Engineering Success

Introduction

Our project started with a simple goal in mind: introduce defluorinating enzymes to create an e.coli that can live in fluorinated environments. According to this study(1) , E.coli are susceptible to membrane damage when grown under high concentrations of perfluorooctane sulfate and perfluorooctanoic acid. With this in mind, our efforts focused on the optimization and implementation of various defluorinating enzymes for useage. Our first objective is to establish a comprehenzive look into dehalogenating and defluorinating enzymes, as to develop a more robust understanding of defluorinating pathways.

Initial Design

With amazing guidance from our past igem team members as well as current members of the lab we worked in, we were able to borrow a cloning vector from a previous project. The original plasmid can be seen here(2) . The plasmid is set up for optogenetic control but needed two extraneous BBs1 sites removed in preparation for future Golden Gate assembly with BBsI. In order to resolve this issue, we consulted our advisors and devised an initial strategy.

1. Domesticate the pDusk plasmid to remove the extraneous BBsI sites
2. Insert mScarlet into the domesticated plasmid
3. Use GoldenGate to insert ordered oligonucleotides from IDT
4. Test the survivability of E.coli in fluorinated environments after transforming plasmids

Domestication of pDusk

Given the pDusk plasmid, we first started by identifying the ROI. There were two BBsI sites flanking a 360 bp region that lacked any functionality, so there was no danger in removing them. To tackle the problem, we took the approach described in this paper(3) . We believed this approach to be the best because the region that needed to be removed was too large. As such, we began with two sets of primers, G1.1 and G1.2. These primers face 5' to 3' and 3' to 5', respectively. They attach near the BbsI sites, with a few base pair mismatches such that the BBsI recognition sites would be dirupted. Then, the primers direct the amplification of the region between them with Taq, which results in the 350bp region amplified, with no BBsI sites. To confirm this, we ran the result on a gel, and confirmed the MegaPrimer at ~350bp. Next, we re-attached the megaprimer to the original backbone, as a way of seamlessly replicating our altered BbsI recognition sites back into the plasmid. Finally, we performed a BbsI digest to confirm that there was no more megaprimer, and only the insert of interest remained on the gel. As such, we confirmed, from the two bands on the gel, that there was only two BbsI cuts, and not the original four.

Insertion of mScarlet

After the initial success with MegaWhop, we quickly moved forward with a new strategy. To avoid needing to order a gBlock, we decided to order primers that contained BbsI cut and recognition sites, as well as base pair overhangs with our pDusk backbone(G1) and a separate construct that contains mScarlet. mScarlet is a constitutively fluorescent protein(4) that will help us identify whether bacterial colonies contain our plasmid of interest or not. mScarlet, combined with optogenetic controls on our plasmid ensures that we have a good degree of understanding of our construct without needing to sequence at every step. After our primers (G2.1 and G2.2) arrived from IDT, we moved forward with the amplification step. After an advisor graciously provided us with a construct containing mScarlet, we performed an amplification reaction to gain our insert. The insert amplified not only contained mScarlet, but also contained a constitutive promoter (5 cit. needed) and terminator from the iGEM parts database. Then, we once again used MegaWhop, with mScarlet as our megaprimer this time. To confirm the success of this step, we decided to plate our PCR products. The plate bloomed with beautiful pink colonies, and we picked 8 and sent one in for sequencing. Sequencing confirmed that our mScarlet plasmid was successfully transformed into the bacteria.

First Objective: Generation of a Dehalogenating Enzyme Library

After the extensive literature review and collaborator support, we selected five potential candidates for dehalogenation.

1. G5.1- DeHa1 an enzyme tested by USAFA 2022 iGEM team. This enzyme came from Delfitia acidivorans
2. G5.2- DeHa2, from D.acidivorans.
3. G5.6- Haloacetate dehalogenase H-1, from Moraxella sp. B
4. G5.7- Fluoroacetate dehalogenase from Bukholderia sp.
5. G5.8- Fluoroacetate dehalogenase from Rhodopseudomonas palustris CGA009.

