Results

Cloning Strategy

Given our metabolic engineering approach in yeast, a well-designed cloning strategy was indispensable. Following the official iGEM rules, general cloning was performed according to the RFC1000 modular cloning system (MoClo) (Fig. G1). For all required gene constructs and plasmids, this was accomplished in Escherichia coli Turbo. Initial genes, as well as plasmids, originated from internal research groups of our institute, from this year’s iGEM teams from Dresden and Bochum or have been provided by Integrated DNA Technologies IDT. In addition, some were obtained from the official iGEM distribution and the Free Genes Project.

The MoClo system relies on IIS-type restriction enzymes to specifically cleave DNA beyond its recognition site, resulting in 5′ or 3′ overhangs. Generation of diverse overhangs allows efficient assembly of multiple DNA fragments in one simple ligation. A suitable overhang design ensures fragments only to be assembled in a predefined order (Engler et al., 2008). This gave us the great advantage of designing standardized overhangs to establish individual building blocks. In this way, we have established a library of genetic parts that can be easily arranged as desired, making them accessible to future iGEM teams.

All genes have been amplified by initial Polymerase Chain Reaction (PCR) using specific primers with the appropriate overhangs attached. For promoter sequences, the forward primer (fw) contained TACAGCTCTTCATCG and fusion site GGAG. The reverse primer (rv) included the 3′ fusion site AATG and the overhang CGAGTGAAGAGCCGAT leading to a final construct without 5′ UTRs. All coding sequences were equipped with the 5′ overhang GACGGCTCTTCATCG and the fusion site AATG, and the 3′ fusion site GCTT and the overhang CGAGTGAAGAGCATAC. Finally, all terminator sequences featured 5′ overhang TGCAGCTCTTCATCG and fusion site GCTT and 3′ fusion site GGCT and the overhang CGAGTGAAGAGCATAA.

Since metabolic engineering of certain metabolic pathways targeted the peroxisome, proper localization within cells had to be considered when cloning the corresponding genes. To achieve this, we added a tag comprising nine specific base pairs to the overhang of the matching rv primer. According to Dusséaux et al., 2020, the C-terminal SKL peroxisomal targeting signal-1 (PTS1) tag consisted of TCCAAGCTG encoding serine, lysine, and leucine, leading to reliable peroxisome localization.

After attaching the specific overhangs to all genes, Golden Gate Assembly into iGEM-derived pSB1C00 backbone resulted in the level 0 parts. These served to design the level 1 constructs where a promoter, coding sequence, and terminator formed single transcription units (TU). We achieved varying levels of gene expression by using different promoters. It was important to choose between pSB1K01-04 as a backbone in view of the later level 2 composition. Up to four level 1 constructs can be assembled into the pSB3C01 backbone as a final level 2 construct. For each position, a spacer was designed in case less than four TUs are to be assembled. All plasmid backbones carried a constitutively monomeric red fluorescent protein (mRFP1) to allow selection of transformed colonies ensuring successful integration of the desired gene construct.

Yeast transformation was to be accomplished either by Clustered Regularly Interspaced Short Palindromic Repeats-CRISPR-assotiated protein 9 (CRISPR-Cas9)-mediated stable genome integration or by using extrachromosomal shuttle vectors. In addition, four different level 2 backbones pSB3C01 were each equipped with His, Leu, Trp, or Ura auxotrophies (pSB3CY-TRP1/LEU2/URA3/HIS3) to allow for subsequent selection. In addition, pSB1KY-TRP1/LEU2/URA3/HIS3 were used as level 1 shuttle vectors. We developed an alternative backbone using cyan fluorescent protein (CFP), pSB3CY-CFP as a second selection marker instead of auxotrophy. For uniform gene expression, all shuttle vectors contained the yeast 2µ ori, which in contrast to the single copy integration, leads to a high-copy number.

---

Cytosolic production of α-pinene in Saccharomyces cerevisiae

Summary

To create a chassis organism that produces the monoterpenoid α-pinene, we engineered the mevalonate pathway of the model organism Saccharomyces cerevisiae. We used the MoClo strategy for efficient engineering. In total we cloned 27 different constructs using Golden Gate Assembly. Thereby, we inserted heterologous genes like PptAPS and upregulated endogenous genes like ScERG13 or SctHMGR in S. cerevisiae with a strong promoter - ScpTEF. Further, we tried to knock out two genes which are the bottleneck of the mevalonate pathway (MVA) using Clustered Regularly Interspaced Short Palindromic Repeats-CRISPR associated protein 9 (CRISPR-Cas9). Overall, we created two yeast strains that produce the monoterpenoid α-pinene in concentrations up to 4.1 mg/l.

Brief project description

The lab group S. cerevisiae cytosol engineered the MVA that takes place in the cytosol of the baker’s yeast. The integration of all genes was plasmid based. For genomic knockouts we used a CRISPR-Cas9 system. Thereby, we engineered a synthetic pathway with a final product: α-pinene. By silencing ScROX1 and upregulating ScERG13 as well as SctHMGR with a strong promoter, we planned to increase the mevalonate concentration and thus the concentration of subsequent intermediate substrates of the pathway. The knockout of ScROX1 was favourable for increasing the mevalonate concentration since it inhibits enzymes in the MVA (Bröker et al., 2018). Further, we introduced the heterologous genes GgmFPS144 from Gallus gallus and AgtGPPS2 from Abies grandis to increase the concentration of geranyl diphosphate (GPP), a precursor molecule of all monoterpenoids. By knocking out ScERG20 and simultaneously inserting GgmFPS144 we tried to inhibit but not completely imped the vital reaction of GPP to farnesyl diphosphate (FPP) (Brown et al., 2015). Lastly, we introduced the heterologous α-pinene synthase (PptAPS) from Pinus ponderosa to facilitate the production of α-pinene within the cytosol of S. cerevisiae (Fig. Scc 1). A schematic representation of the transformed level 2 constructs is displayed in Figure Scc 2.

