Optimization of Saccharomyces cerevisiae central carbon metabolism for biomaterials production

(1)

François Seys

Master programme in Molecular Biotechnology Degree project 45 hp, 2016

Biology Education Center and Visolis Inc, Lawrence Berkeley National Laboratory, University of California Berkeley

Supervisors: Dr. Brian Lee and Dr. Deepak Dugar

Optimization of Saccharomyces cerevisiae central carbon metabolism for biomaterials

production

(2)

(3)

i

Abstract

Shifting our oil-based economy to a bio-based economy is a critical component in the fight against climate change and for ensuring energy security. Improving the scalability of bio-based products by reducing production costs is an important part of this objective. We report here an attempt to increase the yield of mevalonate production, as mevalonate is a chemical of industrial interest. It can indeed be transformed into several high-value chemicals traditionally obtained from fossil resources. Some of these chemicals can be used as fuel additives or polymerized into compounds such as rubber.

We used a kit of standardized genetic parts to build a gene deletion system based on CRISPR-Cas9 to delete several genes in one single step. Four parameters of the method were tested for improvement: (1) the length of the donor DNA; (2) the number of double-strand breaks per target gene; (3) the transcription mechanism of the guide RNAs; and (4) the delivery method of the donor DNA.

The said gene deletion system was then used to delete eight genes hypothesised to generate by- products of the central carbon metabolism (ethanol and glycerol) in a Saccharomyces cerevisiae strain already engineered to produce mevalonate. No more than three genes were successfully deleted in the same strain. However, we report significant increases in the titer of mevalonate production (13.3 g/L final concentration, a 54% increase) and in the molar yield (7.94% of theoretical mass yield, a 50% increase) by a adh1Δ gpd1Δ gpd2Δ triple mutant strain. Continued improvement in the same strain with further deletions of the other alcohol dehydrogenases is expected to lead to a bio-based production of mevalonate in S. cerevisiae that is competitive with current fossil-based methods.

(4)

ii

(5)

iii

Domesticating microbes to produce bio-based chemicals

Popular Science Summary François Seys

Oil is a finite resource that will soon be depleted. We use it to move our cars and fly our planes, but we also need it to produce all kinds of plastics and rubbers. Burning oil produces gases that slowly heat up the planet, and if the planet heats up too much, scientists predict that catastrophic things will happen. We are speaking about the flooding of entire cities, mass extinctions, and major droughts.

As if that were not enough, oil and gas is not fairly distributed around the globe: some regions have much more than others. As a result, people fight for control over the regions with a lot of oil – such as the Middle East. It leads to a lot of pain and violence in the world.

People have thus been trying to find other means of moving our cars, flying our planes and producing plastics and rubbers. In this project, we try to use sunlight, as this is an abundant resource that will not run out for billions of years, and which is for the most part evenly distributed around the globe. We collect sunlight using crops, and then we feed the crops to tiny microbes. We can domesticate the microbes to produce the chemicals that are normally produced from oil. We do it much faster than when we domesticated the dog thousands of years ago: we use a brand new genetic engineering method called CRISPR-Cas9.

In this project, we tried to make the CRISPR-Cas9 method faster and easier by changing a few things. We found out that some tiny modifications could be made, but we also suggested a bunch of new tests that could improve the method in the future.

Then we finally used the CRISPR-Cas9 method to domesticate the microbe. It worked as predicted, and we managed to produce 54% more of our target chemical than usual. The chemical we want to produce is called mevalonate; we can use it to make car tires and many more things. The microbe is now also 50% better at transforming crops into mevalonate and not something else. In practice, that will not be enough to compete with the companies that produce the same chemical from oil. We think that the microbe could do at least 5 times better if we continue to tinker with it, so that one day we will be able to replace oil.

Degree project in Molecular Biotechnology, 2016

Examensarbete i molekylär bioteknik 45 hp till masterexamen, 2016

Biology Education Centre and Visolis Inc, Lawrence Berkeley National Laboratory, University California Berkeley

Supervisors: Dr. Brian Lee and Dr. Deepak Dugar

(6)

iv

(7)

v

Acronyms and Symbols

Acronyms used in the present report and their respective meaning.

Acronym Meaning

Adh Alcohol dehydrogenase

CO2 Carbon dioxide

ConL Left connector

ConR Right connector

CRISPR Clustered regularly interspaced short palindromic repeat DHAP Dihydroxyacetone phosphate

DO Dissolved oxygen [%]

DSB Double-strand break

dsDNA Double-stranded DNA

FD Forward

G3P Glyceraldehyde 3-phosphate

GG1 vector Vector resulting from the first Golden Gate assembly step GG2 vector Vector resulting from the second Golden Gate assembly step GG3 vector Vector resulting from the third Golden Gate assembly step Gpd Glycerol 3-phosphate dehydrogenase

gRNA Guide RNA

HPLC High Pressure Liquid Chromatography

HR Homologous recombination

KO Knock out

LB medium Luria-Bertani medium

NADH/NAD⁺ Nicotinamide adenine dinucleotide

OD Optical density [abs]

ORF Open reading frame

PAM Protospacer adjacent motif

cPCR Colony Polymerase chain reaction Pre-crRNA Precursor CRISPR RNA

RV Reverse

sfGFP Superfolder Green Fluorescent protein snoRNA Small nucleolar RNA

tRNA Transfer ribonucleic acid YPD yeast extract peptone dextrose

(10)

2

(11)

3

1. General Introduction

Bio-based Economy, Climate Change, and Energy Security

The ongoing global warming is now an established fact in the scientific community. The Earth’s climate is well on its way to reach 4°C above the pre-industrial temperature average by 2100. CO2

emissions, and other greenhouse gases, are the main drivers of this dramatic change (IPCC 2014).

The Paris agreement from 2015 vowed to limit the temperature rise to 2°C above the pre-industrial temperature (United Nations 2015). This objective would require drastic changes in the structure of modern societies and the inner workings of our industry. Mainly, it would require stopping the use of fossil fuels to meet our energy consumption, as these fuels are the main emitters of greenhouse gases (IPCC 2014).

The development of renewable energies is thus required in order to replace fossil fuels. While our electricity consumption can eventually be replaced by technologies such as solar panels, geothermal plants, or wind farms, over 80% of worldwide energy consumption is based on fuels (IEA 2015). As such, it is urgent to develop fuel alternatives to oil. The sun’s photons can be processed by biological agents to store chemical energy in the form of solid, liquid, or gaseous fuels. So-called biofuels are an excellent candidate to replace fossil fuels, as their raw material – light – is present everywhere on earth in abundance. A large range of biofuels can be developed to suit the particular conditions of each country (Naik et al. 2010, Taylor 2008).

