Engineering Success

×

On this page, you can read how our team implemented iterations of the engineering cycle to complete a successful project.

Introduction

Purpose of Our Project

The main objective of this project is campesterol production by using waste cooking oil (WCO) as a carbon-source. This is accomplished by engineering the Yarrowia lipolytica, a type of oleaginous yeast. To do so, we aimed to metabolically engineer Yarrowia lipolytica and disrupt the ergosterol synthesis pathway to replace it with 7-dehydrocholesterol reductase (DHCR7), which leads to campesterol production (Figure 1)1,2.

Because metabolically engineering Y. lipolytica to produce campesterol is a project tackled in academic communities at large scale and would have not been feasible within the time frame of this project, we focused on the expression of biosurfactants. This increases the yeast’s access to WCO, which can increase campesterol yield 3. Biosurfactants have amphiphilic properties, meaning they mix well both with oil and water. This allows them to interact with the oil-water interface, reducing surface tension and stabilizing emulsions. By engineering Y. lipolytica to secrete biosurfactants into the cell culture media, WCO uptake can be improved. Our goals are as follows:


  1. Expressing natural hydrophobins as well as mutated hydrophobins and characterizing this expression.
  2. Future work would entail metabolic engineering and thus replacing the ERG-5 enzyme in the terpenoid production pathway with the DHCR7 enzyme to produce campesterol.
  3. Future implementation of our work as a start-up would also lead us to looking into filtering protocols for waste cooking oil and potential technical issues with the surfactants such as foaming.

Figure 1. The mevalonate pathway and examples of terpenoids, such as campesterol, produced by engineered Yarrowia lipolytica  strains1. The mevalonate pathway occurs in the yeast cytosol. Many terpenoids accumulate either in lipid compartments or the extracellular matrix, although the terpenoid transport mechanisms in Y. lipolytica remain poorly understood. Reproduced from a paper with permission from the authors1.

Design Principles

With these goals in mind, we devised a modular approach for three different chassis to produce biosurfactants, that we would execute in the following order:

  1. Optimization and expression in E. coli.
  2. Transformation into and expression in S. cerevisiae.
  3. Transformation into and expression in Y. lipolytica.

First, we optimized hydrophobin sequences in E. coli to create E. coli hydrophobin plasmid constructs. We then devised cloning strategies for the yeast species S. cerevisiae as well as Y. lipolytica.

Another important design aspect we explored were the different protein secretion signals in S. cerevisiae and Y. lipolytica. This was done to ensure an optimal production and transport of biosurfactant proteins in the engineered strains.

The DBTL Cycle

iGEM engineering cycle Figure 2. iGEM Engineering Cycle

To successfully bring our design principle into reality, we’ve referred to the DBTL cycle from the iGEM engineering guideline: Design → Build → Test → Learn (Figure 2).

Our wet lab project presented some challenges that gave us the opportunity to make ample us of all components of the cycle. Hence, we have completed three full iterations of this cycle. We gave them the following names:

  1. Sequence design and our first attempt at cloning.
  2. The transition to a step-wise cloning strategy and a different plasmid system.
  3. Final cloning and expression evaluation.

These iterations are subdivided into four sections:

  1. Design: our fundamental design idea was already explained above, but these sections go into further detail regarding our logic on how we chose a certain backbone and parts to rationally sketch our plasmid constructs.
  2. Build: the actual methods that we used to build our assembly.
  3. Test: we explained which measurement we chose to test our build and showed corresponding results.
  4. Learn: we discussed possible reasons why the first design went wrong, as well as which improvements should be taken for the next DBTL cycle.

The DBTL cycle was incredibly helpful when engineering our microorganism because it motivated us to learn from previous mistakes and accomplish several iterations to achieve true engineering success. However, we only got to implement full iterations for E. coli. We devised cloning strategies for Y. lipolytica and S. cerevisiae in silico, but we eventually didn’t get to build them.