At this point, we had finished construction with half of the project, but we still needed to further confirm that all of our sequences were correct, and that all constructs looked the way we wanted to. Unfortunately, it was discovered that there existed an N-Terminal his tag in our G2 golden gate plasmid, as well as the fact that our ordered defluorinating enzyme inserts had the wrong BbsI cut sites. To rectify this misatake, we designed a set of primer that would replace the BbsI cut sites. Then, we performed an array of PCR reactions with each insert, resulting in "fixed" inserts with the correct BbsI cut sites. Finally, to remove the His tag, we designed a set of primers for inverse PCR; the idea was to anneal on either side of the his tags, amplify the rest of the plasmid, and then re-ligate the plasmid together with the sticky ends the primers held.

On our first attempt at removing the N-terminus his tags, it was revealed that seeing the actual change in the plasmid would be incredibly difficult without sequencing, as the region we are removing is only ~50bp in a 6kb+ plasmid. So, we decided to look at flanking digestion sites, and decided on NheI and XbaI, as the difference in length would be the full 50bp, and the cut out insert woudl only be 50bp long. So, we moved forward with the inverse PCR, and then digested and ran the result on a 2% gel. We did not see any bands at all, and there was self-priming, so this attempt was considered a failure. Moving forward, we decided to re-evaluate and take two different approaches. First, due to a small overhang region as well as a low annealing temperature, we would need to be incredibly careful with our PCR Reaction design. Second, due to the minimal amounts of discernable change, it would be incredibly difficult to determine a "success" vs. a failure. And finally, it seemed that the small digestion insert would be invisible on any gel regardless, so the best idea would just be to send in samples for sequencing. So, we used two different PCR reaction approaches: the same inverse PCR, and a novel technique called Single strand PCR. The idea behind single stranded PCR was to first anneal one primer at a time, and amplify indepdendently. Then, the two reactions would be mixed and annealed together with ligase. From our two approaches, we saw minimal growth from the resulting PCRs, and even saw the growth of bacteria that did not express the mScarlet present in the original backbone. Fortunately, the DNA we sent in returned with positive results, and we were able to successfully remove the his tags.

To fix the plasmid inserts, we first diluted our oredered DNA 10x, and then performed a series of amplification reactions, with the hope of fixing the insert. On our first attempt, we accidentally used the wrong primers. After the PCR amplification step, we ran the results on a gel, which returned a large amount of parimer at the bottom of the gel, which is a sign of self-priming. This means the primers are not landing, so we ran a second PCR, this time with a gradient, in case our annealing temperatures were incorrect. This also proved unsuccessful, so we decided to take a step back and re-evaluate the issue. It turns out that we had been using the primers for removing the His-tags on our inserts incorrectly, and that it was a relatively easy fix. Finally, on our last attempt, we used the correct primers this time for PCR amplification, and ran the results. It was successful, and we saw bands at every expected lenght, with minimal self-priming. The next step was to Gel Purify, so we excise the gel at the right band, and purified the results. Finally, to test if the PCR was susccessful, we digested aliquots of our inserts with BBsI, and ran the result on a gel. This resulted in the expected ~1kb band, as well as two small bands > 50bp. This was a success, so we moved on to cloning in the inserts.

In order to get preliminary defluorination results, G5.1-G5.2:G5.6- G5.8 were plated on 10mM difluorohexanoic acid on plain LB. By plating our constructs on non-antibiotic resistant plates, we hoped to see some level of selective pressure for the E.coli to keep the plasmid. After 3 days of growth there was minimal growth on G5.6-G5.8 and WT plates. We hypothesized a couple potential reasons why - After three days of growth, the WT adapted to a fluorinated environment, the difluorohexanoic acid is not actually toxic to E.coli, or the method in which we applied the saponified difluorohexanoic acid was insufficient to test for viability. A colony PCR of the survivors showed that G5.7 and G5.8 retained the plasmid suggesting that it conferred some advantage, but it's unknown to what degree. A subsequent liquid culture with 1mM of Difluorohexanoic had no growth on the WT but G5.7 and G5.8 had growth. Our next step was a gradient assay that was conducted with varying dilutions of 10mM saponified 2,2-difluorohexanoic acid.

