Wet Lab Cycles

WL Overview

The goal of Oncurex is to create a synthetic biosynthesis pathway for ursolic acid (UA), an anticancer and antibacterial triterpenoid naturally found in fruit peels, such as apple and loquat, and herbs such as rosemary and sage. In plants, the precursor for ursolic acid, α-amyrin, is produced by α-amyrin synthase (AAS). By expressing AAS in Saccharomyces cerevisiae, α-amyrin can be made which ultimately allows for the production of UA by Cytochrome P450 and CRP. For biomanufacturing, it is necessary to increase yields to create a production process that can result in higher yields than plant extraction. Additionally, drug substances sourced from plant extracts are highly heterogeneous and difficult to regulate, so using synthetic biology streamlines production and allows for easier FDA approval. To this end, knocking out a gene encoding the mitochondrial transporter POR2 decreases the flow of acetyl-CoA into mitochondria, where it can be used for other anabolic purposes. Since acetyl-CoA is an early precursor in the ursolic acid synthesis pathway, this increase in cytosolic acetyl-CoA concentration intends to increase yield since ursolic acid production occurs in the cytoplasm by simultaneously decreasing the cellular usage of acetyl-CoA for the production of other compounds.

Instead of solely transforming the yeast with the plasmid of interest, we have chosen to integrate the genes of interest directly into the yeast genome. This method was chosen because yeast has a relatively high rate of plasmid loss, meaning that a culture of yeast with the target plasmid would eventually lose the plasmid over generations [1]. It is crucial for our project that the gene of interest is not lost because we are using a CSTR to produce our target compound, and a high rate of plasmid loss leads to more downtime in the CSTR. The genes of interest will be incorporated into the yeast through homologous recombination. This occurs between two DNA sequences with similar nucleotide sequences and results in the exchanging of genetic information. This process can be leveraged during transformation to allow S. cerevisiae to incorporate the transformed DNA into its genome for insertions and knockouts [2]. Our plasmid sequences were homologous to a targeted region of the S. cerevisiae genome, allowing for controlled gene knockouts and integration of the genes of interest into the yeast genome.

The proof of concept was established by creating a plasmid construct via Gibson Assembly with the AAS gene isolated from Malus domestica (MdOSC1) and a pRS306 backbone. This construct was then transformed into S. cerevisiae cells, where the production of ursolic acid was determined. Characterization of our construct occurred by testing its performance with the presence of a POR2 knockout in addition to various growth conditions such as the absence of uracil.

WL Cycle 1: 2-Step Single Crossover Homologous Recombination

Design

This method differs from the double crossover homologous recombination as it's composed of two separate single crossover processes with two separate plasmids: KanMX and pRS306. KanMX is a commonly used gene knockout and replacement S. cerevisiae plasmid that contains kanamycin resistance genes for selection. pRS306, which was donated to us by the Fromme Lab, is another commonly used plasmid used for integration of genes into the SEY6210 strain of S. cerevisiae and contains a URA3 sequence that is used for selection on media lacking uracil.

Figure 1a, 1b. Plasmid Design for KanMX (left), Plasmid Design for pRS306 (right)

Build

Plasmids were received in Escherichia coli stab cultures, which were plated (Figure 2a) and then liquid-cultured with ampicillin. Following the liquid culture, spectrophotometry was used to verify growth before a miniprep was performed. After mini-prepping the pFA6a-kanMX6 plasmids (Figure 2b), the team noticed low 260/230 ratios, so DNA cleanup was performed to purify plasmid content.

Figure 2a, 2b. Plate Prep for KanMX-plasmid-containing E. Coli (left), Nanodrop for KanMX MiniPrep pre-Clean Up (right)

The pRS306 plasmid was plated (Figure 3a) and grown in a liquid culture with ampicillin. Miniprep was then conducted, and the purity of the sample was confirmed through nanodrop, with a 260/280 ratio of 1.87 (Figure 3b).

Figure 3a, 3b. Plate Prep for pRS006-plasmid-containing E. Coli (left), Nanodrop for pRS0306 MiniPrep (right)

With purified DNA, a gel confirmation (Figure 4) was performed to verify the correct plasmid length and to check for target amplicons. As seen below, we see our PCR product at their corresponding sizes as shown by the ladder (1st well), with the pFA6a-kanMX plasmid (3rd well) at the 4kb and 1.5kb positions and the pRS306 plasmid (2nd well) at the 4kb position.

Figure 4. Gel Confirmation of pRS306 and KanMX Plasmid PCR. Ladder in lane 1. pRS306 in lane 2. KanMX in lane 3.

1.1 KanMX Crossover

To allow the transformation of the pFA6a-kanMX plasmid to have the function of knocking out POR2, it is necessary to insert a homologous POR2 sequence into the pFA6a-kanMX plasmid. To insert the homologous POR2, we performed PCR on the plasmid with primers containing homologous regions to the yeast POR2 gene to add the POR2 homology region (Figure 4).

Figure 5a. KanMX insertion into yeast genome and simultaneous POR2 knockout.

After the PCR, the plasmid was introduced into the S. cerevisiae through transformation. When the engineered plasmid is inserted into the S. cerevisiae, its internal homologous recombination mechanism will recognize the POR2 homologous sequence on the KanMX plasmid and align it with the corresponding homologous POR2 gene sequence in the S. cerevisiae genome. Once this crossover occurs and the KanMX cassette is inserted into the POR2 gene, it will knock out the POR2 gene.