Escherichia coli

Sequence Design and Our First Attempt at Cloning in pET28

  • First Design:

    We decided to assemble parts into a pET28 backbone for transformation into E. coli. To find the best biosurfactant, we screened a diverse library of sequences from three classes of biosurfactants: HGFI (hydrophobin class I), HFBI and HFBII (hydrophobins class II)4,5 and the MBSP1 protein6. In addition, mutations at the cysteine residues were introduced into the hydrophobin sequences to promote higher protein expression and secretion7. Starting from fundamental parts for our vector, such as the terminator sequence, selection marker (KanR2), Ori (pBR322), repressor (lac I), and His-tag, we designed fusion AcGFP-linker-hydrophobin constructs. Each of the hydrophobin sequences was preceded by a GFP signal (AcGFP) linked with a flexible polypeptide sequence to track successful expression of hydrophobin by measuring fluorescence. As we screened four different biosurfactants (HGFI, HFBI, HFBII, and MBSP1) with different cysteine mutations (HGFI_half_mut, HGFI_full_mut, HFBI_mut, HFBII_mut), we created a total of eight AcGFP-hydrophobin assemblies using Benchling (Table 1).

    Figure of pET28 Figure 3. pET28_AcGFP_MBSP1_060723.
    Hydrophobins Class Length (bp)
    HFBI Class II hydrophobin 231 bp
    HFBI mutation * Class II hydrophobin 231 bp
    HFBII Class II hydrophobin 294 bp
    HFBII mutation * Class II hydrophobin 294 bp
    HGFI Class I hydrophobin 324 bp
    HGFI half mutation ** Class I hydrophobin 324 bp
    HGFI full mutation ** Class I hydrophobin 324 bp
    MBSPI Biosurfactant 897 bp
    Table 1. Hydrophobins used for pET28_AcGFP_linker_hydrophobin constructs.
    * mutations = cysteine residues replaced with serine.
    ** HGFI half/full mutation = As HGFI hydrophobin contains 8 cysteine residues, we mutated first 4 cysteins to serine and named it as HGFI half mutation; in case of HGFI full mutation, all cysteines were replaced with serines
  • First Build:

    By using the Assembly Wizard function in Benchling, we generated primer pairs to make the backbone, AcGFP, and hydrophobin sequences compatible with Golden Gate cloning. We calculated annealing temperatures for all the primer pairs using the NEB Tm Calculator tool. We then PCR amplified our pET28 backbone, AcGFP reporter and insert sequences.

  • First Test:

    The PCR for AcGFP as well as for the mutated HGFI sequence worked, while that for the pET28 backbone did not, even with various changes such as the use of different polymerases or by gradient PCR. The PCR result was checked using 1% agarose gel electrophoresis. The result can be seen in Figure 4.

    Figure 4. Agarose gel showing the result of our first iteration of the DBTL cycle.
    Lane 1, 3: 1kb ladder
    Lane 2: pET28 plasmid: 5500 bp
    Lane 4: AcGFP fragment: 730 bp
    Lane 5: HGFI_full_mutated fragment: 324 bp
    Lane 7: Low MW DNA ladder
  • First Learn:

    After repeated failed trials of amplifying our initial desired pET28 backbone, we decided to shift our focus onto a slightly different plasmid backbone (pET29). This backbone already contained a His-tagged GFP sequence. We also dedicated special attention to designing primer pairs with a more appropriate annealing temperature. Another improvement we made was to generate a two-step cloning process, in which a short linker was initially introduced upstream of the sfGFP sequence on the plasmid backbone. Then, the coding sequences for biosurfactant proteins would be introduced in a second cloning step, allowing us to only work with one insert at a time.

    • Transition to a Step-wise Cloning Strategy and a Different Plasmid System (pET29)

  • Second Design:

    Based on the conclusion of the first engineering cycle, we decided to shift to a different backbone: pET29. The result of this can be seen in figures 5 and 6.

    Moreover, the “Learn” from our previous iteration also mentioned improving annealing temperatures. Thus, before building our constructs, we performed an in silico PCR through benchling to see if primers were functional to make amplicons for the backbone. By using a different backbone, we were able to obtain primers with lower melting temperatures to create inserts at the desired site within the vector.

    pET28_AcGFP_MBSP1_060723 Figure 5. pET29_link_MBSP1_200723.
    Figure 6. pET29_linker_sfGFP_200723_5547bp.
  • Second Build:

    To create an intermediate pET29-linker-sfGFP plasmid, we first amplified the original backbone (pET29-sfGFP) using Q5 Polymerase. As correct PCR amplification was confirmed by gel electrophoresis, we moved onto the next step: the annealing of linker oligonucleotides. The amplified plasmid backbone digested with BsaI was ligated to the oligonuclotide duplex containing compatible overhangs using a T4 ligase. The purpose of this cloning step was to introduce a short flexible linker upstream of the GFP sequence, in frame with our expression cassette. After performing T4 ligation, we transformed the ligated intermediated plasmid backbone (pET29-linker-sfGFP) into NEB 10-β competent E. coli cells. We amplified our assembly through miniprep and sent samples for sequencing to check the integration of the short oligonucleotide linker. Figure 7 summarizes what we did for this iteration.