Results

As a strategy for introducing our genes of interest into S. cerevisiae, we used modular cloning. We successfully cloned all level 0, level 1 and level 2 plasmids that were needed, as well as three level 2 shuttle vectors with Golden Gate cloning. We used a mutated ScpTEF as a strong promoter. Since the promoter naturally contains a BsaI restriction site and we were using BsaI restriction enzymes for restriction of our level 1 constructs, we introduced a mutation. Hereinafter the mutated promoter is called ScpTEF. The correct cloning of level 0 and level 1 constructs was confirmed by transforming them into Escherichia coli using antibiotic selection, colony Polymerase Chain Reaction (cPCR) (Fig. Scc 3) and sequencing. The finished level 1 plasmids were used to clone level 2 constructs. Additionally, successful transformation of our level 2 constructs into E. coli was confirmed by selective media.

The genomic knockout of ScROX1 was carried out using CRISPR-Cas9. We tried to replace ScROX1 with ScERG13 and SctHMGR to create S. cerevisiae_MS_1. Thus, the upregulated transcription of ScERG13 and SctHMGR led to an increase of subsequent metabolites which further results in an increased GPP concentration (Bröker et al., 2018). We planned two different strains to further optimize the metabolic flux towards GPP and finally towards α-pinene .

In the first approach, we used S. cerevisiae_MS_1 and tried to knock out ScERG20 by using CRISPR-Cas9. Simultaneously, we wanted to replace this gene with two heterologous genes: AgtGPPS2 and GgmFPS144. Both catalyze the reaction to GPP. The latter, also catalyzes the reaction of GPP to FPP and therefore ergosterol synthesis would still occur, enabling the yeast to endure. In a second approach, we planned to replace ScpERG20 with a copper dependent promoter (pCTR3) and insert SltNPPS1. This would have resulted in a downregulated FPP synthesis as well as in an upregulated NPP production (Brown et al., 2015; Ignea et al., 2019).

The genomic targets within ScROX1, ScERG20 and ScpERG20 were predicted using online tool CRISPOR. Like Laughery et al., (2015) we used the pML104 plasmid containing Cas9, AmpR and URA3. We restricted and ligated the sgRNA for ScROX1, ScERG20 and ScpERG20 into pML104 and designed homology flanks for these genes to fuse them to the level 2 constructs which held the transcription units we wanted to insert (Fig. Scc 4).

However, we were not able to amplify these flanked constructs. While the reason for this hindrance remains unknown, we decided to introduce a new strategy to stay within our schedule. Instead of inserting our constructs into the genomic DNA, we cloned the constructs into shuttle vectors with different auxotrophies. However, this strategy is not applicable for the genomic integration of the promoter ScpCTR3 upstream of ScERG20. Due to time limitation, we had to relinquish the plan to produce NPP.

Another part of our project that had to be left out is the introduction of the limonene-3-hydroxylase (ApL3H), an enzyme which catalyzes the reaction of α-pinene to verbenone. Again, the fusion of homology flanks to the level 2 construct failed. As we had to change our transformation strategy, we were limited by the number of auxotrophy markers. Since there was no auxotrophy marker left to select for the ApL3H, this idea had to be abandoned.

Nonetheless, we used CRISPR-Cas9 to insert a disruption into ScROX1 and ScERG20. Simultaneously, we transformed shuttle vectors that held transcription units of our genes of interest. The transformation process resulted in three different yeast strains (Tab. Scc 1). However, due to time limitation it was not possible to verify the transformation success with cPCR. Further, verification of a successful frameshift was unachievable. We therefore solely relied on the yeast’s growth on selective SD media as well as on SD medium containing 5-Fluoroorotic acid.

Table Scc 1: Engineered yeast strains.Three different yeast strains were generated. In S. cerevisiae_MS_1 we tried to knockout ScROX1 while ScERG13, SctHMGR and PptAPS were inserted. S. cerevisiae_MS_2 and S. cerevisiae_MS_3 lack ScROX1 and ScERG20 while AgtGPPS2, mFPS144, ScERG13, SctHMGR and PptAPS.

For quantification of α-pinene GC-MS was performed. The engineered strains were inoculated for 48 hours at 30 °C in liquid minimal medium, followed by a subsequent lysis and extraction with ethyl acetate for analysis. A standard curve of α-pinene covering a concentration range of 0.2 mg/l to 200 mg/l containing eugenol and para-xylene (p-xylene) as internal standards was recorded. The chromatogram exhibit elution time of α-pinene at 6.5 minutes, para-xylene at 5.3 minutes, and eugenol at 12.0 minutes.

We expected α-pinene production for S. cerevisiae_MS_1 since we inserted PptAPS. However, GC-MS measurement resulted in a chromatogram, in which α-pinene peak was detectable. The internal standards eugenol and p-xylene eluted at expected times. We assume the extraction of α-pinene did not work properly since we used an extraction protocol that included cell lysis with one bigger metal bead instead of a mix of many small glass beads.

Nonetheless, Figure Scc 5 shows the chromatograms and mass spectra of S. cerevisiae_MS_2 and S. cerevisiae_MS_3. In the chromatograms, the retention time in minutes is plotted against the absolute intensity. The mass spectra show relative intensity against m/z. Both gas chromatograms display a clear peak at a retention time of 6.5 minutes. These were verified by a distinct α-pinene pattern within the mass spectra. Quantification of α-pinene resulted in concentrations of 4.1 mg/l produced from S. cerevisiae_MS_2 and 3.5 mg/l produced from S. cerevisiae_MS_3. Further, S. cerevisiae CEN.PK2-1C was included as a negative control. No peak can be observed in the chromatogram at 6.5 minutes. Therefore, we conclude no α-pinene production in this strain. Based on the GC-MS results, we verified the functionality of the heterologous α-pinene synthase. Furthermore, we verified the functionality of our self-designed and build shuttle vectors.