Initially, we only got bands at around 1.5kb for the KanMX PCR product, even though the expected length is 4.5kb. After reviewing each step of our PCR protocol, we realized that this may be a more complex region of DNA, so we increased the elongation time. This improved our results immediately, resulting in the target 4.5kb band as seen in Figure 4. However, a 1.5kb band appeared alongside the 4.5kb band for each PCR. The primers themselves were very long (~60bp), and secondary structures were very prevalent. To troubleshoot this and to increase specificity, we added 3% DMSO. This did not lead to any changes in the bands formed after PCR. We further troubleshooted by performing touchdown PCR at different ranges around the recommended annealing temperature. However, this did not result in a more specific PCR amplification.

1.2 pRS306 Crossover

For the transformation of pRS306, primers were used to linearize the plasmid while adding homologous ends to the ends of α-amyrin synthase to prepare it for Gibson assembly. The α-amyrin synthase was broken up into two parts, AAS1 and AAS2, because the complexity was too high to be constructed. Then to construct one functional AAS coding region, the AAS1 and AAS2 genes were digested by an overlapping EcoRI region and ligated together using T4 ligase. PCR of the α-amyrin synthase gene with primers designed to match the ends of the gene insert with the ends of the linearized pRS306 amplified the gene insert and ensured its ends overlap with the ends of the linearized plasmid backbone. This allowed for Gibson Assembly to be used to assemble the pRS306 plasmid with the α-amyrin synthase insert, creating a new circular plasmid with both the α-amyrin synthase gene and the URA3 sequence that was already present on the pRS306 plasmid.

This plasmid was cut at the URA3 homology site with SbfI before the transformation into yeast. The internal mechanism of yeast would recognize the homologous URA3 sequence, which resulted in crossover. This crossover occurred at the nonfunctioning URA3 gene while inserting the α-amyrin synthase gene and a functioning URA3 gene. URA3 encodes Orotidine 5’-phosphate decarboxylase (ODCase), which would enable yeast to grow without the presence of uracil or uridine. Thus, a successful transformation in which the URA3 gene is disrupted would result in a lack of growth in a medium without uracil. Further success of the transformation would be confirmed by Ni-NTA chromatography as our design of AAS includes a 6x His-tag at the end of the AAS coding sequence. This is to ensure that not only is the AAS gene fragment integrated into the yeast genome but to also confirm successful production of AAS itself.

Figure 5b. pRS306 crossover with AAS and transformation into yeast using URA3 homology

The AAS1 and AAS2 as well as the pRS306 vector were successfully amplified via the PCR primers we designed. The base pair length of AAS1 and AAS2 was confirmed through our gel confirmation shown below.

Figure 5c. Left to right: Lane 1 - Ladder, Lane 2 and 3 - AAS1 (2.3kb), Lane 4-7 - AAS2(1.1kb)

However, when it came to performing the EcoRI digest and the T4 ligase ligation, we ran into many issues. We troubleshooted with different EcoRI and T4 ligase concentrations, digest and ligation time, as well as incubation temperatures. Ultimately, we were unable to obtain the ligated product of AAS1 and AAS2. The only parameter that affected results was the concentration of starting DNA as expected, observed by the increase in nanodrop concentration and slightly brighter bands on the gel. All wells contained the ligation product achieved through different parameters. Wells 4 and 5 had increased starting concentrations of DNA, hence the slightly visible bands.

Figure 5d. AAS1+2 digest and ligation gel confirmation

Test and Learn

Characterization of the construct will occur through lysis and purification of ursolic acid. Given that this method relies on separate single crossover recombinations, we expect a greater success rate compared to the double crossover homologous recombinations. Based on the results, we would iterate the more efficient method allowing for further improvements on the construct and its yield.

For the KanMX, we were initially unable to obtain the target band, but we were able to overcome this by lengthening the extension step. We have learned to do this first thing when amplifying large PCR fragments such as pRS306. However, we were unable to get rid of the off-target amplicon at 1.5kb, even after many iterations of adjusting the temperature parameters. To reduce the possibility of off-target sites as well as the complexity of the target fragment, we decided to amplify a much smaller fragment from the pFA6a-kanMX plasmid containing the KanMX gene for our next cycle.

Although each genetic component was successfully amplified for the AAS insertion, we were ultimately unable to produce the target pRS306 vector with the AAS. After consulting our graduate advisors, we troubleshooted different parameters but were unable to produce the target part. We then settled on a different solution bypassing the need for the EcoRI digest and ligation step involving a three-part Gibson Assembly. This method was considered at the beginning, but our team has had poor luck with Gibson Assembly in the past, leading to the initial use of digestion and ligation instead.

Using two plasmids is not the most efficient design choice since it involves nearly double the lab work with more opportunities for mistakes. Thus, a new build design was utilized in Cycle 2 to allow each transformation step to take place at once. By utilizing gene fragments, we can leverage the ability of S. cerevisiae to easily perform homologous recombinations [3] without the need for plasmids resulting in faster transformations. Despite this ability, the procedure may have a higher failure rate given the nature of a triple crossover, making this procedure a trade-off between time and quality. As we are working with both cycles in parallel, the team will move forward and iterate the more efficient one once the cycles succeed.

WL Cycle 2: Triple Crossover Homologous Recombination

Design

We seek to accomplish two goals: to knock out the POR2 gene and to insert AAS into the yeast genome. In this iteration, we aimed to accomplish both goals in one experiment by transforming two gene fragments into yeast. One fragment was AAS, which would produce α-amyrin, and the other fragment was KanMX - a common selection marker for yeast - to confirm the insertion of the AAS gene into the POR2 coding region. Yeast cells have the mechanisms to be extremely apt at homologous recombination [3], which is the mechanism used for this cycle, but due to the necessity of three homologous recombination events occurring for the insertions to be successful, it might present a challenge to accomplish.

