Jasmine Bracher

Nearly 200 parties signed and committed to the Paris agreement in 2017, which comprises the long-term goal to keep the average global temperature increase well below 2 degrees above pre-industrial levels. Virtually all possible scenarios drafted to reach this goal include a strongly increased use of biofuels for transport by land, sea and air. Bioethanol, whose production could, in principle and in contrast to fossil fuel production, involve a closed carbon cycle, is generated by microbial fermentation of sugars from plant-derived starch or agricultural waste. This liquid transport fuel provides a readily implementable alternative to fossil fuels as it combines the advantages of sustainable fuel production and compatibility with existing combustion engine technologies, without a requirement for time-consuming and expensive changes in our current infrastructure.

To date, bioethanol is the largest volume product of industrial biotechnology. 99% of this ethanol is generated via 1st generation processes, largely derived by fermentation of hydrolysed sugar cane or corn starch by bakers’ yeast (Saccharomyces cerevisiae). So-called ‘2nd generation’ bioethanol, for which the first commercial-scale plants are now starting up, is made by fermentation of sugars present in lignocellulosic biomass, typically harvested from agricultural waste streams, such as wheat straw or sugar beet pulp. Whilst such feedstocks enable a “food and fuel” scenario, their industrial implementation brings along additional challenges for yeasts and biotechnologists. Hydrolysis of lignocellulosic biomass, in particular the cellulose and hemicellulose fractions, releases a mixture of different sugars as well as inhibiting compounds that impair growth and viability of S. cerevisiae. Whilst glucose is the most abundant fermentable carbon source, the pentose sugar d-xylose can cover up to 30% of the total sugar content. The fraction of the pentose l-arabinose typically varies between 2 – 20%, depending on the feedstock used. Although pentose sugars cannot be fermented to ethanol by wild-type S. cerevisiae strains, international research efforts over the past two decades yielded metabolic engineering strategies to enable anaerobic conversion of d-xylose and l-arabinose to ethanol by S. cerevisiae.

The expression of heterologous, d-xylose- and l-arabinose-isomerase based pathways from fungi or bacteria, together with over-expression of genes of the non-oxidative pentose phosphate pathway (PPP) and deletion of the unspecific aldose-reductase gene GRE3 within S. cerevisiae allows this yeast to aerobically metabolize both sugars. The recent advances in metabolic engineering tools, such as CRISPR-Cas9-assisted genome editing, greatly advanced the construction and characterization of metabolically engineered S. cerevisiae strains with improved yields, kinetics and robustness in 2nd generation ethanol production processes.

CHAPTER 1 of this thesis summarizes the history and recent advances in metabolic engineering of S. cerevisiae for the production of bioethanol from an academic and industrial perspective.

Cellular uptake of the native substrate glucose by S. cerevisiae is highly efficient and involves complex regulation of the expression levels and in vivo activities of 20 hexose transporters, which together offer an adaptable range of transport affinities and capacities suited for various extracellular glucose levels. A similar system for the uptake of pentose sugars has not evolved within S. cerevisiae. Whilst d-xylose can enter the cells via multiple hexose transporters, l-arabinose transport depends on the single galactose transporter Gal2, which however exhibits a low affinity for l-arabinose and whose expression is strongly repressed by glucose, which is abundantly present in lignocellulosic hydrolysates. Hence, glucose is a potent competitive inhibitor of l-arabinose transport via Gal2. The combination of both characteristics delays the total fermentation time of a mix of glucose and l-arabinose due to sequential uptake of those sugars and causes a prominent “tailing effect”, reflected by very slowly decreasing l-arabinose concentrations towards the end of the fermentation. A low ethanol productivity, impairs inhibitor tolerance of the cells and negatively affects process economics.

The objective of CHAPTER 2 of this thesis was to increase the affinity of l-arabinose transport in engineered, l-arabinose fermenting S. cerevisiae strains. To this end, the analysis of chemostat-based transcriptome data of l-arabinose-grown Penicillium chrysogenum cultures yielded a selection of candidate l-arabinose transporter genes. Their separate expression in an engineered S. cerevisiae strain lacking GAL2, and growth tests on l-arabinose led to the identification of one transporter that enabled growth within the described setting. In contrast to Gal2, this protein, termed PcAraT, was shown to allow l-arabinose transport even in presence of glucose, theoretically enabling co-fermentation of those two hydrolysate-borne carbon sources. Uptake experiments with radiolabelled sugars revealed a very high l-arabinose affinity of this proton-symporter and the absence of transport activity for both, glucose and xylose. Four hundred fifty-fold lower residual l-arabinose concentrations in chemostat cultures with engineered strains expressing PcAraT, as compared to similar strains expressing Gal2, corroborated those results and confirmed the value of PcAraT as a promising addition to engineered, l-arabinose fermenting S. cerevisiae strains.

PcAraT conferred glucose-insensitive, high-affinity l-arabinose uptake which is specifically relevant for “mopping up” remaining l-arabinose towards the end of a fermentation. However, the transport capacity of this transporter alone did not allow fast consumption of high quantities of l-arabinose. The endogenous galactose transporter Gal2, confers l-arabinose transport with a satisfying transport capacity, but suffers from severe glucose repression.