    Figure 7. Flowchart for intermediate plasmid assembly.
  • Second Test:

    The sequencing results did not confirm the insertion of the linker (Figure 8). However, we observed BsaI sites still present in the construct, which led us to believe the plasmid backbone digestion with BsaI was inefficient.

    Figure 8. Sequencing result for linker plasmid.
  • Second Learn:

    Since we suspected incomplete BsaI digestion, we tested restriction activity of Bsal on a plasmid with Bsal sites. The results (not shown) demonstrated that the Bsal enzyme successfully cut the plasmid into two fragments (6kb +1kb fragment). Accordingly, we decided to repeat experiments and re-digest amplified pET29 vector with Bsal, while changing the incubation time (30 minutes to 1 hour).

    • Final Cloning Strategy and Expression Evaluation

  • Final Design:

    The design we used for this iteration is identical to the one from the previous cycle.

  • Final Build:

    As our subsequent sequencing result showed that the linker was successfully inserted, we started PCR amplification of hydrophobin sequences (HGFI, HGFI_mut_half, HGFI_mut_full) for both Golden Gate and Gibson assemblies. After PCR purification, these hydrophobin amplicons went through a Gibson Assembly reaction to create a “pET29-sfGFP-hydrophobin” plasmid construct.

  • Final Test:

    As part of our final test, the constructs were miniprepped and sequenced to check if the hydrophobin sequence was incorporated or not.

    Since we suspected incomplete BsaI digestion, we tested restriction activity of Bsal on a plasmid with Bsal sites. The results (not shown) demonstrated that the Bsal enzyme successfully cut the plasmid into two fragments (6kb +1kb fragment). Accordingly, we decided to repeat experiments and re-digest amplified pET29 vector with Bsal, while changing the incubation time (30 minutes to 1 hour).

    GFP expression in E. coli Figure 9. GFP expression in E. coli.

    Unfortunately, we came to learn that our expression was leaky, even in the absence of the IPTG inducer. While this observation validated our approach of fusing the proteins of interest to a sfGFP sequence for an easy and visual readout of expression and leakiness, we would like to highlight the need for better transcriptional control in the future. To that end, our team proposes the usage of tight inducible promoters such as those derived in the E. coli “Marionette” strains8.

    • GFP expression in E. coli Figure 10. Our idea for improving our promotor and ensuring inducible expression of our target protein.
  • Final Learn:

    Upon finalizing our documentation, we realized that the wrong inducer was used in the protein expression experiments in our E. coli system. Therefore, the induction with IPTG is most likely not responsible for the driving of protein expression. This realization also highlighted the leakiness of our system and the need for a more tightly controlled system.

    This being said, we still showed that our system leads to hydrophobin expression. This means that fixing the promoter will only result in expression that is actually inducible.

    • Saccharomyces cerevisiae and Yarrowia lipolytica

    Plasmid Figure 10. pBEVY-nat_HFBI_210823, the plasmid we would have used for cloning in S. cerevisiae.

    We successfully cloned MBSP1, HFBI, and mutant HFBI in E. coli and derived strategies to clone the others in S. cerevisiae and Yarrowia lipolytica. We have plasmid backbones for the two yeast species and designed theoretical plasmid maps using Benchling. However, we could not finish them in the lab, as there was a limited time before the deadline. See Figures 10 and 11 for our constructs for S. cerevisiae and Y. lipolytica respectively.

    Plasmid Figure 11. NDV-URA3-P1TEF-Lip2_secretion_MBSP1_Lip2_terminator, the plasmid we would have used in Y. lipolytica.

    Genetic switch

    Two separate parts of our project can be distinguished: campesterol biosynthesis and bioemulsifier expression. Performing both at the same time could be a heavy burden for the cell. Thus, we designed a genetic switch in silico to help mediate this. With a use of such a switch, the cell could switch between producing campesterol and biosurfactants without overworking itself.