The KanMX gene was extracted from the pFA6a-kanMX plasmid (Figure 6a), and primers were designed to add homology regions to the POR2 gene on one end and homology to the AAS gene on the other. The AAS gene was PCR amplified with primers designed to have homology to the KanMX gene on one end and the POR2 gene on the other (Figure 6b). These fragments were transformed into yeast, and we selected for colonies that grow on G418, which the KanMX gene gives resistance to. These colonies should have had the three crossover events to result in the gene fragments inserted into the POR2 part of the genome, knocking out the gene as shown below.

Figure 6a. pFA6a-kanMX Plasmid design

Figure 6b. Model of the triple homologous crossover and simultaneous POR2 knockout in the yeast genome.

Build

Based on our design, the assembly of our recombination would consist of the integration of the KanMX sectioned plasmid and the AAS coding plasmid into the POR2 locus effectively knocking out the expression of POR2. Both plasmid components were first built separately to allow for individual testing, as this allows us to ensure their functionality before combining them into the final homologous construct design.

2.1 Assembly of KanMX sectioned plasmid

Similar to the 2-step single crossover homologous recombination, the extraction of the KanMX gene from the pFA6a-kanMX plasmid consists of plating the Escherichia coli stab culture containing the original plasmid and liquid cultured with ampicillin. The KanMX was obtained from a miniprep performed on the liquid culture used in Cycle 1. This was then followed by a DNA cleanup procedure.

Following the purification procedure, PCR was performed on the isolated plasmid to ensure proper amplification and ample concentration while adding the homologous ends. After ensuring ample concentration, a gel confirmation was performed to verify the length of the pFA6a-kanMX plasmid (1.5kb).

2.2 Assembly of AAS coding plasmid

To ensure the AAS gene fragment isolated from Malus domestica (MdOSC1) was able to successfully crossover with the POR2 homology region, PCR with the necessary forward and reverse primers containing the homologous regions for the POR2 gene was performed. Once more, to ensure the AAS gene fragment was at its accurate length, a gel confirmation was performed as shown in Figure 5c.

2.3 Integration of KanMX sectioned and AAS coding plasmids into POR2 homologous region

Following the gel confirmation and PCR amplification of both AAS and pFA6a-kanMX plasmid, the transformation of the S. cerevisiae was performed through a two-day Lithium acetate (LiAc) method. The first day consisted of prepping the necessary components needed in the transformation mix, such as Lithium Acetate, PEG MW 3350, and Single-Stranded Carrier DNA (Salmon Sperm DNA) which are all essential for the facilitation and uptake of the plasmids by the yeast cells. The cells were then inoculated in the 2XYPAD medium overnight at 30°C on a rotary shaker at 200 rpm in preparation for the main transformation procedure the following day.

Figure 7a. Components of the transformation mix needed for the S. cerevisiae transformation [13]

To verify the growth of the cells, a spectrophotometer was used to determine the optical density at 600 nm. Once the desired cell density was reached, the harvesting of the yeast cells was performed by centrifugation at 3,000 ×g for 5 min and thereafter, washed and resuspended in sterile water. Finally, the cell suspension was transferred into a 1.5 mL microcentrifuge tube and further centrifuged to pellet the cells and remove any supernatant before the transformation process.

To begin the transformation process, the transformation mix prepared in advance is added to the microcentrifuge tubes containing the yeast cells and further vortexed to ensure the cells are appropriately mixed with the transformation media. Afterward, the cells are incubated in a 42 °C water bath to induce heat shock on the yeast cell walls to facilitate the uptake of the plasmids. Following the heat shock, the cells are pelleted again by centrifugation for 30 seconds and any leftover transformation mix is carefully removed to avoid losing any competent cells. Resuspending the cell pellet in 1.0 mL of sterile water, the cells are plated onto the YPD media plates containing the G418 antibiotic selective medium at three different volumes (2, 20, and 200 μL). Lastly, the plates are incubated at 30 °C for 3 - 4 days in preparation for screening and verification.

To verify the transformation of the S. cerevisiae, cells plated on the YPD media containing the G418 antibiotic were screened for resistance. Only cells that have successfully integrated the KanMX gene fragment are capable of surviving due to their resistance to G418. A colony PCR could also be performed by lysing the transformed yeast and using the plasmid released as a template to be PCR and eventually verified through gel confirmation.

Test and Learn

Like the 2-step single crossover, the characterization of the construct will occur through lysis and purification of ursolic acid. Through using chromatography, the ursolic acid can be separated from any other contaminants or non-ursolic acid products to be quantified.

In this cycle, we were not able to produce the full AAS gene through digestion and ligation, similar to Cycle 1. We talked to our graduate student advisors and changed various parameters including EcoRI and T4 ligase concentrations, digest and ligation time, as well as incubation temperatures. We were however unable to obtain the ligated product of AAS1 and AAS2. Similar to Cycle 1, we will construct the AAS fragment through Gibson Assembly into pRS306 and perform PCR with the target primers from there.

