The Engineering Cycle
The Design-Build-Test-Learn (DBTL) cycle of biological engineering outlines the workflow necessary for solving a specific problem within the realm of synthetic biology. This may be the optimal production of a biological molecule, the detection of a specific biomarker with high fidelity, or the optimization of proteins via the incorporation of noncanonical amino acids. The cyclical nature of DBTL describes its iterability for optimization. Each of the steps within the process has its own purpose and is described below:
Design: The process of creating and formulating the experiments needed to achieve the desired result. This step includes the definition of the problem, selection of hosts and pathways, designing any necessary modifications to the host, and modeling, for this greatly informs the design approaches at all levels.
Build: The first step of experimentation where modifications are made to the organisms of interest and assembly begins. This step includes the assembly of all parts (ex. synthetic DNA, pathways, etc.), the combinatorial assembly of components, and quality control mechanisms such as sequencing at multiple steps throughout the building process.
Test: Testing the engineered organisms for the desired modifications. This step may include the screening of various colonies or organisms for the highest expression of a pathway, or other analytical methods such as various chromatographic assays and mass spectrometry.
Learn: The analysis of data produced from the previous steps, specifically the testing step. This analysis provides the information necessary for the improvement of all systems and allows for more robust design and improvements to future experiments.
Iteration One
Design
Menopause is a condition linked to diminishing levels of estrogen that affects women as they enter their mid-forties. Symptoms exist on a spectrum and are manageable for many, but some women suffer from severe symptoms including intense hot flashes and cramping. For those who need treatment, HRT is the most common option and has been shown to effectively lessen the symptoms they experience. However, the main components of HRT – estrogen and progesterone – have been linked to an increased risk for developing conditions such as breast cancer, heart disease, and blood clots when taken in the amounts necessary for symptom relief. Our aim is to provide an alternative treatment utilizing (S)-equol, a phytoestrogen derived from soy. The first step of the design process was conducting a genomic analysis of (S)-equol-producing bacteria. Sequences for the (S)-equol biosynthesis pathway were found by preliminarily selecting bacteria that are found in the human gut microbiome that also contain all four genes of the pathway. Further literature mining was performed to eliminate the organisms that did not produce a significant amount of (S)-equol. The final three species were Adlercreutzia equolifaciens, Lactococcus garvieae, and Slackia isoflavoniconvertens. These sequences were downloaded from the NCBI GenBank database and cross-checked for slight variations in the sequences to allow for downstream testing of the daidzein-equol conversion rate. These sequences were then uploaded to [cad-sge.com](http://cad-sge.com) for codon-optimization and generation of a full operon. The lengths of the optimized genes in the biosynthesis pathway ranged from ~600 to ~2000 base pairs. Thus, to allow for efficient sequencing they were broken up into eBlocks ranging from 300 to 900 base pairs in length. A total of seven eBlocks were designed for each pathway: one eBlock for the DRC gene, one for the DDR gene, two for the TDR gene, and three for the DZR gene (the exact length of each eBlock varied among both genes and pathways). Additionally, the sequences were virtually assembled into a cargo plasmid via the insertion of the optimized operon into a specialized plasmid backbone known as pCargo. This backbone contains a kanamycin kinase to confer resistance to the kanamycin antibiotic, a specialized T7 promoter for downstream transcription applications, and the PhiC31 integrase sequence for the delivery of the operon into the hosts. This virtual assembly allowed for the design of primers for both assembly via Transformation-Associated Recombination (TAR) in yeast and for the screening of junctions to ensure that the plasmid was properly assembled.As the sequence of the cargo plasmid was designed, hosts were selected and modified in parallel. Four host organisms were selected for various purposes. The first is a simple E. coli strain and was selected as a positive control to show that both (S)-equol production via our methodology does work and to provide a comparison between our attempts and previous studies that have been done expressing (S)-equol in low amounts in E. coli. The second host was E. coli Nissle (EcN), as it is a strain of E. coli that has been used as a probiotic and therapeutic agent for over a century, allowing for both ease of study and a potential downstream application. The third host selected was L. casei