However, oil-based fuels are still predominant, as their industry is fully developed (Morone and Cottoni 2016). The biofuel companies attempting to integrate with the market must thus meet the price set by fossil-based fuels. However, biofuel companies do not benefit from the economy of scale developed by the petrochemical industry over the past century. Production costs for biofuels are thus usually higher than oil-based fuels. As a consequence, most biofuel companies need government support to survive (Morone and Cottoni 2016).

Nevertheless, other products can be made based on the same bio-based technology. Some chemicals not used as fuels – isoprene for example – have a sufficiently high market price that the production costs from bio-based processes might become competitive (Ye et al. 2016). A logical business model would thus be to start a company producing those valuable chemicals, and afterward progressively develop the plant to lower the production costs and increase the production volume. Only then does it become economically possible to switch to biofuel production and compete with the economy of scale enjoyed by the petrochemical industry.

Chemicals produced from biological sources, or biomaterials, are thus an important component of the fight against climate change. Their production is in many cases carbon-negative, as they are not destined to be combusted. The abundant biomass they generate as a by-product is non-toxic, carbon- neutral and can be recycled in many ways – such as into fertilizers, feedstock, or even combustibles.

More importantly, they relieve our dependency on oil and oil-producing countries, which is an important aspect of energy security (Herman et al. 2007, Miller and Sorrel 2014).

A given biomaterial is often more expensive than its fossil-based counterpart. It is thus critical to reduce the production costs of bio-based materials. The most obvious way of achieving this goal is to

(12)

4 reduce waste by transforming more of the substrate into the desired chemical. This can be achieved by process engineering and/or strain engineering. The work presented in this report focuses on strain engineering.

Structure of the Project

This report begins with the characterization of the mevalonate producing strains of Saccharomyces cerevisiae that have been engineered to produce less side-product in order to increase the yield of mevalonate production. The nature and the impact of each gene deletion will be detailed. The genetic system that was used to delete several genes at once is also described, and its effectiveness is evaluated.

Additionally, attempts and directions to optimize the construction of the genetic system are reported, so that a similar system can be built more effectively in the future.

(13)

5

2. Optimization of Central Carbon Metabolism in S.

cerevisiae

Introduction

2.1.1. Mevalonate Production in S. cerevisiae

The chemical targeted for bio-based production in the present project is mevalonate. It is a six- carbon hydroxy fatty acid that can be transformed by an efficient chemical process (>90% yield) into various other valuable chemicals traditionally derived from oil, e.g. rubbers or fuel additives.

(Jakočiūnas et al. 2015a, Ye et al. 2016).

S. cerevisiae was the microorganism chosen for production of mevalonate, as it presents several advantages versus other industrial microorganisms such as Escherichia coli. Namely, it is cheap to grow, robust in large-scale fermentation, and has a lower risk of contamination (Li and Borodina 2015).

The maximal mass and molar yields of mevalonate production from glucose for S. cerevisiae central carbon metabolism are of 55% and 67% respectively. Indeed, one mole of CO2 per pyruvate (3C) is lost after glycolysis. As such one mole of glucose (6C) yields only two moles of acetate (2C) and 3 moles of acetates are required for 1 mole of mevalonate (6C), which gives a maximal molar yield of 1.5 moles of glucose for every mole of acetate, or 67%. Since mevalonate and glucose have the same amount of carbon atoms, carbon and molar yields are used interchangeably over the course of this report.

The main strain used during this project has already been engineered to produce mevalonate as a valuable product of the central carbon metabolism (Figure 1). It will be referred to as the WT strain for the rest of the project.

Figure 1: Schematic of the central carbon metabolism of mevalonate producing S. cerevisiae strains ( yVS373).

Blue: genes that are responsible for a carbon positive or carbon neutral step. Red: genes and chemicals that are responsible for a carbon negative step. Purple: non-native genes integrated by genetic engineering to enable S. cerevisiae producing

mevalonate.

(14)

6 The strategy followed here to increase the yield of mevalonate in S. cerevisiae consists of deleting side products of the central carbon metabolism. Indeed, not all of the carbon fed to the yeast will be transformed into mevalonate. Some of it will be consumed to ensure cell survival, and some of it will also be integrated into chemicals only marginally beneficial to a yeast grown in an artificial environment. Such chemicals reduce the carbon yield of the whole process and are wastes from an industrial point of view.

As can be seen in Figure 2, ethanol and glycerol are two molecules produced abundantly by our WT strain, and since they stem from the same pathway as mevalonate, their deletion is expected to significantly improve the yield of mevalonate. The metabolisms responsible for their production will thus be shut down. To assess the effects of these deletions on yeast’s central metabolism, the levels of mevalonate, ethanol, glycerol, and acetate produced by the engineered strains will be monitored.

Figure 2: Mevalonate, ethanol, and glycerol productions of a strain of S. cerevisiae engineered to produce mevalonate. Fermentation at 10 mL-scale in YPD medium (A) or at 1 L-scale in Westfall medium complemented

with leucine and yeast extract (B). Depending on the conditions, ethanol or glycerol is the major product of the fermentation instead of mevalonate.

(15)

7 2.1.1.1. Ethanol Metabolism

Ethanol production is catalyzed by alcohol dehydrogenases (Adh), an enzyme that catalyzes the following reaction, in which an aldehyde is reduced into an alcohol:

acetaldehyde +NADH ↔ ethanol + NAD⁺

It takes place right after glycolysis during sugar fermentation in yeast and many other organisms to recycle NADH back into NAD+. This reaction maintains the redox balance of the cell by preventing accumulation of NADH and depletion of NAD⁺ (De Smidt et al. 2008).

Several such Adhs are known to be expressed in S. cerevisiae. Five are directly involved in ethanol metabolism (Adh1 to Adh5) (Bennetzen and Hall 1982, Young and Pilgrim 1985, Smith et al. 2004), while two others (Adh6 and Adh7) are thought to have a different function and have a low specificity (De Smidt et al. 2008). All Adhs are cytosolic except for Adh3 which is active in the mitochondrial matrix. This might also be the case with Adh4 (Dreawke and Ciriacy 1988, Huh et al. 2003).

ADH1 is undoubtedly the gene principally responsible for ethanol production. However, not only do adh1Δstrains still produce ethanol but triple adh1Δ adh3Δ adh4Δ (Drewke et al.1990) and adh1Δadh3Δ adh5Δ (Smith et al. 2004) knock-out mutants also generate a significant amount of ethanol when grown on glucose substrate. As a consequence, it is suspected that the different Adhs are able to substitute for each other to some extent, depending on the cellular context (De Smidt et al. 2008). Clearly, the regulation and the activity of all the ADH genes are still not completely understood.