---

Peroxisomal production of α-pinene in Saccharomyces cerevisiae

Summary

This part of MonChassis comprised metabolic engineering to target the mevalonate metabolism of S. cerevisiae and the monoterpene synthesis to the yeast peroxisomes. Our main objective aimed to microbially produce the monoterpenoid precursor α-pinene in yeast. To achieve this goal, we cloned more than 25 level 0, 1, and 2 constructs and designed multiple yeast expression vectors using Golden Gate Assembly. Yeast expression vectors contained heterologous genes and upregulated endogenous genes equipped with a C-terminal SKL PTS1. Final GC-MS analysis revealed that our generated yeast strain S. cerevisiae_MS_8 produced about 3.6 mg/l α-pinene and S. cerevisiae_MS_9 with additional ApL3H 12.8 mg/l verbenone.

Brief project description

The budding yeast Saccharomyces cerevisiae is undoubtedly one of the most established model organisms in applied biotechnology (Ostergaard et al., 2000). Extensive knowledge of cultivation, genomic modification techniques, and downstream processing enable a convenient experimental design. In addition to the cytosol, the peroxisome became a promising target for metabolic engineering, as its physiological properties, such as substrate availability, were thought to significantly enhance product synthesis (Dusséaux et al., 2020). Thus, metabolic engineering targeting the peroxisome comprised cloning of plasmids to drive heterologous gene expression. We used yeast expression vectors to introduce numerous genes into S. cerevisiae. For targeted peroxisome localization in yeast cells, we equipped all our genes of interest with the C-terminal SKL PTS1 previously established by Dusséaux et al., 2020. For the verification of functional peroxisomal tagging, we additionally fused a green fluorescent protein (GFP) to PTS1 and performed fluorescence microscopy of corresponding yeast cells.

Since mevalonate metabolism does not occur naturally in peroxisomes (Zhang et al., 2020), we incorporated the required PTS1-fused genes ScERG13, SctHMGR, ScERG12, ScERG8, and ScERG19 into each yeast strain under the control of the constitutive pADH promoter for uniform gene expression (Fig. Scp 1). In a first approach, we additionally introduced the synthase SltNPPS1 from tomato which resulted in the production of the monoterpenoid precursor neryl diphosphate (NPP). In a second approach, we further inserted the heterologous genes GgmFPS144 and AgtGPPS2 from Gallus gallus and Abies grandis, respectively, which led to the generation of the NPP structural isomer geranyl diphosphate (GPP). For both approaches, we finally introduced PptAPS which catalyzes the reaction of NPP or GPP to α-pinene. Two specific genes are involved in the bottleneck phase of the mevalonate pathway, ScERG13, and SctHMGR, that are considered rate-determining enzymes in synthesis. Therefore, we upregulated gene expression using ScpTEF instead of ScpADH, which was expected to result in increased product yields due to stronger promoter activity of ScpTEF. The same promoter sequences were used in the other yeast part of the project. ScpTEF promoter has also been used to control expression of PptAPS, SltNPPS1, GgmFPS144, and AgtGPPS2. For the comparison of the cell-free production of verbenone in our in vitro approach, we also generated a yeast strain containing cytochrome P450 monooxygenase (CYP) limonene-3-hydroxylase (L3H) in addition to GPP and NPP production. Thereby, we were able to obtain microbially-derived verbenone in addition to the electrically-driven CYP product. The heterologously produced monoterpenoids were quantified by gas chromatography-mass spectrometry (GC-MS).

Within the partnership of MonChassis and the iGEM Team TU Dresden, we attempted to produce therapeutical thymol for the treatment of chronic wounds (Costa et al., 2018). This project required the insertion of additional genes including TvtGTPS1, TvCYP71D179, and TvSDR1 to synthesize thymol. This monoterpenoid is also obtained by three-step catalysis of GPP (Krause et al., 2021).

Results

For targeted metabolic engineering of yeast peroxisomes, the corresponding gene constructs were initially cloned up to level 2. Each construct was transformed into E. coli and confirmed by antibiotic selection, cPCR, and all parts of levels 0 and 1 were additionally sequenced. Final level 2 shuttle vectors were used for transformation to achieve a high plasmid copy number in the resulting yeast. The arrangement was designed to functionally combine multiple genes into a single module (Fig. Scp 2). First, ScERG10, ScERG13, and SctHMGR composed the mevalonate module. Second, ScERG8, ScERG12, ScERG19, and ScIDI1 formed the downstream module, and finally, PptAPS and SItNPPS1 built the NPP module (Tab. Scp1). In planning our experiments, we considered using the genome editing system CRISPR-Cas9 to integrate the required genes into the S. cerevisiae genome. Compared to plasmid-based systems, genomic integration is usually more reliable in terms of gene copy number and stability (Reider Apel et al., 2017). Due to time limitations, the stable integration was not implemented and is pending for future projects.

Table Scp 1: Successfully generated yeast strains of S. cerevisiae with metabolically engineered peroxisomes. Listed yeast strains were modified by gene insertion via shuttle vectors to produce NPP and verbenone. Abbreviations: afraGFP = Agaricia fragilis-derived green fluorescent protein; NPP = neryl diphosphate; PTS1 = peroxisomal targeting signal-1; ApL3H = limonene-3-hydroxylase.

Some of the yeast strains that we planned to deliver in MonChassis could not be implemented due to difficulties encountered during the cloning process and limited time. Although the establishment of the yeast strain designed to produce GPP-derived α-pinene is still pending, we succeeded in cloning the level 0 construct AgtGPPS2 including PTS1. For the thymol-producing yeast strain a correct level 2 shuttle construct was obtained (Fig. Scp 3).

