3.2 | Biosynthesis of oleochemicals
We expanded our model to include various oleochemicals by adding Michaelis-Menten parameters for new enzymes (SI Methods). These parameters, which are commonly reported in the literature, are reasonable when enzyme-substrate complexes maintain steady-state concentrations during oleochemical production—a reasonable assumption in stationary phase, where protein concentrations are approximately constant, and rates of cell formation and death are approximately equal (Li et al., 2014; Pletnev et al., 2015). With these parameters in hand, we assembled models for the biosynthesis of alcohols, alkanes, methyl ketones, fatty acid methyl esters (FAMEs), and fatty acid ethyl esters (FAEEs) by carrying out two steps: (i) We added estimated values of kcat and Km for the requisite enzymes (Fig. 1), and (ii) we adjusted enzyme concentrations and kinetic parameters by fitting our model to the product profiles of engineered strains (Figs. 2 and 3). We note: In all fits, we scaled titers reported from experimental cultures to titers expected for in vitrosystems at 12 minutes (SI Methods). Our models provided good fits to experimental data with two notable exceptions: Tow models for methyl ketone biosynthesis underpredicted the production of either (i) 2-nonanone by a pathway containing Umbellularia californicaUcFatB1 (BTE), a thioesterase specific for short chains (strain 5 in Fig. 3A), or (ii) palmitic acid by TesA, a thioesterase specific for medium chains that is native to E. coli (Fig. 3B). The underprediction of 2-nonanone matches the narrow product profile that we used to parameterize BTE; the associated study may have overlooked minority products. For palmitic acid production by TesA, the large experimental concentration may result from contamination by membrane lipids, which are hydrolyzed in common extraction procedures (Grisewood et al., 2017). Importantly, we recreated the experimental profile by reducing kcat for fatty acid-CoA ligase (FadD) on palmitic acid by four-fold (Fig. S8), but this adjustment is inconsistent with the reported substrate specificity of FadD (i.e., it requires the kcat for C16 acyl-CoAs to be lower than for C14 and C18 substrates (Arora et al., 2005)). In the absence of a strong experimental justification for re-parameterizing the enzymes in question—that is, a rationale for using one experimental dataset over another—we left their kinetic parameters unchanged.
Our modeling analysis allowed us to classify oleochemical pathways by the availability of kinetic data and the required fitting procedure. We created three groups: (i) Pathways with