The aim of CHAPTER 3 of this thesis was to improve l-arabinose transport capacity of engineered, l-arabinose consuming S. cerevisiae in glucose-xylose-arabinose media in order to allow co-fermentation and ultimately reduce total fermentation times of mixtures of those sugars. To this end, a glucose-phosphorylation-negative, PcAraT expressing, S. cerevisiae strain that could transport but not metabolize glucose, was generated and subjected to adaptive laboratory evolution in sequential anaerobic batch reactors. The latter contained glucose-xylose-arabinose media, thereby generating a selective advantage for cells capable of taking-up l-arabinose in presence of high concentrations of competing sugars. Whole- genome sequencing identified duplications of GAL2 as well as mutations within this gene in multiple independently evolved single-colony isolates. The relevance of individual mutations was assessed both via reversion to wild-type alleles in evolved strains and by introduction of mutations into the native parental strain, as well as by quantitative transport assays using radiolabeled sugars. Whilst the mutated GAL2 alleles revealed novel Gal2 characteristics in favour of l-arabinose transport, duplication of GAL2 alone was shown to, in itself, not confer instantaneous anaerobic growth on l-arabinose. However, transport assays with Gal2N376T,T89I revealed glucose-insensitive l-arabinose transport characteristics and showed that this transporter completely lost its ability to transport glucose. The transporter encoded by a second evolved allele, GAL2N376T, showed high l-arabinose transport capacities at the cost of reduced glucose transport abilities. Lastly, Gal2T89I showed a decreased glucose affinity while increasing overall l-arabinose affinity at the cost of transport capacity of both sugars. Restoring this allele to wild-type within an evolved strain nearly doubled the total fermentation time of 20 g L-1 l-arabinose within a glucose-arabinose-xylose mixture from 25 to 45 h. Deletion of PcAraT within the same evolved strain reduced overall growth rate from 0.12 h-1 to 0.1 h-1 and consequently increased the overall fermentation time. With the discovery and functional expression of PcAraT in S. cerevisiae, and the selection for and characterization of mutated Gal2 alleles with improved l-arabinose transport characteristics in presence of glucose, valuable elements for improved l-arabinose fermentation were provided for the generation of novel yeast strains for 2nd generation bioethanol production.

With d-xylose accounting for up to 30% of the total sugar content of lignocellulosic biomass, anaerobic fermentation of d-xylose is paramount to ensure an economically viable process. Not surprisingly, a significant part of the research effort to engineer S. cerevisiae for biofuel production has been dedicated to anaerobic d-xylose fermentation. Remarkably, reports on metabolic engineering of d-xylose-fermenting, xylose-isomerase based S. cerevisiae strains disagree as to whether extensive laboratory evolution is necessary to enable anaerobic growth on d-xylose or not. In particular, two previous publications from the TU Delft’s industrial microbiology group, using the same laboratory strain and a similar metabolic engineering strategy, in one case reported instantaneous anaerobic xylose fermentation after targeted metabolic engineering while, in the second case, anaerobic growth on this pentose required anaerobic evolution and an additional mutation.

The objective of CHAPTER 4 of this thesis was to systematically reassess the genetic requirements for anaerobic growth on d-xylose of engineered, CEN.PK-derived S. cerevisiae strains. To this end, reconstruction of d-xylose-metabolizing strains allowed for the identification of subtle differences in cultivation procedures that strongly affected anaerobic growth. Inoculum density was found to be a particularly important factor. Cultures of a particular engineered strain inoculated with 0.2 g biomass L-1 readily picked up anaerobic growth on d-xylose upon inoculation of the anaerobic bioreactor while, at an inoculum density of 0.02 g biomass L-1, the same strain required an anaerobic lag phase of 7-8 d prior to initiation of anaerobic growth. Mimicking the higher initial CO2 concentrations in high-inoculum density cultures by sparging low-inoculum-density cultures with CO2-enriched nitrogen gas or by using l-aspartate instead of ammonium as nitrogen source, reproduced the phenotype of the former. Finally, omission of over-expression cassettes of the PPP paralogs NQM1 and TKL2 allowed the (re-)construction of a solely metabolically engineered strain that instantaneously fermented xylose when inoculated at low inoculum densities and sparged with pure nitrogen gas. Together, these observations resolved apparent contradictions in the literature on metabolic engineering strategies required for anaerobic growth of S. cerevisiae on d-xylose and expanded our knowledge on the potential relevance of CO2 availability and anaplerotic carboxylation reactions on anaerobic d-xylose fermentation by metabolically engineered S. cerevisiae strains.

To ensure standardization and reproducibility of scientific results, yeast growth media in academia as well as industry are routinely supplemented with vitamins to prevent potential vitamin limitations from affecting the studied phenotype. Careful investigation of the exact vitamin requirements and prototrophies (i.e the ability to synthesize certain vitamins intracellularly), has the potential to significantly reduce media costs. Previously, whole genome sequencing of the laboratory strain S. cerevisiae CEN.PK113-7D revealed the presence of all biosynthesis genes assumed to be required to synthesize biotin (vitamin H/B7). Nevertheless, this strain showed only barely detectable growth in biotin-free media.

CHAPTER 5 of this thesis was inspired by the academic interest in unravelling the as yet unknown aspects of the biotin biosynthesis pathway in S. cerevisiae, and by the attractivity of biotin prototrophic S. cerevisiae strains for industrial and academic use. To this end, CEN.PK113-7D was selected in media lacking this expensive supplement. Using parallel sequential batch cultures and accelerostat regimes yielded biotin prototrophic strains, some of which grew as fast in the absence of biotin as in its presence. Whole-genome sequencing and reverse-engineering studies identified a massive amplification of BIO1 to be predominantly responsible for improved growth in biotin-free media. The additional deletion of the genes encoding the transporters Tpo1 and/or Pdr12, which were found to be inactivated in evolved prototrophic strains, further improved the growth rates of a BIO1-overexpressing strain. The identification of metabolic engineering targets for obtaining biotin prototrophic yeasts in CHAPTER 5 has the potential to improve process economics of a possibly large number of biotechnological products and to inspire future researchers to expand this method to other expensive growth factors.