Engineering Success

Engineering Process

---

The engineering cycle — design, build, test, learn — is imperative when aiming to engineer biology and to characterize any biological component of interest. This year, our team’s project aimed to develop a proof-of-concept to support bacterial remediation of the 1,4-dioxane plume in Ann Arbor. Our goal use case is to integrate our modified bacteria into existing infrastructure to improve the current water treatment process's sustainability, efficiency, and efficacy. To characterize the degradative capacity of our bacteria, we conducted extensive testing and optimization of our biodegradative and quantification systems. Accordingly, our team employed the engineering process throughout our project to achieve an optimally functional product.

THFMO Construct Formation

---

Goal: Clone THFMO gene sequence into our backbone plasmid.

Cycle 1

Design and Build

We planned to use the Gibson reaction to clone IDT-ordered sections of the THFMO gene sequence into our pJN105 plasmid (a gentamicin resistance-conferring plasmid backbone shown to work well with P. putida). We acquired the pJN105 plasmid from the Bardwell Lab with a NicA2 gene inserted, which P. putida had used to degrade nicotine. We ordered the THFMO gene sequence in three gene blocks from IDT, using IDT resources to find the optimal split points within the gene and include overlaps for the Gibson reaction. We used the NEBuilder Tool to create primers for the Gibson reaction, aiming to have overhangs of 20 nucleotides. We planned to excise the NicA2 gene with the restriction enzymes XbaI and EcoRI-HF.

Test

To test whether the Gibson reaction was successful, we transformed the Gibson reaction plasmid product into E. Coli using heat shock. In our first Gibson attempt, colonies did not grow. We can see this in Figure 1.

Learn

Because colonies did not grow on our first Gibson transformation we decided to revise our protocol to optimize the Gibson product formation.

Cycle 2

Design and Build

To optimize the Gibson reaction, we analyzed potential factors which could contribute to reduced yield. To reduce variability, we used a PCR cleanup kit to purify our PCR products before ligation. This would remove any leftover components of the PCR reaction (polymerase, dNTPs), giving a purified PCR product. Moreover, we noticed that the three THFMO DNA fragments in the Gibson reaction were not added at the optimal 1:2 ratio with the backbone insert due to calculation errors. Thus, we recalculated the correct concentration of fragments to add to the reaction mixture and performed a Gibson assembly again.

Test

After utilizing PCR cleanup and optimal primer concentrations, we observed colony growth after heat-shocking E. coli DH5α cells with the pJN105-THFMO plasmid and plating them on a gentamicin agar plate, as seen in Figure 2.

This indicates one of three possible results:

1. Our Gibson reaction successfully ligated the 3 THFMO fragments and ligated the construct into our plasmid.
2. A portion of the original plasmid’s NicA2 gene was not completely digested prior to the ligation and this was transformed into our DH5α E. coli, transferring their gentamicin resistance despite an unsuccessful Gibson reaction.
3. The NicA2 fragments that were excised were relegated into the plasmid backbone before this plasmid was transformed into the E. coli.

To verify the success of our Gibson reaction, we conducted a screening of the colonies which grew. This entailed growing a 5 mL culture of each of nine separate colonies from the heat shock and then purifying the plasmid for gel electrophoresis. Since the THFMO gene is 4.3 kb, the THFMO-containing plasmid should be noticeably larger and thus run slower on an agar gel. If the Gibson reaction was successful, the new plasmid would be ~10.3 kb, whereas the original plasmid was 7.5 kb. From the gel run, we determined three candidates for sequencing due to visible plasmid banding at approximately 10.3 kb as seen in Figure 3.

Upon observing that each plasmid migrated slower than the original pJN105-NicA2 plasmid and displayed variable positions on the gel, we decided to evaluate the first three mini-prepped plasmids further. These samples were located in Lanes 3, 4, and 5 of Figure 3. To achieve this, we amplified their THFMO inserts using PCR. We began by using 1 μL of each plasmid as a template. For the amplification, we employed the forward Gibson primer of the THFMO fragment 1 (Part BBa_K5196002) and the reverse Gibson primer of the THFMO fragment 3 (Part BBa_K5196004). Once PCR was complete, we ran these amplified products on a gel, which yielded the following gel in Figure 3.

While the second plasmid in Figure 4 seems to lack at least one of the fragments of THFMO (later confirmed to be the second fragment), the first and third plasmids have inserts between 4 and 5 kb, exactly as expected. Knowing this, we confirmed the unaltered insertion of the THFMO sequence by Sanger sequencing both plasmids from lanes 1 and 3.

Learn

Based on the gel and sequencing of our Gibson products, we had complete confidence in the success of our reaction. We were able to confirm that while the plasmid from lane 3 of Figure 3 had an extra duplication of an overlap region of the first THFMO fragment, the plasmid from lane 5 of Figure 3 had the correct, unaltered sequence (100.0% identity). As we confirmed the integrity of the THFMO gene, the arabinose-inducible promoter, and the gentamicin resistance gene, we decided to use this product for our future degradation testing after using electroporation to transform the plasmid into Pseudomonas putida S16.