We thus chose to delete all the ADH genes from ADH1 to ADH6 and study the impact of such deletions on ethanol production and the final yield of the central carbon metabolism. The following points cover each targeted Adh separately in more details.

 Adh1

Adh1 is the main enzyme responsible for acetaldehyde degradation (Leskovac et al. 2002). It is only repressed when S. cerevisiae is grown on non-fermentable carbon sources such as glycerol or ethanol (Denis et al. 1983). As a consequence, S. cerevisiae produces ethanol even in the presence of oxygen (Lagunas 1979, Lagunas 1986).

 Adh2

Adh2 amino acid sequence is 95% identical to that of Adh1 but catalyses the reverse reaction by oxidizing ethanol back into acetaldehyde. By doing so, it enables S. cerevisiae to use ethanol as a carbon source. Adh1 is also able to fulfill this function to a lesser extent, if cytosolic ethanol concentrations are low and acetaldehyde is removed fast enough. Adh2 is repressed by glucose and anaerobic conditions (Ciriacy 1975, Russel et al. 1983, De Smidt et al. 2008).

 Adh3

Adh3 is very similar to Adh1 but is only expressed in the mitochondrial matrix. As such, it is responsible for the re-oxidation of mitochondrial NADH (Young and Pilgrim, 1985; Bakker et al.

2000). It is not able to rescue S. cerevisiae in anaerobic conditions if ADH1 is knocked out.

However, it decreases alcohol production if knocked out in conjunction with ADH1 in aerobic

(16)

8 conditions (Smith et al. 2004). It is also repressed by glucose (although to a lesser extent than Adh2) (Leskovac et al. 2002).

 Adh4

Adh4 is not homologous with other eukaryotic ADH genes (De Smidt et al. 2008, Williamson and Paquin 1987) and is not usually expressed in laboratory strains (Drewke et al. 1990, Drewke and Ciriacy 1988). It has however been shown to rescue adh1Δ strains under antimycin A selection (Dorsey et al. 1992, Walton et al. 1986). Antimycin inhibits respiration, so adh1Δ strains subjected to it usually cannot grow.

 Adh5

The open reading frame (ORF) of Adh5 shares 76% identity with ADH1, but to the best of our knowledge the putative protein has never been isolated or characterized (De Smidt et al. 2008). Its effect on ethanol production could not be detected, except maybe if ADH1 and ADH3 were already knocked out in strains grown aerobically (Smith et al. 2004). Its transcript is however upregulated in strains engineered to use xylose as a carbon source in the absence of oxygen (Sonderegger et al.

2004).

 Adh6

Adh6 has limited identity with Adh1 (26%), and is strictly NADPH dependant. Its function is not known and could range anywhere from NADPH homeostasis to lignin degradation (De Smidt et al.

2008). It is very similar to ADH7, which was not targeted for deletion in the current study.

(17)

9 2.1.1.2. Glycerol Metabolism

The glycerol synthesis pathway stems directly from glycolysis, using glyceraldehyde 3-phosphate (G3P) as a substrate (Figure 1). The triose-phosphate isomerase maintains equilibrium between G3P and dihydroxyacetone phosphate (DHAP). Then, the NAD⁺-dependant glycerol 3-phosphate dehydrogenase reduces DHAP into L-glycerol 3-phosphate, which is the immediate precursor for glycerol and other lipids (Nevoigt and Stahl 1997).

DHAP + NADPH ↔ G3P + NAD⁺ → glycerol + phosphate

There are two G3P dehydrogenases expressed by the genes GPD1 and GPD2, respectively. The enzymes expressed by both genes have essentially the same activity, but each gene is regulated very differently. Glycerol accumulation is indeed either triggered by osmotic stress or by the need for an electron sink. According to Ansell et al. (1997) “Mutants deleted for both GPD1 and GPD2 do not produce detectable glycerol, are highly osmosensitive and fail to grow under anoxic conditions”.

Ansell et al. also showed that both genes could substitute for each other at some extent, as the sum of the glycerol produced by a gpd1Δ strain and a gpd2Δ strain is higher than the amount of glycerol produced by a WT strain.

 Gpd1

Gpd1 expression is regulated by the HOG pathway, which is activated in the case of osmotic stress.

Glycerol accumulates inside the cell to prevent water loss from the cytosol if the environment is rich in osmolites (Ansell et al. 1997, Albertyn et al. 1994).

 Gpd2

Gpd2 is expressed in the case of accumulation of NADH in the cytosol. This typically happens under anoxic conditions, for example. It serves as an electron sink, helping to maintain the redox balance of the cell. Indeed, gpd2Δ mutants grow poorly under anoxic conditions (Ansell et al. 1997, Eriksson et al. 1995).

(18)

10 2.1.2. Multiplex Gene Deletion Constructs

Genetic engineering has for a long time relied on homologous recombination (HR) to knock out genes or insert new ones into the yeast’s genome (Longtine et al. 1998). Recently, genetic engineering techniques involving clustered regularly interspaced short palindromic repeat (CRISPR) and RNA-guided Cas9 nucleases dramatically improved the yield of such genetic modifications (Storici et al. 2003, DiCarlo et al. 2013). CRISPR-Cas9 systems cause a double strand break (DSB) at a locus targeted by customized non-coding guide RNAs (gRNA). This localized DSB actively recruits the HR machinery. If DNA parts homologous to the locus of the DSB are available to the HR machinery, then a scar-free recombination will occur with great yield (Pâques and Haber 1999, Symington et al. 2014). Such homologous DNA parts are named “donor DNA” in the rest of this report. For example, Jakočiūnas et al. (2015b) assembled in vivo 15 DNA parts into 3 different loci in one single transformation step, leaving no scar or marker in the yeast’s genome.

CRISPR-Cas9 systems have such good DSB yields that integrating selection markers into the host’s genome is indeed not required for genetic engineering anymore. This in turn enables the modification of strains that were very difficult to engineer with traditional methods. Furthermore, once the system is built, the effort of engineering one or several strains at once is roughly equivalent.

Finally, the system is built in such a way that old genetic parts can be reassembled into new constructs. In conclusion, the CRISPR-Cas9 system can be used to quickly and easily evaluate different combinations of genetic engineering in different strains.

We report the use of a CRISPR-Cas9 construct where the guides of 2 to 6 different DNA targets were embedded into a host vector expressing a Cas9 nuclease (Figure 3). The Cas9 nuclease gene and the gRNAs have previously been modified to integrate respectively a nuclear localization signal (Ryan et al. 2014) and RNA structural elements allowing the guides to bind to Cas9 (Jinek et al.