BL23, as it is also been studied and used extensively for its probiotic properties and favorable safety profile. (Additionally, this strain, in particular, has been shown to not produce (S)-equol with conventional biotechnological methods, and we hope to show that it indeed can, via the cross-kingdom expression of synthetic genetic elements.) The fourth and final host selected was P. putida, as it has the potential for industrial bioproduction of (S)-equol if probiotic applications prove infeasible. For the proper insertion of the cargo plasmid, these strains needed to integrate the v43 landing pad as described by Patel et al$^2$. Fortunately, both E. coli and P. putida have been shown to express the v43 landing pad in previous experiments by our lab, thus stocks were available to us. This meant that we needed to integrate the landing pad only into EcN and L. casei BL23. This integration was carried out via conjugation for EcN in which strains of E. coli already expressing the landing pad were used as donors, and formed a sex pilus with the Nissel cells that did not contain the landing pad. For *L. casei* BL23, electroporation methods were used instead in which cells were made electrocompetent, then shocked, allowing the landing pad to enter and potentially be expressed. In both cases, GFP screening was conducted, as the v43 landing pad expresses a GFP nanoLuc reporter which is eventually replaced with the synthetic operon; thus, by measuring the expression of GFP, we could determine which colonies have the highest expression of the landing pad. The following images outline the data obtained from the GFP screening for EcN as an example. The seven screens with the highest GFP expression were selected for the build step.
Build
First and foremost, entry vectors needed to be assembled from the eBlocks with one gene per entry vector. Chemically competent cells, sourced from NEB, were transformed with the unmodified pGGAselect plasmid utilizing NEBuilder HiFi, following instructions provided by the manufacturer; the only modification made was increasing the amount of assembly product from 2 μl to 5 μl to ensure as many cells as possible would uptake the plasmid. Cells were spread onto chloramphenicol-containing selection plates (since the pGGAselect plasmid backbone contains a chloramphenicol acetyltransferase gene, thus conferring chloramphenicol resistance) and incubated overnight at 37°C. After growth on the selective media was shown, colonies were picked and inoculated into liquid chloramphenicol-containing media and allowed to incubate at 37°C within a shaking incubator overnight.
Plasmids were extracted from the inoculation utilizing a QIAprep Spin Miniprep Kit. Purified pGGAselect was then linearized using a BamHI-HiFi restriction enzyme digest. The results of the digest were analyzed on a 1.2% agarose SybrSafe gel. Gel extraction and purification was carried out with a QIAquick Gel Extraction Kit and followed the instructions provided by the manufacturer to obtain a purified solution of linearized pGGAselect backbone.
eBlocks were amplified by combining 1 ng of DNA with 7.5 ng of each primer for the eBlock and enough water to form a 25 μl PCR reaction. Reactions were placed into the thermocycler set to an extension time of one minute per cycle for 35 cycles. PCR products were run on a 1% agarose ethidium bromide gel to validate the lengths. For samples that did not display the proper band length, a gradient PCR was run using various lower annealing temperatures. After all PCR products were verified, PCR purification was done using a QIAquick PCR Purification Kit.
Two-, three-, four-part Gibson assemblies were performed using NEBuilder HiFi Assembly Master Mix and protocols which followed instructions provided by the manufacturer. Plasmids were then transformed into NEB chemically competent cells utilizing the same methodology as in the transformation of pGGAselect. Colonies were grown on chloramphenicol-containing selective plates and incubated overnight. Utilizing Kappa2G Fast PCR Master Mix sourced from Sigma Aldrich, colony PCR was run on the products of the incubation. Products of the colony PCR were run on a 1% agarose ethidium bromide gel to validate the length of the insert — and thus the length of the entry vector — to ensure that it was as predicted. After validation, PCR products were sent to Quintara Biosciences for Sanger Sequencing. Sequencing data were aligned against the virtually created constructs to view if any nonsynonymous mutations occurred during the PCR amplification. Only those without such mutations could be used in further assemblies.
After entry vectors were sequence verified, the next step was to carry out TAR assemblies and clone the cargo plasmid containing the synthetic operon into *S. cerevisiae*. The pCargo backbone was sourced from NEB. To linearize the plasmid for insertion into *S. cerevisiae*, a double digest using BSAI and SBFI was carried out. The now-linearized plasmid was run on a 1% agarose ethidium bromide gel to validate the length of the cut plasmid.