WL Cycle 3: Simulation

Design

Our lab was required to move to a different lab space in a newly constructed building, but due to logistics, all wet lab work had to be paused until further notice. To maintain productivity during this time, we performed SnapGene simulations of the wet lab genetic engineering of our parts so that we can ensure success when we regain access to a lab space. At the time the wet lab was paused, both the Cycle 1 and Cycle 2 groups had accomplished successful minipreps of the KanMX and pRS306 plasmids with much troubleshooting of the PCR.

Since we already experimented with different temperatures for the PCR, we decided to build and order new primers for AAS, pRS306, and KanMX. The first part of the simulation involved confirming successful PCR through the virtual gel. Before designing the new primers, we decided to test the old ones for KanMX to see if it could yield a successful result. The products labeled G3 were the Cycle 2 products and the rest were the Cycle 1 products. Interestingly, all of the PCR simulations resulted in the target results with bands indicating the right length.

Figure 8a. Simulation of PCR using Cycle 1 and Cycle 2 primers

Next, the same simulation was performed on the new set of Cycle 1 KanMX and AAS primers. Since we had difficulties ligating the AAS1 and AAS2, we decided to do a three-piece Gibson Assembly, so new primers were designed accordingly. The KanMX primers were designed to target shorter fragments flanking the KanMX gene region to decrease the possibility of off-target sites. For this, three sets of KanMX primers were designed. These new primers were designed to result in around 1.6kb amplicons each instead of the original 4kb. All of the designed primers had the PCR amplicons at the correct lengths, so these primers should work once we return to the lab.

Figure 8b. Simulation of PCR using new Cycle 1 and Cycle 2 primers.

Since the DNA fragment is shorter, we decided to go through a double crossover cassette recombination between the yeast genome and the transformed fragment as opposed to the original single crossover recombination with a larger fragment. This may raise the failure rate due to the nature of double crossovers compared to single crossovers. However, we think that the shorter length of the target fragment would increase PCR efficiency and integration efficiency.

Figure 8c. KanMX insertion into the POR2 coding region of the yeast genome

Build

For the second part of the simulations, we ensured that the designed primers produced PCR products that would be successful for Gibson Assembly. As in the design phase, Gibson Assembly simulations of both the old and new designs were run on SnapGene. As seen below, the PCR products yielded from both the old and new primers should result in the desired Gibson Assembly construct. However, we were unable to successfully ligate AAS1 and AAS2 together leading to the redesigning of the primers to do a three-fragment Gibson Assembly.

Figure 9a. Simulation of new Gibson Assembly on SnapGene

Figure 9b. Simulation of old and new Gibson Assembly with visualization through gel

Test and Learn

Once we return to the lab, we will be performing all of the simulated experiments. Since the simulations confirmed the successful PCR and Gibson Assemblies, everything should go relatively well in the lab. This will allow us to quickly get to the transformation stage so that we can provide the PD team with the target yeast liquid culture for the CSTR. At the same time, it will allow us to further test the yield of UA.

However, the simulations have revealed that not everything that should work in practice would work in the lab. One great example of this is the KanMX PCR that we were troubleshooting in Cycle 1. Even after experimenting with annealing temperature with touchdown PCR, extending elongation time, and adding DMSO, the KanMX PCR was unsuccessful even though it was successful on the SnapGene simulations. Knowing now that this is a possibility, we adjusted best practices to prepare several sets of primers, especially for more complex and longer PCR fragments.

WL Cycle 4: New Lab Space and Future Steps

Design and Build

After our SnapGene simulations, we looked into designing extraction and encapsulation methods after producing ursolic acid in yeast. In regards to extraction, we looked in the literature for potential methods to extract a molecule similar to ursolic acid. We found researchers had developed a signal peptide system where a signal peptide sequence from an extracellular transport protein was integrated into a carrier protein, such as supernatant protein factor (SPF), which binds to squalene. This would essentially transport the drug extracellularly, allowing for the extraction of the bound molecule (4). Since squalene and ursolic acid are both hydrophobic molecules, we looked into the affinity of the lipid-binding domain of SPF (tSPF) and ursolic acid via MolModa, a molecular docking software. We found the Gibbs free energy of the binding event to be negative, conveying that it's spontaneous and indeed would occur. This model gave us confidence in this extraction method working with ursolic acid.

Figure 10a. Plasmid design of the signal peptide extraction of UA utilizing pMV-LEU2 and integrating the Suc2 signal peptide with tSPF to bind to UA for extracellular export.

Next, we considered designing a method to quantify its production in the bioreactor that the PD team was producing. Through a literature review, we found that ursolic acid inhibits, through allosteric regulation, α-glucosidase, an enzyme that hydrolyzes starch and oligosaccharides into glucose (5). Since our bioreactor has a glucose sensor built in, we theorized a system where we would have ursolic acid bind to α-glucosidase and test glucose levels as a proxy to quantify the amount of ursolic acid produced in the bioreactor.

Finally, once the ursolic acid is isolated, encapsulating the ursolic acid in a method that would increase its bioavailability would be the issue at hand. The bioavailability of ursolic acid is a relevant focal point for implementing ursolic acid in clinical settings. Through a literature review and an interview with Dr. Shaoyi Jiang, we compiled three methods for encapsulating ursolic acid to maximize its bioavailability. One method takes advantage of β- or γ-cyclodextrin, oligosaccharides that form a complex wrapping around ursolic acid, increasing its hydrophilic properties and simultaneously having a great affinity to the hydrophobic ursolic acid at its inner core (6). Then, we could encapsulate this inside sEVs, fuse this with a liposome to increase drug loading capacity, and form sEV-liposome fusions with β-cyclodextrin (β-CD) bound to the UA, as observed in Figure 10b.