2012). The construct was built in a modular fashion, in three consecutive steps of Golden Gate assembly (Engler 2008) using standardized parts. It was then transformed along with customized donor DNA into S. cerevisiae.

The method used to assemble the multiplex gene deletion has been designed by Lee et al. (2015). It consists of four sets of genetic parts: protospacers, host vectors, connectors, and finally spacers.

Additionally, a donor DNA was designed for each targeted gene to replace it during homologous recombination. Each set of parts is detailed separately in the following sections. Most of the parts feature some restriction sites (BbsI, BsaI and/or BsmbI) recognized by type II restriction enzymes.

Type II restriction enzymes produce a DSB outside of their recognition site, which has been exploited in synthetic biology to design modular genetic parts (Lee et al. 2015). Each restriction enzyme used in this project produces a 4bp sticky end whose sequence depends on the particular locus of the restriction site.

Figure 3: Final multiplex gene deletion construct. Each of the 6 gRNA cassettes is composed of its own promoter and terminator, so is the Cas9 cassette. The plasmid bears two origins of replications (ColE1 & Cen6/Ars4) and two resistance markers (KanR & KanMX). One of each is used for selection in E. coli or in S.

cerevisiae respectively.

Final multiplex gene deletion vector

(19)

11

Material and Methods

2.2.1. Strains and Culture Condition 2.2.1.1. S. cerevisiae

All the S. cerevisiae strains were grown at 30°C in yeast extract peptone dextrose (YPD) medium (either liquid or on Petri dishes with 2% agar) (see section 7.3.1.2.). When grown on liquid medium, the flask was agitated at 250 rpm except otherwise stated. When required, the following antibiotics were used: geneticin (200mg/L), nourseothricin (100 mg/L), and hygromycin B (200 mg/L).

One of the S. cerevisiae strains used for deletion is the wildtype CEN.PK2 (genotype MATa/α ura3- 52/ura3-52 trp1-289/trp1-289 leu2-3_112/leu2-3_112 his3 Δ1/his3 Δ1 MAL2-8C/MAL2-8C SUC2/SUC2). It is referred to as yVS10 in the rest of this report.

However, except otherwise stated the assays and the deletions have been carried out on a CEN.PK2 background that had been previously engineered to overexpress mevalonate (Figure 1). This strain is named yVV373 and is additionally leu2Δ.

 10 mL-scale Fermentation

Cultures were started at an OD600≃0.2 and an initial volume of 10 mL YPD without any antibiotic.

They were agitated at 60 rpm on a 15 cm vertical wheel. 1mL of culture was extracted on each day of observation (up to 4 days) to collect OD600 and High Pressure Liquid Chromatography (HPLC) data – including on Day 0, i.e. at the very start of the incubation. OD600 was measured using an adequate dilution factor (2x to 40x, so as not to exceed OD600≃ 0.4 in raw measurements), and the undiluted sample was then centrifuged at max speed for 5 min. 800 µL of the supernatant was collected and stored at 4°C for subsequent HPLC measurements. For important results, the strains were studied in biological triplicates.

 1L scale Fermentation

For the final characterization of strains of interest, the cultures were up scaled to a 1L Dasgip^TM fermentor. For each strain, an inoculum of ~30mL was grown overnight in YPD at 30°C and it was used to seed the bioreactor with a starting OD600~0.1. The fermentation was run at 30°C in 500 mL of Westfall medium complemented with 10 g/L yeast extract for faster initial growth and 0.3 g/L leucine as the yVS373 background is auxotrophic for leucine (see section 7.3.1.4.).

The reactors are run in Fed batch to maintain a Dissolved Oxygen level (DO) of 30% and a pH of 5.

When DO reaches the threshold of 70%, glucose is fed to the reactor so as to reach a concentration of 5 g/L. These parameters were respectively adjusted to 40% and 3 g/L later in the run in order to reduce DO fluctuations. pH is monitored in real time and adjusted with an 8M KOH solution by the Dasgip software. Glucose levels are not measured in real time but calculated based on the volume of the reactor and the concentration of the feed (600 g/L). DO is also regulated by a stirrer agitating between 400 rpm and 1200 rpm.

OD800 is used to monitor the concentration of cells in real time. However, the calibration of the individual probes does not allow comparing the OD800 of the reactors with each other. Some manual samples of OD600 have been taken to this effect, most notably to calculate mevalonate productivity per OD600 as normalization for cell mass.

(20)

12 1mL of cell culture is automatically extracted every 5h for 135h. A manual sample was taken at 139h when the fermentation was terminated. Cells were removed from the sample by filtration and ran on HPLC.

2.2.1.2. E. coli

The constructs are all cloned into Escherichia coli DH5, grown at 37°C overnight in LB medium (either liquid or on petri dish with 2% agar) (see section 7.3.1.1.). Depending on the antibiotic resistance carried by the vector being cloned, the LB medium is either inoculated with chloramphenicol (25 mg/L), carbenicillin (100 mg/L), or kanamycin (50 mg/L). Plates are kept at 4°C when not in use.

2.2.2. Cloning in E. coli

After each Golden Gate assembly step, half of the reaction mix (5 µL) is transformed into E. coli for cloning (see sections 7.3.9. and 7.3.10.).

The colonies are then plated on the appropriate antibiotic to select only the cells that took up a plasmid, and only the non-fluorescent cells are picked for colony PCR (cPCR) (see section 7.3.12.2.). The primers used for the cPCR anneal with the vector, so that a signal is still delivered even if the integration of the insert failed. The primers are designed to be 20 bp, ~50% GC content, and Tm ≃ 55°C. The size of each amplicon is compared to the WT and to the size expected from the theoretical construct.

If the size of the amplicon is correct, the colony from which it came is grown overnight and the plasmid is isolated by miniprep with a Qiagen kit (see section 7.3.8.1.) to be sent for sequencing. The sequencing primers are either the cPCR primers, or sometimes primers that anneal to the insert. The sequencing is delegated to Quintara Biosciences. Sequencing results are read and aligned using the ApE software (Davis 2013).

If the sequencing result matches the theoretical construct, the colony is stocked in 15% glycerol for permanent archive at -80°C and the cloning, sequencing details, plasmid sequence, and plasmid map are entered into a FileMaker database for permanent record keeping.

2.2.3. S. cerevisiae Transformation

The gene deletion constructs were transformed into S. cerevisiae using fresh competent cells (see section 7.3.11.). After transformations, cells were plated on YPD petri dishes complemented with the adequate antibiotic for selection of the transformants after two to four days of growth.

