This action might not be possible to undo. Are you sure you want to continue?

MT980104

**Metabolic Engineering from a Cybernetic Perspective: Aspartate Family of Amino Acids
**

J. Varner* and D. Ramkrishna School of Chemical Engineering, Purdue University, West Lafayette, Indiana 47907 E-mail: varnerÄbiotech.biol.ethz.ch; ramkrishÄatom.ecn.purdue.edu Received January 27, 1998; revised May 13, 1998; accepted May 18, 1998

Using the modular cybernetic framework developed by Varner and Ramkrishna (Varner and Ramkrishna; 1998a, b) a cybernetic model is formulated that describes the time evolution of the aspartate family of amino acids in Corynebacterium lactofermentum ATCC 21799. The network model formulation is employed in the role of a diagnostic tool for the overproduction of threonine. More precisely, having determined a parameter set that describes the time evolution of a base strain (lysine producer), the model predicted response to genetic perturbations, designed to enhance the level of threonine, are simulated using an appropriately modified cybernetic model and compared with the experimental results of Stephanopoulos and Sinskey (Colon et al., 1995a, Appl . Environ. Microbiol . 61, 74 78) for identical genetic perturbations. It is found that the model predicted response to enzymatic over-expression in the aspartate pathway agrees, for the most part, with experimental observations within the experimental error bounds. This result lends credence to the hypothesis that cybernetic models can be employed to predict the local response of a metabolic network to genetic perturbation, thereby, affording cognizance of the potential pitfalls of a 1999 Academic Press particular genetic alteration strategy a priori. Key Words : cybernetic models; metabolic engineering; amino acid production.

1. INTRODUCTION Millions of years of selective pressure have forced microorganisms to be strongly adaptive and reactive to environmental and genetic perturbation, so as to ensure survival. This trait follows from the unique ability to modulate physiological function via regulatory signals emanating from elaborate control architectures that have been programmed by evolution to meet new and diverse nutritional challenges. Thus, it is not surprising, that

* Present address: Institute of Biotechnology, ETH-Honggergerg, Zurich, Switzerland CH-8093. To whom correspondence should be addressed.

metabolic networks are often seen to ``resist'' alteration through adjustments of functionality by way of shifts in protein expression which in turn impacts enzymatic activity. Indeed, there exists a seemingly innumerable assemblage of experimental developments with the shared objective of exploring system response to altered levels of enzymatic machinery in the hope of mapping system regulatory action or obtaining organisms with desired features (Brown, 1997). Clearly, however, given the ill-defined, intricately complex nature of metabolic networks, such a carefree approach to metabolic design is in reality only slightly better than a hit or miss proposition. This is the case because the very same control mechanisms that confer behavioral robustness and adaptability to microorganisms frequently frustrate metabolic engineering efforts (defined as the useful modification of physiological function via recombinant DNA technology (Bailey, 1991; Stephanopoulos and Vallino, 1991)) by proactively resisting attempts to redistribute metabolic flux. With that said, given the level of intellectual involvement currently invested in metabolic engineering efforts, it is true that progress has been made. However, no one would argue that the rational design of metabolic pathways has progressed to the point of being a functional engineering discipline (Bailey, 1991). The biological tools are in place for relatively easy genetic manipulation of metabolic network functionality. Thus, this step in the design process can no longer be considered as the limiting factor, which arguably was not the case in the past. Rather, currently, the rigorous design guidance that mathematical analysis can provide is sorely lacking. Mathematical tools such as Metabolic Control Analysis (MCA) developed by Kascer and Burns (Kacser and Burns, 1973) and Biochemical Systems Theory (BST) developed by Savageau (Savageau, 1976) attempt to describe the control of metabolic networks through the use of sensitivity coefficients. The results of attempts to a priori rationally design metabolic networks using these tools have met with limited success with some notable exceptions (Hatzimanikatis, Floudas and Bailey,

88

1096-7176Â99 30.00 Copyright 1999 by Academic Press All rights of reproduction in any form reserved.

Cybernetic Models and Metabolic Engineering

metabolic engineering 1, 88 116 (1999) article no. MT980104

1996; Hatzimanikatis and Bailey, 1996; Hatzimanikatis et al., 1998). There are a number of reasons why this might be so. Metabolic control analysis is not a modeling framework, rather, it is a set of postulates (following simply from the chain rule of differentiation) that allow the systematic computation of localized network sensitivities to single perturbations in the environmental or network parameters. Thus, its predictive capability is limited by the quality and source of the information that is input into the analysis structure. Surely, an explicit dependence upon experimental input negates any type of a priori predictive potential because the results of the experiment are required for the analysis. If the analysis input stems from a kinetic model of the network, the claim that the sensitivity coefficients reflect the input of metabolic control mechanisms is suspect because traditional kinetic models often lack a closed description of the regulatory component (in the best case a representation of metabolite level regulation may be afforded via the kinetics expressions.) In this instance, the sensitivity coefficients do not describe regulatory effects, rather, they reflect the influence upon system behavior of kinetic competition within the network. Biochemical systems theory does contain a modeling aspect, and as such, theoretically it possesses the potential of being predictive in scope. However, it also depends upon sensitivity arguments to develop a conception of network regulation. The explicit dependence upon sensitivity coefficients, which in general are valid only locally, limits the predictive scope of this framework as well. Perhaps a more fundamental reason for the limited success of the mathematical tools discussed above is the founding belief that metabolic networks are ``static'' reaction systems. In other words, often enzyme level as well as activity are treated as static parameters (influencing V max in some ad hoc manner), an assumption which is clearly questionable for cases in which the makeup of the metabolic machinery is being significantly modulated. On the other hand, very detailed representations of metabolic regulation at the genetic level and moreover, its impact upon protein expression, such as that provided by the pioneering work of Lee and Bailey (Lee and Bailey, 1984) while being an ideal alternative on a smaller scale are not generally tractable for large networks or in cases where a very detailed mechanistic understanding is absent. The application of the cybernetic framework, developed by Ramkrishna and co-workers, for the description and analysis of metabolic networks is a new approach and may offer a distinct advantage over traditional methodologies. The cybernetic framework is the only modeling framework in existence that affords a systematic representation of the influence of metabolic regulation, both the control of enzyme activity as well as at the genetic level. The framework hypothesizes that metabolic systems have evolved

89

optimal goal oriented regulatory ``programs'' as a result of evolutionary pressures to cope with environmental andÂor genetic perturbation. The optimality hypothesis implies the system directs the synthesis and activity of network enzymes such that a nutritional objective is achieved in an optimal manner. The inclusion of a goal oriented regulatory strategy gives the cybernetic description of a metabolic network the key feature of regulatory reactivity, an element that is missing from all other contemporary metabolic network analysis and modeling frameworks. In other words, the feature that a cybernetic realization of a metabolic network offers, that is absent from all other contemporary approaches, potentially, is the ability to determine the manner in which a metabolic network redirects enzymatic expression and activity in the face of environmental perturbations or genetic alterations to network function. Thus, a cybernetic model of a metabolic network has the capacity of describing how the network adjusts itself upon perturbation. The advantage gained from such knowledge is obvious. If we knew, a priori, how a network would respond to perturbation, such an understanding could be incorporated, along with other key information, into the design of the genetic alteration, thereby, offsetting or even reversing the potential pitfalls stemming from ingrained regulatory reactivity. The conceptual extension of classic cybernetic doctrine to metabolic engineering has been presented in detail elsewhere (Varner and Ramkrishna, 1998a, Varner and Ramkrishna, 1998b) so we dispense with a formal first principles discussion, instead we present only a brief review of the salient conceptual features. Thus, our primary focus is not a reiteration of the basis of the extension, rather, it is to evaluate the potential of the framework by analyzing a realistic case study. In particular, we consider the overproduction of members of the aspartate family of amino acids. In Corynebacterium spp., lysine, methionine, threonine, and isoleucine derive part or all of their carbon from aspartate. The pathway flux that leads to the formation of these intermediates is controlled by regulatory architecture associated with key metabolic branch points. In particular, the aspartic-;-semialdehyde and homoserine branch points control the distribution of the carbon flux used for lower pathway synthesis. Therefore, control of the flux distribution at these key branch points is crucial to the overproduction of valuable pathway intermediates, such as lysine, threonine, and isoleucine. This metabolic network is a natural extension of the previous cybernetic investigation of branch point control architectures (Varner and Ramkrishna, 1998b) and serves as an illustration of the application of the framework to a complex, realistic pathway system. On a deeper conceptual level, this development is an evaluation of the ability of a cybernetic model

dashed arrows denote repression.25(gÂgDw } h) 0. 3. in a limited sense.01 (gÂgDw) 0. e 6 acetohydroxy acid synthase. p 5 threonine. e bb 3 homoserine kinase. Heavy solid feedback arrows denote enzymatic inhibition. p 8 isoleucine.0024(LÂgDw) 0.01(gÂgDw) 0. j K1 ba K3 K5 a K3 a Kg K ei \i. p 6 :-ketobutyrate.metabolic engineering 1. Unless otherwise noted. p 2 DL-asparticb ba . Metabolite legend: p 0 aspartate.001 (gÂgDw) 0. :k j 0.25 (gÂgDw } h) 0. Scope of the Investigation Using the modular concept introduced in Varner and Ramkrishna (Varner and Ramkrishna.05(gÂgDw } h) 0.01(gÂgDw } h) .01 (gÂgDw) 0.01 (gÂgDw } h) 0. e b 2 bb homoserine dehydrogenase. 1995a. All other metabolite and enzymatic species denote lumped abstracted quantities.05 (gÂgDw } h) 0. Jetten et al.4 (gÂgDW } h) 0. The complete set of differential equations that govern the time evolution of the metabolic network is presented and the model system response to various genetic alterations is analyzed and compared with experimental results. 1998). j j :j 0.01(gÂgDw) 0.265 (gÂgDw } hr) 0. 88 116 (1999) article no. In particular. max +a 3 + max 9 K0 b K2 bb K4 K7 K8 K ej \ j Y xÂSg 1.85 (gÂgDw } h) 0.-semialdehyde. 1. whereas. b. p a 3 lysine.001 (gÂgDw) 0. e 5 threonine dehydratase.0219(gDwÂg) + max 1 . We begin the development by outlining the process equations and kinetics that constitute the model framework.01(gÂgDw) 0..01 (ggDw } h) to describe the local regulatory reactivity of a metabolic network and contributes. p 3 homoserine. to the proof of the conceptual basis of the extension of the cybernetic framework to metabolic engineering applications and especially recombinant systems as put forth by the postulates presented in Varner and Ramkrishna (Varner and Ramkrishna. p 4 methionine. we compare the model predicted response to genetic perturbations with those obtained experimentally by Stephanopoulos and Sinskey (Colon et al. .01(gÂgDw) 0. max +a 2 bb.001 (gÂgDw) 0. 1995) and co-workers for the overproduction of lysine and threonine. MT980104 Varner and Ramkrishna TABLE 1 Sample Parameter Set Aspartate Family Model + max 0 .05 (gÂgDw } h) 0. 1. FIG. Enzyme legend: e 0 aspartokinase.05(gÂgDw } h) 0.01 (gÂgDw) 0.3 (gÂgDw } h) 0. The model equations were numerically evaluated 90 using the ODE15s routine of Matlab.01 (gÂgDw) 0. Ij a K2 bb K3 K6 bb K4 K T.1 (gÂgDw } h) 0.. 2. max + ba 4 k d. max +b 2 max +5 .01 (gÂgDw) 0. 1. e 4 threonine synthase.01(gÂgDw) 0.01 (gÂgDw) 0. 1998a) a cybernetic model is constructed which describes the time evolution of the aspartate family of amino acids. the simulation parameter set was that shown in Table 1.1.001 (gÂgDw) 0. MATERIALS AND METHODS All models discussed within the development were constructed and simulated within the Simulink environment of Matlab.25 (gÂgDw } h) 0. Once this is accomplished. max +3 + max 7 + max 8 + T .25(gÂgDw) 0. Model framework aspartate family of amino acids. max + ba 3 max +6 . Ij . This model system is used to investigate genetic alterations designed to enhance the level of amino acids present within the network. PROCESS EQUATIONS AND KINETICS In this section we outline the process equations and kinetics which constitute the model framework shown in Fig.80(gÂgDw } h) 0.01 (gÂgDw) 0. we turn our attention to making the model system complete by formulating the appropriate cybernetic variables that modify the various catalytic rates and the rates of enzyme synthesis. We then make our model system complete by formulating the cybernetic variables that describe the regulation of the synthesis and activity of the network key enzymes.25 (gÂgDW } h) 0.