The second and third methods involve utilizing sEVs and liposomes to encapsulate UA. One technique is to obtain sEVs with biomarkers of interest for specific cellular targeting, then perform electroporation to eliminate its inner contents. Then, liposomes encapsulating ursolic acid can be fused with these sEVs, creating a sEV-liposome hybrid that not only contains ursolic acid but has specificity to the cellular biomarkers we are interested in (7). An alternative technique is to mix polylactic-co-glycolic acid (PLGA), a hydrophobic polymer, with ursolic acid to form nanoparticles via precipitation. Once the PLGA-UA complex is formed, we can mix them with sEVS, and self-assembled sEVs containing the hydrophobic complex will be formed (8). A fascinating feature of PLGA is that based on the amount of PLGA used, we can determine its usage since it is biodegradable, so the dosage can be controlled via the release of the drug.

Figure 10b. A diagram summarizing the UA encapsulated in β-cyclodextrin, followed by the encapsulation of this complex in sEVs, then fused with liposomes to form sEV-liposome hybrid nanoparticles.

To build the pMV-LEU2-Suc2-tSPF construct, we would first design primers to integrate the Suc2 signal peptide sequence with the tSPF sequence via PCR-mediated homologous recombination. Then, we would integrate Suc2-tSPF to pMV-LEU2 through Gibson Assembly, similar to how we inserted the AAS sequence into the pRS306 plasmid as described earlier in Cycle 2. This would proceed with the plasmid being cut at the LEU2 homology site with EcoRI before transformation into yeast. The internal mechanism of yeast would recognize the homologous LEU2 sequence, resulting in a crossover event. This will occur at the nonfunctioning LEU2 gene while inserting the Suc2-tSPF sequence (Figure 11a). LEU2 encodes for β-isopropylmalate dehydrogenase, an enzyme crucial for the biosynthesis of leucine. Thus, a successful transformation in which the LEU2 gene is knocked out would result in bacteria growing in media without leucine.

To perform the transformation of pMV-LEU2 into yeast, primers will be used to linearize the plasmid while adding homologous ends to the ends of Suc2-tSPF to prepare it for Gibson Assembly. PCR of the Suc2-tSPF sequence with primers designed to match the ends of the gene insert with the ends of the linearized pMV-LEU2 will amplify the gene insert and ensure its ends overlap with the ends of the linearized plasmid backbone. This provides the opportunity for Gibson Assembly to be used to assemble the pMV-LEU2 plasmid with the Suc2-tSPF insert without requiring restriction enzymes. This provides the opportunity to move forward with the transformation of the new circular plasmid into yeast.

To test for successful transformation, cells plated on the YPD media containing yeast drop-out medium supplements without leucine will be screened. Only cells that have successfully integrated the Suc2-tSPF gene fragment will not be able to grow on media without leucine. A colony PCR could also be performed by lysing the transformed yeast and using the plasmid released as a template to be PCR and eventually verified through gel confirmation.

Figure 11a. Proposed insertion of signal peptide sequence Suc2 into yeast genome using pMV-LEU2 plasmid, which knocks out the LEU2 coding region upon insertion.

Once the transformation has been tested to be successful, we can transform the plasmid into the yeast cells that have been engineered to produce ursolic acid to test if the extraction is successful. One way to test if the extraction is successful is to measure the glucose levels where the ursolic acid should be excreted since ursolic acid is known to inhibit α-Glucosidase, an enzyme that converts carbohydrates to glucose [9]. When ursolic acid and α-Glucosidase are present in a solution, the ursolic acid will bind to the enzyme via hydrogen bonding, inhibiting its action and causing glucose levels to be lower when complex carbohydrates are added in the solution.

Successfully exporting ursolic acid to the extracellular environment will segue into the next step: encapsulation. Figure 11b below describes the four encapsulation methods we have determined will be the most effective delivery systems for ursolic acid considering its low bioavailability due to its strong hydrophobic properties.

Figure 11b. Pros and Cons of various ursolic acid encapsulation methods for effective delivery.

Test and Learn

By incorporating the tSPF-Suc2 into yeast, we'd expect to see an increase in UA secretion compared to yeast without the signal peptide complex. Prior research achieved a squalene secretion of 225 mg/L by incorporating the tSPF-Suc2 into yeast [4]. Since squalene has a slightly lower binding energy to tSPF, -10.5 kcal/mol, compared to compounds similar to ursolic acid like its precursor, α-amyrin (-6.8 kcal/mol), and its isomer, oleanolic acid (-6.3 kcal/mol), we’d expect a yield around or lower than the one found for squalene.

We expect to encapsulate UA in β-CD through a stirring method involving the solubilization of β-CD in water and an excess of UA. This method shown to have achieved successful encapsulation of UA by β-CD according to research [11], which was confirmed by comparison of differential scanning calorimetry (DSC) curves of the physical mixture of UA and β-CD and its inclusion complexation, in which peaks in the UA curve were found in the UA and β-CD physical mixture curve. Additionally, a virtual screening by [12] measured UA’s binding affinity to be -4.2 kcal/mol, which indicates that β-CD encapsulation would be favorable. Further encapsulation of the UA-β-CD complex in a cancer-derived sEV-liposome hybrid fusion is expected to increase specificity to cancer cell biomarkers, as demonstrated by research by [7] showing that lung cancer cell A549-derived sEV-liposome hybrids displayed a 14-fold higher cellular uptake to A549 cells than non-hybrid sEVs.