For every antibiotic used in a transformation, a negative control is made to test for contaminations and proper quality of the antibiotic plates. This negative control is a batch of competent cells not transformed with any vector or donor DNA, and thus remains WT. It is expected to yield no surviving colony on the antibiotic plates.

When two vectors are transformed in the same strain, three negative controls are made: one batch of competent cells is transformed with no vector, one other batch is transformed with one vector and the last batch transformed with the remaining vector.

(21)

13 2.2.4. Genetic Parts Description

2.2.4.1. Protospacers

The term “protospacer” is borrowed from the nomenclature of the CRISPR/Cas-based immune system discovered in many prokaryotes (Barrangou and Marraffani 2014). In the context of this adaptive immune system, a protospacer is a DNA or RNA sequence from a virus that is located immediately upstream of a 3 bp sequence called protospacer adjacent motif, or PAM (Sternberg et al. 2014). The microorganism integrates the sequence of the protospacer into its own genome (into the CRISPR locus), but they do not integrate the PAM. Once inserted into the genome, the protospacer sequence is called a spacer. Later, Cas proteins will use RNA transcribed from their spacers to cut into the protospacer upstream of the PAM in the invading virus’s genome.

What we call protospacer in this report is a 20 bp sequence directly upstream of a PAM located inside the yeast genome. Each of the chosen protospacers targets one gene of the yeast genome for deletion (ADH1-6, GPD1-2 or the selection markers cassette). Once assembled with a structural RNA sequence to enable interaction with the Cas9 nuclease, these protospacers will be used by the Cas9 nuclease to cut into their respective locus.

A first attempt of protospacer design has been made using the online resource of DNA2.0 (www.dna20.com, 20160125) with limited success. The parameters used were:

Another attempt has been made using the CRISPR guide design tool of Benchling (benchling.com, 20160404), also with limited success. The parameters used were:

Independently of the method to identify adequate protospacers in the targeted genes, each protospacer was flanked with a 5’GATC and a GGGC 3’ using customized primers. These 5’ and 3’

tails are necessary for the subsequent assembly steps, where they will be used as ligation sites to the first host vector (see section 2.2.4.2.). As much as it was possible, a protospacer located in the middle of its gene was chosen to maximize the chances of homologous recombination at both ends of the gene.

All the protospacers were designed based on the DNA sequences found on the Yeast Genome Database (yeastgenome.org, 20160125)

In the supplementary materials, Table 4 displays all the information relative to the final protospacer designs for most experiments. Table 5 discloses the associated primers.

Species S. cerevisiae S288C (R64-1-1)

PAM NGG

I want WT Cas9

Design Type Single guides

Guide Length 20 bp

Genome ASM76626V2 (Saccharomyces cerevisiae SK1)

PAM NGG

Max A/T/C/G % 80%

(22)

14 2.2.4.2. Host Vectors

Three host vectors are used consecutively in three consecutive Golden Gate assemblies. They all contain a sf GFP expression cassette used to spot false positives after transformation.

 GG1 Vector – pVS237

The first vector (a.k.a GG1 vector) accepts a protospacer in place of sfGFP to create a whole guide RNA (gRNA) expression cassette. By doing so it also ligates the sequence of structural RNA elements to the 5’end of the protospacer. Those structural elements include an RNA sequence which will be bound and recognized by the Cas9 nuclease. This vector is referred to as pVS237 and is marked by chloramphenicol (Figure 4). It is used during the first Golden Gate assembly (GG1), where a BbsI digestion exposes two sticky ends, CATG5’ and 5’GTTT.

 GG2 Vector – pVS127

The second vector (a.k.a GG2 vector) is meant to assemble a gRNA expression cassette with modular connectors (see section 2.2.3.4.). Those connectors will later enable several gRNA expression cassettes to be ligated next to each other in a predetermined order. This 2^nd vector is referred to as pVS127 and is selected by carbenicillin (Figure 5). It is used during the 2^nd Golden Gate assembly where a BsaI digestion exposes a GGGA 5’ and a 5’TACA sticky end.

Figure 4: Configuration of the first host vector (a.k.a GG1 vector), pVS237.

The replication origin ColE1 is followed by a yeast snoRNA promoter pSNR52. This promoter is followed by a super- folder GFP (sfGFP) transcription cassette composed of an E. coli promoter, the sfGFP ORF and a T7 phage terminator.

The sfGFP transcription cassette is delimited by two BbsI restriction sites. Following this cassette are found the RNA structural elements (sgRNA) and a tSUP4 terminator. A last lies a chloramphenicol resistance cassette (terminator, ORF,

promoter).

pVS237 – GG1 host vector

Figure 5: Configuration of the second host vector (a.k.a GG2 vector), pVS127.

The replication origin ColE1 is followed by a super-folder GFP (sfGFP) transcription cassette composed of an E. coli promoter, the sfGFP ORF and a T7 phage terminator. The sfGFP transcription cassette is delimited by two BsaI

restriction sites. At last lies a ampicilin/carbenicillin resistance cassette (terminator, ORF, promoter).

(23)

15

 GG3 Vector – pVS98, pVS166, pVS167, pVS336, pVS354

The third and final vector hosts the final construct (a.k.a GG3 vector). It is this vector that is transformed into S. cerevisiae after assembly. As such, it is the only vector of the set that possesses both a bacterial and a yeast origin of replication (a.k.a. a shuttle vector). This vector will be assembled with a minimum of two parts during the third and last assembly step (GG3): a Cas9 expression cassette and one gRNA expression cassette. We use Cas9 from Streptococcus pyogenes, that produces a blunt end DSB between the 4^th and 3^rd base upstream from the PAM sequence 5’NGG3’ (Ran et al. 2013).

There is no theoretical limitation to the number of gRNA expression cassettes that can be added to this design. This is part of the appeal of the method, as it allows deleting many genes in one single step, once the final construct has been built. In practice, assembling more than 6 guides in this last vector becomes challenging. Depending on which connectors are chosen during the previous step, spacers might also be necessary to bridge two sets of connectors that happen to be incompatible (see section 2.2.4.4.). Several GG3 vectors have been used in this project in order to accommodate different sets of selection markers in the yeast strains. One of them is detailed in Figure 6. Their selection marker is assigned as follows:

pVS98 geneticin

pVS166 nourseothricin

pVS167 hygromycin

Figure 6: Configuration of one of the third (and last) host vectors (a.k.a. GG3 vectors), pVS98.