a K 2 +p2 (3.1) b. Note because the structure of the lower metabolic pathway is not included in the model framework. max b . b+ p K e2 2 (3. 88 116 (1999) article no. K3 denote the rate and saturation constants where + ba 3 ba. max denotes the maxigoverning the formation of p a 3 and e 2 a mum level of key enzyme e 2 .9) where : 1 .10) . However. K2 denote the rate and saturation constants where + b 2 b. K e0 denote the rate and saturation constants governing the synthesis of e 0 .Cybernetic Models and Metabolic Engineering metabolic engineering 1. K p 3 denote the rate and saturation constants governing the utilization of p a 3 . max r ba 3 =+3 \e e ba 3 ba. The key enzyme e a 2 is assumed to be induced by the intermediate p 2 and is expressed according to the specific rate . max denotes the maxgoverning the formation of p ba 4 and e 3 ba imum level of key enzyme e 3 . max ba .4) pa 3 . max a .5) ba. K 1 are the rate and saturation constants governwhere + max 1 ing the synthesis of p 2 and e max denotes the maximum level 1 of key enzyme e 1 . We consider the representation of the interaction of the lower pathway metabolic pathway and the nutritional state of the organism with the aspartate pathway in the metabolic regulation section. The formation of p 3 catalyzed by key enzyme e a proceeds according to the 2 specific rate r =+ a 2 a. K 0 are the rate an saturation constants governwhere + max 0 ing the consumption of p 0 and e max denotes the maximum 0 level of key enzyme e 0 . The key enzyme e 0 is assumed to be induced by the presence of p 0 at the specific rate r e0 = : 0 p0 . K e0 + p 0 (3. this enzyme level must be responsive to changes in the nutritional state of the microorganism and is not considered a constant. ba K3 +pb 3 (3. K2 are the rate and saturation constants where + a 2 a. K ea + p2 2 (3. max 2 a where + a 3 . The key enzyme e b 2 is assumed to be induced by the presence of p 2 and is expressed at the specific rate b b= : r e2 2 where : 0 . K + p2 b 2 (3.6) 91 ba ba = : r e3 3 pb 3 b.2) . MT980104 a where : a 2 . max 2 eb 2 p2 . The key enzyme e ba is 3 assumed to be induced by p b 3 and is expressed at the specific rate =: a r ea 2 2 p2 .8) b where : b 2 . K0 + p0 (3. max 1 1 1 e1 p1 (3. max rb 2=+2 \e + b. The branch metabolite p b 3 can be conbb ba sumed to produce both p ba 4 and p 4 . K pa + pa 3 3 (3. The branch intermediate p a 3 is assumed to be consumed for lower pathway metabolic activity at the specific rate \e + K + p . K e1 are the rate and saturation constants governing the expression of e 1 . max governing the rate of formation of p b denotes the 3 and e 2 b maximum level of key enzyme e 2 .11) . K e2 denote the rate and saturation constants governing the expression of e b 2. no regulation can be assigned to the key enzyme which mediates the utilization of p b 3 . max 2 ea 2 p2 . max 3 + pb 3 . ba + p K e3 3 (3. The intermediate p 1 is consumed to produce the metabolite p 2 via the key enzyme e 1 at the specific rate r 1 = + max 1 p2 .3) . The key enzyme e 1 is assumed to be induced by p 1 and is expressed at the specific rate r e1 = : 1 p1 . K e 2 denote the rate and saturation constants a governing e 2 expression. The formation of p b 3 is catalyzed by key enzyme e b 2 and proceeds according to the specific rate The intermediate p 0 is supplied from upper pathway metabolism at the rate R 0 and is consumed via the key enzyme e 0 at the specific rate r 0 = + max 0 \e + max 0 e0 p0 . The branch point precursor metabolite p 2 can be conb a sumed to produce metabolite p a 3 or p 3 .7) . The formation of p 4 is ba catalyzed by e 3 and proceeds according to the specific rate \e + a. max =+a R pa 3 3 . K e1 + p 1 (3.

The formation of p bb 4 catalyzed by the key enzyme e b 3 and proceeds according to the specific rate Varner and Ramkrishna . MT980104 ba where : ba 3 . max 6 6 6 e6 p6 (3. K8 + p8 (3. The intermediate p 7 is consumed to form p 8 via the key enzyme e 7 according to the specific rate r 7 = + max 7 \e + max 7 e7 p7 . is assumed to be induced by the The key enzyme e bb 4 presence of p bb 4 and is expressed at the specific rate . max r bb 3 =+3 \e + b. K4 denote the rate and saturation constants 4 bb. K3 denote the rate and saturation constants 3 b.17) p8 .14) ba where + ba 4 . the dynamics of the key enzyme that catalyzes are considered in the downstream utilization of p ba 4 metabolic regulation section. K e7 denote the rate and saturation constants which govern the synthesis of e 7 . K 4 denote the rate and saturation constants a that govern the utilization of p ba 4 . K + p ba 4 ba 4 (3. K7+ p7 (3. The key enzyme e 3 is assumed and is expressed according to the to be induced by p b 3 specific rate where : 5 .13) \e + K + p . The intermediate p ba is assumed to be utilized by 4 downstream metabolic processes according to the specific rate where + max . K 7 denote the rate and saturation constants where + max 7 governing the formation of p 7 and e max denotes the maxi7 mum level of key enzyme e 7 .23) .16) where : 7 . The intermediate p bb 4 is consumed to form p 5 via key enzyme e bb according to the 4 specific rate bb 4 bb. b K3 + pb 3 (3. 88 116 (1999) article no. max r 4 = r bb 4 =+ 4 \e e + p bb 4 . max 5 5 5 e5 p5 (3. The key enzyme e 5 is assumed to be induced by p 5 and is expressed at the specific rate r e5 = : 5 p5 . K e7 + p 7 (3. K e3 denote the rate and saturation constants governing the expression of e b 3. max b where + bb . K denote the rate and saturation constants governing the expression of e bb 4 . K e6 + p 6 (3. K 5 denote the rate and saturation constants where + max 5 denotes the maximum which govern p 6 formation and e max 5 level of key enzyme e 5 .19) b where : b 3 .22) bb bb = : r e4 = r e4 4 p . max 4 where : 6 . The metabolite p 5 is consumed to produce p 6 via the key enzyme e 5 according to the specific rate bb e4 bb 4 r 5 = + max 5 \e + K + p . max bb where + bb . K e5 denote the rate and saturation constants governing the synthesis of e 5 .12) . The intermediate p 6 is consumed to form p 7 via the key enzyme e 6 according to the specific rate r 6 = + max 6 b b=: r e3 3 pb 3 .21) bb.20) ba R ba 4 =+ 4 p ba 4 . K eb + pb 3 3 (3. K e5 + p 5 (3. K 6 denote the rate and saturation constants 6 governing the formation of p 7 and e max denotes the maxi6 mum level of key enzyme e 6 . The key enzyme e 7 is assumed to be induced by p 7 and is expressed at the specific rate r e7 = : 7 p7 . max 3 eb 3 pb 3 . K e6 denote the rate and saturation constants which govern the synthesis of e 6 . The key enzyme e 6 is assumed to be induced by p 6 and is expressed at the specific rate r e6 = : 6 p6 .15) .metabolic engineering 1. K e3 denote the rate and saturation constants is governing enzyme synthesis. max and e 4 denotes the maximum level of key enzyme e bb 4 . The intermediate p 8 is assumed to be utilized by downstream metabolic processes according to the specific rate R 8 = + max 8 92 where : . bb bb + p K e4 4 bb 4 (3. K + p bb 4 bb 4 (3.18) bb. max denotes the maxigoverning the formation p bb 4 and e 3 b mum level of key enzyme e b 3 . Following the p 3 discussion.

8. As per the previous discussion. G is assumed to be induced by the growth substrate S g and is expressed at the specific rate r eT. j =3a.29) where + T . Sg Sg . G . G = + max T. K eT. 4ba.26) (3.e. G Sg . 1 and discussed above are specific.27) where + max . When considering the overexpression of system enzymes. Because we seek to maintain generality. G + S g (3. (3. K eg + p g (3. We denote the abiotic phase level of the j th promoter inducer by I j . The inducer is assumed to be assimilated by the carrier protein e T at the specific rate r T. G Sg . The model processes shown in Fig. In particular.. Kg + pg (3. G = : T.30) where : T. Metabolic Regulation: Network In what follows the modular approach for the formulation of the cybernetic variables governing the synthesis and activity of enzymatic machinery is summarized. Additionally. MT980104 where + max . 2. G . Note the induction by the growth substrate S g implies inducer transport is a consequence of the presence of the growth substrate. G e T. j ( p j & . K eT . K eg denote the rate and saturation constants which govern the expression of e g . T. Model framework lumped structured cybernetic growth model. j = k d. a representation of the growth process must accompany the description of the metabolic network. 4bb. Ij = + max T where + max T. Once present within the cell. i. K g denotes the maximum specific rate of p g g consumption and the saturation constant governing the rate of biomass formation. 3.Cybernetic Models and Metabolic Engineering metabolic engineering 1. the use of a tac or lac promoter that is sensitive to levels of Isopropyl-. 5. which in this case is glucose. e g is assumed to be induced by p g and is expressed at the specific rate r eg = : eg pg . respectively. the intermediate p g can be consumed to produce biomass via the lumped key enzyme e g at the specific rate r g = + max g where : T . We assume that the j th metabolic intermediate can diffuse from the cell at the specific rate r d. This is equivalent to.1. G denote the maximum rate and saturation constants that govern the formation of the growth precursor pool p g and e max T.24) FIG. we assume the use of both artificial and native promoters.P j ). K 8 denote the rate and saturation constants 8 which govern the utilization of p 8 . 3b. denotes the specific cellular volume. the dynamics of the key enzyme which catalyzes p 8 utilization are discussed in the metabolic regulation section. The key enzyme e T. G . The artificial promoter (in this case) does not allow transcription to occur unless a particular chemical inducing agent is present within the biotic phase. are written with respect to the biotic phase. Ij denotes the rate and saturation constants governing the assimilation of I j and e max denotes the maxiT mum level key enzyme e T .28) where P j denotes the abiotic phase level of p j and . we postulate a simple structured cybernetic growth model to describe the time evolution of the biotic phase and macroscopic variables such as growth substrate and biomass level. G at the specific rate r T.25) where : eg . is assimilated and converted to the lumped growth precursor pool p g via the key enzyme e T. G \e + K max T. G denote the rate and saturation constant governing the synthesis of e T. The lumped transport enzyme e T is assumed to be induced by the presence of S g and is expressed at the specific rate r eT = : T.-D-thiogalactopyranoside (IPTG). for example. we briefly discuss the theoretical basis for the extension of the cybernetic framework to genetically altered 93 \e + max g eg pg . Kg + S g (3. G . The growth model framework is shown in Fig. The key enzyme . and e max denotes the g maximum specific level of key enzyme e g . Accordingly. G + S g \e + K max T eT T . we postulate the growth substrate S g . K T. Ij Ij . +Ij (3. K g denote the rate and saturation constants governing the synthesis of e T . G denotes the maximum specific level of the key transport enzyme e T. 88 116 (1999) article no. 2. K T.

we postulate that every elementary pathway in the ``library'' has a cybernetic resource allocation problem associated with it. Varner and Ramkrishna. This is true because each of the elementary pathways is a discrete cybernetic subunit. 1994). we postulate that metabolic regulation broadly consists of three interacting layers. it is possible to construct the topological realization of an arbitrary metabolic network by simply assembling the members of this library in an appropriate fashion.metabolic engineering 1. of generalizations of the elementary building blocks proposed by Straight and Ramkrishna (Straight and Ramkrishna.'' In particular. that reflect all three postulated levels of metabolic control are termed the complete cybernetic variables and are given functionally by uj # ` u k j k \ m( j ) +\ n( j ) ` U j. q q + vj# ` v k j k \ y( j ) +\ ` V + . q .q q v jg # ` V j.e. The regulatory portrait of the system then follows from the method of assembly.'' The local level then follows from the interaction of the elementary pathways upon assembly.. z \j.e. (3. 3. Let us denote the objective and resource constraint associated with the k th elementary pathway by max[ f k( x k ) ] subject to g k = : x k z #Xk . The functional forms of U j. follows from the solution of the elementary cybernetic resource allocation problem(s) associated with key enzyme e j . v j . Thus. This postulated level of regulatory input governs the interactions of the elementary pathways stemming from assembly. and lastly. (the q th global signal influencing the expression and activity of the j th enzyme) follow from the specifics of the global regulatory communication.31) k where u k j . q denote the elementary and global cybernetic variables that regulate the synthesis and activity of the j th enzyme. The second founding postulate then goes as to how the individual regulatory influences (formally termed elementary regulatory signals) are ``fused'' into a ``complete'' regulatory signal. i. in other words. Our elementary pathway ``library'' consists. i. each elementary pathway has a particular regulatory ``flavor'' stemming from its isolated objective and it is the interplay between these different modes of regulation that determine the complete regulatory character of the network.32) are defined as the local regulatory input. In a mathematical context the cybernetic or control variables reflecting all three levels of control. The functionality of u k j and v k j . as alluded to previously. v lj denote the local cybernetic variables. To accomplish the systematic integration of all signals into a complete regulatory signal. i.1. but the manner in which seemingly unconnected network elements communicate.34) . where u lj . Review : Modular Approach At the core of the concept is the postulate that metabolic networks are segmented.1. are defined as 94 Definition 3.) The product(s) n( j ) x( j ) u jg # ` U j. 1998c). 1998a. termed elementary. local and global. termed the complete cybernetic variables. is that influence directly associated with the individual elementary pathways in our ``library. each member of the pathway ``library'' possess an objective function which is subject to constraint on the level of resources that are allocated for pathway operation. j. The set of control. the elementary cybernetic variables reflecting the k th control signal influencing the expression and activity of the j th key enzyme. MT980104 Varner and Ramkrishna networks.e. in this case.1. q .. V j. Note that the local component consists of the product of the elementary components because of the possibility that e j may be written as a member of multiple elementary pathways.) The converse of this postulate implies that given a ``library'' of elementary pathways. The product of the elementary regulatory contributions m( j ) y( j ) u lj # ` u k j k v lj # ` v k j w (3. competes for resources from the resource pool associated with each elementary pathway (overlapping assembly.2. the global level acts as a continuous switch upon local and elementary metabolic activity by translating the ``higher meaning'' of physiological state or network topology into regulatory action. Definition 3.. q . For a more detailed presentation the reader should consult (Varner and Ramkrishna. for example the pieces of a jigsaw puzzle (where interaction between elementary pathways can be via material flux or regulatory communication. respectively. V j. The elementary level. This understanding implies that networks can be decomposed into a set of interacting elementary pathways. they are composed of a collection of interacting elementary pieces. The elementary cybernetic allocation problem is the optimal resource allocation problem associated with the elements in the pathway ``library. where ``fusion'' consists of not only the amalgamation of elementary regulatory flavors. 88 116 (1999) article no. q q x( j ) (3.33) are defined as the global regulatory input influencing the expression and activity of e j . or cybernetic variables. akin to. q q (3. U j.