Our preliminary trials were also unsuccessful in determining if there was any kind of toxicity or selective pressure, as all plates showed considerable growth. Our way forward would be to research future methods of media preparation or even move to solely growing in liquid cultures. Analyses with LC-MS, Fluorine-NMR, or GC-MS would have been able to confirm degradation for any of our trials, but it was unfortunate that we did not get to that point.

Potential next steps include adjusting the media preparation by adding the saponified difluorohexanoic acid to a M9 minimal media agar once cooled. We would test various methods of adding the fluorinated compound to media. After that our next steps would have been to transform G5.7 and G5.8 plasmids into electromcompetent G4 cells. We would then test this construct in aerobic and anaerobic conditions with nitrate supplemented M9 minimal media with increasing concentrations of 2,2-difluorohexanoic acid, by doing so we would be able to see if there was metabolization of the difluorohexanoic acid.

Obtaining quantitative data for our defluorination enzymes relied heavily on the planned usage of LC-MS. Our plates would have been prepared according to the LCMS facilities instructions, and sent for sampling. We would send in a plain plate with just LB/10mM difluorohexanoic acid, our wild type plate, and our 5 constructs; G5.1-G5.2, G5.6-G5.8 to analyze the substrate using LCMS.

However, for LCMS to work, it has to be calibrated to detect PFAS. Fortunately for us, Dr. Rittman's lab was also planning to do LCMS assay of PFAS chemicals, but unfortunately for us the facility was not calibrated in time for us to analyze our samples before the project deadline.

Our preliminary trial did offer a positive outlook in that G5.7 and G5.8 had minimal growth on the difluorohexanoic acid LB while the wildtype control had zero growth, inferring that we confered some level of advantage to the construct. Further analyses with LC-MS, Fluorine-NMR, or GC-MS would be able to confirm any degradation.

Our next steps would have been to transform G5.7 and G5.8 plasmids into electromcompetent G4 cells. We would then test this construct in aerobic and anaerobic conditions with nitrate supplemented M9 minimal media with increasing concentrations of 2,2-difluorohexanoic acid. By using minimal media, we restrict the E.coli to only using the fluorinated hexanoic acid as a carbon source. This would enforce selective pressures on bacteria that successfully coupled Beta-Oxidation and defluorination.

From Single Genes to Metabolic Pathways

After finally finishing the construction of the first expression vector; we decided to take a step back and assess our stance. To begin designing constructs to break down PFAS, we must first understand the problems we are facing. With the varied and different forms that PFAS describes, we sought first to understand any commonalities between fluoringated compounds. Through extensive literature review and comprehensive research, we came to realize that the functional groups of PFAS can serve as an "Achilles Heel". From here, one of our contributors created a script that could use structural information available on an EPA Comptox database to determine which functional groups were most common amongst PFAS. Amongst other things, it showed that the most common perfluorinated compounds were either sulfuric acids, carboxylic acids, or aromatics. Due to safety concerns with testing and ordering highly toxic substances, we decided to focus mainly on carboxylic acids. Specifically, we decided on hexanoic acid, and 2,2-difluorohexanoic acid. We chose a less fluorinated hexanoate because literature suggests that the main C-F bond that creates issues with structural strength is the beta carbon in carboxylic chains. As such, we decided on a lower degree of fluorination without sacrificing important structural similarities, as well as taking into consideration the limited funding and the safety of our team.