The replication origin ColE1 is followed by a super-folder GFP (sfGFP) transcription cassette composed of an E. coli promoter, the sfGFP ORF and a T7 phage terminator. The sfGFP transcription cassette is delimited by two BsmbI restriction sites and two connectors (ConLE and ConRE). It is followed by a yeast selection marker cassette (in pVS98, KanMX selecting against geneticin) composed of an Ashbya gossypii promoter and terminator (AgTEF2_p/t) framing the KanMX ORF. Next is found replication origin

named Ars4 and a yeast centromere (Cen6). At last lies a kanamycin bacterial resistance cassette (terminator, ORF, promoter).

(24)

16 2.2.4.3. Connectors

During the second assembly step (GG2), each whole guide RNA coding cassette is linked to two connectors of around 200 bp. Different gRNA coding cassettes are assigned to different sets of complementary connectors so that at the next assembly step, the connectors-gRNA cassettes will be able to find each other and daisy-chain into a single construct composed of several gRNA cassettes and one Cas9 cassette (see section 2.2.5.2.).

One can distinguish two classes of connectors as a function of the end of the gRNA cassette to which they fit: the Left-connectors (ConL) fit the 5’-end (after BsaI digestion, their sticky ends are 5’CCCT and TTGC5’) and the Right-connectors (ConR) fit the 3’ end (after BsaI digestion, their sticky ends are 5’GCTG and ATGT5’). Before digestion by BsaI, each connector is incorporated into a plasmid with chloramphenicol resistance and a bacterial origin of replication.

When the GG1 vector (pVS237) is cut with BsaI during GG2, it releases an insert with 5’AACG and CGAC5’ for sticky ends. As a consequence, those two sticky ends are respectively complementary with every ConL connectors and every ConR connectors digested with the same restriction enzyme.

Every connector also incorporates a BsmbI restriction site. The sticky end produced by a BsmbI digestion during GG3 is different for each connector. However, the BsmbI-sticky end of each ConL is complementary to the BsmbI-sticky end of one single cognate ConR. This feature allows the simultaneous restriction and ligation of many different parts in an organized manner. For example, ConR2 is complementary to ConL2; ConR3 is complementary to ConL3, and so on. Every gRNA cassette is thus assembled with conLs of the same order but conRs of the next order, as generalized in the following expression.

ConL_𝑛 − gRNA_𝑛− ConR_𝑛+1 where 𝑛 𝜖 ℕ₀ In the supplementary materials, Table 3 details the features of all the connectors used in this project.

2.2.4.4. Spacers

During the last assembly step, when all the connectors-gRNA cassettes are brought together into one single construct, the last cassette must bridge the GG3 vector. Since the number of cassettes can vary from construct to construct, the last ConR of the string of connector-gRNA cassettes also changes.

As a consequence, several parts have been designed solely to connect any ConR to the 5’ end of the GG3 vector. All spacers begin with a different ConL, but all end with the same ConR, labelled ConRE. The sequence between the two connectors (a.k.a. the spacer) codes for nothing at all.

To connect the first connectors-gRNA cassette to the 3’end of the host vector, the Cas9 cassette is used. Indeed, it is already assembled with its own set of connectors (ConLS and ConR1) to make this link. It can also happen that the construct design misses a set of connectors. In this case, the same spacer is flanked with the missing connectors and is used to fill in the gap in the sequence of connectors. For example, to assemble ConL1-gRNA1-ConR2 with ConL3-gRNA3-ConR4, the spacer ConL2-spacer-ConR3 will be required in the assembly mix.

In the supplementary materials, Table 2 summarizes all the spacers used during this project. All the spacers share the 37 bp sequence ACTAGAGATCTATGTGAGGATCCTAACTCGAGATCGA and are around 75 bp long.

(25)

17 2.2.4.5. Donor DNA

Once the homologous recombination machinery has been recruited by the DSB caused by the Cas9- guides construct, they must be repaired by a knock-out sequence in order to effectively remove the target genes. This knock-out sequence is called donor DNA and is simply designed by putting together the outer boundaries of a targeted gene (Figure 7). Between the sequences of the two outer boundaries of a given gene, a random DNA sequence of 20bp (called the barcode) is inserted to enable confirmation of the genomic integration, if necessary. The hypothesis behind this course of action is that each outer boundary will recombine with its homologous sequence on the donor DNA, and the target gene will thus be deleted as a consequence of its replacement by the DNA barcode.

AT-or GC-rich regions have been avoided as much as possible. As a consequence, most of the donor DNAs are homologous to a 42 bp region ~200bp upstream of their target genes and another 42 bp region ~200bp downstream.

The barcode is generated with the Random DNA Sequence Generator of UC Riverside (www.faculty.ucr.edu, 2016) for a GC content equal to 50%. The barcodes are then BLASTed against the S. cerevisiae genome to avoid random homology in the barcode (Altschul et al. 1990).

All the primers are synthesised by Integrated DNA technologies Inc (www.idtdna.com).

In one instance, a traditional polymerase chain reaction (PCR) mediated yeast gene deletion method was used (Longtine et al. 1998). This particular procedure required designing a 1.4 kb donor DNA consisting of a geneticin resistance gene flanked by two 42bp sequences homologous to the 5’ and 3’

outer boundaries of ADH1. A geneticin resistant plasmid (pVS81) was used as a template for amplification.

In the supplementary materials, the primers used for amplifying the donor DNAs are presented in Tables 6 & 12, and the ones used for amplifying the geneticin resistance gene are displayed in Table 7.

Figure 7: Schematic of the gene deletion process hijacking the homologous recombination machinery. The donor DNA is designed with sequences homologous to the 5’ and 3’ outer boundaries of the gene to be deleted (target gene). The barcode of the donor DNA has no other purpose than confirming the gene deletion or detecting off-target insertions, e.g. by using it as a primer annealing site for sequencing. To recruit the HR machinery, Cas9 produces a DSB inside the target gene. In traditional gene deletion the barcode is replaced by a resistance marker such as geneticin. This last technique exploits randomly occurring DSBs within the target gene and uses the resistance marker to select the few knock-out mutants.

(26)

18 2.2.5. Golden Gate assemblies

All the parts described in the section “Genetic parts description” are connected to each other by Golden Gate assembly (Engler et al. 2008). This process is based on type II restriction enzymes and the use of a thermocycler. Indeed, the recognition site of type II restriction enzymes is not present on the ligated fragments anymore. As a consequence, if the reaction mix goes through serial restriction and ligation steps, none of the targeted ligations will be subject to subsequent restrictions. However, the fragments that re-ligate to their original source will restore the recognition site, and will thus be cut again at the next digestion round. Eventually, after enough restriction and ligation rounds, most of the strands are correctly ligated.