Fresh strains of *S. cerevisiae* sourced from NEB were grown overnight in YPAD media. Some of this culture was back diluted into 10 mL of YPAD and was allowed to incubate in a shaking incubator for 4 hours at 30°C. The cells were centrifuged and washed in 10 mL 0.1 M LiAc before repeating the centrifugation. Cells were then resuspended in 300 μl 0.1 M LiAc and allowed to incubate without shaking for 10 minutes. No greater than 20 μl total of the pCargo vector and four gene inserts (amplified from the entry vectors utilizing primers with long overhangs) suspended in water were added to 100 μg of freshly boiled SSS carrier DNA. This solution was then added to 100 μl of the yeast cells. A master mix containing 70 μl 1M LiAc, 70 μl 10x TE, and 560 μl 50% PEG 3350 (pH 7.5) was made, then added to the previously cellular solution. The product was left to incubate without shaking for 30 minutes at 30°C before adding 85 μl DMSO. After thorough mixing, the cells were heat shocked for 7 minutes at 47°C within a water bath. Cells were allowed to remain at room temperature for 3 minutes before harvesting via centrifugation at 5,000 rpm for 1 minute. The supernatant was aspirated, and cells were resuspended in 1ml of sterile water. These steps of centrifugation and resuspension were repeated twice more, lowering the conditions to 3,000 rpm for 10 seconds. After the final wash, cells were plated on non-selective plates and allowed to incubate overnight. These plates, after displaying colony growth, were stamped onto plates containing carbenicillin. Colonies were verified by colony PCR to validate both that the entire biosynthesis pathway was incorporated into the cargo plasmid and to ensure that the junctions were as predicted by utilizing both flanking and junction primers. This was then sent to Plasmidsaurus for full plasmid sequencing.
Unfortunately, various nonsynonymous mutations occurred in all the constructs sent for sequencing, appearing in different locations within the plasmid. Thus, various restriction digests were carried out on three different constructs to cut out locations containing a nonsynonymous mutation and to ligate the associated correct sequence from other plasmids. This “cut and paste” methodology produced plasmids with the full, unaltered optimized sequence for (*S*)-equol production for each of the selected species listed in the design section.
Finally, the plasmids needed to be integrated into the landing pad within each of the host organisms. This was done by conjugation, a process of horizontal gene transfer between bacterial strains. To accomplish this, we used two strains. The donor strain is auxotrophic upon the amino acid Diaminopimelic acid (DAP) and carries chloramphenicol resistance the through target plasmid while the recipient strain carries no antibiotic resistance and no auxotrophy.
We grew both strains to equivalent optical densities and suspended them into the same LB+DAP solution. We then plated this solution onto conjugation filters placed onto LB+DAP plates and allowed for incubation for an hour at 30°C. During this time, the donor strain transferred the plasmid to the recipient strain, creating a new transconjugant strain which carries chloramphenicol resistance and but is not auxotrophic. We then suspended these colonies onto LB and plated this solution onto LB+CAM plates. The donor cannot survive on these plates due to the lack of DAP, and the recipient cannot survive due to the presence of CAM, so only the transconjugant strain survives.
Through this protocol, we are able to confidently select for transconjugant host organisms with the landing pad integrated into them.
Test
There are two main methods for testing whether our approach worked - the first is theoretical validation via sequence verification and the second is empirical validation using liquid chromatography-mass spectrometry (LC-MS).
The former relied on data from all points mentioned thus far along with sequence verification via full plasmid sequencing. For the heterologous hosts to produce (*S*)-equol, it needs to have successfully integrated the landing pad and it must express a correct plasmid. The former is proven by the GFP assays of the design step, and the latter is proven via sequencing data. We can then run theoretical validation by checking off all of the necessary boxes for the expression of our plasmid:
- Has the landing pad successfully been integrated into the host cell?
- If so, which colonies that were screened had the highest GFP expression?
- Does the plasmid contain all of the necessary components?