With a reformed and clear goal in mind, we thankfully collaborated with Dr. Jordan Steele, the PI for the 2021 USAFA iGEM team. In his project, his team identified possible defluorinating enzymes through soil columns that were taken from highly polluted regions, which contained PFAS concentrations over 6ppt. From the results gathered by his team, it was discussed that the N-terminus his-tagging of his selected enzymes may have affected coupling ability. The issue was presented that perhaps the enzymes themselves needed to recouple with each other, and the his tags involuntarily disrupted enzymatic activity. Furthermore, Dr. Jordan Steele proposed that the actual solution to the degradation pathway was more of a microbiome issue rather than independent bacterial species. This coincides with the efforts of the Rittman Lab, and gave us an idea that more of a complete degradation pathway is necessary rather than just focusing on the defluorinating aspect of enzymes. While fluorinated compounds are toxic, we could bolster the pathway by providing more of a metabolic pressure, rather than just selective pressure on our bacteria to survive. As one of our solutions already centered on Beta-oxidation, we decided that the integration of our enzymes with a robust beta oxidation cycle would provide the solution we are looking for. In terms of the first paper we cite for our experimental design, the further goal of having an augmented beta-oxidation cycle allowed us to further understand why PFAS are toxic. It turns out the PFAS's structure is so similar to LCFAS, such that E.coli begins to substitute in fluorinated compounds in the fatty acid metabolism cycles. The toxicity lies in the addition of fluorinated fatty acids into the phospholipid bilayer of the bacteria, which leads to membrane structure degradation, as Fluorines are too electronegative to be in close proximity to other fatty acid chains. The resulting degradation builds up, until eventually the bacteria lyses from a breakdown in membrane structure. With our reformed goal in sight, we dove into literature, and identified the primary target: the chain-length specificity of E.coli. It turns out that, due to slight miscommunication, chemicals had been ordered before experiments had been designed. As such, we were forced to perform experiments on medium and short chain fatty acids, which E.coli is unable to metabolize aerobically. Therefore, in inteest of both bolstering Beta Oxidation efficiency and ciucrumventing the issue of chain-length specificity, we decided to turn to the Beta-oxidation cycle of the Salmonella Enterica Enterica Typhimurium Serovar LT2 strain. E.coli's beta oxidation enzymes are direct homologues of thise in Salmonella, so we decided that the most similar enzymes would net the best results. We expressed genomic Salmonella enzymes, which was done through direct translation from NCBI; converting Amino Acids into nucleotides. This paper(5) directly cites the three three enzymes we chose as major metabolic contributors to the differences between Escherichia coli and Salmonella Enterica beta-oxidation cycle.

Objective 2: Ectopic Metabolic Construct

Due to the necessity of two separate plasmids to express both the metabolic construct as well as the defluorinating enzymes, we needed a second plasmid with a different antibiotic resistance. Thus, we looked towards pRSET-A, as a backbone for our enzymes. We did not order this plasmid, and instead opted to try to clone into a pre-existing backbone. The one available to us from our PI was pRSET-iGluSNFR, which included a ~4kb insert. To begin, we noted the enzyme cut sites around the plasmid, and decided to digest with HindIII and NdeI, which left us without the 4kb insert and a 1kb backbone. Unfortunately, there was an issue with the design. When we ordered our enzyme inserts, we did not realize that the inserts themselves were ~2kb, whereas our backbone was only 1kb. Originally, the plan with the inserts was to ligate them together and then MegaWHOP them into the backbone. With the large base pair length discrepancy, we could no longer use MegaWHOP, and we have to turn to Gibson assembly. Fortunately, the inserts were designed with 50bp overhangs, which allowed for a specific annealing location both between the inserts and between the inserts and the backbone. Finally, after plating and sequencing our results, we were able to insert and express the plasmid in our bacteria.

Anaerobic Expression

After much deliberation, we decided to move forward with the consideration that one or more of the above may be true. So, our solution was to address all of the issues at the same time. First, we decided to saponify the hexanoic acid with NP-40. Second, we remade the m9 plates with ATCC trace minerals. Third, we circumvented the expression issues with a different approach: Using an Anaerobic Chamber.

So, it turns out that E.Coli has two Beta-oxidation cycles: Aerobic and anaerobic. According to this paper (6), the E.coli anaerobic beta-oxidative cycle is better suited for short-to medium chained fatty acids. The enzyme responsible, fadK is less regulated by fadR, and, serves an identical role to that of fadD, which is the acyl-coA synthetase. In anaerobic conditions, the regulative pressure of fadR decreases, and more allows for the processing and usage of fatty acids. This helps us circumvent most of these issues, and reinvigorated our efforts moving forward. To simulate an anaerobic environment, we used a rubber-sealed plastic container that had gas intake valves. We first added oxygen absorption packs to the bottom, along with an oxygen indicator. Then, we loaded the plates, and sealed the container with some petroleum jelly. Then, we pumped the container full of nitrogen to flush out any remaining oxygen. Finally, the container is carefully transferred to an incubator.