The Golden Gate assembly allows for scar-free assembly of several parts at the same time (see section 7.3.5.).

2.2.5.1. 1^st Golden Gate assembly (GG1)

The first Golden Gate assembly (GG1) is illustrated in Figure 8. It aims at connecting the protospacer to the gRNA expression cassette. This cassette comprises a snoRNA promoter and terminator, as well as gRNA structural elements that will help the Cas9 to use the gRNA to cut the target gene (Ryan et al. 2014).

Prior to the Golden Gate assembly, the FD and RV primers designed to form each protospacer are annealed together in a thermocycler (see section 7.3.3.). This procedure gives a double-stranded protospacer with two sticky ends, 5’ATCG and CAAA5’.

The GG1 host vector (pVS237) releases its sfGFP expression cassette upon digestion by BbsI, and exposes CATG5’ and 5’GTTT sticky ends. Since the sticky ends of the protospacer and of the GG1 vector are complementary, they can anneal together and form a single circular DNA molecule upon ligation by T7 DNA ligase. The resulting vector bears no sfGFP expression cassette, which allows for rapid screening of the GG1 vectors that have not been correctly ligated. Indeed, after GG1 the reaction mix is directly transformed into E. coli for cloning and plated on chloramphenicol plates.

All the gRNA cassettes that have been assembled during this project are listed in the supplementary materials in Tables 9, 15 and 22. They are divided in three groups depending on the experiment in which they are involved.

pVS237 – GG1 host vector gRNA cassette vector

Figure 8: Schematic of the first Golden Gate assembly (GG1).

Using BbsI type II restriction enzyme, a protospacer is integrated into the GG1 host vector to form a gRNA expression cassette.

(27)

19 2.2.5.2. 2^nd Golden Gate assembly (GG2)

The second Golden Gate assembly is illustrated in Figure 9. It aims at ligating the gRNA expression cassettes to a set of two connectors. Those connectors will enable several gRNA cassettes to be assembled together at the next assembly step.

Each connector is flanked by two BsaI restriction sites inside a chloramphenicol resistant plasmid.

After BsaI digestion, ConLs expose 5’CCCT and TTGC5’ sticky end; similarly, every ConR exposes 5’GCTG and ATGT5’ sticky ends after BsaI digestion.

The vector resulting from GG1 also bears BsaI restriction sites at each end of the gRNA expression cassette. After BsaI digestion, the gRNA expression cassette has a 5’AACG and a CGAC5’ sticky ends. Each is thus complementary to one of the sticky ends of ConL and ConR, respectively.

At last, the GG2 host vector exposes a GGGA5’ and a 5’TACA sticky end after BsaI digestion. Each sticky end is thus complementary to the remaining sticky end of ConL and ConR, respectively.

After GG2, these three sets of parts (gRNA cassettes, connectors and GG2 host vector) will thus be ligated into one single molecule. Again, these are then transformed into E. coli and plated on carbenicillin plates. Colonies with green fluorescence are discarded as containing plasmid that did not correctly complete the assembly.

All the connectors-gRNA cassettes that have been assembled during this project are listed in the supplementary materials in Tables 10, 16, and 23.

gRNA cassette vector pVS127 - GG2 host vector

ConR vector ConL vector

gRNA-connectors vector n

Figure 9: Schematic of the 2^nd Golden Gate assembly.

Using BsaI type II restriction enzyme, a gRNA cassette, a ConL and a ConR are brought together inside the GG2 host vector to form a gRNA-connectors cassette.

(28)

20 2.2.5.3. 3^rd Golden Gate assembly (GG3)

The third and last assembly step is illustrated in Figure 10. It aims at connecting several gRNA- connector cassettes together with a Cas9 expression cassette into one single shuttle vector. After cloning, this same vector will be transformed into S. cerevisiae.

Each gRNA-connector cassette has two BsmbI restrictions sites. Each connector of the reaction mix has one single complementary connector after BsmbI digestion. The sticky ends resulting from a BsmbI digestion for each particular connector are all summarized in Table 3. At least one spacer is integrated to the reaction mix to enable ligation of the last ConR of the connectors-gRNA cassettes to the GG3 host vector. The Cas9 expression cassette is already connected to its own set of connectors (ConLS and ConR1) that bridge the other end of the host vector to the first gRNA- connector cassette. At the end of the cycles of digestions and ligations, most connectors have found their complement and an array of connectors-gRNA cassette is formed behind Cas9 in the GG3 host vector.

All the final multiplex gene deletion vectors that have been assembled during this project are presented in the supplementary materials in Tables 11, 17 and 24.

Figure 10: Schematic of the last Golden Gate assembly (GG3).

Using BsmbI type II restriction enzyme, a gRNA-connectors cassette, a Cas9 cassette and a spacer are integrated into the GG3 host vector to form the final multiplex gene deletion vector. Using the exact same procedure and using the adequate connectors, several gRNA cassettes can be integrated at once

in the same construct.

gRNA-connectors vector n pVS98 – GG3 host vector

Cas9 cassette vector Spacer vector n+1

Multiplex gene deletion vector

(29)

21 2.2.5.4. One-step Protospacer Integration

A rapid single gene deletion is often necessary to quickly assess the effect of the mutation. For this reason, and because the assembly method going through all three Golden Gate assemblies can be cumbersome, an alternative procedure has been designed to produce single deletion plasmids in one step. It is illustrated in Figure 11. The principle is the same, except that the Cas9 cassette and one single gRNA cassette have already been inserted into a GG3 plasmid. However, the BbsI restriction sites of the gRNA cassette have been kept intact so that the plasmid still expresses sfGFP – and still lacks a protospacer sequence. The protospacer can then be integrated into the final construct in one single Golden Gate step using a BbsI restriction enzyme. It is then directly ready to be transformed into S. cerevisiae.

This procedure has been used in several instances to test individual guides. However, it was only available to us late in the project, so it has not been used systematically on each guide. All single- guide gene deletion vectors that have been built over the course of this project are detailed in Tables 8, 9 and 21 in the supplementary materials.

Single-guide gene deletion vector

Figure 11: Schematic of the One-step protospacer integration into the final construct.

The protospacer is ligated into the final deletion construct after a BbsI digestion. Only one single gRNA can be integrated per construct using this procedure.

pVS370 – One-step integration host vector

(30)

22 2.2.6. Gene Deletion in S. cerevisiae

To proceed to a gene deletion (multiplex or not) using the Cas9-mediated system, 300 ng of the final gene deletion vector is transformed into the target S. cerevisiae strain. At the same time, 1 µg of donor DNA is also transformed in the same reaction mix.