In all five samples, we succeeded in performing a correct level 0 assembly of AgtGPPS2 into pSB1C00 including the C-terminal PTS1 fusion, which consists of approximately 900 bp. Additional fragments of varying sizes that were apparent in the Midori Green-stained agarose gels were probably synthesized due to nonspecific primer binding or represent the template fragments. cPCR of the level 2 thymol module targeted 6 kb which was partially confirmed by gel electrophoresis. However, the corresponding bands showed rather low intensity, whereas off-target fragments were more prominent. Consequently, successfully cloned vectors can be used in further experiments after the iGEM period to obtain transformed yeasts for α-pinene production. All generated yeast strains from Table Scp 1 were analyzed by cPCR (Fig. Scp 4) by amplifying a region within the inserted construct using specific primer combinations.

Suitable primer combinations were selected to confirm both plasmid and gene insertion. Together with the auxotrophy selection markers, we assumed that we had successfully introduced these constructs into the yeast strains. In contrast, we could not successfully verify the mevalonate module because it was not reliably amplified. Further experiments would require confirmation of the construct by repeated cPCR. Since the generated yeast strain S. cerevisiae_MS_9 possessed an additional ApL3H, confirmation of the latter was also necessary. However, this could not be done due to time constraints within the iGEM regulations.

The final evaluation of our genetically modified yeast strains was performed by quantification using GC-MS analysis (Fig. Scp 5). For this, α-pinene and verbenone have been considered important monoterpenoids to infer synthesis efficiency. Yeast strains S. cerevisiae_MS_8 and S. cerevisiae_MS_9 were grown in liquid culture for 48 hours and then extracted with ethyl acetate and small glass beads for analysis. P-xylene for α-pinene and eugenol for verbenone and verbenol were chosen as internal standards. For comparison, the S. cerevisiae CEN.PK2-1C wild type was also included.

The GC chromatogram shows the absolute intensities of detected peaks along their specific retention time. The mass spectra show relative intensity against m/z. The internal standards p-xylene and eugenol displayed peaks at approximately 5 and 12.0 minutes, respectively (Fig. Scp 5 A). Obtained results of S. cerevisiae_MS_8 demonstrated the requested α-pinene expected to eluate at 6.5 minutes. A review of the MS spectrum reveals a distinct pattern that indeed confirmed α-pinene, subsequently quantified at 3.6 mg/l. GC output of S. cerevisiae wildtype did not show a targeted α-pinene peak, indicating that no production occurred (Fig. Scp 5 B). Secondly, S. cerevisiae_MS_9 yielded in 12.8 mg/l verbenone eluting at 10.8 minutes indicating proper function of inserted ApL3H catalyzing α-pinene-derived verbenone that is additionally confirmed by the MS spectrum (Fig. Scp 5 D). In addition to catalytic conversion, autooxidation of α-pinene to verbenone was also likely in line with previous studies (Moore et al., 1956). The peaks for α-pinene and verbenone showed identical retention and fragmentation patterns as the previously measured authentic standards (Fig. Scp 5 C). Consistent with our prediction, both strains indeed appear to exhibit targeted synthesis, supporting our original approach of microbially produced high-value precursors.

Cytosolic production of α-pinene in Yarrowia lipolytica

Summary

Yarrowia lipolytica is a promising organism for monoterpenoid synthesis, since it contains a high concentration of acetyl-CoA, is able to efficiently accumulate lipids, and shows high tolerance toward monoterpenoids (Ji et al., 2020, Wei et al., 2021). We aimed to introduce the Saccharomyces cerevisiae genes SctHMGR and ScERG13 to eradicate the bottleneck of the mevalonate pathway. Additionally, we tried to introduce the AgtGPPS2 and GgMFPS144 genes into the YlERG20 locus of Y. lipolytica to reduce the amount of produced farnesyl pyrophosphate (FPP) and increase the amount of geranyl pyrophosphate (GPP). In a second approach, we sought to introduce the SltNPPS1 to produce neryl pyrophosphate (NPP), and insert the copper repressible promoter YlpCTR1 upstream of the YlERG20 gene to reduce the amount of FPP produced, decreasing competition for the reactants isopentenyl pyrophosphate and dimethylallyl pyrophosphate (IPP and DMAPP). Lastly, we wanted to introduce the PptAPS to produce α-pinene from NPP or GPP (Fig. Ylc 1).
To demonstrate the versatility of the NPP and GPP producing strains we also planned on introducing genes responsible for the synthesis of thymol (TvGTPS1, TvCYP71D179, TvSDR1)s an alternative to the α-pinene producing PptAPS.
We were able to assemble all (multi -) transcription units on level 1 and level 2 plasmids. The creation of linear fragments containing the (multi -) transcription units with fitting homology flanks up and downstream caused some difficulties. Therefore, the CRISPR-Cas9 mediated gene insertion or gene knockout in Y. lipolytica could not be achieved in the timeframe of MonChassis.
We are confident to have found some of the potential key problems concerning the gene insertion that can be tackled in the future.

MoClo-based assembly of transcription units

We could create all designed level 1 plasmids. The correct assembly was confirmed via colony-PCR (cPCR) (Fig. Ylc 2). Although white colonies, that suggest successful gene insertion, were found on plates for the level 2 plasmid assemblies, cPCR appeared to be negative. Later, we were able to amplify the entire multi-transcription unit (MTU) from the pSB3C01-Yl-SctHMGR-ScERG13 and pSB3C01-Yl-AgtGPPS2-GgmFPS144 plasmids using primers binding to the promoter and terminator, confirming the correct assembly of these two level 2 plasmids (Fig. Ylc 3).