GC-FID/GC-MS Quantification

After designing our construct and running degradation testing, developing an efficient quantification system was the next essential step toward determining degradative efficacy.

Quantification Cycle 1

Design and Build

In our initial literature review into 1,4-dioxane quantification, we encountered two systems, GC-FID and GC-MS, suitable for determining 1,4-dioxane concentration. GC-MS was found to have a lower detection limit than GC-FID. However, due to accessibility and budgetary concerns regarding working with a GC-MS machine, the relatively simpler protocol associated with GC-FID, and because we worked with dioxane concentrations well over the FID limit of detection, we decided to focus first on developing an applicable GC-FID protocol in conjunction with the Manz lab ^1,2.

Due to issues with water miscibility and similar volatility, we needed to determine a method to separate 1,4-dioxane in our samples from water. We decided to use a frozen liquid-liquid extraction method developed by Li et al. to shift 1,4-dioxane into a methylene chloride (DCM) organic phase for more straightforward GC-FID quantification ¹. We used our own GC-FID parameters, based on the mass spectrometry parameters developed by Li et al. to develop our own working GC-FID protocol with the assistance of Rachel Klein and Dr. Katherine Manz ¹.

Test

The first parameter we needed to test was the elution point of 1,4-dioxane in the GC-FID. For our initial experiment, we used an extended sample elution period summarized below to verify that we were eluting all sample components and to observe which peak was 1,4-dioxane. Instead of conducting a Liquid-Liquid Extraction (LLE) for our calibration experiment, we opted to use varying volumes of 1,4-dioxane standard (methanol solvent) in 500 uL of methylene chloride. Table 1 summarizes these parameters and purposes.

The results of the initial run of methylene chloride, methanol, and varying volumes of standards (of concentration 100 μg/mL) are shown in Figure 5.

We were able to create a standard curve from the data we obtained, with a R² of 0.9883, as seen in Figure 6.

Learn

From our initial run, we characterized the elution of 1,4-dioxane in an environment analogous to our LLE product. We determined that the 1,4-dioxane elution peak occurs at 15.5 minutes during our protocol. Based on the results of the initial run, we determined that the lower limit of detection for GC-FID analysis was 4.76 ppm (25 μL standard + 500 μL) as the 10 μL standard + 500 μL methylene chloride standard (1.9 ppm) did not show a 15.5-minute peak. As determined by EGLE (the Michigan Department of Environment, Great Lakes, and Energy), this value is almost 1000 times higher than the Michigan residential 1,4-dioxane drinking water cleanup criteria of less than 7.2 ppb³. Thus, we decided to divert our efforts toward a higher fidelity method. We decided to pursue GC-MS based on the lower limit of 1.6 ppb attained by Li et al. in their work (2011).

Quantification Cycle 2

Design and Build

The first parameter we needed to test was the elution point of 1,4-dioxane in the GC-MS. It is important to note that because GC-MS is a quantification method with a more accurate analysis process, we tested the elution point and lower detection limits of authentic and internal standards, or 1,4-dioxane and 1,4-dioxane-d8. We expect that deuterated compounds will have a higher mass and, thus, a quicker elution time.

For our initial experiment, we used an extended sample elution period summarized below to verify that we were eluting all sample components, which were 1,4-dioxane and 1,4-dioxane-d8. Instead of conducting a Liquid-Liquid Extraction (LLE) for our calibration experiment, we opted to use varying volumes of authentic and internal standards (methanol solvent) in 100 uL of methylene chloride. Table 2 summarizes these parameters and purposes.

We also set a mass range for SIM (selective ion monitoring) of 25 to 300 g/mol. This will capture the authentic and deuterated dioxane. The results of the initial run of methylene chloride, methanol, and varying volumes of standards (of concentration 100 μg/mL) are shown in Figure 7.

Test

Learn

We noticed that the methylene chloride peak overlapped with the dioxane peaks, which were eluting much earlier than expected. We also noticed that the ramping rate was too slow and thus the peaks were quite flattened.

Design and Build

GC-MS Parameters for Second/Final Calibration

We noticed that the dioxane peaks were eluting around 4.82 minutes, with the authentic standards eluting slightly earlier than the internal standards. As such, we decided to hold the oven temperature at 35°C for 4 minutes rather than 5 minutes in order to elute out the methylene chloride first. Then, after implementing a solvent delay to ignore the methylene chloride, we ramped the temperature by 35°C/min instead of 20°C/min for the next 3 minutes before ramping at 50°C/min for the next 2.7 min and holding for another minute. Table 3 displays this information.

Test

Figure 8 shows the results of the second run of methylene chloride and standards. The authentic standard eluted at around 4.83 minutes, while the internal standard, being 8 g/mol heavier, eluted in around 4.81 minutes. While the internal standard peak areas were larger in this second run, the authentic standard peak areas were still much larger in comparison to the internal standard peak areas.