A control is made with a reaction mix containing the same number of competent cells. In this control reaction mix, no donor DNA is added. As a consequence, the survival rate is expected to be much lower than with the donor DNA. Indeed, if the construct works the target genome is cut but not repaired as efficiently as when the donor DNA is available for repair (Lee 2015).

In one instance, a traditional yeast gene deletion was performed using a donor DNA that replaced ADH1 with a geneticin selection marker. Over 100 µg of donor DNA were transformed into 100 µL of competent yVS373 gpd1Δ gpd2Δ strain. In this case, recombination (and the gene deletion) is thought to occur through random DSB in the target gene. The recombined colonies are then selected on geneticin YPD plates.

2.2.7. S. cerevisiae Colony PCR

The colonies growing on the transformation plates are genotyped by PCR (see section 7.3.12.4.).

This genotype is then compared to the genotype of the WT to identify mutants. The primers are designed to be 20 bp long, ~50% GC content, and Tm ≃ 55°C.

2.2.8. S. cerevisiae Glycerol Stocks

If the size of the amplicon matches the size of a deleted gene, the colony from which it was amplified is re-streaked on YPD plates with no antibiotic so as to isolate singles colonies. After one day of incubation at 30°C the plate is replicated with a velvet stamp on an antibiotic plate. After one additional day of incubation both plates are compared to select a single colony which has lost the resistance to the antibiotic – and thus has lost the multiplex gene deletion vector. This colony is re- patched for a second cPCR, and if the second cPCR confirms that the colony is deleted for the gene of interest, the single colony is grown overnight in liquid YPD medium for storage at -80°C in a 15%

glycerol solution. The cloning details are then entered into a FileMaker database for permanent record keeping.

2.2.9. High Pressure Liquid Chromatography

Glucose, glycerol, acetate and mevalonate titres of liquid cultures were measured using a Shimadzu High Pressure Liquid Chromatography (HPLC) and its associated software. After loading of a sample in the column, an eluent of 5mM H2SO4 was run at 0.6 mL/min for 25 min to 28 min. Peaks were then automatically labelled and quantified by the software on the basis of calibration curves calculated with three standards of known concentrations and composition.

(31)

23 2.2.10. Calculations for 10 mL-scale and 1L scale Fermentations

Based on the HPLC and OD600 or OD800 data collected during fermentation, several characteristics of each strain were calculated using the following formula.

The following notations are used:

𝑡(𝑡_𝑛) [ℎ] = 𝑡_𝑛− 𝑡_𝑛−1 [ℎ]

𝑚_Mvl( 𝑡_𝑛) [𝑔] = 𝑉_𝐵( 𝑡_𝑛) [𝑚𝐿] ∗𝐶_Mvl( 𝑡_𝑛)[𝑔/𝐿]

1000 [𝑚𝐿/𝐿]

𝑚_Glc( 𝑡_𝑛)[𝑔] =𝑉_𝐹(𝑡_𝑛)[𝑚𝐿] ∗ 𝐶_𝐹 [𝑔/𝐿] + 𝑉_𝐵(𝑡₀)[𝑚𝐿] ∗ 𝐶_Glc(𝑡₀)[𝑔/𝐿]

1000 [𝑚𝐿/𝐿]

𝑀𝑣𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑜𝑛 𝑝𝑒𝑟 𝑂𝐷600 10 𝑚𝐿 𝑠𝑐𝑎𝑙𝑒 ( 𝑡_𝑛) [ 𝑔

𝑎𝑏𝑠] =𝐶_Mvl( 𝑡_𝑛)[𝑔/𝐿]

𝑂𝐷600 ( 𝑡_𝑛)

𝑀𝑣𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑝𝑒𝑟 𝑂𝐷600_{1 𝐿 𝑠𝑐𝑎𝑙𝑒} ( 𝑡_𝑛) [𝑔/𝐿

𝑎𝑏𝑠] =𝑚_Mvl ( 𝑡_𝑛) − 𝑚_Mvl( 𝑡₀) [𝑔]

𝑂𝐷600 ( 𝑡_𝑛) 𝑀𝑣𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 ( 𝑡_𝑛) [𝑔/ℎ] =𝑚_Mvl( 𝑡_𝑛) − 𝑚_Mvl( 𝑡_𝑛−1) [𝑔]

𝑡(𝑡_𝑛) [ℎ]

𝑀𝑣𝑙 𝑚𝑎𝑠𝑠 𝑦𝑖𝑒𝑙𝑑 ( 𝑡_𝑛) [%] =𝑚_Mvl ( 𝑡_𝑛) − 𝑚_Mvl( 𝑡₀) [𝑔]

𝑚_𝐺𝑙𝑐 ( 𝑡_𝑛) [𝑔] ∗ 100 [%]

𝑀𝑣𝑙 𝑚𝑜𝑙𝑎𝑟 𝑦𝑖𝑒𝑙𝑑 ( 𝑡_𝑛) [%] =𝑚_Mvl( 𝑡_𝑛) − 𝑚_Mvl( 𝑡₀)[𝑔]

𝑚_Glc ( 𝑡_𝑛) [𝑔] ∗ 𝑀𝑀_Glc [𝑔/𝑚𝑜𝑙]

𝑀𝑀_Mvl[𝑔/𝑚𝑜𝑙]∗ 100 [%]

The mevalonate production per OD600 at the 10 mL-scale has been calculated on the basis of mevalonate concentration (and not mass of mevalonate) to avoid the error on the volume of the bioreactor. Indeed, at this scale the volume of each sample extracted from the reactor becomes significant. Furthermore, the setup at this scale potentially allows important water evaporation.

Mevalonate productivity per OD600 [gmevalonates^-1OD600-1] was not calculated because of the length and variability of the time intervals between OD600 measurements (9 samples separated by 3h to 44h intervals). Indeed, most of the exponential phases of biomass growth and mevalonate production could not be witnessed by the OD600 measurements, leading to inconsistent productivity calculations.

Instead, mevalonate productivity [gmevalonate/s] and mevalonate production per OD600

[gmevalonate/OD600] were plotted individually.

t(tn)= duration of the time interval n [h] tn= time interval n [h] 𝑛 ⊂ [0; 28]𝜖ℕ VX (tn)= total volume of X at tn [mL] MMX= molar mass of X [g/mol] B = Bioreactor mX(tn)= mass of X produced at tn [g] MMGlc= 180.16 g/mol Glc = Glucose OD600(tn)= optical density at 600 nm at tn [abs] MMMvl= 148.16 g/mol Mvl=Mevalonate CX(tn)=concentration of X in the bioreactor at tn

[g/L]

CF=concentration of F= 600 g/L F= Glucose feed