CRISPR-Cas9-mediated gene insertion or knockout

While completing the cloning of our regular level 0, 1, and 2 plasmids, we started cloning our CRISPR plasmids. We decided on using the pJME4472 plasmid (Larroude et al., 2020) since it contains the URA3 gene as a selection marker. When positive transformants are plated on media containing 5-fluoroorotic acid (5-FOA), the yeasts eliminate the plasmid, since 5’FOA is converted into a toxic compound when the uracil synthase coded by URA3 gene is present. This way, we can later introduce a different pJME4472 plasmid for subsequent gene insertions or knockouts.

We introduced crisprRNAs (crRNAs), hybridized from oligonucleotides, into the pJME4472 via Golden Gate Assembly using BsmBI, which excises a red fluorescent protein gene (mRFP1) from pJME4472. Thus, positive transformants are white, while E. coli colonies with the original plasmid turn red.

The genomic loci for the insertions were chosen based on Schwartz et al., 2011. YlERG20 was also chosen as a locus since the YlERG20 gene was a target for deletion or downregulation.

To improve the efficiency of homologous recombination, we aimed to knockout the KU70 gene. The KU70 protein is part of the non-homologous end joining complex, which is the preferred repair method for DNA strand breaks in Y. lipolytica (Ji et al., 2020, Larroude et al., 2020). The genomic target sequences were chosen to be close to the starting sequence of the gene, and with high efficiency, specificity, and no off-targets based on CRISPOR web software.

Three different target sequences were selected for the YlKU70 knockout, which displayed a high probability of a frameshift (Fig. Ylc 4). Successful cloning was confirmed via cPCR.

After overcoming initial problems with the transformation method of Y. lipolytica , we were able to transform all three of the plasmids using an electroporation-based approach. A sequence of 777 bp was amplified from the genome, purified and sequenced, but no mutation was found.

Overlap-Extension PCR to obtain fragments for CRISPR-Cas9 mediated gene insertion

Gene insertion required the addition of 1 kb homology flanks (HFs) complementary to the desired gene locus upstream and downstream of our transcription units (Fig. Ylc 5). The genomic loci were selected based on the work of Schwartz et al., 2016.

We amplified the HF from Y. lipolytica gDNA using primers with a 20 bp overhang to the transcription units (Fig. Ylc 6). For further information on our strategy, see above.

Simultaneously we amplified the genes and transcription units from different level 0, level 1 and, level 2 plasmids. Amplification of the YlpCTR11 from the level 0 plasmid and of the transcription units from level 1 plasmids was achieved using primers with a 20 bp overhang to the HF (Fig. Ylc 7).
For amplification of MTUs from level 2 plasmids primers that bound to the promoter or terminator as well as part of the backbone were chosen, to achieve a specific binding site. This method proved to be difficult and inconsistent, and the following overlap-extension PCRs to add the HF were not successful, although we tried several different approaches to enable specific amplification, such as gradient PCR or adding dimethylsulfoxid (DMSO) to prevent steric inhibition.
Finally, we settled on amplifying the single transcription units composing the MTUs from level 1 plasmids and linking the individual transcription units with overlap-extension PCR. We focused on the SctHMGR-ScERG13 and AgtGPPS2-GgmFPS144 MTUs.

For this purpose, we designed specific oligonucleotides targeting only the promoter and terminator sequences. Using these primers, the transcription units from the level 1 plasmids could be amplified, and even amplification of MTUs from level 2 was successful. The amplified MTUs were not used for overlap-extension PCR but this could confirm the correct assembly of the level 2 plasmids (Fig. Ylc3).
The amplification and subsequent overlap-extension PCR to obtain the MTUs were successful for both the SctHMGR-ScERG13 construct as well as the AgtGPPS2-GgmFPS144 construct (Fig. Ylc 8).
The addition of the overhangs to the single transcription units, as well as the MTUs, was also successful, except for the ERG20_HF1-AgtGPPS2-GgmFPS144-ERG20_HF2 construct. (Fig. Ylc 8).

Strain generation

Subsequent transformation of Y. lipolytica with the fully assembled, gel-excised fragments and the complementary pJME4472 did yield colonies, that appeared to replicate the plasmid. cPCR never showed a successful gene insertion, but rather fragments of the length that would be expected from the original genomic sequence.Due to time constraints, we were unable to continue with our work on the generation of Y. lipolytica strains with an optimized cytosolic GPP/NPP pathway.
To improve the gene insertion efficiency, repeating the YlKU70 knockout should be considered. It is also possible that the electroporation-based transformation method does not allow for optimal uptake of linear fragments (Wang et al., 2011). A lithium-acetate based transformation method, or the introduction of plasmids containing the MTUs and fitting HFs could be a solution.

---

Peroxisomal production of α-pinene in Yarrowia lipolytica

In addition to the large cytosolic acetyl-CoA pool of Y. lipolytica (Wei et al., 2021), its extensive fatty-acid metabolism makes Y. lipolytica an interesting chassis organism for the peroxisomal production of monoterpenoids. The peroxisomes of Y. lipolytica have already been utilized to produce the sesquiterpene α-humulene and this strategy achieved a higher humulene titer than cytosolic expressions (Farhi et al., 2011). Here we tried to implement the designed metabolic pathways that utilize either GPP or NPP as the substrate for the monoterpene synthase PptAPS in the peroxisomes of Y. lipolytica (Fig. Ylp 1). Within MonChassis the respective level 1 parts of the mevalonate pathway with the PTS1 tag were cloned. Genes that were supposed to be strongly expressed were under the regulation of the YlpTEF promoter, while genes that required weaker expression were assembled with the YlpGAP promoter.

