Skip to main content
eScholarship
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Application of 2-Scale 13C Metabolic Flux Analysis to Growth Phenotypes in S. cerevisiae

Abstract

Fluxes are useful because they are the most phenotypically relevant output we can infer. The dominant methods of obtaining these profiles, FBA and 13C MFA, have advantages and disadvantages. 2S-13C MFA combines the advantages of both to obtain more reliable genome-scale flux profiles that are self-consistent with both the carbon transition model and data used to infer them. This makes them a better basis for troubleshooting phenotypic changes and for making predictions.

We do not fully understand the role of Sip1 in glucose repression systems. A better understanding of it and the general phenomenon of carbon catabolite-repression could result in better ways to engineer cells and utilize various feedstocks. Previous work in our lab re- sulted in an unreported growth phenotype upon knockout of SIP1 in mixed glucose/galactose medium. To better understand the relative roles of galactose and knockout of SIP1, we constructed base and sip1∆ mutant strains in a CEN.PK113-7D ura3∆ gal1∆ background and compared their 2S-13C MFA-derived flux profiles in both glucose-only and mixed glu- cose/galactose media. Our original hypothesis, that deletion of SIP1 was necessary to see an effect from the presence of galactose, was incorrect. The presence of galactose was nec- essary to see a phenotypic difference in growth rate upon knockout of SIP1. Both the base and mutant strains exhibited, upon addition of galactose, increases in specific growth rate, decrease of PPP pathway activity to 1/100 of its initial value, shift from use of NAD- to NADP-dependent malic enzyme, and redistribution of flux towards branched-chain amino acid biosynthesis. Despite the expected lack of change in growth rate upon knockout of SIP1 in glucose repressing conditions, extracellular ethanol flux decreased, mitochondrial flow completely shut down, and flow was directed toward a part of the network involving arginine and threonine biosynthesis. Regardless of the noted differences, all strain/condition pairs were accompanied by certain normal glucose-repressing phenotypes (i.e. ethanol pro- duction and repression of TCA/glyoxylate cycle activity). Additionally, a cycle from cytosolic pyruvate through cytosolic malate and the mitochondria back to cytosolic pyruvate through NAD-dependent malic enzyme occurred. Clearly, glucose repression and the role of Sip1 is more complicated than we realized. Recent work makes it plausible that galactose enters the cell at the galactose/glucose ratio in this study. The more-reliable and self-consistent flux

12

profiles generated by 2S-13C MFA were instrumental in studying this problem at unprece- dented detail. However, more work is necessary to fully elucidate the role of Sip1 in carbon catabolite-repressing conditions.

Carbon lost to non-target metabolites (e.g. ethanol) represents a missed opportunity. It might be possible to redirect this lost flux towards biomass and/or target-molecule produc- tion. Additionally, the question of whether we can use an NADH oxidase to correct cofactor imbalances is an interesting one. We attempted to rescue the growth of an ADH1 null mutant via low-copy plasmid expression of 15 different promoter/gene pairs. These promoter/gene pairs consisted of all combinations of 5 mutated TEF1 promoters and 3 species-specific NADH oxidases from L. lactis, S. pneumoniae, and A. capsulatus. The growth of these 15 variants were characterized, along with that of base and adh1∆ mutant strains, in low- replicate shake flask, 96-well, 24-well, and high-replicate shake flask experiments. Expression of the NADH-oxidase appeared to result in a growth rescue in the initial low-replicate flask and the 96-well plate experiments. There was no clear difference in 24-well plate experiments and the high-replicate shake flask experiment definitively showed that heterologous expres- sion of one particular variant resulted in no growth rescue. More experiments are necessary to determine if growth rescue by balancing cofactor utilization is a viable strategy. Higher replicate experiments with the same variant, high-copy expression, chromosomal integration, and/or codon-optimization could all be tried as extensions of this work. It’s interesting how there appeared to be a difference in the 96-well experiment, admittedly with about half the replicates of the final one. Differing rates of aeration at these two scales may be responsible for this observation, since all NADH oxidases used in this study converted oxygen to water.

Flux profiles are the most phenotypically relevant output we can infer from experimental data. Any advance in the speed or accuracy of flux profile inference from 13C labeling data could increase the range of its applications. The optimization inherent to 13C MFA is a high-dimensional, nonconvex least squares minimization subject to third-order polynomial constraints. Nonconvex systems tend to be harder to solve due to local minima. We were interested to see if we could find bounds or solver starting points and/or even recover the same global minimum solution using a convex semidefinite program relaxation with forms of rank- sparcity encouragement. For a six-reaction toy model, we found that the SDP-relaxation resulted in a lower bound on the objective and the rank-sparcity encouragement methods had their advantages and disadvantages. The weighted-trace objective method resulted in an optimal solution and argument closer to that found in the paper, albeit requiring more user supervision and resulting in a less-feasible point. The eigenvalue inflation method was more automatable, faster, and possessed better mass balance constraint satisfaction. Though these results are encouraging, more work remains to be done to determine whether convex relaxations could improve overall 13C MFA flux profile inference. The space of solutions possesses a semi-linear nature so, it is plausible that convex relaxations could be useful. The SDP relaxation attempted here and the rank-sparcity encouragement methods, among others, should be attempted on larger networks.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View