Learn

From our initial GC-MS run, we were not able to characterize the elution of 1,4-dioxane and 1,4-dioxane-d8 due to various factors such as a long initial hold as well as overlap with the methylene chloride peak in an environment analogous to that of our LLE product. After adjusting our methodology by shortening the initial hold by a minute and increasing the ramping rate, we determined that the 1,4-dioxane elution peak occurs at 4.83 minutes and the 1,4-dioxane-d8 elution peak occurs at 4.81 minutes. Based on the results of the initial run, we determined that the lower limit of detection for GC-MS analysis was at least 17.62 ppb, as we were able to detect a peak with our 0.2 μM authentic and internal standards. This value is much closer to the Michigan residential 1,4-dioxane drinking water cleanup criteria of less than 7.2 ppb as determined by EGLE ³. As we were trying to degrade much higher concentrations of dioxane to the ppb range, we were satisfied with the fidelity of this analytic approach and to move forward with analyzing our samples with GC-MS.

Design and Build

Optimizing a protocol came with a few setbacks that required the application of engineering principles to overcome. Initially, a sample preparation protocol was prepared as follows:

1. Collect 0.2 mL media sample, and filter it through a PTFE 0.2 μm, pre-slit septum green cap filter.
2. Transfer a 147.5 μL aliquot of the filtered sample to a clean Agilent screw cap 1.5 mL glass vial by pipette.
3. Add 2.5 μL of a 1:50 methylene chloride-diluted internal standard 1,4-dioxane-d8 of concentration 40 mg/L into the glass vials for a final concentration of 0.667 mg/L.
4. Add 150 μL of methylene chloride to this glass vial and let it sit on the counter for 5 minutes.
5. Place in a -80°C freezer inclined at an angle of 45° from the horizon to reduce the potential of cracking of the vials once the water freezes. Freeze at −80°C for 20 min, so that only the water phase is frozen, but not the methylene chloride solvent with the extracted dioxane.
6. Remove the liquid phase (~150 μL) of methylene chloride solvent and place it into a 250 μL glass insert in a fresh screw top vial. The extract was then ready for GC-MS processing and stored at 4°C until analysis.

When we ran a select number of samples after following this methodology, we obtained the following data, visualized in Figures 9 and 10.

The initial protocol was the source of a few issues with the obtained data. For one, the internal standard peak area counts should be consistent (less than 10% variation), but we can see that they range from nearly 4,000 to 26,000; this is much too large of a range to yield accurate results and is likely the result of an inaccurate LLE procedure as well as spiking internal standards individually rather than combining that step with adding the methylene chloride to reduce variation. Moreover, we can see peak trailing in the 1,4-dioxane panels, likely due to a large sample concentration. This warrants diluting the original media sample by a factor of 30x. This also solves the issue of a high AS:IS ratio, which we ideally want closer to 1 than 50.

Test

Following the aforementioned results and potential solutions, we decided to implement the currently used protocol:

1. Collect 0.2 mL media sample, and add 10 μL of this sample to 290 μL DI water.
2. Filter 200 μL of this 30x dilution through a PTFE 0.2 μm, pre-slit septum green cap filter.
3. Transfer a 150μL aliquot of the filtered sample to a clean Agilent screw cap 1.5 mL glass vial by pipette.
4. Add an equal volume (150 μL) of extraction solvent (methylene chloride spiked with an internal standard 1,4-dioxane-d8 of concentration 1.333 mg/L) into the glass vials.
5. Pipette this mixture up and down 5 times.
6. Place in a -80°C freezer inclined at an angle of 45° from the horizon to reduce the potential of cracking of the vials once the water freezes. Freeze at −80°C for 20 min, so that only the water phase is frozen, but not the methylene chloride solvent with the extracted dioxane.
7. Remove the liquid phase (~150 μL) of methylene chloride solvent and place it into a 250 μL glass insert in a fresh screw top vial. The extract was then ready for GC-MS processing and stored at −20°C until analysis.

This yielded the following results for three samples, as visualized in Figure 11.

These results gave us confidence that our method had been finalized, as the AS:IS ratio is much closer to 1, peak trailing was vastly reduced, and internal standard peak areas were much more consistent. As such, we decided to proceed with a full sample analysis, as seen in Figures 12-16.

Learn

As we can see from Figure 11, we were able to get equivalent peak integration area counts for internal peaks, meaning our new protocol was successful in standardizing the internal standard area counts. We also observed better AS/IS ratios and less peak trailing in the authentic dioxane panels.

Figure 12 shows the end result of our iGEM season: P. putida S16 transformed with THFMO and induced with arabinose can substantially degrade 1,4-dioxane. We can see from Figure 13 that without arabinose induction, degradation of 1,4-dioxane is minimal meaning the araBAD promoter expression is not “leaky.” Figure 14 shows that P. putida S16 does not naturally uptake (or degrade, as expected) 1,4-dioxane to a significant extent. Figure 15 confirms that 1,4-dioxane evaporation is minimal.

Figure 16 shows a comparison of one biological replicate selected from each condition. As we can see, while the control replicates showed minimal drops in 1,4-dioxane concentration, the arabinose-induced P. putida S16 replicate degraded ~55.471% of the 1,4-dioxane in 10 hours, and ~58.592% of the 1,4-dioxane in 35 hours.