All genes of the NPP utilizing pathway were successfully assembled into level 1 parts (Fig. Ylp 2). For the GPP using pathway, however, it was not possible to successfully assemble the pSB1K04-Yl-AgGPPS_PTS1 part. Moreover, even though assembled level 2 constructs were transformed, and white colonies were observed, verification of the correct assembly proved to be difficult because of the large size of the constructs. We did not perform analytical cPCRs of level 2 constructs, but tried to amplify the respective multi-transcription units for the subsequent overlap-PCR to assemble multi-transcription units with homology flanks (Fig. G1). However, no bands were visible after the PCR of level 2 constructs to add the overlaps for the planned overlap-PCRs (data not shown). Thus, the test of our metabolic designs for peroxisomal monoterpenoid production is pending. To overcome the limitations of MoClo-based genetic engineering of Y. lipolytica , we propose a new vector design for MoClo-compatible level 2 vectors for cytosolic and peroxisomal-tagged expression strategies. More information on the proposed design of the vector can be found here.

---

Proposed MoClo compatible vectors for faciliated engineering of Yarrowia lipolytica

The establishment of MoClo-compatible level 1 and level 2 plasmids containing the HF up and downstream of the (multi-) transcription unit could be a feasible alternative to linear fragments. This could be achieved by Gibson Assembly, similar to the creation of our level 2 shuttle plasmids. The design of the pSB3C01 vector containing HFs for the AXP gene is shown below (Fig. Yl 1). If necessary, the transcription unit, including the HF, can be excised with PmeI.

This improved strategy would utilize most of the already cloned plasmids, maintain the modular and versatile MoClo approach, while increasing gene insertion efficacy to generate a highly efficient monoterpenoid-producing Yarrowia lipolytica strain.

---

E-driven CYP

Our platform for the production of various monoterpenoids, MonChassis, consists of two components; in the in vivo part, beneficial precursors are produced by the optimal yeast strain. The in vitro part aims to catalyze these precursors cost-efficiently to profitable monoterpenoids. To this end, MonChassis uses a self-designed BioReactor for Enzymatic ElectroSynthesis (BREES) and a modified cytochrome P450 monooxygenase BM3. As a proof-of-concept, we aimed to catalyze the reaction from α-pinene to the bark-beetle repellent verbenone electrically-driven. Thus, we also developed a colorimetric assay for the rapid screening of verbenone-producing BM3 mutants.

BREES

Aiming to synthesize monoterpenoids electrically-driven, we mainly built our platform on the research by Zernia et al., 2018. They demonstrated that the recombinant fusion of an indium tin oxide (ITO) binding peptide and the BM3 can be used for electrically driven catalysis. Thus, we aimed to reconstruct this set-up to produce verbenone as our proof-of-concept. We managed to clone, express, and purify a recombinant BM3 and built BREES with which we conducted initial tests.

Cloning, Expression & Purification of the enzyme BM3

First, we attached an N-terminal 6x HisTag for purification and a C-terminal ITO binding peptide with the amino acid sequence (RTRHK)₄ to a BM3 variant. We received a plasmid including BM3 with three mutations (A74G, F87V, L188Q) showing high activity on several fluorescence assays (Zernia et al., 2016) from the working group of Sarah Zernia from Leipzig University. Using self-designed primers with the ITO oligonucleotide sequence attached, we amplified the BM3 via PCR. After this, we inserted the gene into the vector pET28b which contains a 6x HisTag at the N-terminal end of the multiple cloning site. The successful insertion was verified by sequencing using only primers for the border regions of the multiple cloning site (Fig. E-CYP 1). We transformed the expression strain Escherichia coli BL21 DE3with the completed plasmid.

To obtain large and clean amounts of enzyme for application in BREES, we expressed and purified the desired protein BM3 with the ITO binding peptide. For the purification, a column with Nickel-NTA was used, which binds the N-terminal 6x HisTag of the expressed BM3. We performed a Bradford assay to determine the protein concentration of the eluate. Therefore, we created a standard curve of protein with bovine serum albumin (BSA) in known concentrations. The eluate containing BM3 had a total protein concentration of 4.25 mg/ml. Subsequently, we performed an SDS-PAGE to verify the presence of the protein and check the quality. Thus, we checked samples from the pellet and supernatant of disrupted cells, flow through, washing step, and the final eluate.

The SDS-PAGE confirms the occurrence of a protein with ~125 kDa in the pellet, supernatant, and the flow through, as well as in both dilutions of the eluate (Fig. E-CYP 2). This is most probably the protein of interest, the recombinant BM3 with 124 kDa. In the pellet (2), a relatively high amount of a protein with ~35 kDa is present, indicating the presence of the in E. coli BL21 DE3 overexpressed membrane protein OmpT with 33 kDa. Additionally, a protein of ~55 kDa is detected in both dilutions of the eluate (5 & 6), which can be due to binding of amino acids like histidine to the Ni-NTA. Thus, we have a high amount of BM3 although there is also another protein in our eluate. A second washing step with an even higher concentration of imidazole could lead to higher purity of the protein BM3.

Activity tests of BM3

We wanted to test the general enzyme activity as well as the immobilization on the ITO-electrode of the purified BM3. Since the variant of BM3 we use in MonChassis shows high activity in the deethoxylation of 7 ethoxycoumarin to 7 hydroxycoumarin (Zernia et al., 2016) due to the three mutations, we chose a fluorometric assay to examine the general enzyme activity. We tested the purified BM3 with the ITO binding peptide in two set-ups. In set-up I, we diluted BM3 in assay buffer and added substrate as well as nicotinamide adenine dinucleotide phosphate (NADPH) to start the reaction. For set-up II, BM3 was incubated with the electrode for 1 h to attach to the ITO surface before the assay (Fig. E-CYP 3). After that, we discarded the supernatant and added substrate and NADPH to the set-up. In both set-ups, we extracted 7-hydroxycoumarin from the reaction solutions with ethyl acetate and measured the fluorescence with excitation wavelengths of 340 nm and 360 nm and an emission wavelength of 465 nm using a plate reader.

Plotting the data against a standard series of known amounts of 7-hydroxycoumarin, we saw no activity of purified BM3 either in solution (set-up I) or on the ITO-electrode (set-up II). This could be due to mutations in the BM3 leading to a defective folding of the native protein since the protein sequence was not entirely analyzed before transformation. Additionally, no activity in set-up I can be due to low affinity of the ITO binding peptide to the electrode resulting in the discard of the BM3 before the substrate and NADPH were added. For this, the binding affinity should be tested again.