e. and g s + q( x s+q ) # X s + q . In the case of a seemingly contradictory objectives.. The influence of global control then acts as an additional constraint upon local allocation policy by modifying the local regulatory element in the complete cybernetic variable... u k j stems from the solution of the cybernetic resource allocation problem associated with the k th elementary pathway in isolation. this process is altered slightly..e. although such an objective may be operating at a suitably global level. because of evolutionary pressure. The metabolic control system.1. and max[ f k( x k ) ] and . In particular. in fact the introduction of genetic changes to metabolic networks and moreover the subsequent ``reaction'' of the network is intimately integrated with the hypothesis that metabolic pathways are goal oriented.35) where e j is a member of q elementary pathways. (3. The core hypothesis of the cybernetic framework is that.1.. when formulating abstracted models of microbial processes. The control action takes the form of the regulation of expression and activity of the enzymes that constitute the network. The functional form of the elementary cybernetic variable u k j . The functional form of the elementary cybernetic variable that describes the regulation of the activity of e j stemming from the k th elementary pathway follows from the cybernetic proportional law. The objective of a metabolic network follows as a consequence of the structure of the local regulatory aspect. and as such.2. the microorganism has 95 evolved control machinery that steers the nutritional state toward an objective in an optimal manner. influence the performance of the pathway. . in spirit remains unchanged. the network objective follows from the interaction of the individual elementary pathway objectives. x 2 .. for example in the case of storage metabolism. and moreover.1. which particular enzymes are expressed and the degree of promotion in each case because now the microorganism faces a modified set of operational alternatives.) In these instances the objective of nongrowth related metabolic networks is still a function of the topology. i. x n ] denotes the resource allocated to the z th alternative competing in the k th elementary pathway.. i. in the context of metabolic engineering. Thus. must cope with an expanded or . the metabolic control system that regulates the network. At first glance these concepts may seem incompatible. Postulate 3. This implies that the local regulatory element represents the `` fusion'' of elementary objectives and moreover that the objective associated e j expression is the disjoint objective set max[ f s( x s ) ] and max[ f s +1( x s+1 ) ] and .. We postulate that for a genetically modified network. In particular. however. Overexpression or deletion of enzymatic machinery does of course. energetic metabolites.Cybernetic Models and Metabolic Engineering metabolic engineering 1. Postulate 3.e. Often the product of metabolic function is required for growth.. The objective of a base network is still in place for a genetically modified version of the same network. specific intermediates. the network is subject to global nutritional control. MT980104 k k k k where x k z # x k # [x 1 . etc. a postulated objective of a metabolic network is local in the sense that it is a suitable refinement of the macroscopic growth rate postulate. 3.e. the choice operational functionality. it must be remembered that global objectives may be nutritional state dependent (as is true of storage metabolism or maintenance function. Whichever the case may be.2. and g k( x k ) # X k and ... x z . the final item that must be addressed before formulating and evaluating the aspartate model system is the extension of cybernetic principles to recombinant systems.. and max[ f s + q( x s+q ) ] subject to the disjoint constraint set g s( x s ) # X s and g s +1( x s+1 ) # X s +1 and . Lemma 3. 88 116 (1999) article no. In other words. Previous cybernetic investigators have assumed. i. We present a series of postulates that form the basis for the extension. the base network objective is in place because of the force of millions of years of evolutionary pressure.. However. .36) (3. the nutritional state in some measure influences the network startup. i.34). the maximization of growth rate as an evolutionary objective. This postulate implies that overexpression or deletion of existing pathway enzymes does nothing to change the objective of the network. evaluates the existing nutritional alternatives and directs the synthesis of the network enzymes for the set of alternatives that yield the optimum nutritional outcome.e.. We then reexamine these postulates in light of overexpression and deletion of existing pathway enzymes. These two schools of thought are not inconsistent. Extension of the Cybernetic Framework to Recombinant Systems Now that we have reviewed the basis of the modular approach. the control variable that governs the allocation of critical resources from the k th elementary resource pool for the expression of e j follows from the matching law optimization of (3.. we begin by presenting the basic concepts that underlie the optimality of the cybernetic framework. can not simply be immediately erased by a small number of perturbations. even when they appear to be contradictory. however... in the case of the modified network.. the local objective of a metabolic network is postulated to be a non-unique function of the pathway topology.. however. i.

Now suppose the gene that encodes for the j th enzyme has been deleted. within the present context. The j th enzyme no longer competes for key cellular resource from the k th resource pool or any other which implies uk j # 0.2.4) or remark (3. the intact base regulatory machinery still directs function toward the base network objective in an optimal manner.5. Clearly. the expression profile and the activity of e j are altered. . The deregulation of existing network enzymes can also be described within the framework.1) because we assume e j is still expressed but has no activity. This implies that the alteration that leads to decreased activity is not in the promoter region. k j =0) is not equivalent to postulate (3. Functionality shifts observed in response to genetic perturbation are an attempt on the part of the network control machinery to maintain the base objective as best possible in the face of changes forced upon the system by genetic alteration.3. thus. In a formal sense. . Note that postulate (3. Postulate 3. k j =1 denotes the wild type activity. the allocation of critical resources for network operation is still taken to be optimal. then in addition to postulate (3. 88 116 (1999) article no. The modified elementary cybernetic variable that reflects this contingency. the result of genetic modifications to the network (deletion or overexpression) is the contraction or expansion of the set of alternatives that are available to the microorganism.40) . denoted by v Äk j . In a cybernetic sense.metabolic engineering 1. . i. However. If this is not true. the completely blocked case ( .e. it does so in light of the genetic perturbation. inclusion of this class of genetic manipulation into the current framework is straightforward. the expression of e j does not compete for cellular resources because allocation is based upon the rate of return.39) where . In short. the enzyme is capable of only a fraction of its native activity. the available set of alternatives that are open to the organism for the case of the genetically modified network is different. not the alteration of the objective itself..5) assumes a normal expression profile. however. Let v k j denote the elementary cybernetic variable that regulates the activity of the j th enzyme of the k th elementary pathway. k j denotes the fraction of maximum of activity that the modified e j is capable of. these signals are classified as global nutritional signals.1.e. the deletion of existing pathway enzymes reduces the number of alternatives that compete for key cellular resources. Remark 3.38) (3. Thus. the set of alternatives that compete for these resources is modified. An entirely equivalent description of the ramifications of gene deletion follows by assuming the rate constant governing the expression of enzyme e j is defined as : ej # {0 : ej gene encoding for e j present gene encoding for e j deleted.k j 1. Remark 3. MT980104 Varner and Ramkrishna contracted set of possible alternatives reflecting overexpression or deletion of network enzymes when regulating the synthesis and activity of existing network enzymes. Suppose the enzyme e j is a member of the k th elementary pathway and is subject to q global control signals. Postulate 3. In simpler terms.6. Note. 96 Remark 3.e. as measured by reaction rate. however.3. Another possible route of genetic manipulation is the mutation of a gene which encodes for an enzyme such that a normal expression profile is observed.37) This representation implies e j =0 \t O r j =0 if the gene that encodes for e j has been deleted. it is hypothesized that the objective function of the base network is still in place even in the case of the recombinant system. (3. A value of . Accordingly. thus. i.5) the rate and saturation constant(s) that govern e j expression can be altered to reflect the new expression profile. i. however. Given the assumption of a normal expression profile. (3. Thus. Let u k j denote the elementary cybernetic variable that governs the allocation of critical resources form the k th pool to the j th network enzyme. the removal of metabolite feedback inhibitionÂrepression. we maintain Postulate 3. is of the form k k v Äk j #. 0 .. however. Postulate 3. . k j =0 indicates a completely blocked enzymatic step. the expression of the j th gene that encodes for e j is identical to that of the wild-type network. from a resource allocation perspective. we frame the discussion around the more general topic of removal of global regulatory sensitivity. whereas. (3. these arguments imply that base network control machinery directs metabolic function towards the base objective despite the genetic alteration. More precisely.. Suppose the gene which encodes for e j is altered such that e j is only capable of a fraction of its wild type activity.4. the base case and the altered network are not the same. U q ). U r .j v j . The complete cybernetic variable that regulates the expression of e j is given by u j = u lj ( U 1 U 2 ..

p. This alteration is reflected by assuming U r # 1.46) whereas enzyme expression via the plasmid is measured as rp ej . the return on investment can be measured as the rate of enzyme synthesis. Postulate (3. dR k \i. Postulate 3. Within the present context. i.49) 97 . 1 .45) Equation (3. 88 116 (1999) article no.43) (3. respectively.7) implies the existence of a cybernetic control variable that regulates the allocation of key resources between possible expression sources. Specifically.. i. . both routes compete for key cellular resource to synthesize a common product. In short. (3. MT980104 where u lj denotes the local regulatory component. it stipulates that the optimum allocation policy is one in which the fractional allocation of critical resources is equal to the fractional return on investment. (3.e.7. as a consequence of an unspecified mutation.e. U q ). Accordingly. enzyme e j .47) whose form is dependent upon the promoter type. k = g. we postulate that the local objective of enzymatic expression (barring any global interaction. p where superscript(s) global control variable U e j g. . and e ij denotes the level of enzyme e j synthesized via the i th route. n denotes the total number of possible routes. Both routes require critical cellular resources to proceed. namely. it follows from Eq. There exists a global cybernetic variable. (3. there exist two routes for the synthesis of the common product.. (3.45) is a restatement of the matching law criteria put forth by other cybernetic investigators. (3. The over-expression of e j is accounted for within the model equations by replacing the term r ej u j in the e j balance with g p p +r e Ue . These statements formally become the constrained optimization problem max : e ij ( R i ) i de k j = n k =1 dR i .41) where R i denotes the resource allocated to the i th route of synthesis of enzyme e j . The k . r ej u j U e j j j subject to g = : R i # R. reflects the substitutable nature of overexpression. (3. The optimality condition of the constrained maximization problem given by de 1 j dR 1 = de 2 j dR 2 = }}} = &1 de n j This implies the modified complete cybernetic variable that regulates that expression of e j is given by u j = u lj ( U 1 U 2 . we are considering the competition between the chromosomal synthesis route and expression via a plasmid of the key enzyme e j .45) that the cybernetic variables which govern the allocation of critical resources for the synthesis of key enzyme e j from the genome and plasmid are given functionally as g = Ue j r ej u j p r ej u j + r e j p Ue = j p re j p r ej u j + r e j . such as nutritional state dependent synthesis) is the maximization of the level of key enzyme e j subject to a constraint upon the amount resources available for synthesis of e j . If the allocation policy is implemented at every instant in time and takes place in time dt. \j (3. k denoted as U e . The removal of the global control of enzyme activity follows by analogy. p denote genome and plasmid. e j is no longer sensitive to the r th global control signal. . In the case in which the synthesis route is chromosomal. Postulate 3. The inclusion of the regulated rate of chromosomal synthesis follows form the postulate that the enzyme synthesis routes are completely functional and independent. Enzyme expression stemming from the genome and plasmid form a substitutable pair in the cybernetic sense. r ej u j . Accordingly. i (3. p denote genome and plasmid. Now suppose. that regulates the allocation of j critical resources to each individual expression route..44) can be rearranged to yield the matching criteria de ij n k =1 Remark 3. respectively.Cybernetic Models and Metabolic Engineering metabolic engineering 1. To derive the form of these control variables we pose a problem similar to the elementary convergent pathway. k = g. The overexpression of existing pathway enzymes is a substitutable process.8. Because the native and plasmid expression routes are unaware of the existence of one another the control action that regulates the choice of synthesis route is classified as a global signal.42) dR n &1 = de n j dR n (3.48) { n = n where the superscripts g. the rate of enzyme expression is given by the regulated rate. The representation of the overexpression of existing pathway enzymes is slightly more involved because of the possibility of multiple expression sources.4. The postulate above describes the removal of a global sensitivity connected with enzyme expression.