After this section of the project, it was well understood that our ectopic Metabolic Chassis conferred heightened metabolic levels, as seen from the preliminary data collected from the hexanoic plates. However, the bacteria we tested the chassis on still contained endogenous expression of metabilc enzymes, and we sought to create a fully controlled and domesticated bacteria, such that its only possible choice is to subsist completely off of 2,2-difluorohexanoic acid.

Objective 3: Genomic Engineering of E.Coli

There are two more enzymes associated with anaerobic beta-oxidation in E.Coli. They are known as fadI and fadJ. These two enzymes are homologues of fadBA, and serve a similar function, just anaerobically. They share a connection with fadE, in that their metabolites produced are still useable by fadE. So, when we insert our plasmid into bacteria, and then grow them anaerobically, we essentially have two sets of enzymes that could be potentially metabolically active. To circumvent this, we decided to begin designing a KO plasmid that would reduce or remove expression of fadIJ entirely. After some genomic investigation through NIH and BioCyc, it was determined that fadIJ are expressed polycistronically- that is, they share one singular promoter. fadJ is a mere 7 base pairs downstream from fadI, so it was understood that, to knock out both genes, it would be best to disrupt the promoter region of fadI. Furthermore, even if only fadI was knocked out, because fadIJ only functions as a tetramer, there would be no functionality to the remaining expressed fadJ. The rationale is to force E.Coli to be dependent on our ectopic metabolic pathway, and completely remove E.Coli's ability to independently express the Beta-Oxidation cycle.

Due to the cycle nearing its end, as well as budget constraints, we decided to test untreaded waters and engineer a novel gene knockout format. Usually, knock out plasmids are designed with incompatible origins of replication to the target bacteria, and cloned into a different species. Due to our limitations, we decided to forgoe an origin of replication entirely. Theoretically, the only necessary conditions provided by the secondary bacteria is a repaired plasmid from any nicks in PCR reactions, as well as higher and easier replicate counts due to the origin of replication. To combat this, we designed a plasmid, containing: a 600bp homologous region, two constitutive terminators, and an antibiotic resistance gene. The homologous region is to enable recombination. The two terminators are to prevent transcription of downstream endogenous genes. Finally, antibiotic resistance serves as a selection parameter to test if recombination occured.

To assemble this plasmid and ensure a high yield of DNA from minimal PCR reactions, we assmbled our own circularized DNA from two 1kb fragments using PCR and ligation. These were ordered off of TWIST, each with two regions of homology to the other insert. The design set the regions of homology in the recombination site and the anitbiotic resistance, as to avoid affecting the expression of the promoter/terminator regions. Along with these two inserts, we also ordered primers corresponding to each, as to amplify the inserts. The two inserts from twist are called G6.1 and G6.2, collectively combining to be G6.

Troubleshooting process

---

Following this, we decided to take a step back and try to optimize the whole procedure. From the start, the bands were incredibly faint, so there was not much hope to begin with. Then, we failed to run gels at every PCR step, as to check the progress of our insert. The lack of caution in this assembly proved fatal to the results, so we came out with this set of conclusions:

1. The first amplification reaction should be done at 25µL, with 50 cycles, to ensure more optimal amplification reactions.
2. Gels should be run after every possible PCR step.
3. The antibiotic we chose doesn't actually kill cells, but only inhibits growth. This means:
   1. We should NOT outgrowth overnight
   2. We should set up more controls to test if the media is effective at all
   3. There could be a possible issue with the plates we used.
4. A 1:1 molar ratio SHOULD NOT be used; we were creating long "chains" of our inserts, up to 30kb+ long.

Second Attempt

---

After the new stance we gained from last time's failures, we decided to continue using the leftover amplification reaction, as there was nothing inherently wrong with it. This time, we decided to not use psuedo-MEGAWHOP, but use Gibson Assembly to perform fold ratio calculations. We ran the assembly, but then failed to run a gel afterwards due to time constraints. We then transformed the results on the same day, this time with the right plates and controls. Unfortunately, there was no growth on any plates, except for the positive control. Seeing this, we decided to run a gel on the Assembly Reaction in post. We saw a band slightly above 2kb, as circularized DNA runs a bit higher than linear fragments. This suggests that the PCR reaction is working, but somehow the transformation is not working. It seems asthough the cells aren't dying from the transformation procedure, yet our DNA was not entering the bacteria either. This is a conundrum, and we chalked it up to some kind of unfortunate luck.