Product Development Cycles

PD Overview

The goal of Product Development was to create a continuous stirred-tank bioreactor (CSTR) to produce Ursolic Acid using the genetically engineered yeast from Wet Lab. This will also increase efficiency compared to batch reactors by shortening downtime and providing a more energy efficient model that can correct parameters falling below ideal conditions in real-time to preserve yield.

Furthermore, an electrochemical biosensor was designed with the aim to be integrated with the reactor system for live UA tracking without the need for offline and expensive UA test kits on the outflow of the bioreactor. Tracking the concentration of UA is crucial for assessing reaction efficiency and controlling industrial production yield.

The methodology and modeling of the bioreactor/biosensor can be found on the Modeling page on our website, while this page will detail the DBTL cycles and progress of developing the bioreactor along with initial and future steps on the biosensor.

PD Cycle 1: Initial Design and Gathering Individual Components

Design

After weighing the options and confirming the goal of building a continuous stirred tank reactor, the design started. All the first-round designs showed clear consistency. The reactor generally consists of a chassis, a main reaction chamber, and inflow and outflow reservoirs.

Figure 12: One of the first round design sketches

The main reaction chamber and the reservoirs that require temperature control are placed inside the chassis. A sous vide will be used to control the temperature of the water inside the chassis and thus the internal temperature of the containers. The containers are preferably made of glass to increase thermal conductivity. Different containers are connected through PVC tubes and peristaltic pumps. Several sensors are fixed inside the main reaction container, including a thermometer, a pH meter, a dissolved oxygen sensor, and a glucose sensor. An agitator connected to a motor will also be fixed inside the main reaction container. Finally, an Arduino connected to a computer will control the entire reactor. All circuit boards and some circuitry will be stored next to the chassis, and this part needs to be watertight.

Figure 13: Final Reactor Design

We had more in-depth discussions on some of the details. First, we discussed how many reservoirs we need. It is known that outflow and inflow are necessary. After discussion, we decided to store the engineered yeast and glucose-based nutrients separately to avoid premature reactions that are difficult to control. Then, oxygen will be supplied using a bubbler placed at the bottom of the main reaction container. Secondly, we discussed the outflow position. Although there is an agitator, reaction products may still deposit at the bottom of the container. Our original design had the outflow from the top of the container, but after consideration, we decided during the design phase to have the outflow from the side at the bottom of the reaction container. Another option is to use a pump to overcome the gravity and allow outflow extracted from the bottom area of the reaction container. We finalized the design of outflow during the building.

Figure 14: Final design of the outflow of the bioreactor

Build

During the building phase, we completed the testing of all our pH meters, dissolved oxygen sensors, and thermometers. Additionally, all pumps, including two high-flow pumps and several low-flow pumps, as well as the sous vide, have been tested and are functioning correctly. We have also decided to use a plastic box as the chassis and several glass jars as containers. During testing, none of the three pH meters we had provided accurate readings, so new pH meters were ordered. The remaining sensors worked properly, and we are familiar with the related Arduino setup, with the relevant code already saved.

Figure 15a: Testing the thermometer

In the next steps, once the new pH meters and calibration solutions arrive, we will complete the final testing. Then, we will begin assembling the different parts together. Anticipated challenges include mounting the sous vide. Because the edges of the chassis are not flat, the sous vide clips cannot hold it steadily. Plastic plates are planned for fixing. Additionally, we need to fix the pumps onto the containers, which requires drilling holes in the glass jar lids. Secondly, sterility of the reaction containers needs to be ensured. We will ensure the integrity of the seal for each container, and after building, we also need to test the seal of the entire liquid flow area. Additionally, we will use plastic wrapping to ensure the cleanliness of the internal environment. Finally, a plastic box is prepared for watertight storage of the Arduino boards, but after fixing the circuit boards, we will need to ensure watertight sealing with plastic cover.

Figure 15b: Testing the bubbler and motors

Figure 15c: Running all 5 motors at the same time using Arduino, controlling different speeds

Figure 15d: Rough sketch of the circuit design of the motors

Test

The testing process of our CSTR will incorporate different stages. First, we would check if the CSTR would function overall, meaning, the fluid flows throughout the bioreactor. If this means that we need to make sure that the flow going from the input reservoir to the main tank to the output reservoir is all even so that there are no overflows. This will be mainly determined from the pumps, so we will prioritize the consistency of all the pumps that we use to build our bioreactor. We can also use food dye and add it to the input reservoir, and see if the food dye gets transferred to the output reservoir to test if the water flows through the entire bioreactor. Next, we will test the systematic control of fluid temperature with sous vide, making sure that the temperature remains constant for the reaction to occur continuously. This can be tested easily by checking the sous-vide, as we tested already if the reading on the sous-vide is the same as a thermometer’s reading. To follow that, we will test the pH, oxygen, fluid glucose levels with base/acid inflow and pH meters, a bubbler, and glucose inflow, respectively. This can be done once all the sensors arrive and are tested to show that they read accurate measurements. Once all the components have passed necessary tests, we next can go through a major test to check the continuity of the flow ( ) to make sure that the production does not stop until we manually pause it. We can first test if the bioreactor runs for a day, and then test it longer through proof of battery; this way, we do not have to waste time just observing the bioreactor run for days. We will note any important observations and use them to make adjustments in the learning process.