In other words.2. and hence the network reaction rates. For example. the shift in the relative levels of pathway enzymes is the imputes for shifts in the level of network intermediates.5. enzyme expression from a plasmid is not under the direct control of the microorganism. p4 network objective. This implies that no direct regulatory interaction exists between the branch metabolites.. rather. Suppose e j catalyzes r q j and is a member of q elementary pathways. The local component is derived from the topological nature of the pathway. is the maximization of the elementary linear pathway end ba products. the shared elements between elementary pathways compete for resources from multiple pools. accordingly... p =0 which implies Before addition of inducing agent. shifts in metabolite level cannot compensate for the artificially high level of flux through the j th network step. 3. Thus. the activity of e j or fluxes that feed e j ). given this regulatory formulation. if changes in metabolite level can blunt the effect of overexpression (by altering r j . The interesting biological implication of such an observation is that the synthesis and the activity of the key enzymes that contribute to the formation of the network intermediates are coupled via this resource sharing. The overexpression e j implies an artificially 98 enhanced level of this key enzyme. i. Thus. the cybernetic regulation will consist of three levels.) As a result of overexpression. the makeup and activity of the network enzymatic machinery is predicted to be modulated with respect to the base pathway makeup because resource distribution is altered. expression is controlled via the characteristics of the plasmid borne promoter. with end ba and p 8 . metabolite feedback sensitivity.e. Derivation: Elementary Cybernetic Variables As per the discussion above. We model the aspartic-. the plasmid.e.7. These act as a global control mechanism upon enzyme synthesis that regulates the source of synthesis of e j .6. given a particular level of chemical induction agent. It follows from above that e j competes for cellular resources from each of the q elementary resource pools. Remark 3. the microorganism possess no regulatory recourse that could halt expression from the plasmid (barring plasmid structural stability arguments. the qualitative picture of resource allocation to the elementary pathways may be substantially altered. In effect. the ramifications ripple . it is allocated a larger portion of the critical resources earmarked for e j expression. Note because the elementary pathway overlap. inducing agent. the qualitative picture of resource allocation is not significantly altered by overexpression. local and a global regulatory component. Rather. the native expression route is the only active synthesis route for e j therefore it consumes all resources allocated to e j expression. the plasmid route is faster than synthesis from the genome.metabolic engineering 1. If however. u j =1. MT980104 Varner and Ramkrishna where the second synthesis rate stems from expression via k ..-semialdehyde and homoserine branch points as flexible as discussed by Stephanopoulos and Vallino (Stephanopoulos and Vallino. However. p a 3 . Note the presence of the control variables U e j k = g. they interact only through kinetic competition for the common branch precursor metabolite. U e is equivalent to the fraction of cellular j resources allocated to e j expression being directed to the k th synthesis source. suppose the synthesis rate from the plasmid is much greater than chromosomal synthesis rate and under the control of an artificial promoter. all pathway fluxes catalyzed by this key enzyme are initially increased. as resource competitiveness of key enzymes is modulated.e. this deregulates the expression of the plasmid borne enzyme. 1991). r e j p g U ej =0 and U ej =1. because of the assumption r e j the majority of enzyme synthesis is from the plasmid.e. Remark 3. Because the elementary allocation policy follows as a consequence of reaction rate. namely. i. This implies that the local control structure dictates what the exact meaning of ``maximum'' is relative to the objective of the pathway. suppose the native route (in the absence of synthesis of e j from the plasmid) is highly favored. i. the level of resources allocated to the other enzymes which compete from the same pools must necessarily decrease in the short term. an elementary. Note the assumption that the objective of enzyme expression is to maximize the level of e j is consistent with previous cybernetic developments because the return of the chromosomal route of enzyme expression is measured as the regulated rate. whereas. 88 116 (1999) article no. The control variables U e j j as a continuous global switch whose purpose is the selection of the particular synthesis route or combination of synthesis routes that yield the maximum level of key enzyme e j . Furthermore. i. After addition of the p r r ej u j . This allocation imbalance is reflected by the values of the control variables p g and U e and expression of e j is steered away from the Ue j j p g and U e acts chromosomal route. 1998b)) that the local regulatory component consists of three overlapping elementary linear pathways. p 4 and p 8 subject to the level of resource allocated for function of the corresponding elementary pathway. Thus. namely. follows from the set of elementary building blocks used to realize the pathway topology. As postulated. Remark 3. k In simpler terms. p. respectively. the global component describes the higher regulatory intentions of the organism. The metabolic products p a 3 . It follows from the previous development regarding the cybernetic representation of metabolic branch points (Varner and Ramkrishna (Varner and Ramkrishna. in this instance.

the cybernetic variable that governs the allocation of critical resources from the k th elementary resource pool to the synthesis of j th key enzyme belonging the k th elementary linear pathway. 1994). 1.55) k k k k k max[ p k n( R n &1 . j =0 \k (3. (3. b. Moreover. respectively. k r k Ä n.51) (3. k dp = dR k n &1 k n k n &1 +a Ä n &1. (3. . These postulates imply k vk j t *k r j . j (3. n &1.. . n &2.52) denotes the stoichiometric coefficient(s) relating the conversion ej &1 The elementary cybernetic variable that governs the activity of the key enzyme catalyzing the j th metabolic transformation of the k th elementary pathway. k dp k n &2 + } } } + a 1 (3.. consistent with Straight and Ramkrishna (Straight and Ramkrishna. from p 0 to ba pa 3 and p 0 to p 4 are secondary elementary branch pathways and are denoted as a.. k dp k n+a n &1 + a n &2 + } } } + a 1 = where r dR k 0 dR k 0+ n &1 y =1 dR k y . j (3. is given by uk j = a Ä n . this law postulates that enzyme activity is proportional to reaction rate. n &1. n &2. Note that each of these pathways is linear..Cybernetic Models and Metabolic Engineering metabolic engineering 1. \k . n &1. \k . k dp k 1 b=b dp k Ä n. s. then the return on investment can be measured as the reaction rate. The network under consideration is composed of three overlapping elementary linear pathways.57) p j &1 ww Ä pj in the k th elementary linear pathway. j. k dp k Ä 2. k # ` j = s +1 a j. the constrained optimization problem associated with the k th n( k )-dimensional linear pathway dp k n dp + a Ä n..e. p 1( R 0 ))) ] subject to g k = : R k j #Rk . The two remaining elementary pathways. 88 116 (1999) article no. respectively. to be the maximization of the end product subject to a constraint upon the level of resources allocated for elementary pathway function. = }}} =` j =2 dp k j dp k j &1 dp k 1 dR k 0 \k . k dp k Ä n &1. Accordingly. 1 r k n &1 + a n &2 + } } } + a j + } } } +a 0 j =0. Accordingly.58) . k dp k Ä 2. extends from p 0 to p 8 and is denoted as elementary pathway c. it is postulated that the key enzyme which catalyzes the maximum rate amongst the set of rates competing in the k th elementary pathway possess the maximal activity. n &1. \ k .. denoted as u k j . p n &1( R n &2 . 1. denoted as vk j follows from the cybernetic proportional law.. thus.54) a Ä r.51) can be rearranged to yield the set of matching criteria 99 The cybernetic variable v k j is constrained by definition to lie in the interval 0 vk j 1. termed the backbone pathway. 1.53) dR k n &1 &2 k + n y =0. k # . n &1 dR y a Ä 2. the elementary cybernetic problem associated with each is identical. The global regulatory component which describes metabolite repressionÂinhibition then acts to constrain the local allocation policy or local enzyme activity. In short.50) possess the optimality condition k dp k dp k dp k dp k dp k n n n &1 n n &1 dp n &2 = = k k k dR k dp k dp k n &1 n &1 dR n &2 n &1 dp n &2 dR n &3 n The matching condition(s) shown above simply state that optimum pathway operation with respect to resource allocation comes about when the fractional return on investment is equal to the fractional allocation of pathway resources. Optimality condition (3. j. j +1. MT980104 throughout the network because of competition for key cellular resources. If we assume that the optimum allocation policy is implemented at every instant of time and that allocation takes place on the time scale of dt. k . i. k r k Ä n. j &1. The primary elementary pathway. k r k j rk Ä n. 1.56) where dp k j dp k j &1 a j. . j &1. The objective function of the elementary linear pathway is postulated. (3. (3.

65) The local cybernetic variable which governs enzyme activity follows from above and is given by b c v l0 = v a 0v0v0. rk j where the local competition set [r k( j ) ] is given by *k \k . pa 3 e0 . r 2 . k( c ) q =4 b 2 ba 3 V a e0 . max max where V pa and V denote the minimum e0 . p3 V0 t * pa ( pa 3& 3 a : rk=r0 +r1+r +r . 88 116 (1999) article no. r 3 . r 1 .63) 100 . k( a ) The key enzyme e 0 is sensitive to concerted feedback inhibition by the network metabolites p a 3 and p 5 .. .60) The local cybernetic variable that governs the expression of e 0 is the product of the elementary variables. (3.e. [r q ] q =4 ]. p5 p 5 3 a level of p 3 .. r b 2 . This follows from the possibility of allocation from multiple resource pools. p3 . p5 3 5 a maximum biotic phase level of p 3 and p 5 . respectively. r a 2 ].61) 3. Namely. p 5 at which e 0 has zero activity where 0< V . c .59) \j in the k th elementary pathway. 5 (3. Derivation: Complete Cybernetic Variable The key enzyme e 0 is a member of all three elementary linear pathways which implies that it can receive resources from three different independent resource pools. .. j . i. 1 r 2 . The a terms V e0 . This signal is represented as a global control upon local activity. p 5 Ä 0.66) r0 k( j ) r k . j . The local cybernetic variable which governs e 0 synthesis is given by b c u l0 = u a 0u0u0.metabolic engineering 1. b. r b 2 . when p a 3 = p 5 =0 e 0 experiences no j inhibitory effects.64) vk j = r k j k k k max( r k 1 .. \ k .67) The elementary cybernetic variable which governs the activity of e 0 (note that e 0 is a member of three elementary pathways) is given by the set r0 .3. thus. b . ba [r k( b ) ] = [r 0 .62) where : rk=r0 +r1+r a 2. r 3 ]. (3. c (3. r n ) *k = (3. r 1 . i.. In other words. which implies v k j is of the form (3. r n ) .. Additionally. (3.59) The proportionality constant * k must obey equation (3. i. max pa ) 3 p5 V0 t * p5 ( p 5 & V e 0 .. 0 (3. V 0 Ä 1 as p a 3 . Clearly.. j = a. r 1 . 0< V 1 and p a and p max denote the e 0 . (3. r j . the complete inhibition threshold. max a 1.. j = p 3 . k( b ) 8 bb : r k= r0 + r1 +r b 2 + r 3 + : rq . we postulate the inhibition of e 0 is proportional to the level of the intermediates p a 3 and p 5 ... MT980104 Varner and Ramkrishna which implies 1 .e. p 5 are defined as the minimum fraction of the maximum biotic phase metabolite level at which complete inhibition of e 0 occurs.. The derivation of the global control variables follows from the arguments presented in the previous members of this series. the key enzyme e 0 is subject to concerted feedback inhibition by the branch metabolite p a 3 and the intermediate p 5 . j. * k takes the form 1 k k max( r k . j =0 when p a and p are greater than or equal to the V0 5 3 . [r k( a ) ] = [r 0 . The elementary cybernetic variable governing the allocation of critical resources for the synthesis of e 0 from the j th pool is given functionally as j u0 = Note because expression of e 0 is not sensitive to any global regulatory signals u l0 describes all regulatory input to the synthesis of e 0 .e. Furthermore. bb 8 [r k( c ) ] = [r 0 . p3 e0 . . max[r k( j ) ] j = v0 j = a . p5 p max ). the complete cybernetic variable that controls the expression of e 0 is given by u 0 = u l0 .

because e 1 is not sensitive to any global regulatory signal. c. ba [r k( b ) ] = [r 0 . The local cybernetic variable that governs resource allocation is the product of the elementary pathway allocation variables given by j u1 = Because e 1 is not subject to any global regulatory signal. p3 . r 1 . r0 +r1+r a 2 (3.77) It follows the complete cybernetic variable that controls the synthesis of e a 2 takes the form u 2a = u l2a = u a 2.79) Moreover. because the key enzyme e 1 is a member of three elementary pathways the local cybernetic regulatory component is the product of elementary components. (3. [r q ] q =4 ]. c.75) The key enzyme e 1 is a member of all three elementary pathways.73) Thus. k( c ) q =4 The key branch point enzyme e a 2 is a member of only a single elementary pathway and is not subject to any global regulatory signal. MT980104 complete inhibition threshold.80) The functional form of the cybernetic variable that controls enzyme activity follows along the same lines as shown 101 . the complete cybernetic variable that describes the activity of e 1 is given by v 1 = v l1 . b . (3. The complete cybernetic variable governing the activity of e 0 is the product of the local and global activity components and is given by p3 5 Vp v 0 = v l0 V 0 0 . (3. k( a ) ba : rk=r0 +r1+r b 2+r3 . r b 2 . r a 2) (3. The elementary cybernetic variable governing the synthesis of e a 2 is of the form u l2a = u a 2= ra 2 .72) The activity of e a 2 also follows solely from the elementary regulatory component which is given by v l2a = v a 2= ra 2 . r 1 .69) [r k( a ) ] = [r 0 . a The local regulatory component that describes enzyme activity is given by b c v l1 = v a 1v1v1. (3. r 3 ]. max a p e 0 . 5 where the local competition set [r k( j ) ] is given by (3. r a 2 ].68) above. r 1 . r 1 . max pa 3 (3. j = a. is not sensitive to any global regulatory input. p5 p 5 p5 < p5 V e 0 . p3 V a e0 . the complete cybernetic variable which describes the activity of e a 2 is of the form v 2a = v l2a = v a 2. (3. max 3 and r1 . (3. b. the regulation associated with e a 2 is composed of only the elementary regulatory component. r 3 .Cybernetic Models and Metabolic Engineering metabolic engineering 1.70) (3.74) p5 V0 = { \ 0 1& p5 V max e0 . r b 2 . p5 p max 5 p max .76) r1 k( j ) rk . the complete cybernetic variable that controls the expression of e 1 is given by u 1 = u l1 . Thus. p5 V e0 . max[r k( j ) ] j = a . (3. 88 116 (1999) article no. The elementary cybernetic variables which govern the activity of e 1 are given by the set j v1 = p a. p3 3 + + pa 3< p a 3 V a e 0 . however. More exactly. k( b ) 8 bb : r k= r0 + r1 +r b 2 + r 3 + : rq . bb 8 [r k( c ) ] = [r 0 .71) where : rk=r0 +r1+r a 3. These statements imply the functional form a p3 V0 = { \ 0 1& pa 3 V a.78) It follows that the local cybernetic variable that controls the synthesis of e 1 take the form b c u l1 = u a 1u1u1. (3. max( r 0 .