- Gene encoding for T7 DNA Polymerase
- Specialized origin of replication
- Unmutated DNA integrase
- Unmutated biosynthesis pathway
- Kanamycin resistance gene
- Is the host cell able to uptake the substrate (daidzein)?
All of the above points have been validated by the data already collected. The landing pad was shown to have successful integration into the hosts, the plasmid was - after some additional cleanup work - shown to have the correct pathway with all necessary elements, and previous studies in the literature have shown that at least *E. coli* cells are able to uptake daidzein from growth media without complication. As this project followed methodology from various papers that have been proven effective, it is believed that expression will occur.
Additionally, LC-MS can be used to experimentally verify these results. Unfortunately, due to a lack of time and resources, the LC-MS assays could not be completed before the “Wiki Freeze” deadline, but we hope to have the results in before the Jamboree. We will use standard protocols in which daidzein will be added to the cell culture and allowed to incubate for the suggested amount of time - we will be conducting this step with *E. coli* to ensure quality and to compare to the previous (*S*)-equol production outlined in the literature.
This protocol was adapted from Gaya et al. 2016, who successfully quantified (S)-equol production in a bacterium from the order Lactobacillales$*^3$*. Because we wish to quantify this product in EcN, some modifications were made. In particular, 10mL of culture supplemented with daidzein will be grown to saturation before centrifuging at 5,000g for 5 minutes. The supernatant will then be removed for use in all downstream steps, as it can be assumed that (*S*)-equol diffuses out of the cells during centrifugation. Then, organic extraction will be performed a total of four times - twice with 2mL diethyl ether and twice with 2mL ethyl acetate. When this step is complete, we can then proceed with the evaporation of solvents. This can occur at room temperature using a stream of air. A residue will be left and must be dissolved in a 300µL 50:50 v/v methanol/water solution. This can then be filtered through a 0.22-µm cellulose acetate filter and transferred to an LC-MS vial. The analytical conditions for LC-MS analysis are based on a protocol described by Dueñas et al. 2009$*^4$* and used by Gaya et al. 2016.
Learn
At the time of writing this document, we are still in the process of preparing/conducting LC-MS analysis on the bacteria with the integrated landing pad and synthetic genetic element provided by the pCargo-Equol plasmid. Even so, we are able to learn from the steps described in the preceding sections. Firstly, it was found that the different pathways analyzed only contained about 85% similarity between amino acid sequences. Though the only plasmid that was completed at this time was that for *A. equolifaciens*, we would like to test the (*S*)-equol production between all three aforementioned species to determine how the differences at the amino acid level alter the production of (*S*)-equol. This would provide insight into which proteins from each pathway are most efficient and may allow for kinetic assays on each of the proteins in each of the pathways. This could, in turn, set the stage for combinatorial analysis of enzymes from multiple species, It could also set the stage for protein engineering and further modifications based on the trends of the kinetic assays for further optimization of this biosynthesis pathway.
Additionally, it was clear to us early in the process that something was inhibiting the rate of (*S)*-equol production and that the rate of the reaction could be further optimized. Thus, we combed through our results and returned to the literature, finding a paper by Lee et al. that described the introduction of a mutation within one of the genes of the *S. isoflavonconvertens* pathway$*^5*$. We decided to implement this mutation as well — in-parallel with Iteration 1 — in hopes of creating an even further optimized pathway for our second round of the DBTL cycle.
Iteration Two: Introduction of the P212A Mutation into the S. isoflavonconvertens Pathway.
Design
After further mining the literature, we found that the DDR gene in the (S)-equol biosynthesis pathway can be activated one step too early, converting the (R)-dihydrodaidzein into (3R,4R)-cis-tetrahydrodaidzein before it can be properly racemized into (S)-dihydrodaidzein (Lee et al., 2016). (3R,4R)-cis-tetrahydrodaidzein cannot be converted into (S)-equol and thus takes up both time and space that could be used for the synthesis of the proper molecule. Thus, a P212A mutation, as described by Lee et al., 2016, was introduced into the DDR gene for S. isoflavonconvertens. Primers were designed to incorporate one C to G mutation at the 212th amino acid in the sequence, converting it from proline to alanine, and were ordered from IDT. Using the purified plasmid containing the DDR gene for this pathway and these new primers, the entire plasmid was replicated with this site-specific mutation. This would be the only alteration to allow for proper comparison between the wild-type pathway and the P212A mutant.