    The mechanism would be a simple toggle switch 7,8 that when triggered, changes from one state to the other. Coming up with the idea for the trigger turned out to be more challenging than we expected. The easy approach would be to create an oscillatory circuit​9​ that depends on the concentration of campesterol and biosurfactants. Unfortunately, due to the properties of these molecules, detecting their concentration would be hard or even impossible. After looking into other possibilities such as a recombinase addressable data (RAD) module10​, we gathered a few ideas for our bio switch. One of them consisted of a thermal switch which would be relatively easy to apply.

    Normally, during steroid biosynthesis, temperature rises, which would require constant cooling. We could easily control temperature in a bioreactor by stopping the cooling for a moment and letting the temperature rise. This would then cause the switch to go from one state to the other. An additional advantage of such an implementation is that the temporary stops in cooling would decrease the energy consumption. We also talked about the possibility of exploring the morphological transition from yeast-to-hyphal growth in Y. lipolytica as the trigger for the switch. This could act as a “one time” switch where in the yeast growth phase, the cell would produce biosurfactant and in the hyphae form, it would switch to campesterol production. Some other ideas consisted of using metabolic load as the trigger, or utilizing a quorum sensing system, but these avenues would need to be explored more thoroughly.

    Genetic switch

    Future Engineering Concerns

    As can be seen from the points above, our team did a lot of engineering work already. However, this process isn’t quite done, and we anticipate further engineering challenges down the line. Here, we go over them quickly.

    We acknowledge that the use of potentially toxic waste may come with challenges, such as inconsistent chemical composition of the raw material. Another concern are impurities that could affect the proper growth of our yeast as well as patient safety if they are included in the final medicinal product. It is for this reason that we investigated strategies of purifying the WCO substrate for better experimental control. From the literature, we were able to find that using a polypropylene (PP) hollow fiber ultrafiltration (UF) membrane or a combination of esterification, adsorption, solvent extraction, and distillation have been successfully employed in research and industrial settings9.

    Moreover, as foaming can become an issue when using surfactants, we acknowledge this limitation and plan to address it through tight transcriptional control of the bioemulsifier sequence eventually produced by our yeast. By separating the emulsifier production phase from the metabolization of waste cooking oil, Y. lipolytica would only produce the proper amount of biosurfactant that it needs to efficiently grow on the oil substrate. To see how we are planning on doing this, check out the next chapter!

    1. Arnesen, J. A., Borodina, I. (2022) Engineering of Yarrowia lipolytica for terpenoid production. Metab. Eng. Commun. 15: e00231, ISSN 2214-0301.
    2. Du H-X, Xiao W-H, Wang Y, Zhou X, Zhang Y, Liu D, et al. (2016) Engineering Yarrowia lipolytica for Campesterol Overproduction. PLoS ONE 11(1).
    3. Meyer, A.J., Segall-Shapiro, T.H., Glassey, E. et al. Escherichia coli “Marionette” strains with 12 highly optimized small-molecule sensors. Nat Chem Biol 15, 196–204 (2019).
    4. Cheng, Y., Wang, B., Wang, Y., Zhang, H., Liu, C., Yang, L, et al. (2020) Soluble hydrophobin mutants produced in Escherichia coli can self-assemble at various interfaces. J. Colloid Interface Sci 574, 384-395.
    5. Aliwarga, L. et al. Impurity Removal of Waste Cooking Oil Using Hydrophobic Polypropylene Hollow Fiber Membrane. J. Eng. Technol. Sci. 51, 216–230 (2019).
    6. Cárdenas, J. et al. Pre-treatment of used cooking oils for the production of green chemicals: A review. J. Clean. Prod. 289, 125129 (2021).
    7. Gardner, T. S., Cantor, C. R. & Collins, J. J. Construction of a genetic toggle switch in Escherichia coli. Nature 403, (2000).
    8. Leon, M., Woods, M. L., Fedorec, A. J. H. & Barnes, C. P. A computational method for the investigation of multistable systems and its application to genetic switches. BMC Syst Biol 10, (2016).
    9. Niederholtmeyer, H. et al. Rapid cell-free forward engineering of novel genetic ring oscillators. Elife 4, (2015).
    10. Bonnet, J., Subsoontorn, P. & Endy, D. Rewritable digital data storage in live cells via engineered control of recombination directionality. Proc Natl Acad Sci U S A 109, (2012).