87) (3.88) The complete cybernetic variable governing the activity of key enzyme e b 2 is the product of the local and global regulatory components and takes the form 5 . c u l2b = u b 2b u 2b . r 1 . r 1 . 102 . p4 p ba.82) The local component is subject to the repressive influence of the intermediate p ba 4 . 88 116 (1999) article no. v 2b = v l2b V 2b p (3. p5 V b e2 . p5 p p max 5 + p5< p5 V b e2 . The elementary regulatory component governing e b 2 expression is described by the elementary cybernetic variables j u2 b= The complete cybernetic variable that describes the control of the synthesis of e b 2 is the product of the local and global regulatory components given by 4 u 2b = u l2b U 2b . 4 (3. Additionally when p ba U2 4 =0. Let e2 denote the minimum level of p ba .85) rb 2 k( j ) rk .metabolic engineering 1. the b b key enzyme e 2 should feel no repressive effect. Thus. 5 (3. MT980104 Varner and Ramkrishna The key enzyme e b 2 is a member of two elementary pathways and is subject to a inhibitory signal emanating from p 5 and a repressive signal emanating from p ba 4 .. j = b . Accordingly. r b 2 . max p ba ) 4 (3. (3.86) where the competition set [r k( j ) ] is given by ba [r k( b ) ] = [r 0 . c v l2b = v b 2b v 2b .e. The local regulatory component that describes the regulation of eb 2 expression is the product of the elementary regulatory influences. p ba (3. p4 . Thus. ba r0 + r1 + r b 2+ r 3 (3. that the local cybernetic regulatory component is the product of the elementary components i. we postulate the global control variable is proportional to the level of p ba 4 i.83) and after some simple algebraic manipulation we arrive at the functional form The key enzyme e ba 3 is a member of only a single elementary pathway and is sensitive to the level of p ba 4 . max b ba above. p4 p 4 4 b where e 2 expression is completely repressed. r b 2 . max( [r k( j ) ] ) j = b . p5 p max 5 p max . c. the allocation policy of critical resource for the synthesis of e b 2 must be modified by a global component to make it sensitive to the repression. k( b ) 8 bb : r k= r0 + r1 +r b 2 + r 3 + : rq . (3. Because e b 2 is a member of two elementary pathways the local regulatory component is the product of the elementary cybernetic variables given by j v2 b= where bb : rk=r0 +r1+r b 2+r3 . p4 U b ba e 2 .e. The elementary regulatory component governing the allocation of critical resources for the synthesis of e ba 3 is given by u l3ba = u b 3ba = r ba 3 . max p ba .90) U p4 2b ba = { \1& 0 p ba 4 U b ba e2 . The functional form of the global control variable follows from arguments presented earlier and is given by V = p5 2b { \1& 0 5 V b e 2 .89) U b ba e2 . max 4 + p ba 4 < p ba 4 U b ba e2 . max p ba 4 . because e b 2 can receive resources from two independent elementary pools. It follows that at this level of p ba 4 . ba 4 ba( p U 2b t * p4 4 & p ba The activity of the key enzyme e b 2 is sensitive to the levels of the pathway intermediate p 5 . c. The form of the global control variable follows from the discussion ba. r 3 . k( c ) q =4 rb 2 . [r q ] q =4 ]. r 3 ].. (3. the global control variable denoted as ba p4 assumes a value of zero. It follows that the local regulatory component is given by It follows.84) Note because e ba 3 is a member of only a single elementary pathway the local and elementary components are identical. p4 . bb 8 [r k( c ) ] = [r 0 .81) The control of enzyme activity follows from the discussion above. the local cybernetic component governing enzyme activity must be modified by a global control variable so as to account for this sensitivity.

(3. p5 p p max 5 + p5< p5 V bb e 3 . (3. [r q ] q =4 ] . (3. the local cybernetic regulatory component that governs e bb 3 synthesis is given by u l 3bb The key enzyme e 4 is a member of only one linear pathway and is not subject to any global regulatory signals.92) The activity of the key enzyme e ba 3 is sensitive to the level of the intermediate p ba .91) The elementary cybernetic variable governing the activity of e bb 3 is given functionally as v l3bb = v c 3bb = r bb 3 . Following the previous development. In other words. MT980104 The complete cybernetic variable governing the synthesis of e ba 3 is given by u 3ba = u l3ba = u b 3ba ..102) 8 k =4 rk . r b 2 . The elementary component that governs e 4 expression is given by the cybernetic variable uc 4= r4 r 0 + r 1 + r + r bb 3 + b 2 =u c 3bb . 5 (3. because e bb 3 expression is not sensitive to global signals. (3. r 1 .100) V p4 3ba ba = { \1& 0 p V ba ba e3 . the global control variable that describes the inhibition of e bb 3 with respect to p 5 is given functionally as (3. (3.99) (3. v l3ba = v b 3ba = r . max p ba . 88 116 (1999) article no. It follows because e ba 3 is a member of only a single linear pathway that the local and elementary regulatory components are identical i. The form of the global control variable 4 which describes this sensitivity follows from the previous discussion and is given functionally by Because e bb 3 is a member of only a single linear pathway the local regulatory element that governs e bb 3 activity is the elementary component i.e. p4 V ba ba e3 . p4 .. The elementary cybernetic variable which governs the synthesis of e bb 3 is given by uc 3bb = r r0 + r1 + r + r + b 2 bb 3 bb 3 (3. r 3 . (3.93) The complete regulatory component is the product of the local and global regulatory elements and is given by 4 . (3.94) The complete cybernetic variable which governs the activity of e bb consists of the product of the local and global 3 regulatory components and is given by p5 v 3bb = v l3bb V 3 .97) 103 Because e 4 is member of only a single linear pathway the local regulatory element consists solely of the elementary component.98) The activity of e ba 3 is subject to both local as well as global control. r 1 . max p ba 4 .101) p ba (3.96) 8 k =4 rk . the local cybernetic regulatory element is given by u 4bb = u 4 = u l4 = u c 4. p5 p max 5 p max . v l3bb = v c 3bb (3.Cybernetic Models and Metabolic Engineering metabolic engineering 1. p4 ba 4 p ba.104) . max 4 + p ba 4 < p ba 4 V ba ba e3 . r b 2 .103) Moreover.e. 4 The activity of e bb 3 is also sensitive to the level of p 5 . v 3ba = v l3ba V 3ba V p5 3bb = { \ 0 1& 5 V bb e3 . the complete cybernetic variable governing the synthesis of the key enzyme e bb 3 is given by the local regulatory component u 3bb = u l3bb . max( [r k( c ) ] ) (3. r3 ) ba 3 where the competition set [r k( c ) ] is given by bb 8 [r k( c ) ] = [r 0 .95) is a member of only a single elementary Because e bb 3 pathway. bb The key enzyme e bb 3 is a member of a single elementary pathway and is sensitive to levels of p 5 . p5 V bb e3 . ba max( r 0 .

117) The key enzyme e 5 is a member of only a single elementary pathway and is sensitive to levels of the intermediate p 8 . (3. Following the previous discussion.120) . max( [r k( c ) ] ) (3. u l5 = u c 5.e.107) p8 = V5 { \1& 0 8 V max e5 . the complete cybernetic variable governing the expression of e 5 is given by u 5 = u l5 .114) where the competition set [r k( c ) ] is given by bb 8 [r k( c ) ] = [r 0 . (3. v 5 = v l5 V 5 (3. Therefore. r b 2 . (3. p8 p max 8 p max 8 (3. 88 116 (1999) article no. v = max( [r k( c ) ] ) c 5 U = p8 6 { \1& 0 8 U max e6 . r 1 . l 4 The complete cybernetic variable which controls the activity of the key enzyme e 5 is the product of the local and global regulatory components.111) The expression of e 6 is sensitive to levels of the intermediate p 8 .116) (3. the global cybernetic variable that describes the repression of e 6 synthesis by p 8 takes the form The elementary cybernetic variable that controls the activity of e 5 is given by r5 . i. p8 8 p p + p8 < p8 V e5. r 3 . p8 8 p p + p8 < p8 U e 6 .105) The local regulatory component consists of solely the elementary cybernetic variable and is given by v l5 = v c 5.110) Because e 6 is a member of only a single linear pathway the local regulatory component consists solely of the elementary component. (3. u l6 = u c 6. (3. p8 U e6 . [r q ] q =4 ].e. [r q ] q =4 ]. (3. (3.113) 104 (3.e.109) 8 k =4 rk 8 k =4 rk The local regulatory element governing the expression of e 5 is composed solely of the elementary component.115) . The elementary regulatory component that governs the expression of e 5 is given by uc 5= r5 r0+r1+r b + r bb 2 3 + . i. r 3 .metabolic engineering 1. p8 p max 8 p max 8 (3. Accordingly. because e 4 is not subject to any global regulatory signals the complete cybernetic variable that governs the activity of e 4 is given by v 4bb = v 4 = v .106) The activity of the key enzyme e 5 is sensitive to levels of the metabolite p 8 . u 6 = u l6 U 6 (3. Moreover. the global control variable that describes the inhibition of e 5 by the metabolite p 8 takes the form Because e 4 is a member of only a single linear pathway the local regulatory element is composed solely of the single elementary component. MT980104 Varner and Ramkrishna The elementary cybernetic variable that governs enzyme activity is given by vc 4= r4 . The complete cybernetic variable that controls the expression of e 6 is the product of the local and global regulatory components and is given functionally as p8 .119) .112) where the competition set [r k( c ) ] is given as bb 8 [r k( c ) ] = [r 0 ... the complete cybernetic variable which controls the activity of e 5 is given by p8 . p8 V e 5 . i. (3. r 1 .. (3. following the previous discussion. The local cybernetic regulatory component which controls the expression of e 6 is given by uc 6= r6 r0+r1+r b + r bb 2 3 + .108) The key enzyme e 6 is a member of only a single linear elementary pathway and is sensitive (both inhibition and repression) to levels of the intermediate p 8 . r b 2 . v l4 = v c 4.118) Because e 5 is not sensitive to global regulatory signals.

(3. g (3. max( [r k( c ) ] ) (3. (3.134) The complete cybernetic variable which controls the activity of e 6 consists of the product of the local and global regulatory elements and is given functionally as p8 v 6 = v l6 V 6 .e.e. (3. Formally. 1994).133) where yields the optimality condition dc dp g dc = .130) Because e 6 is a member of only a single elementary pathway.132) V = p8 6 { \ 0 1& 8 V max e 6 .. we must derive cybernetic variables that regulate the synthesis and activity of the lumped transport and growth enzymes e T. (3. this implies the constrained optimization problem max[c( R g .4. g )) ] subject to g = R g + R T. g = R.122) bb 8 [r k( c ) ] = [r 0 . The elementary cybernetic variable which controls the expression of e 7 is given by r7 u = b r 0 + r 1 + r 2 + r bb 3 + c 7 8 k =4 rk .Cybernetic Models and Metabolic Engineering metabolic engineering 1.123) Because e 7 is a member of only a single elementary pathway. p g( R T.126) Because e 7 is a member of only a single linear pathway.. the local regulatory element consists solely of the elementary component.e. max( [r k( c ) ] ) (3. r b 2 .121) The elementary cybernetic variable which controls the activity of e 7 is given by vc 7= r7 . because e 7 activity is not subject to a global regulatory signal. p8 8 p p + p8< p8 V e6 . g and e g . [r q ] q =4 ].124) 3. dR g dp g dR T. Following the development of Straight and Ramkrishna (Straight and Ramkrishna. u 7 = u l7 . i. i. respectively. (3. r 3 . the local regulatory component is composed solely of the elementary regulatory element. p8 V e 6 . v l6 = v c 5. 88 116 (1999) article no. (3. r b 2 . r 1 . Following the previous discussion.129) where the competition set [r k( c ) ] is given by bb 8 [r k( c ) ] = [r 0 . (3.127) Moreover. where the competition set [r k( c ) ] is given by (3. r 1 . the global control variable which describes the inhibition of e 6 by p 8 takes the form Furthermore. the local regulatory component is identical to the elementary component..e. the return on investment of structural resource can be measured as the reaction rate.128) 105 The coefficient dcÂdp g denotes the yield of biomass per unit of p g consumed and is considered constant. i.131) The activity of e 6 is sensitive to levels of the pathway intermediate p 8 . v l7 = v c 7. (3. the cybernetic variable that controls the .125) The key enzyme e 7 is a member of a single linear pathway and is not subject to any global regulatory signal. the complete cybernetic variable which controls the activity of e 7 is given by v 7 = v l7 . u l7 = u c 7. 8 (3.. e 7 synthesis is not subject to a global regulatory signal which implies the complete cybernetic variable that controls the expression of e 7 is composed of the local control element. Metabolic Regulation: Growth Model The lumped growth model proposed above is also a cybernetic model thus. p8 p max 8 p max . i. [r q ] q =4 ]. MT980104 The elementary cybernetic variables that control the activity of e 6 is given functionally as vc 6= r6 . If we assume that the allocation of critical structural resource takes place in time dt and the optimal allocation policy is implemented at every instant of time. Accordingly. we postulate the objective of the linear growth pathway is the maximization of biomass subject to a constraint on the level of structural resource available for the pathway operation. r 3 . (3.

we postulate flux control variable(s) of the form p0 t *( r T.6.140) It follows that the flux control variable(s) are given functionally by 0 = V flux p r T. we discuss the hypothetical regulation associated with the network input and output flux and the manifestation of this regulation via linkage of the network model to the abstracted growth model formulated above. We further postulate that 0 r T. i. $ large implies the lag between an upper pathway metabolic perturbation and the corresponding response in the network input flux is long. More precisely. as well as. MT980104 Varner and Ramkrishna allocation of critical resources to j th step of the growth process. and p are functions of the growth phase of the culture. p 4 . respectively. g Y cÂpg r T. g v T.. Thus. as the system moves toward balanced growth. The coupling is manifested via two sources. biomass level and temporal concerns. In the present context. hence. g( t-$ ) + max T. until eventually. In this way we incorporate some sense of upper and lower pathway regulation as it relates to temporal external abiotic phase phenomena. each state variable is subject to dilution due to growth. (3. coupling of the abstracted growth model to the metabolic network does offer the ability to qualitatively link the biotic phase network dynamics to easily observable external abiotic phase phenomena such as nutrient uptake. Complete Model System The time evolution of the model system is governed by the following set of differential equations: Nonenzymatic state variables . The functional forms of the cybernetic v-variables that regulate the activity of the key enzymes that catalyze the growth pathway reactions follow from the proportional law and are given by the set v T. r g ) vg= rg max( Y cÂpg r T. the enzyme level is low. The second source is the hypothetical dependence of input and output flux upon the glucose uptake and growth 106 ba j=pa 3 . In particular. g[ t-$ ] ) V flux j V flux t # j ( r g v g ).137) where $ denotes the length of time required for upper pathway metabolic perturbations to be felt by the aspartate network. 8 4 In other words. during the initial stages of the culture. However. and p 8 . Specifically. is given as u T. (3.141) These variables modify the rate of input and output flux so as to mimic the changes in the levels and activity of the key enzymes that catalyze these fluxes. the catalytic enzyme levels also moves towards a maximum level. Temporal Linkage: Growth and Network Model(s) Lastly.e. g Ä + max T. we lump the dynamics of the key enzymes that catalyze the input and output flux of the network into a set of control variables that are a direct function of growth rate and glucose uptake rate.metabolic engineering 1. p8 .. g #j= 1 + max g . p8 . perhaps changes in dilution rate in a continuous culture. Y cÂpg r T.139) 3 . p 4 . where the yield coefficient Y cÂ pg # dcÂdp g . p 4 . we are interested in transition between the balanced growth regime and the onset of accumulation.5. before proceeding to the formulation of model equations and the subsequent analysis. in general. direct calculation of the ba utilization of p a 3 . Upon exhaustion of a system growth substrate.135) rate(s). decaying to zero. denoted as u j . + max g ba j=p a 3 . g = Y cÂpg r T. ba j=pa 3 . (3. g j V flux = rg vg . p8 . g max( Y cÂpg r T. This implies the proportionality constant(s) take the form *= 1 + max T.138) ba j=p a (3. g v T. an explicit formulation of the growth process as it relates to the aspartate family is absent. 3. g v T. p ba . As the growth rate slows. or growth substrate concentration. because the network is written with respect to the biotic phase. p8 . p4 . g + r g (3. g . whereas. r g ) . However. (3. metabolite levels increase if the system maintains activity. j O V flux Ä 1.e. for example the carbon source. the importance of the dilution term decreases. We postulate that the key enzymes that catalyze the formation of p 0 and the consumption of p a 3. the level of catalytic enzymes is postulated to fall because of the lack of metabolic activity. g . the maximum level of input and output flux into the network occurs under balanced growth conditions. g O V flux Ä 1 p (3.136) 3. Firstly. these perturbations are limited to alterations in the growth rate. g = Y cÂpg r T. other intermediates for biomass synthesis is not possible. 88 116 (1999) article no. g + r g ug = rg . If $ is small. In the present model system. given the present development. then perturbations in upper metabolism are felt by the network quickly. p 4 . r g v g Ä + max g i.

dt dP 8 = r d. 3 dt de b * bb u = r e3 3bb &( r g v g + . dt j ba dp p4 ba ba ba = r ba 3 v 3ba & R p4 V flux & r d. = r 1 v1 & : r 2 dt j j = a. MT980104 dp 0 p0 &r 0 v0 & rg vg p0 . g & dt Y cÂ pg de T. as e j Ä e max the complete as e j approaches e max j j cybernetic u-variable governing the allocation of critical resources to the synthesis of e j moves towards unity. g c . ) e T. These assumpj g tions imply + . dt dp 8 p8 & r d. g = r eT. Firstly. 7 k = a. p5 c. =& dt Y cÂsg dp g 1 rg vg = r T. ) e g + r* eg dt dc = r g v g c. = R p0 V flux dt dp 1 = r 0 v0 & r 1 v1 & r g vg p 1 . = r 7 v 7 & R p8 V flux dt Balance on network enzyme levels: de j = r ej u j &( r g v g + . dt de k 2 k * ku = r e2 2k( r g v g + .Cybernetic Models and Metabolic Engineering metabolic engineering 1. 3 dt de bb 4 bb * bb u = r e4 4bb &( r g v g + . = r d. dt dp 6 = r 5 v5 & r 6 v6 & r g vg p 6 . dt dp 2 j v 2j & r g v g p 2 . g u T. given a few simplify assumptions. dt where r* ej denotes the rate of constitutive enzyme synthesis defined as r* ej = de ba 3 ba * ba u = r e3 3ba &( r g v g + . Secondly.142) where * ej denotes the basal level of expression from the plasmid.e. (3. b de T = r eT &( r g v g + . 5. dt ba 4 dp bb 4 bb = r bb 3 v 3 & r4 v4 &rg vg p4 . dt dp 5 = r 4 v 4 & r 5 v 5 & r d. we assume that e j & e max when r g v g & + max . g + r* eT. p2 & r g v g p 2 . is the rate constant governing the first order decay of enzymatic machinery. i. = r d. ) e T + r* eT . p4 & r g v g p 4 . The parameter . a dp a p2 3 a a a =r a 2 v 2a & R p2 V flux & r d. dt Balance and abiotic phase network intermediates: j dP 3 a c. ) e max + r e* &0 : ej &( + max g j j (3. g &( r g v g + . p8 & r g v g p 8 . we assume that enzyme synthesis is saturated with respect to inducer metabolite(s) . Lastly. p3 & r g v g p 3 . 1. p3 dt j = a. Ij & r g v g p I j . 88 116 (1999) article no. * ej is a measurement of the extent of promoter leakage. the maximum specific level of the j th key enzyme is readily determined from the e j balance. 6.. ) e 3 + r e ba . g v T. p5 & r g v g p 5 . 4 dt Balance on growth model species: dS g 1 r T. p8 c. dt dp Ij dt = r T . dt Balance on promoter inducer and transport species: dI j =&r T. ) e 3 + r e b . b de g = r eg u g &( r g v g + . dt dp b 3 bj b b =r b 2 v 2b & : r 3 v 3bj & r d. g dt 107 b 3 {r*+ * ej ej r* ej p Ij =0 p Ij >0. 2 dt j =0. Ij c.143) . Lastly. p 4 dt dP 5 = r d. b dP ba 4 ba c. ) e 4 + r e bb . g v T . dt dp 7 = r 6 v6 & r 7 v7 & r g vg p 7 . ) e 2 + r e k . The yield coefficient Y cÂ pg denotes the yield of biomass (gdw) per unit p g . ) e j + r* ej .

4\0. the no . however. we focus upon the qualitative aspects involved.1 2. 1995) and coworkers dealing with lysine and threonine overproduction in C. however. we simulate the exact experimental alterations made by Stephanopoulos et al. Thus..1 Threonine Homoserine Isoleucine <0.2 0.1 1. Specifically.0 4. \ j. batch culture) of t22 gÂL. from G. We examine this point subsequently.. the increased flux toward threonine was stalled at homoserine rather than moving all the way to the target. Specifically. The construct ATCC 21799 (pJD4) is designed to eliminate this situation by employing a push pull strategy around the accumulating intermediate homoserine ( p b 3 ). 4.1 Reproduced.8\0.1 <0. Colon et al. the model predicted course of action did result in diversion of flux away from lysine. al.. (Nakamori et al.3 <0. MT980104 Varner and Ramkrishna which can be easily solved for e max to yield j : ej + r * ej + max +. Interestingly.. examine the system given the previous branch point development.1. Given the scope of our investigation.9\0. the predicted course of action is the overexpression of the corresponding branch metabolite key enzyme. however.metabolic engineering 1. the strain ATCC 21799(pM2) has incorporated into the chromosome a deregulated aspartokinase which is no longer sensitive to inhibition by lysine. and contrast the model response with experimental observations. Experimental results reproduced from (Colon et al. this configuration results in an unexplained drop in lysine formation for the no induction case and a nonlinear dependence upon the level of the promoter inducer IPTG.2 5. More exactly. 1987) and Reinscheid et al. Specifically. glutamicum strain ATCC 21799 (pM2) ATCC 21799 (pJD4) ATCC 21799 (pGC42) No induction 1.5\0. Rather. some flux still found its way to lysine. whereas. Specifically.6\0. homoserine dehydrogenase (e b 2 ) is under its native promoter.9\0. (Reinscheid et 108 Lysine 22. given the predictions of the branch point analysis. that the kinetic competition for p 2 is not entirely in the favor of p b 3 formation. Sinskey. we begin the analysis portion of the development. Specifically.1 1. ANALYSIS AND DISCUSSION Now that we have outlined the model framework. we dispense with discussing the specifics of the genetic manipulations involved in the threonine overproducer construction. This implies. Thus. g TABLE 2 Table of Experimental Results: ATCC 21799 (pM2) Conversion to Threonine Overproducer a e max & j .1 1. the objective is to divert metabolic flux away from lysine ( p a 3) synthesis toward threonine production ( p 5 ). Bearing in mind the p 2 branch point is flexible.9\0. Specifically. This review is followed by examination of the model response to genetic perturbation.144) Strain: Corynebacterium lactofermentum ATCC 21799 Carbon Source: Glucose (80 gÂL) Sample Time: 90 h Amino acid concentration (gÂL) C.5 +molÂ100 ml 5 +molÂ100 ml a 4. Jetten et al. lactofermentum ATCC 21799.1 0.1 6. Along this line Nakamori et al.0\1.8\0.3\0. 88 116 (1999) article no.6 9.7\0. by permission.1 5. Experimental Results of Stephanopoulos and Sinskey The focus of the series of papers authored by Stephanopoulos. pGC42 assumes homoserine kinase ( e b 3 ) is under a tac promoter.0\0.3\0. (1995a). this configuration is clearly not optimal because large levels of lysine as well as homoserine still remain. and moreover.3 11. The construct ATCC 21799 (pGC42) approaches the problem of threonine formation from much the same perspective as pJD4 with the exception of the promotion. b. The logic of the experimental development follows closely the predictions of the branch point analysis. 1994) have overexpressed a feedback insensitive homoserine dehydrogenase ( e b 2 ) which did result in homoserine accumulation. Our goal in this endeavor is to evaluate the cybernetic models ability to predict the manner in which pathway metabolic function reacts to alteration in local genetic structure..1 0.2 0. (3. Moreover. and co-workers has been the construction of a threonine producing microbe using as a base a lysine overproducing strain. the push pull strategy does result in the formation of a substantial homoserine conversion. The objective of the experimental exercise is the conversion of the lysine producing ATCC 21799 (pM2) into a threonine overproducer. The experimental results show homoserine ( p b 3) accumulation accompanied by residual lysine formation ( pa 3 ) in addition to threonine formation.1 <0.0\0. 1995a. pJD4 contains the genes that encode for a feedback insensitive homoserine dehydrogenase ( e b 2 ) as well as a wild-type homoserine kinase under their respective native promoters. large amounts of residual lysine also accumulated. and Sinskey et al.9\0. This affords a microbe that has lysine titers (90 h. we review the experimental results of Stephanopoulos and Sinskey (Colon et al.1 <0. 1995a) are shown in Table 2.

Before moving to the evaluation of the genetic perturbations we take a few a moments to discuss the cybernetic significance of the overexpression of existing pathway enzymes given a frozen parameter set. Upon induction with 5 +molÂ100 ml of IPTG the level of threonine accumulation falls well below the level resulting from induction at 1. given the model formulation. This includes both macroscopic external variables such as growth substrate and biomass levels in addition to specific levels of metabolic intermediates. Unfortunately.5 +molÂ 100 ml of IPTG the bulk of the homoserine is converted to threonine and some lower pathway metabolites. description of the external operating environment is key to the temporal aspect of the model prediction. respectively. thus.75 mmolÂL-h) compares well with that predicted by the flux analysis . isoleucine. we are interested in verifying the concept that a cybernetic model can be used to predict the local network reaction to genetic perturbation. The model predicted glucose uptake rate during balanced growth (14.75 h &1 and the g biomass yield Y cÂSg =0. Thus. More exactly. we determine a model parameter set such that pM2 evolution is captured. It was reported that all the glucose was consumed by t =70 h yielding a final biomass titer of t1. This involves not only the description of biotic phase phenomena such as intermediate accumulation. 1989) and Kiss (Kiss. Our interest in this respect is two-fold. 88 116 (1999) article no. 1993) for an organism similar to pM2 producing lysine to get a handle on the parametric nature of glucose uptake. We make a special point in this regard because of our model structure. the prediction of metabolic network response to genetic perturbation. we have assumed the input and output flux of the network model to be directly proportional to the glucose uptake rate and system growth rate.0219 (gDwÂg). The experimental results of Stephanopoulos et al. and Sinskey et al. allows the cybernetic framework to capture the manner in which 109 enzyme synthesis and activity is redirected in the light of genetic perturbation. we fit the model parameters so as to describe the time evolution of ATCC 21799 (pM2) and then freeze the parameter set. From this data as well as the experimental information. methionine and isoleucine. This follows from the postulate that upper metabolic activity drains in the direction of the aspartate family $ hours after growth has ceased. In the present context. we determine the parameter set that allows the cybernetic regulatory framework to describe the goal oriented regulatory behavior that is inherent to this construct. but the prediction of abiotic phase evolution as well. the local objective of the aspartate family network is postulated to be the maximization of the levels of lysine. the metabolic engineering problem of how to convert a lysine producer into a threonine producer has merit in its own right. we are able to determine system parameters such that abstracted growth model is consistent with experimental observations. Investigation of Network Response We now subject the model framework to the identical genetic perturbations discussed in the previous section. we examine the model equivalent of pM2 (which is our base structure).Cybernetic Models and Metabolic Engineering metabolic engineering 1. Specifically. a predictive model framework could aid in the design and optimization aspects of the construction.e. 4. consists of only a single point taken at t =90 h. First. As an additional input we employ the flux analysis results of Stephanopoulos and Vallino (Vallino and Stephanopoulos. The microbes were aerobically cultured at 30% C (250 rpm) in 500 ml baffled flasks (100 ml of C. this is not our primary objective. namely. Freezing the parameter set around the description of pM2.2. Specifically.. Specifically. The maximum specific growth rate from the precursor pool p g was predicted to be + max =0. by fitting the base model to pM2. This perspective is much the same as the penicillin V biosynthetic network model formulated and evaluated in the next paper. Comparison of the predicted response with experimental data affords a means to this end. and then subsequently subjecting this framework to genetic perturbation. MT980104 induction case is marked by low levels of lysine accumulation coupled with relatively high levels of threonine and homoserine accumulation. The algorithm is as follows: Firstly. 1991)) which contained kanamycin (50 mgÂL) and ampicillin (50 mgÂL) and glucose as a carbon source (80 gÂL). specifically. our object is to compare the model predicted level of metabolic intermediates at t =90 h with experimental values. The cybernetic framework is a goal-oriented methodology that assume a microbe regulates the synthesis and activity of key enzymes so as to direct its nutritional state toward an objective.5 +molÂ100 ml of IPTG.8 (gDwÂL). where the input flux is delayed by $ h. As such. rather. it affords a description of how pM2 responds to genetic perturbation because we assume local wild type objectives are still in place. However. despite the genetic manipulation. this is the extent of the experimental observations at hand. glutamicum minimal media adapted from von der Osten and Sinskey (von der Osten and Sinskey. Thus. In what follows we examine the response of the model system to the genetic perturbations described above. a simple abstracted growth model drives the temporal aspects of the aspartate network model so as to describe both the external and internal culture time evolution. Upon induction with 1. The significance of this is subtle at first glance but becomes apparent when viewed from the perspective of our objective. pJD4 and pGC42. i.

95 mmolÂh-L.e. The inflection indicates a point where the production of lysine is becoming substrate limited. Simulations results for abiotic lysine level of pM2. This implies the same growth model. Stephanopoulos and Vallino predict 2. 1993) to get a handle on the input flux into the aspartate network. The construction of the base strain ATCC 21799(pM2) (which also carries a shuttle vector encoding for antibiotic resistance) results from the incorporation of a lysine feedback insensitive aspartate kinase ( e 0 ) into the chromosome of ATCC 21799 (Jetten et al. i. the growth performance of the microbe is not qualitatively effected by the presence of the plasmid.metabolic engineering 1. the lower pathway results. Note the predicted lysine level agrees well with experimental results. is in good agreement with the predicted flux analysis value. MT980104 Varner and Ramkrishna FIG. unlike plasmid borne alterations. Wild-type network (pM2)(pJD4)(pGC42): External growth model predictions. i.) The V r 1. under their respective native promoters. the input flux into the p0 aspartate network.. because we have assumed a direct proportionality with the growth rate. feedback insensitivity to the level of threonine is not a . can be used throughout our study. In addition to the experimental results discussed above. there does not exist multiple enzyme expression sources. it was reported that the growth characteristics of each of the constructs was similar.) Using these parameters. parametrically identical. from the plasmid or the chromosome.e. in the case of e b 2. 4 where the abiotic phase lysine level is plotted versus time. Thus. 110 FIG. for the levels of the various other metabolic intermediates sampled by Stephanopoulos et al. although not shown. This is represented a within the model framework by assuming V e0 . These temporal regions clearly stem from the influence of the macroscopic growth model and the assumption of upper pathway drainage. i. Near t =65 h the uptake and growth rate(s) slow and eventually drop to zero which implies the aspartate family output flux toward biomass formation. the model simulation of the abiotic state of the system is shown in Fig. p0 = the simulation predicts (during balanced growth) R 0 V flux 4. 3. accumulation stoppage. p 5 Because both enzymes are native to ATCC 21799 (pM2). The construct pJD4 assumes the gene for a threonine feedback insensitive homoserine dehydrogenase ( e b 2 ) and a wild-type homoserine kinase ( e b 3 ) are incorporated. i.. Note. denoted by R 0 V flux . the influx into the p 2 branch point slows and eventually stops. ( t15 mmolÂL-h..e.. The simulation results are shown in Fig. 88 116 (1999) article no. we again turn to the flux analysis of Stephanopoulos and Vallino (Vallino and Stephanopoulos. decays to zero. assumption of a feedback insensitive e b b 2 implies e2 . This results in intermediate accumulation because we assume the upper pathway metabolic drainage occurs for $ h after the cessation of glucose uptake. Moreover.. also can be shown to agree within experimental error.e. This is all that is required for the description of a genetic change incorporated directly into the chromosome because. on a plasmid inserted into the base pM2 construct (in place of the shuttle vector. At this point we freeze the parameter set used in the pM2 simulation and examine the ramifications of the genetic alterations. whereas. 1995). In particular. 3. and eventually. Notice the onset of the accumulation regime approximately at t =65 h followed by an inflection near t =75 h.. there exists multiple sources of enzyme expression. p3 r 1. Additionally.23 mmolÂh-L. 4. Note.

j =2. However. these parallel routes consist of the chromosomal and plasmid expression sources of e b 2 . 3.. respectively. Accordingly. (4.4) where : o .e. k = g. to account for the possibility of multiple expression sources for the key b b enzymes e b 2 and e 3 the synthesis source terms in the e 2 and b bu b e 3 balances. This is perhaps a questionable assumption. Thus. whereas. p from a single plasmid and N denotes the copy number. The superscript g and p denote genome and plasmid.3) 111 FIG. parallel routes competing to produce a common product. thus. (4. r e2 2b and r e3 u 3b must be replaced by g p r ejb u jb U jb + r eb U jb . as it is known that copy number varies as a function of culture phase. . the maximum level of expression of e b j from a single plasmid is equal to the maximum rate of expression of eb j from the chromosome.p 3 (4. (4. Model prediction of pJD4 dynamics. However. i. potentially important are neglected. 3. This understanding allows us to treat these cases using the standard assumptions described in the first paper of this series. The control variables k Uk 2b . The model prediction of the overexpression of b a feedback insensitive e b 2 and a wild-type e 3 versus experiment is shown in Fig. We.2) where the rate of enzyme expression from the plasmid takes the form p2 b+p K e2 2 pb 3 b. (4. which is assumed to be N =50. e b 3 overexpression can be described using the standard development. MT980104 feature of the wild type enzyme. however. In particular. 3. we have assumed the competition that regulates synthesis source is substitutable. although. Within the current context. assumption of substitutable competition is not strictly valid. we assume N to be approximately constant over the course of the fermentation. we have postulated that the multiple expression sources of enzyme e b j are substitutable. Of course. in a short period of time the majority of the e b 2 stems from the plasmid expression route and the ratio of b feedback insensitive e b 2 to wild type e 2 is strongly in favor of the former. e .. 5.e. p j p b ej j =2.1) p p where the second expression term r e b U jb denotes the expres- sion of enzyme e b j from the plasmid. i.Cybernetic Models and Metabolic Engineering metabolic engineering 1. the chromosomal source yields a wild type e b 2. :o =:b b 2 e . p denote the global cybernetic variables that regulate the choice of synthesis source of the enzyme e b j . thus.5) The parameter k denotes the sensitivity of macroscopic enzyme expression to changes in copy number. U 3b . N denote the maximum rate of e b b j expression ej . In short. Note. the global control variable that regulates the choice of expression source take the form r ebj u 2b r ebj u 2b + r eb j j U = g jb p U = p jb rp b e j r ebj u 2b + r eb j p .. b+p K e3 3 r eb = : eb 2 2 p p r eb = : e b 3 3 p p (4. acquiesce from such issues and assume the chromosomal route is slow in comparison to the plasmid route.e. These two enzymes are clearly not identical. the gene encoded on the plasmid yields a feedback insensitive e b 2. Because. 1992) we assume k <1. thus. The gene that encodes for a feedback insensitive e b 2 was isolated from a mutant strain and incorporated onto pJD4. b e . Consistent with the copy number heuristic of Shuler (Schuler. 5. The importance of such a detail follows from the derivation of the global control that regulates the choice of synthesis source. i. enzymatic overexpression is assumed to be under the control of their respective native promoters. The form of the global control variables was derived previously so we dispense with a formal derivation. 88 116 (1999) article no.p 2 :o =:b b 3.4) should be considered as a first approximation. such considerations. The simulated overexpression of e b 2 p j =2. j The rate constants : eb take the power law form j p : =:o ( N ) k.

e2 p b e2 (4. The low level of isoleucine indicates flow into the lower pathway below threonine is blocked by inhibition of e 5 by isoleucine.. Additionally. their activity is very low. Moreover. The observations are borne out through simulation of the network model. a sizable portion of the flux that has been diverted is wasted in the accumulation of homoserine ( p b 3 ).6) p p where : e b .3443 eb 2 b . i. flux from upper pathway drainage has nowhere to go other than p 5 . The lysine accumulating pM2 construct shows that e a 2 is clearly present at higher levels and is substantially more active than e b 2 . Accordingly.e. The various temporal regions of accumulation as well as the effect of system feedback interactions are clearly visible from the simulation. given the redirection at the branch point. max e2 v 2b \ + 0. j =2. Although the agreement with the experimental information is good at t =90 h. Note. and more importantly. in the case of artificial TABLE 3 b Level of Amplification of e b 2 and e 3 for Construct pJD4.3340 0.1736 1. these findings clearly illustrate that redirection of the flux away from lysine formation toward threonine is composed of two aspects. The rate p constant : e b takes the power law form given above and the 2 maximum rate of expression from a single plasmid. b K + p I.3537 5. the diversion of flux at the p 2 branch point and the elimination of homoserine accumulation. 3 are increased.6049 0. K b denote the rate and saturation constants that e2 2 govern the expression from the plasmid of e b 2 . the rationality of such an observation can be derived upon examination of Table 3. however.4608 1. As a thought experiment (parameter set slightly different than pM2) to test our hypothesis that a stronger promoter governing feedback insensitive e b 2 expression could eliminate lysine accumulation. however. some lysine formation is observed even in the amplified pathway. our measure of promoter strength is left as a parameter in the artificial case.1315 15. homoserine levels at longer times fall below the experimental values.0466 0. This also explains the large threonine accumulation during this stage. namely. Simulated enzyme level and activity HD LYS HK Construct \ + 0. The onset of accumulation occurs as the system growth rate decays to zero consistent with the discussion above.8304 0.metabolic engineering 1. The construct pJD4 does divert flux away from lysine formation.1168 0. These results are consistent with the experimental findings of Stephanopoulos and co-workers (Colon et al. Upon Varner and Ramkrishna amplification.1) where the rate of expression from the plasmid is given by the form b=: b r e2 e 2 p b p I.e. In the case of pGC42 induction is initiated at t =30 h. and conversely.1245 0. 1995a) and are tabulated in Table 3.0662 0.2910 eb 2 b . The model predicted response agrees well with that observed experimentally.4906 0.4215 1. max e2 \ + 0.2246 eb 3 b . Interestingly.3913 0.0272 0.4977 0. threonine is accumulating to much higher levels than reported experimentally..1438 ea 2 a. some residual lysine is still present along with a substantial level of homoserine. 112 . However.3664 0. the levels of e b j . the activity of these enzymes increases dramatically affording e b 2 an increased share of the precursor metabolite p 2 . the majority of the branch point precursor p 2 is converted to lysine due to the competitive advantage of e a 2 . max e3 \ + 0.7094 12. the drain of upper pathway metabolites decays to zero by approximately t =80 h and homoserine is converted to threonine.4118 7.. max e2 v 2a \ + 0. From a mechanistic perspective. MT980104 b and e b 3 results in an t7-fold increase in the level of e 2 .5 +molÂ100 ml 5. (4. we consider the overexpression of a feedback insensitive homoserine dehydrogenase under artificial and native promoter(s).4040 1.4215 0. and moreover. pGC42 as Compared with pM2 at t =50 h. 88 116 (1999) article no. clearly. max e2 \ + 0. i.9600 eb 3 b . max e3 v 3b pM2 pJD4 pGC42 No Induction 1.0169 0. To account for the possibility of multiple sources of e b 2 we replace the rate of enzyme synthesis in the e b balance with 2 Eq.2727 0.and b an t6-fold increase in the level of e 3 .0387 ea 2 a .0 +molÂ100 ml Note. the enzyme levels on the homoserine side of the branch are relatively high. e 2 .0078 0. This indicates that eb 2 is not being overexpressed strongly enough to completely pull p 2 away from p a 3 formation.

