Scope of this chapter with the proviso that methods adopted in the early stages of development should ideally

be robust and adaptable to manufacturing scales. Our chapter mainly focuses on pilot-scale fermentation and downstream process development. In its simplest form, scale-up can be achieved by merely increasing the scale of compound production and subsequent purification. There are times when this approach is the most practical route to getting a compound quickly. There is an inevitable risk, however, that any increase in scale, without sufficient process development, will not be successful. An alternative and generally preferable approach is to increase the concentration of the target compound relative to unwanted compounds (impurities) in the material to be processed. This in turn can significantly improve the ease of downstream purification and ultimately result in enhanced yields. The latter course is usually undertaken with microbial product in the form of an integrated program of process development to improve both the fermentation and DSP processes. Cost and process efficiency become more important as the scale increases, and there will always be some scope for improving the purification method itself. The initial compound isolation will, in most cases, have been bioassay guided, with speed of identification the major objective, and is therefore highly unlikely to have been efficient in terms of either compound production levels or purification step yields. As the scale of operation increases, for fermentation products at least, so does the proportion of the overall scale-up process taken up by DSP activities. The importance of the relationship between the production and the DSP of the target compound, and the potential for the latter activity by improving the former, cannot be stressed too highly. 2.methods 2.1 assays and product quantitation The importance of reliable, quantitative assay as tools for directing the development of scale-up process, by measuring target compound concentration in fermentation and in concentrated extracts or eluates during various stages of purifications, should not be understated. They allow the accurate determination of step yields, thus providing a means of measuring and improving efficiencies. Appropriate assays are usually chromatographic and should be capable of high throughputs (typically 10-100 of samples per day). Reversed-phase highperformance liquid chromatography (HPLC) method with UV/ visible or occasionally mass spectrometric detection are used most frequently. Evaporative light scattering detector (ELSDs) allow the detection of all compound in a sample, including those with no UV absorbance, and add an extra dimension when assessing purity. These chromatographic methods are preferred to assays based on biological activity, through which many natural products are first detected and discovered , as such assays measure a total response to all of the active compounds present in a sample. Given that the majority of secondary metabolites are produced mixtures of

closely related compounds with varying degrees of activity, biological activity assays will deliver a composite response that could be misleading in respect of a particular compound. There is also scope for synergistic or antagonistic effects between components of a mixture in such biological assays. Compound-specific chemical assays are thus useful for separately quantifying different compounds within a series. There is a danger, however, in relying too much on an assay that is highly product specific. For fermentation development, at least, there is merit in running in parallel a product-specific HPLC assay with a steep gradient elution method that allows gross changes in the overall metabolite profile to be monitored. It is relatively common in process development to implement a desirable improvement in process characteristic (e.g., higher titer, better medium composition, to name a few) that also produces some unwanted impact on the process (e.g., a change in the profile of minor components, morphological change in the producing organism, and so on ). It is essential, therefore, that the analytical techniques applied be broad enough to pick up both desirable and undesirable change during the development phase. 2.2 fermentation development The general principles for process development were outlined in the first edition of this volume (1). Hence, we do not reiterate there principles here. Instead, we update and expand upon these initial comments and illustrate with recent references where appropriate. For instance, we cite a case study that exemplifies the more standard approaches to process development, and in addition, illustrates the value of having a taxonomically well-characterized culture collection, which can be explored not only for new secondary metabolites but also for other interesting biological properties such as organisms that may produce related compounds or possess useful enzymic activities that can be utilized to modify certain molecules. We also provide a brief basic introduction to the increasing application of recommendation techniques to process development. 2.2.1 reproducibility and inoculums The first step in process optimization typically involves some investigation into medium composition. Yields of will-tipe organisms are often around 1 mg/L or less, and medium optimization can be an effective means to rapidly improve yields to greater than 100 mg/L. These early stage optimization studies are typically carried out in shake flasks and tend to be based around statistical experimental methods such as Placked Burman, surface response, multivariate, and principal component analysis. Process reproducibility is a necessary prerequisite before process development and optimization can begin, or more properly it might be considered the first step, as it provides a reliable baseline against which to gage future improvements. This is particularly important when using statistical

experimental design techniques, where process variability may be misinterpreted as a statistically significant improvement. It is uncommon for organisms producing natural products to display a certain degree of physiological and metabolic instability which, in practice, may manifest itself in a lack of reproducibility-for instance, in activity in a downstream bioassay. Precious resources and much time cab be tied up both in fermentation and chemical isolation by trying to track down these lost or variable activities from wild-type isolates. It is therefore desirable in the first instance that the initial activity be reproduced in the same way as the original fermentation. A confirmed activity from such a refermentation will provide the scale-up scientist with the confidence to take the hit further. The first scale-up fermentation should preferably be performed in a system as close as possible to the fermentation system in which the hit was first detectedusing multiple units if necessary-to avoid unnecessary complications with scale-up issues. For example, if the initial hit was first detected in an extract obtained from 50 mL shake flask fermentation, then multiple shake flasks could be employed to produce volume up to 5-10 L. if the referment is successful, it adds to the growing confidence in the activity and increases the likelihood that the activity will scale-up into stirred fermenters. If the referment is unsuccessful, then this is likely indicates some issue with poor reproducibility rather than a scale-up issue. Alternatively, if the equipment is available, bench top fermenters may be used-if this stategy is adopted it is desirable that a confirmatory