Third Attempt

---

We ran out of the original amplification reaction, so we decide to re-amplify, this time using multiple 25µL reactions, each running for 50 cycles. This resulted in an amazing gel result, and we excised the DNA and gel purified. Them, we decided on performing two reactions, not just assembly or MegaWhop.

1. Assembly Reaction: 1:9 molar ratio
2. psuedo-MEGAWHOP: 1:9 molar ratio, 50 cycles. (Unpublished technique Flores 2023)

This time, we learned from past mistakes, and ran the result on a gel. We saw faint bands for both reactions at around 2kb, so we determined this was a successful PCR step. Then, we transformed the G6 Product, this time splitting up the outgrowthed media into a very specific set of aliquots.

1. MegaWhop and Assembly: From the 500µL outgrowth, we split it into:
   1. 200µL onto two antibiotic resistance plates
   2. 200µL into a 4mL antibiotic resistance liquid culture
   3. 100µL into 4mL of LB only liquid culture tube (Tranform positive control)


2. Controls: WT BL21 cells, split into:
   1. 100µL onto an LB only plate: served both as a positive control as well as a plate for making WT glycerol stocks
   2. 100µL onto a LB antibiotic plate: negative control
   3. 200µL LB-tet liquid culture: negative control
   4. 100µL into 4mL of LB only liquid culture tube (Tranform positive control)

This control split assay would help us determine useful information about our transformation without having to backtrack and use up more time to perform troubleshooting, and we hoped to prevent any failures.

From this experiment, a very intriguing result occured. The positive controls all worked, but all of the negative controls did not. While we determine that the tranformation is not killing the cells, it still remains that, despite perfect DNA going in, nothing is coming out. However, even more interesting, one liquid culture succeeded. We believed that this was a successful KO recombination, so we decided to replate parts of that liquid culture onto antibiotic resistance plates, as well as perform colony PCRs on the liquid culture.

Unfortunately, the colony PCR showed no annealing, with large amounts of background and inconclusive results. We surmised that recombination must be a messy procedure, and decided that we needed a WT control for the next colony PCR. After running a second colony PCR with two controls (WT and Water), we saw that all the noise was in fact background, and we have not succeeded in any transformation or recombination events.

Troubleshoot 2

---

At this point, we felt incredibly discouraged, and could not understand what could be going wrong with our procedures. "good" DNA goes in, but nothing comes out? There must be a different issue at hand. It was briefly discussed whether this part of the project should be given up, but fortunately, our PI Dr. Benjamin Bartelle was able to spot this issue. In the benchling, it turns out that we had designed the insert with a promoter both missing basepairs as well as facing the wrong direction. This was integral, as this meant that antibiotic resistance would never be expressed, unless our construct underwent extreme mutations. As such, it was quickly decided that we needed to design new primers to inverse PCR in a new promoter, as well as removing the old promoter.

New Understandings

Early into the cycle, we, as a team, were overconfident and short-sighted, focusing too much on the defluorinating aspect of the paper read and too little on the actual information being shown. Through multiple experimental cycles and building three separate constructs, we as a team have thoroughly demonstrated the design, build, test, learn cycle asked of iGEM teams. For each and every portion of our project, we made serious decisions and consolidated the failures of our naivety with the new information that was gained from each trial we faced. Through it all, we are proud to display all of the knowledge we have gained and hope that our difficulties can prove to uplift the efforts of future iGEM teams. If given more time, we would definitely continue pursuing the first preliminary results. This would have used the Dehalogenating Enzyme Library from our first onjective, the ectopic Metabolic Cassette of our second objective, as well as direct genomic engineering from our third objective. After gaining a comprhensive understanding of the metabolic activities of our bacteria, we would expand into the development of and creation of a device that could be utilized in bioremediation efforts. Although we hold regrets about this year's cycle, we are overall hopeful and incredibly thankful for this opportunity to show our efforts. For more information on how each experiment was performed, visit the protocols page.