Build
Using the wild-type DDR plasmid as a template, the specialized site-directed mutagenesis (SDM) primers were used to introduce this single-base mutation. After a simple PCR reaction and purification protocol, Sanger sequencing was conducted on the insert portion of the product to ensure that it had only the desired mutation. When this construct was verified, TAR assembly was carried out using the mutated DDR entry vector and the wildtype entry vectors for the other genes of the pathway. All other building steps were identical to that for the wild-type pathway, except for using the mutated DDR entry vector instead of the wild-type vector.
Test
The testing procedure was the same as Iteration 1, including both theoretical analysis and LC-MS quantitative tests. Unfortunately, neither of these tests could be fully completed as the P212A mutation’s complete pCargo-Equol plasmid could not be assembled with the current time and resource limitations. However, it is something that we would very much like to pursue in the future to try and confirm the results from Lee et al. - namely, that the introduction of the P212A mutant does not increase the overall yield of (S)-equol, but does increase the rate at which daidzein is converted into the final concentration of (S)-equol.
Learn
If the Lee et al. paper’s results can be confirmed, it can be inferred that the DDR step is not the rate-limiting step as previously thought. It is possible that (S)-equol bind to the DZR enzyme at the beginning of the pathway, though this is not currently known, nor is it known if the binding is competitive or allosteric. Understanding if this is the case could lead to further protein engineering and the possibility of a greater final yield of (S)-equol. This inspired us to design one more engineering cycle - one for future experiments, of course - to determine the binding affinities of (S)-equol and daidzein in the DZR enzyme. If we can determine both affinities and the mechanisms of binding, we can then continue with the protein engineering proposed in Iteration 1 to optimize the pathway even further, potentially even preventing inhibition completely.
Iteration Three: Future Experiments for Pathway Optimization
Design
The results from Iteration Two confirmed that DDR is not the rate-limiting step; this, combined with other evidence from the literature, enabled us to conclude that the first reaction in the pathway, catalyzed by DZR, is inhibited by (S)-equol in a negative feedback loop that greatly limits the amount of (*S*)-equol produced. Thus, we decided to purify the DZR protein and design biochemical assays to collect kinetic data that would characterize this inhibition. These data would allow us to determine the mechanism of inhibition and would inform future protein engineering efforts.
We began by creating three different plasmids containing the DZR gene in a pET28-a(+) vector backbone: one with a C-Terminal 6xHistidine tag, one with an N-Terminal 6xHistidine tag, and one with a N-Terminal 6xHistidine tag and a thrombin site to allow for eventual cleavage of said tag. When expressed in bacteria, these each produced the DZR enzyme with a histidine tag, allowing for high-fidelity protein purification via immobilized metal affinity chromatography (IMAC) on a Nickel-NTA resin.
We intend to use the successfully purified protein to conduct enzyme assays: a 96-well plate will be used to prepare an array of 200 µL reactions containing varying concentrations of the substrate (daidzein) and the inhibitor ((S)-equol) along with the purified enzyme at a final concentration of 0.5 µM. LC-MS will then be run on the plates to quantify the production of dihydrodaidzein. Results will be used to generate a Lineweaver-Burk plot to determine whether the mechanism of inhibition is competitive, non-competitive, or uncompetitive; this will inform the feasibility of rational protein engineering aimed at eliminating the negative feedback mechanism and optimizing production. If inhibition proves to be competitive, then active site engineering will likely be necessary to remove the feedback mechanism; hence, successfully executing a rational protein engineering strategy will be difficult if not impossible. If, however, inhibition proves to be non-competitive or uncompetitive, then molecular modeling software, like Schrödinger, can be employed to identify potential allosteric binding sites. These sites can then be modified via SDM to create a feedback-resistant DZR mutant capable of driving yield much higher than is currently possible with the feedback-inhibited pathway.
Other images to add once you finish engineering:
