Although many plants with interesting phenotypes have been generated by genetic manipulation, the central metabolic objective of being able to make predictable changes to specified fluxes generally remains elusive. The numerous reports of engineered plants with metabolic phenotypes that are not usefully different from the wild type, for example, in starch metabolism (Fernie et al., 2002), show that the rational manipulation of plant metabolism is far from straightforward, and that in many instances our understanding of plant metabolic networks is insufficient to permit accurate predictions about the metabolic consequences of genetic manipulation. Unexpected metabolic phenotypes are interesting in their own right since they often provide information about the structure and regulatory properties of the network, but from an engineering perspective, they are undesirable since they consume resources and reduce the efficiency of the process.

If the production of unwanted metabolic phenotypes is to be avoided, then metabolic engineering has to be based on a detailed quantitative understanding of the capabilities of the metabolic network. Essentially this requires: (1) definition of the network of reactions, (2) definition of all the molecular interactions in the system that have an impact on the functioning of the network, and (3) specification of the intracellular and external environments in which the network is functioning. Unfortunately, each of these requirements is potentially very demanding: the plant metabolic network is of necessity complex, reflecting the demands placed on sessile organisms that live in a fluctuating environment; this complexity increases the scope for regulation of the network through changes in enzyme level (via changes in gene expression and protein turnover) and enzyme activity (via covalent modification, effector binding, and changes in substrate and product concentrations); and for most purposes, plants have to be grown under non–steady-state conditions, thus complicating any prediction of metabolic performance. The net result of these complications is that models of plant metabolism (Giersch, 2000; Morgan and Rhodes, 2002) tend to be relatively limited in scope and to fall some way short of the virtual cell that is required if accurate predictions are to be made of the impact of genetic manipulation on metabolic fluxes.

Three topics central to the development of a quantitative understanding of the metabolic capabilities of plant cells are discussed in this chapter. First, the complexity of the plant metabolic network is described and the prospects for obtaining a complete description of the network are assessed. Second, a review is provided of some of the tools that are now available for understanding the structure and performance of the network. Finally, to emphasize the level of sophistication that is required for models with real predictive value, we review some landmark studies that highlight the complexity of the system-wide mechanisms that permit the integration of plant metabolism. The emphasis is on the primary pathways of carbon metabolism since these pathways are fundamentally important for the functioning and manipulation of the network.