Amino acid composition (gÂL) Construct WT hom dr-N hom dr-( p 1 ) hom dr-( p 2 ) Lysine 24. Amplification response HD LYS TABLE 5 Intermediate Level Response to Expression of a Feedback Insensitive Homoserine Dehydrogenase as a Function of Promoter Type and Promoter Strength at t =90 h. the resistance to flux alteration still affords lysine accumulation. however. The change in promoter configuration is reflected in the model configuration by the form of the rate p of enzyme expression from the plasmid. p 3 p p o Construct \ + \ + 0. the increased flux toward threonine formation stops at homoserine.724 4.. The construct pGC42 assumes the gene that encodes for a threonine feedback insensitive e b 2 and the gene that are under separate promoters. 1995a). This allows e b 2 to capture a larger share of p 2 and the level of lysine falls accordingly.3989 1. max e2 ea 2 a .) To account for the multib ple sources of key enzyme e b 2 and e 3 we replace the enzyme b b synthesis terms in the e 2 and e 3 balances with those shown in Eq. are identical to those given previously.5559 eb 2 b . This implies. whereas.4545 1.e.4900 0. e b is under its native promoter. Increasing promoter strength by a factor of 2 yields drastically different results. respectively. i.7539 Note. Thus. max e2 v 2b \ + \ + 0.. This simulation result validates the postulate that stronger promotion of feedback insensitive e b 2 can eliminate lysine accumulation. The rate constant : e b is 3 given by the power law form shown above and the maximum rate of e b 3 expression from a single plasmid is an order of magnitude greater than the maximum rate of wild type e b 3 expression. k = g. .e. i. i. Note.7) where : eb . The system response to overexpression both in terms of intermediate level as well as enzymatic level and activity are tabulated in Tables 4 and 5.3069 0.. The overj b expression of the enzyme e 2 remains under its native p promoter which implies r eb is identical to the previous case.0687 <0.6812 1. The artificial promoter case for a promoter strength equivalent to native (hom dr-( p 1 ) case) is nearly the same as wild type. Consider the native promotion case. When applicable inducer (5 +molÂ100 ml) is introduced at t =30 h.1 4.e. (4. b 2 The overexpression of e 3 is now assumed to be under an p artificial promoter which implies r e b is of the form 3 r e b = : eb 3 3 p p b p I .Cybernetic Models and Metabolic Engineering metabolic engineering 1..3315 0. e3 . under an artificial promoter ( tac.7764 0. hence.2380 0.1).1380 ea 2 a .5378 24. however. the still exits the problem of homoserine accumulation which is consistent with the experimental observations of Nakamori et al.6017 0. The level of e b 2 increases t3 fold whereas the activity increases dramatically. 113 The model simulations of pGC42 under three different expression loads are shown in Figs. respectively.1 6.5045 Threonine <0. the level and activity of e a 2 decrease significantly.0137 0. 88 116 (1999) article no. At this level of promotion.1 3. the need for the push aspect of the strategy. p where superscript g and p denote variables U jb genome and plasmid. max e2 eb 2 b .1 3. but not threonine.2278 0. : eb . b K + p I. there is still a substantial lysine accumulation.0199 0.5154 0. e3 p b e3 (4. encodes for a wild-type e b 3 Specifically. K e b denote the rate and saturation constants that 3 3 p govern the overexpression of e b 3 . The functional forms of the global control k . the flux through the p 2 branch point is almost preferentially drawn towards the formation of homoserine. 6 8. even though e b 2 levels and activity have been increased by at least 3-fold. the inducer (5 +molÂ100 ml) is introduced into the media at t =30 h.661 Homoserine <0.0910 0. The results in terms of enzyme level enhancement are shown in Table 3. e b 2 3 is TABLE 4 Enzymatic Response to Expression of Feedback Resistant Homoserine Dehydrogenase as a Function of Promoter Type and Promoter Strength at t =50 h.. Note the latter set of results is qualitatively consistent with the experimental development of Stephanopoulos et al.5844 0. In contrast. (Colon et al. MT980104 promotion. r eb . the model system clearly predicts the failure of a simple deregulation pull approach to redirect flux toward threonine and away from lysine. Note.9377 <0. g # : 3b =10 WT hom dr-N hom dr-( p 1 ) hom dr-( p 2 ) Note. max e2 v 2a ' p. When applicable inducer (5 +molÂ100 ml) is introduced at t =30 h.

Rather. does not give a rationale as to why this might be so. an unexplained reduction in the level of lysine and moreover a nonlinear response to the levels of IPTG. However. Note a biological basis for this hypothesis is currently under investigation. 7. Given this adjustment. because e b 2 overexpression is under the control of its respective native promoter. The model system predicts that the drop in lysine level follows as a consequence of a postulated enhancement of the e b 2 native promoter strength. 8. FIG. it was found by simulation that the native promoter strength of e b 2 must increase by a factor of 2.metabolic engineering 1. . Colo Á n et al. MT980104 Varner and Ramkrishna FIG. 6. 88 116 (1999) article no. it is observed that lysine levels drop dramatically in pGC42. thus. Specifically. Model prediction of pGC42 dynamics: 1. namely. Model prediction of pGC42 dynamics: no induction case. this is found not to be the case.5 +molÂ100 ml IPTG. we employ the model framework to explain these experimental observations. the model predicted system response matches well the experimental observations. FIG.5 to account for the decrease in lysine levels. 114 The pGC42 experimental data has an interesting feature associated with the no induction case. Model prediction of pGC42 dynamics: 5 +molÂ100 ml IPTG. Intuitively. This hypothesis follows straightaway from an understanding of the behavior of flexible nodes presented in the preceding paper of this series. We postulate the no induction response and the nonlinear network response to induction levels follows as a consequence of a cooperation between the artificial and natural promoters. we expect the lysine level in the no induction case to be identical with pJD4.

154 161. it would seem that this explanation is unlikely. the experimentally observed local network response under a number of different perturbations. (1996a). We postulate the rate of expression of e b 2 under its native promoter is influenced by the level of induction of e b 3 overexpression. H. 482 488. U. M. J.Cybernetic Models and Metabolic Engineering metabolic engineering 1.. given the cooperation hypothesis. T. J. G. A. Biotechnol. Biotechnol. the model predicted response(s) shown in Figs. V. Follettie.. the maximum rate of expression of e b 2 must pass through a maximum near the 1.. Control of metabolic flux in yeasts and fungi. ``Metabolic Activity Control of l-Lysine Fermentation by Restrained Growth Fed-Batch Strategies. A. the availability of more experimental information as to the wild-type state and the response of the system to perturbation could definitely aid in improving the quality of prediction. Cambridge. Kiss. Certainly. such as. and moreover. M. Given these adjustments.27 over the no induction case to capture to the experimental dynamics. has not been undertaken.. Optimization of regulatory architectures in metabolic reaction engineering. The deeper objective of this development is to evaluate the cybernetic framework's ability to describe the manner in 115 . V. the current functional power law form could be modified such that p k =: o :e ej . Sinskey. Microbiol. The implication of this set of results is far reaching in terms of the conceptual applicability of a cybernetic metabolic engineering approach. pp. Microbiol. Appl. specific metabolite levels or operating environment. Kascer. M. the proximity of model prediction and experimental observation is sufficiently close to warrant the suggestion that the founding postulates of the framework put forth have conceptual merit. Appl.. Nguyen. Additionally.). G. and Stephanopoulos.. T. The model framework was subjected to genetic perturbations. G. (1996b). and the network model responses were compared to experimental observations. J. (1998). and Sinskey...D.. Biotechnol. Bioeng. Colon. Appl. 74 78. Towards a science of metabolic engineering. 485 500. J. hom dr (under its native promoter) is cloned in front of homoserine kinase. (1995a). In ``Control of Biological Processes'' (D... 61. Sinskey. the native promoter strength must be increased by a factor of 1. Specifically. Jetten. Massachusetts Institute of Technology. 1998a. MT980104 Another obvious possibility for the unexplained decrease in lysine level in construct pGC42 is basal expression stemming from a ``leaky'' artificial promoter. 52. Specifically. it lends credence to the assumption that local genetic alterations do not alter the presumed local network objective.9) where f ( X ) is a function that describes the sensitivity of the promotion of e j to culture conditions. Ed. (4.. C. A. The nonlinear response with respect to induction level can also be explained given the cooperation hypothesis. A. thus. A. Effect of inducible thrB expression on amino acid production in Corynebacterium lactofermentum 21799. It was found that an appropriately modified base model system predicted. however. 1277 192. Analysis and design of metabolic networks via mixed-integer linear optimization.. Floudas. 42. T. Emmerling. 1668 1675. M. Production of isoleucine by overexpression of ilvA in a Corynebacterium lactofermentum theorine producer. As the level of inducer is increased. 88 116 (1999) article no.. This feature has direct bearing on the utility of the metabolic engineering development outlined in briefly here and in more detail in (Varner and Ramkrishna.. should not drastically impact e b 2 levels. Science 252.. The cooperation mechanism can be brought into the current framework through the reformulation of the maximum rate of enzyme expression from a single plasmid. and Sinskey et al. Cambridge Univ. we formulated a cybernetic model that describes the time evolution of the aspartate family of amino acids. Hatzimanikatis. Davies.045 times the no induction case to capture the experimental observations. 445 447. Gubler. j which a metabolic network responds locally to genetic alteration. (1991). G. E. and Burns. Bioeng. Jetten. 58.. (1973). the native e b 2 promoter strength must decrease to 1. 65 104.. Microbiol. M-T. 43. the quality of the simulation results presented could be improved by paying more careful attention to the kinetic forms of the rates involved in the pathway and by formulating more sophisticated representations of the effects of metabolite feedback. and Stephanopoulos. CONCLUSIONS Using the modular approach. leaky expression of the kinase should have no obvious effect on lysine levels. M. The control of flux. temperature. Sauer. 5. Environ. However. R. S. b). V. and Bailey. It is predicted from simulation. and Bailey. 7 and 8 agrees well with experimental observations. 76 82. A. that a recombinant system is still optimal with respect to resource allocation albeit to an expanded or contracted set of alternatives. thesis. J. A. for the most part within experimental error. AIChE J... Jetten. 15. E. and Bailey. J. C. Thus. (1991). Brown. However. Primarily. (1995).e. J. Hatzimanikatis. Biotechnol. E. E. REFERENCES Bailey. Hatzimanikatis. 43. (1997). p f ( X )( N ) .'' Ph. M. A systematic formulation of f ( X ) to date.. Colon. Application of mathematical tools for metabolic design of microbial ethanol production. M. Floudas. Effect of different levels of aspartokinase on the lysine production by Corynebacterium lactofermentum. as outlined by Stephanopoulos et al. these issues aside. (1995b). Nguyen. Trends in Biotechnol. Follettie.5 +molÂ100 ml induction level and then decrease as the inducer concentration increases. i. Press.

Design of a defined medium for growth of Corynebacterium glutamicum in which citrate facilitates iron uptake. Microbiol. Appl. 26. K.. Metabolic engineering from a cybernetic perspective II. (1998a). Biotechnol. J. E.'' pp. (1998b). C. and Ramkrishna. Cybernetic modeling and regulation of metabolic pathways. D. 205 219. (1994). and Sahm. Nonlinear analysis of cybernetic models.. 1372 1382. J. D. J. (1987). S-B. Nakamori. and Sinskey.. Reinscheid. Vallino. 11 16. 51. Biotechnol. Network rigidity and metabolic engineering in metabolite overproduction.. 60. J. Kronemeyer. Submitted for publication.. Biotechnol.. NJ. H. and Ramkrishna.'' Prentice Hall. (1989).. C. Varner. G. S. Ishida. and Stephanopoulos. J. Theoretical preliminaries.. K. 126 132. Englewood Cliffs. J. (1991). ``Bioprocess Engineering: Basic Concepts. 1675 1681. Stephanopoulos. J. J. M.. and Sano.. Lett. Ito. and Bailey... Biotechnol. and Ramkrishna. Tagaki. Guidelines for model formulation. 574 587. and Vallino. Prog. Flux determination in cellular networks: Applications to lysine fermentations. (1992). Gioannetti. H.. (1984). Science 252. Varner.metabolic engineering 1. Chem. 10. Environ.. Bioeng. Improved l-threonine production by the amplification of the genes encoding homoserine dehydrogenase in Brevibacterium lactofermentum. A. Printed in Belgium 116 . Genetically structured models for lac promoter-operator function in Escherichia coli chromosome and in multicopy plasmids: lac operator function. MT980104 Lee. W. Qualitative investigation of nodal architectures and their response to genetic perturbation. 11. (1998c). L. and Ramkrishna. Stable expression of him-1-thrB in Corynebacterium glutamicum and its effect on the carbon flux to threonine and related amino acids. G. J. (1993).. Varner and Ramkrishna Straight. D.. 88 116 (1999) article no. K. Metabolic engineering from a cybernetic perspective I. 87 91. Agric. Shuler. Eggeling. Growth on complimentary nutrients. D. Eikmanns... in press. Varner. D. In ``Frontiers of Bioprocessing. Biol. Submitted for publication. (1994). von der Osten.

- LubrpltXchrts
- Mobil Shc Pf 462
- Mobil Centaur XHP 460 Br v1
- Compressed Air Ref Eng
- GuideToAirCompFlds
- A Guide to Global Warming Harvey (1993a_Energy_Policy_GWPs)
- Artigo Updated Radiative Forcing Estimates of 65
- PossamaiLisiane MP
- Grammar Essentials Graded Exercises in English Robert j Dixson
- 164064-CO2-marine
- ManualTecnicoCaldeiras_2006
- SISTEMAS FIXOS DE EXTINÇÃO DE INCÊNDIO POR AGENTES GASOSOS
- Monografia Carlos Felipe Aschoff CEEST
- Compilacao Legislativa SCIE Www
- Renato Port Hun
- Folder Cascipe Agosto 2013 PDF
- Mobil Product Guide South Pacific May09
- Fundamentos de Ecologia
- Alegria e Triunfo - Lourenço Prado
- Studyguide Bobtail
- 28536136 Manual Abiquim
- Fispq+Glp+Rev+01 11
- Fundação CECIERJ
- Fundação CECIERJ
- Chemical Process Safety, 2nd Ed, Daniel a. Crowl, Joseph F. Louvar

- Lecture 6 - Metabolic Modeling
- 1_5 Anhydro fructose a versatile chiral building block biochemistry and chemistry.pdf
- Kurse DTU
- Flux Balance Analysis
- npjsba20169
- 46_395
- Liver Detoxification Review Liska
- 1-s2.0-S1369527407000513-main
- Enzyme Assays and Kinetics
- Summer Research Fellowships 2015.pdf
- sdarticle-12
- 153678-november-2012-mark-scheme-63.pdf
- Enzyme Activity D
- How to Write an Effective Conclusion
- Biology Course Work
- Estimation of Metabolic Fluxes
- Tips to Answer Paper 3 Biology
- NITROGEN FIXATION
- fulltext[1]
- Bio Transformations Using Plant Cells, Organ
- How to Answer Paper 3 Question CHEMISTRY
- NutraHacker Complete SAS
- Full Text
- IJEB 51(4) 322-335.pdf
- Microbiology and Industrial Biotechnology of Food-Grade
- GATE HELPLINE Bioprocess Engineering MCQ III
- Ghosh D 2004 PhD Thesis
- 1-s2.0-S0167779907002879-main.pdf
- Untitled
- EXTREMOZYMES

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd