Inverse Metabolic Engineering: A Strategy for Directed Genetic Engineering of Useful Phenotypes James E. Bailey,” Adriana Sburlati, Vassily Hatzimanikatis, Kelvin Lee, Wolfgang A. Renner, and Philip S. Tsai Institute of Biotechnology, ETH Zurich, CH-8093 Zurich, Switzerland Received September 29, 1995/Accepted April 8, 1996 The classical method of metabolic engineering, identify- ing a rate-determining step in a pathway and alleviating the bottleneck by enzyme overexpression, has motivated much research but has enjoyed only limited practical success. Intervention of other limiting steps, of counter- balancing regulation, and of unknown coupled pathways often confounds this direct approach. Here the concept of inverse metabolic engineering is codified and its ap- plication is illustrated with several examples. Inverse metabolic engineering means the elucidation of a meta- bolic engineering strategy by: first, identifying, con- structing, or calculating a desired phenotype; second, de- termining the genetic or the particular environmental factors conferring that phenotype; and third, endowing that phenotype on another strain or organism by di- rected genetic or environmental manipulation. This para- digm has been successfully applied in several contexts, including elimination of growth factor requirements in mammalian cell culture and increasing the energetic ef- ficiency of microaerobic bacterial respiration. 0 1996 John Wiley & Sons, Inc. Key words: inverse metabolic engineering hemoglobin cell cycle CHO cell culture culture fluorescence INTRODUCTION Metabolic engineering has been defined as “the improve- ment of cellular activities by manipulation of enzymatic, transport, and regulatory functions of the cell with the use of recombinant DNA technology” (Bailey, 1991) (use of the term “improvement” in this definition connotes an identi- fied goal and, therefore, clearly refers to a directed manipu- lation). The potential applications of metabolic engineering span the entire spectrum of biotechnology, and encompass creation of new processes and products as well as improve- ment of existing processes. Considering the general feasi- bility of introducing any heterologous genes (natural or syn- thetic) and of making any change in the host genome, the set of genetic possibilities available to the metabolic engineer is almost infinite. Only a small number of these infinite pos- sibilities will be effective in achieving the metabolic engi- neer’s goals. This means that metabolic engineering is al- most certain to fail unless powerful algorithms can be iden- tified which greatly increase above random chance the probability of identifying an effective genetic change. * To whom all correspondence should be addressed. Fax: 411-633-10-5 The classical problem in the early emergence of meta- bolic engineering is identifying a flux-limiting step in a specified metabolic pathway. This formulation of the objec- tive embodies an implicit assumption of several layers of knowledge about the pathway. Not only is the identity of the pathway assumed-what steps occur-but also the identity of the catalysts involved should be known. Furthermore, to choose a possible flux-limiting step, other than one at ran- dom, much more information must be available, whether in terms of reaction kinetics, intermediate metabolite concen- trations, or results from well-designed stimulus-response experiments (Cornish-Bowden and Cardenas, 1990; Ga- lazzo and Bailey, 1990; Schlosser et al., 1993). Later developments of metabolic engineering have con- sidered much more complicated metabolic networks and objectives such as selectivity improvement or creation of a pathway new to the original network. However, common to most of these exercises is a rational, deductive approach (Bailey, 1991). Typically, based on knowledge of the meta- bolic system of interest, a genetic manipulation is proposed which in some way has postulated potential benefit based on the expected perturbation within the known biochemical network. This classical way of posing the metabolic engi- neering problem, and of formulating a solution strategy, will here be termed constructive metabolic engineering. Results from the constructive metabolic engineering ap- proach have been mixed. Notable successes have been achieved, for example, in bacterial processes for amino ac- ids and a few other fine chemicals. However, in many cases, the metabolic consequence of the genetic change based on the constructive strategy differs substantially from that de- sired (Bailey, 1991). Perhaps with the benefit of accumu- lated hindsight, such failures are quite likely due to several classes of limitations in knowledge of metabolic networks, effects of which will here be called secondary responses to metabolic engineering. First, any network contemplated in undertaking a metabolic engineering constructive design, now and in the foreseeable future, is always a subnetwork of a much larger, much more complex global metabolic net- work, at least at the level of a cell and usually extending to an interacting multicellular population. The subnetwork of interest is typically coupled with the global network through common cofactors and metabolites and, possibly, more Biotechnology and Bioengineering, Vol. 52, Pp. 109-121 (1996) 0 1996 John Wiley & Sons, Inc. CCC 0006-3592/96/010109-13