 1995; Michels et al., 1979) and it was assumed that similar mechanisms apply for all the acid species in the model. Thus the energy available in the proton gradient generated can be conserved by a membrane bound ATP-ase (Konings et al., 1995). Full efficiency of the interconversions from any form of energy to ATP is assumed in the model. The energy change of transport of a certain species across the cellular membrane is thus due only to the concentration gradient (second term of Eq. 1). Besides the energy-mediated transport of the acidic products, the undissociated forms of the products and substrates are assumed to diffuse freely over the cell membrane without any energy change considered. Energy- mediated transport is considered only for acidic products since energy consumption for uptake of substrates (glucose and ammonium) will not affect the optimum catabolic product spectrum but mainly the biomass yield. Figure 3 gives the scheme of the modeled transport processes, the performance at low and high extracellular pH values is shown. At lower extracellular pH, more energy is needed for transport of the undissociated acid product due to the higher concentration outside the cell (see Eq. 2). Furthermore, the free back diffusion flux of the undissociated form of acid leads to a futile cycle that consumes e