Testing the BioReactor for Enzymatic ElectroSynthesis

MonChassis aims to use enzymatic electrosynthesis for the conversion of α-pinene to verbenone. Therefore, we produced BREES for an inexpensive and easy way to use immobilized enzymes for electrically-driven catalysis. The detailed design as well as further information about the bioreactor can be seen here. We conducted different experiments to test BREES.

At first, we examined the binding affinity of the ITO binding peptide onto the electrode. Therefore, we measured the UV-Vis spectra of purified BM3 with the NanoDrop to determine the wavelength of its absorbance (Fig. E-CYP 4). Several peaks at different wavelengths were detected. Since the SDS-PAGE (Fig. E-CYP 2) revealed the presence of two proteins in the eluate, we decided to choose the peak at 426 nm as indicator for the purified BM3 which may result from the absorbance of its oxidized heme-group. Based on the cross section of BM3 published in literature (Zernia et al., 2018), we determined that the maximal possible amount of BM3 to be immobilized onto the ITO-electrode is 20.83 pmol. Thus, we diluted the eluate gained from purification 1:1000 and added 280 pmol of BM3 onto the electrode to ensure its complete coverage. To determine if the binding peptide attaches to the ITO-surface, we measured the absorbance at 426 nm before and after applying purified BM3 on the electrode using the sample before application as standard. We assessed a difference in absorbance of 9 mAU. This could verify the immobilization of BM3 on the ITO-electrode. But since the small number of maximal immobilized BM3 is very near to the detection limit of the NanoDrop, these data are not significant. The binding affinity of the ITO binding peptide must be examined with other methods, such like antibody adhesion.

Although we measured no enzyme activity of the purified BM3 in the deethoxylation of 7 ethoxycoumarin, we decided to test the enzyme’s ability to convert α-pinene to verbenone in BREES anyways. Therefore, we aimed to immobilize the BM3 on the ITO-electrode in two set-ups with either NADPH or with electricity for reduction (Fig. E-CYP 5). The ITO-electrodes were incubated with the purified BM3 for 30 min and then washed with PBS buffer. Then, we added the substrate α-pinene. We started the reaction by adding NADPH in the set-up with no electricity and by connecting the ITO-electrode with electricity in the other set-up. In both batches the supernatants were then extracted with ethyl acetate and measured with GC-MS.

Unfortunately, we saw no occurrence of either verbenol or verbenone. This confirms that the purified BM3 had no activity in the conversion of α-pinene to verbenone in our set-ups. Same as in the general activity assay, the reasons can be mutations in the BM3 or the immobilization of the enzyme on the ITO-electrode does not work properly. Nevertheless, we were able to do the first tests of our self-designed BioReactor for Enzymatic Electrosynthesis (BREES) and gained experience with its application. We saw that the air pressure sealing allowed no loss of reaction liquid and led to visible stirring and ventilation.

Colorimetric screening of favorable BM3 mutants

MonChassis wants to produce verbenone as a proof-of-concept for electrically-driven monoterpenoid production. To optimize the conversion from the precursor α-pinene to verbenone, we developed a colorimetric assay for the rapid screening of favorable mutants of BM3. We used 2,4-Dinitrophenylhydrazin (DNPH) that reacts with unsaturated ketones (enones) like verbenone and converts them to a colored phenylhydrazone precipitate (Fig. CYP 6). The amount of verbenone that is thereby formed, is proportional to the visible red coloration. Saturated ketones and aldehydes do also react with DNPH, but do not form a colored complex and thus remain yellow-colored (Liu et al., 2016).

The problem with an intracellular enzyme reaction is that α-pinene is toxic in higher concentration and verbenone cannot diffuse through the membrane. To avoid a laborious extraction of verbenone from the E. coli cells, we used the method of autodisplay. In this method, a passenger protein is fused to a signal peptide and a membrane protein, called the autotransporter system. This ensures the passenger proteins display on the surface of E. coli (Jose et al., 2012; Schumacher et al., 2012). In our colorimetric assay, we used a wildtype BM3 as a passenger protein. Due to the simple insertion of different mutants in the autotransporter system, our colorimetric assay allows rapid screening of favorable BM3 mutants for verbenone production.

α-pinene toxicity test

Monoterpenoids including α-pinene are known to be toxic to bacteria such as E. coli (Dunlop et al., 2011; Sarria et al., 2014; Trombetta et al., 2005). Therefore, we aimed to determine an appropriate E. coli strain for autodisplay and examined the highest possible concentration of α-pinene it can be exposed to. In total, we incubated four different E. coli strains with eight different α-pinene concentrations and a negative control. We measured the course of the optical density at 600 nm over 16 h (Fig. E-CYP 7).

As seen in Figure E-CYP 7, most approaches showed slow and steady growth after starting at an OD₆₀₀ of ~0.7. The highest optical density was detected for an α-pinene concentration of 2.5% (v/v) for all strains. Lower concentrations showed a similar course to the control whereas higher concentrations impaired cell growth. Some data rows of BL21 (DE3) codonPlus were irregular and had a high standard deviation, indicating that the growth of the strain was negatively influenced by α-pinene. For the other strains, the results were similar with a reliable standard deviation and the highest possible α-pinene concentration of 2.5% (v/v). Thus, we decided to use E. coli BL21 for the colorimetric assay, since this strain lacks the T7 promoter system as well as the outer membrane protease OmpT (Jose et al., 2012). This makes the strain suitable for autodisplay, and thus, for our colorimetric assay.