Learn

From the test results, we will make proper adjustments to different parts of the bioreactor to further maximize the production of UA. This process can go in many different ways as we will change the conditions of the CSTR depending on the result and the materials we have. During this phase, we will mostly work with what we have in our inventory to keep the testing and learning process moving forward. We will also collaborate with Wet Lab to make sure the UA production is stopped while making adjustments to our bioreactor ( ). If we encounter any contamination even despite well sterilizing before, we will halt the bioreactor and remove any feed. We will troubleshoot what caused the contamination and clean the whole bioreactor thoroughly and fix the problem that caused contamination. Once we learn how to fix any errors we find on the way, the cycle will begin again with new designs for an improved bioreactor. If we find more errors, we will repeat the process again with testing and learning until we produce our desired bioreactor.

PD Cycle 2: Combining the Components Together

Design

After completing the first cycle of DBTL, we initiated the second cycle of design based on the experience gained from the first round. The most significant difference in the design phase is related to the pH reservoir, which was not emphasized in the initial design. Starting from our needs, we need to maintain the pH level of the solution inside the reaction container within a range optimizing the yeast growth, which is pH 4-7. As a result, we need the ability to both raise and lower the pH of the solution. Therefore, we need to set up two containers: one for storing a basic solution and another for storing an acidic solution. This introduces an additional container requirement in our originally designed bioreactor, which also implies the need for an extra set of circuits for the motor.

Figure 16a: Bioreactor focusing on the pH reservoirs

Moreover, the placement of each container and its corresponding motor was also determined during this DBTL cycle. Due to considerations of container shape and stability, we chose to place the motors for the glucose reservoir and the first pH buffer solution reservoir externally, for example fixed on the chassis rather than on the container lid. Additionally, since the main container needs to accommodate many sensors and an oxygen bubbler while also ensuring seal ability, numerous holes need to be made in the lid. The original container lid could not meet these requirements, so we decided to 3D print a lid that would satisfy both the need for holes and seal ability.

Figure 16b: Motors mounted on the chassis

Figure 16c: The 3D printed Lid

Lastly, the circuit design did not undergo many changes compared to the first DBTL cycle. The only differences are that we removed the glucose sensor and separated the oxygen sensor from the Arduino system, as we decided to promote the reaction by providing an excess of glucose. Additionally, since the oxygen bubbler itself cannot control the flow rate, we removed the oxygen sensor from the Arduino system, and purchased a sensor that can display readings independently. Regarding the circuit boards, during this cycle, we optimized the positioning of different boards. This optimization facilitates the arrangement of the wiring and connections to various motors. It is worth mentioning that the design and building processes in this cycle were intertwined, with troubleshooting being our primary method for generating new design ideas.

Build

During the building process, encountering issues is inevitable. Troubleshooting became the main theme of our second building cycle. In this cycle, we integrated the entire motor system and sensor system onto a single Arduino board for the first time. Previously, our tests were conducted separately. After merging everything onto one Arduino, the first issue we encountered was that the pH meter could not obtain readings and consistently displayed “NaN.” To solve this problem, we conducted multiple comparative tests, including disconnecting parts of the motor system, using a different breadboard, and eventually replacing the Arduino board. We discovered that the issue was with the Arduino board. However, a new problem quickly arose—the pH meter could not provide reliable readings. It gave a higher reading in acidic solutions than in neutral solutions, which was obviously incorrect. We then updated the pH meter's code and made it display the initial voltage reading on the probe. This allowed us to determine whether the issue was with the sensor probe or the algorithm that converts the voltage readings. Finally, we succeeded in enabling the pH meter to read reasonably accurate values within the pH range of 4-7.

Figure 17a: Calibrating ph meter with buffer solutions

Next, after integrating the motor circuit system and the sensor circuit system, we made improvements to the overall circuit. First, using just one Arduino board, we employed two breadboards to separate the motor and sensor circuits, preventing potential interference. Then, based on the new design, we optimized the positioning of the L298N motor driver board to ensure that the wires connected to the motors had enough space for proper connection. In terms of wiring, we first distinguished different types of wires by color and then updated the wire types to ensure stable connections, as we had encountered sensor reading abnormalities due to poor wiring connections during testing.

Figure 17b: Final circuit

Finally, in terms of coding, we integrated the motor system code with the sensor code. After ensuring they did not interfere with each other, we developed code to maintain solution balance. The key part was controlling the flow of the two pH solutions based on the pH meter’s readings. When the pH was too low, the alkaline solution motor would activate, and vice versa. We also tested the flow rate of the motors—how much liquid they could draw per second—and integrated this value into the code. Ultimately, the code allowed us to maintain data output from the pH meter and thermometer while all motors were operating. Since the oxygen sensor was not integrated into the Arduino system, its readings were displayed separately.

Test

During this period, the primary testing objective is to conduct a dry run on the assembled bioreactor. The first task is to ensure that all the motors and pumps are functioning correctly. Next, we need to verify through testing whether the motors’ speed and rotation direction can respond to the sensor readings while the sensors are operating properly. Specifically, by modifying the code, we adjust the target pH range to simulate how the motors respond under different pH conditions. Ideally, the motors associated with pH1 and pH2 should activate when the pH value falls below or exceeds the target range. During the testing process, we measured the flow rate of each motor, that is, the volume output per unit of time. This allows us to relate the actual flow rate to the values in the code, giving us a better understanding of control conditions.

In addition, we tested the interaction between the Arduino, the motor circuits, and the sensor circuits. On one hand, we ensured that each motor was connected properly and that no coordination issues arose after all parts were integrated. During this process, the sensors needed to continuously provide up-to-date readings. This was not only to maintain control over the reaction environment but also played a decisive role in regulating the flow rates of the different reservoirs.

