Yeast extract (YE) is commonly used as a key component in the complex media for industrial fermentations. However, the lot-to-lot variation of this raw material frequently requires extensive "use testing" of many lots to identify only the few that support desired fermentation performance. Through extensive fermentation studies and chemical analyses, we have identified adenine and two metabolizable carbon sources, trehalose and lactate, as the principle components in YE that affect the production of a recombinant protein antigen by a yeast strain. Adenine is required for culture growth and the relationship between biomass and measured adenine can be expressed by a Michaelis-Menten model, while the slowly metabolized trehalose serves to maintain the energy supply to the continued antigen synthesis. The rapidly utilized lactate exerts an indirect positive effect by sparing some of the accumulated ethanol from being consumed for growth to being utilized in the product formation. The effects of these YE components are mutually dependent. Based on the database generated from 40 lots at laboratory scale, a relatively high level of carbon sources in YE (trehalose plus lactate, >9.5% w/w) and an intermediate level of adenine (0.14-0.24% w/w) appear to be the minimal requirement of a good lot for this recombinant yeast fermentation. Many poor lots were improved in lab fermenters by rational supplementation of trehalose, lactate, or adenine to compensate for their insufficiencies. At the large production scale, predictions based on adenine and trehalose/lactate contents in various YE lots used correlated reasonably well with culture growth and antigen yield, illustrating the feasibility of such a simple chemical/biochemical analysis as a rapid and reliable initial screening tool. Without incurring any compositional change to an established manufacturing medium, this study demonstrates an effective approach to achieve consistency in fermentations employing complex nutrients and to improve fermentation productivities supported by suboptimal lots of raw material. (C) 2003 Wiley Periodicals, Inc.