Colorimetric assay with DNPH

To ensure that only verbenone reacts with DNPH even in the presence of other substances and/or cells, we have performed several tests. Therefore, we grew E. coli BL21 in LB medium to an OD₆₀₀ of 1.0 and incubated them with known amounts of verbenone. As negative control and for comparison of coloration, we used the same amount of verbenone diluted in water (Fig. E-CYP 8). After incubation, we added DNPH. We also examined the ketone acetone as well as other monoterpenoids such as α-pinene and verbenol on their interaction with DNPH and E. coli BL21.

The results show a rapid determination of verbenone occurrence in samples with or without E. coli BL21 cells is possible by red staining with DNPH (Fig. E-CYP 8). A higher amount of verbenone leads to more intensive red staining, enabling a comparison if known standards are present. Other compounds such as α-pinene , acetone, and verbenol show no red staining and thus, do not interact with cells or DNPH. This supports that a colorimetric assay using DNPH and autodisplay can be used as a rapid screening for favorable BM3 mutants.

These findings rely solely on a subjective determination of verbenone occurrence by visual red staining. Since an objective assessment based on measured data is more reliable, we also wanted to determine the absorbance using UV-Vis spectroscopy.

Since the absorption maximum of the hydrazone formed by verbenone with DNPH was not known, we measured the spectrum against DNPH without verbenone as control (Fig. E-CYP 9). Verbenone-dinitrophenylhydrazone has an absorption maximum at 390 nm as shown in the figure. The data was validated by measuring the DNPH absorption maximum at 352 nm as previously described (Roberts and Green, 1946). Since the absorbance at 450 nm was so high that it was outside the linear measurement range, both approaches were diluted in acetonitrile before the measurements.

To avoid interference with other substances and unreacted DNPH in our further experiments, we chose a higher wavelength for the measurements. At 450 nm, an accurate standard curve could be measured for the verbenone-dinitrophenylhydrazone (Fig. E-CYP 10).

As shown in Figure E-CYP 10 the standard series with verbenone was successful, since the R²-factor is close to 1, confirming the possibility of protein engineering by an objective color assay as the regression line can be used to determine the concentration of verbenone. We also incubated the autodisplay cells with verbenone using the same method. For this purpose, the E. coli BL21 cells must be removed by centrifugation, as they interfere with the absorption (Mesquita et al., 2014). The supernatant was transferred and DNPH was added. After incubation, we measured the spectrum of the samples in the plate reader.

Problematic was that even controls with no verbenone, but NADPH showed high absorbance at 450 nm. This is unusual since the absorption maximum of NADPH is actually around 340 nm and the maximum of NADP+ is even lower (Sugishima et al., 2020). To avoid this problem, we decided to work with an NADPH regeneration system, since in this case overall less NADPH/NADP+ is present, so interference is diminished. These samples showed indeed a lower absorption, but it still influenced the results.

Therefore, mutation screening for autodisplay by UV-Vis spectroscopy still needs to be revised by possible extraction with, for example, ethyl acetate. Nevertheless, finding a BM3 with increased verbenone production is definitely easier with the rapid screening of our developed colorimetric assay. Thousands of variants can be screened within days, which would take weeks for measurements by gas chromatography with mass spectrometry (GC-MS).

---

Conclusion and Outlook

Through extensive cloning of numerous gene constructs and yeast transformation, we have clearly demonstrated that the synthesis of α-pinene and verbenone is possible by introducing a synthetic pathway into Saccharomyces cerevisiae in both the cytosol and peroxisomes.

The results we obtained were indeed promising and offered insights into more advanced possibilities, as we were able to confirm the synthesis of monoterpenoid precursors in yeast. This sets the basis for future experiments regarding advanced strain optimization and exploring the promising organism Yarrowia lipolytica. We considered high throughput screening to substantiate the quantification and functional characterization of individual pathway members of interest. However, quantification is not reliable due to insufficient replicates and lack of improvements in experimental design. Despite our tremendous success in creating different chassis organisms, we had to adjust our expectations in finding the most efficient α-pinene producing yeast strain. Because of the limited time we were unable to exploit the full potential of MonChassis.

To unlock the full potential of metabolic engineering, we developed a powerful software tool for metabolic modeling to reveal further targets for improvement and to identify the optimal carbon source for monoterpenoid production. Accounting for all aspects mentioned the most efficient chassis organism can be determined and utilized for large-scale biotechnological applications. Concerning the in vivo approach, which was to produce monoterpenoids, we were able to demonstrate the functionality of MonChassis.

Finally, MonChassis also aimed at developing an in vitro system for downstream enzymatic electrosynthesis that converts yeast-produced precursors into any desired monoterpenoid. The use of such a cell-free platform eliminates the need for costly cofactors, like NADPH, extends toxicity-related yield limits, and provides a modular method of monoterpenoid production. To demonstrate this aspect of MonChassis, we focused on the conversion from α-pinene to verbenone. We cloned, expressed, and purified a genetically engineered cytochrome P450 monooxygenase BM3 variant, successfully immobilized it on an indium tin oxide (ITO)-electrode, and tested its activity in various experimental setups. We also designed, built, and tested a modular, easy-to-copy bioreactor that can be used by future iGEM teams. This platform also includes a colorimetric assay system for rapid mutation screening.

For industrial application of MonChassis, the next logical step will be to investigate direct purification of BM3 using the ITO-binding peptide. This could vastly decrease the costs since no complex purification methods are necessary and the cell extract can be applied immediately onto the ITO-electrode. The yield of monoterpenoids such as verbenone can be further increased by enzyme engineering. Promising amino acids for increased α-pinene specificity of BM3 have already been mentioned (Lehmann, 2016), for which the colorimetric assay developed by MonChassis can be employed. For industrial scale-up, bioreactors with larger electrode surfaces and multiple reaction chambers need to be designed. For all future steps, our key findings paved the way for the enzymatic electrosynthesis of monoterpenoids. MonChassis thus provides a step towards a sustainable bioeconomy.