Figure 18a: Dry Run

After completing the dry run, we proceeded with the wet run using dyed water. The primary goals of the wet run were, first, to ensure that the water flow in the system was normal, that water could successfully flow from different reservoirs into the main chamber, and could be collected through the outflow. In this phase, we mainly checked the operation status of each motor, i.e., the pumps, and the water flow rate. Specifically, we differentiated each reservoir by using colored water and mixed them in the main chamber, with the mixture flowing into the outflow container. This also demonstrated that the bioreactor had the feature of continuous flow.

Secondly, we filled the chassis with water and conducted a sous vide test, bringing the bioreactor to the target testing temperature, which is crucial for controlling reaction conditions. Another key test involved the sensors. We needed to ensure that the sensors could operate correctly and stably in the real working environment, especially the pH meter and temperature sensors, which needed to coordinate with the entire system. In the end, the wet run went smoothly, and we achieved the expected results.

Moving forward, we plan to conduct more wet runs, including additional tests using dyed water. We will need to pay closer attention to the efficiency of pH regulation. Since we didn’t use a pH buffer this time, we may add one in future tests to examine how well the code and motor adjustments control the pH. Additionally, we plan to perform the final tests using yeast culture, which will be a real test of how the bioreactor operates and will help with future industrial applications. Finally, we will continue to optimize the entire bioreactor system, with a greater focus on sterility level, yield, and the overall operational efficiency.

Learn

One challenge we encountered was in pin resource allocation and cable arrangement when connecting all the wiring to a single Arduino. The placement of the three L298N boards used to control the motors had to satisfy the requirement for short connections to the Arduino while also leaving enough space for the cables to reach each motor attached to the reservoirs. Ultimately, we connected the interfaces requiring continuous signal output to the PWM pins, while connecting others to the basic digital pins. The sensors were connected to the analog inputs. This arrangement ensured that the available pin count was sufficient, and the layout was logical.

Figure 19a: Arduino Pins

Additionally, during hardware integration, we realized that motor vibrations could affect structural stability, so we updated the mounting design for some of the motors on the reservoirs. We secured the motors for the pH1 and glucose reservoirs to the chassis to meet our stability requirements. Finally, we optimized some of the wiring and code naming to ensure that after the dry test, the wet test would also proceed smoothly.

Figure 19b: Motors Running

During the wet run, we encountered several troubleshooting issues. First, we discovered that one of the pumps of pH reservoirs wasn’t functioning, but we found this was related to the logic of how we controlled the pH. When the sensor is active, if the pH remains within range, the motor for the pH buffer reservoir does not activate. After temporarily modifying this portion of the code, we successfully got both pumps for the pH reservoirs running. Additionally, we tested the sous vide functionality and its impact on temperature. It worked properly and corresponded with the thermometer readings. Lastly, regarding the flow rate of the entire system, the current outflow rate is quite significant. However, in order to optimize the yeast culture, we may reduce the flow rate in the future to allow the yeast in the main chamber more time to culture properly.

Figure 19c: Wet Run Prep

In the future, we will focus more on interpreting the results of the wet runs. We will also optimize the testing procedures and attempt to standardize them to minimize differences between trials. Moreover, we plan to correlate this with our modeling for validation, which will aid in scaling up. Finally, we want to emphasize that this will be a long-term process of optimization and scale-up, but the current bioreactor has already demonstrated the feasibility of this strategy.

PD Cycle 3: Biosensor Initial Design

Design

The next step to improve our bioreactor was to add a biosensor as a new component. We began to design our biosensor, but we initially began with designing the housing for the sensor to attach to the reactor. Thus, we first considered our initial sketch of the biosensor from earlier.

Figure 20a: Biosensor Sketch

We wanted the housing to be simple and cheap to keep costs and complexity down. As such, we came up with a simple box-based design in Fusion 360. Each hole was sized to commercially-available electrodes we intended to use for the sensor design with the larger one intended for an inlet.

Figure 20b: Biosensor Housing CAD Model

The biosensor chamber would consist of three different electrodes: counting, reference, and working. Using voltmeter and ammeter, and connecting them to Arduino, the biosensor will recognize Ursolic Acid. This is done by tracking the voltage using Ohm's law. Tracking the voltage change using the transducer will help us recognize the UA activity.

Build

For the building portion, we will focus on making the house for the sensor. With our housing design created, we 3D-printed the biosensor housing using our Prusa Mk.4 printer and PLA filament.

Figure 21: Printed Biosensor Housing

We then attached the outflow hose of our bioreactor to the inlet of the housing and began to move fluid through it. The biosensor would fit right into the housing to detect the UA activity, but for now, we will focus on whether or not there are leaks in the housing.

Test and Learn

To test, we would add water to the reactor and let it flow through it, similar to the previous tests that have been done with water and food coloring. The whole point of this test run is to check for any leaks, with the addition of the biosensor added to our device. All of our other components of the bioreactor should work properly as they were tested in the previous cycles. This cycle focuses on the proper housing of the biosensor that has been attached to the bioreactor. Once we take future steps and complete building the biosensor, we would place them into the housing we tested. We would test the functionality of the biosensor with the yeast and all the necessary components needed to get the reaction started. Using our biorecognition model, we would measure the glucose level to determine UA’s activity. If the detection of UA works, the bioreactor and the sensor will be a success. Otherwise, we will test if there is any problem with the biorecognition model and test it outside of the bioreactor.