Engineering

From DNA to impact, we engineer solutions for our world.

Overview

Design

Use Engineering principles to describe a biological system in an intended model, to improve our initial design, or its posterior modifications, favoring the product development. It must consider what the system should or should not do, and what's vital to the successful product usage.
Build

Describe the process that we previously proposed, highlighting its qualities and requirements. This stage usually includes model considerations, and prototype simulation to pre-aboard testing.
Test

Test the design built to compare results with the expected outcome simulated.
Learn

Identify the main issues and discrepancies between the expected results and the real evidence, denoting the bottleneck and focusing on the main mistakes.
Reflection

Closes the cycle by applying the knowlede acquired through our process.

This page takes you along AureoBos’ journey by capturing the trial and error occurred during the project and how we solved the challenges that came up through iterations of the engineering cycle. By applying this method, we integrate all the stakeholders’ suggestions to obtain an intended product that responds to the world's necessities.

Any engineering project involves things that do not go right the first time, and while this can be daunting, you can learn and adjust details to make it work based on mistakes or blind spots. Considering the iGEM Engineering Cycle and its four phases, our product was developed by three great cycles, ten semi-cycles, and 13 iterations. Also,we decided to complement our stages by adding semi-cycles and a fifth stage: Reflection, which closes the cycle by representing the way to apply the knowledge acquired during all the process, associating it with a complete new iteration cycle. It formulates a new question to address, along with the corresponding method to resolve it. In the same way, it makes possible the extrapolation of knowledge to prevent mistakes.

Main cycles

Click "Go to section" to change between sections

Protein
Induction

Golden Gate Assembly

First sub-cycle:

Second sub-cycle

Third sub-cycle:

First Cycle: DNA Assembly

Our product is composed of three fusion-endolysins: LysCSA13, LysK, and LysSS. The aim of the Golden Gate Assembly cycle was to obtain our expression constructs.

First sub-cycle: Chemocompetence and transformation

Click on the accordions to see details about each iteration

1st Iteration

Design

Through a sponsorship we received Escherichia coli 5-alpha competent cells. After using them all we performed a bacterial reactivation to conserve the cell strain, which was very useful for us in the Golden Gate Assembly phase. We performed a chemocompetent cell’s production to transform our E. coli strains (5-alpha, and BL21(DE3)) for the storage of our constructs and its posterior expression. After inoculating the Escherichia coli 5-alpha strain in a LB agar plate, its competent ability got lost. To use them in the transformation we followed the protocols for chemically competent cells and transformation used by Tec-Chihuahua 2022.

Build

An individual E. coli 5-alpha colony was used to inoculate LB culture medium in order to do the protocols of forming a cell pellet. To continue with the transformation process, we made 50 uL aliquots to add 5 µL of the Golden Gate Assembly plasmid DNA, and mixed by inverting 5 times. After ending with the protocol, we incubated overnight at 37°C.

Test

Figure 1. First iteration’s transformation results.

The E. coli 5-alpha chemically competent cells were transformed with a control plasmid, pUC19, to prove that there was a good transformation efficiency. But nothing meaningful has grown. In the same way, the reaching of the optical density was much longer than that we expected, about nine times more than the 30 minutes described in the protocol. We knew that cells survived the heat shock of the transformation process, because our competent cell control showed cell growth, but we couldn’t know if the complication was in the competent cell protocol or in the transformation one.

Learn

The efficiency of transformation was very low. We assumed that it might be due to rough manipulation of the cells, that the CaCl2 for the cells was not very cold and that the OD was barely between the appropriate range of 0.6 to 0.8. Besides, we considered that could be related to the heat shock temperature, because of the poor water bath thermostat definition.

Reflection

We have evaluated the learning stage, consequently we identify that the apparent issues were about the cell’s manipulation. The new question was related to the temperature required in the process, and how we could be completely assured of their values. In the same way, we have evaluated the incubation times, where we decided to increase their lapses. We conducted an in deep-analysis of the entire process, and discovered that the shaking step wasn't as efficient as it should be. To solve these problems, we decided to analyze the original protocols and suggestions of the NEB website, and review troubleshooting data sheets.

2nd Iteration

Design

Considering our past observations, we found that some of the most common complications in the cell transformation are related to the heat shock temperature, in which an unstable control of the ranges hinder the process. In the same way, the cold temperatures are very important in the cell competent protocol, including the material and consumables used. We also considered the option of using our -80°C stored cryovials.

Build

To perform our second iteration we modified the chemocompetent protocol described in the Experiments Page of the Tec-Chihuahua 2022 Team, assuring that all the equipment required was at 4°C, and all the material and consumables were previously put on ice. In the same way, we have ensured that our water bath was at 42°C by double-checking the temperature with an additional termometer. Finally, we increased the incubation time up to 2 hours, realizing that Eppendorfs position was actually allowing the culture movement.

Test

Our results were increased in comparison to our first iteration, but they were still lower in comparison between NEB cell competent colony forming units. At this time we have positive results to our transformation protocol as well as our cell competent control, but no meaningful result has been achieved for the cryovials samples.

Learn

We actually increased the colony forming units of our transformation protocol, in which we assumed that was related to the assuring of the temperature of the water bath in the transformation protocol, as well as the constancy in the cold temperature of the entire chemocompetent cell process. We also achieved that zero transformation efficiency was obtained from the cryovials previously made, hence we considered that it is crucial to perform a new chemocompetent cells protocol by the time that transformation is required. Although we improved our results by using the Engineering cycle, we considered that these results were not good enough to try with our Golden Gate Assembly construct.

Reflection

We realized that by using the Engineering Cycle was possible to upgrade our results in the number of colony forming units of our transformation protocol, however, we weren’t safe enough to take the risk trying cells' transformation with our Golden Gate Assembly construct, because is well-know that transforming with a generic positive control achieve better yields that which are obtained with other samples. To continue improving our results we analyzed every step of the previously mentioned protocols, evaluating the possibility to increase the cell concentration plated over the agar plate without risking the absorption of bacteria, besides considering the optimization of the times’ process. The next question was related to the optimization of our process, and the solution was planned to appear by investigating the past iGEM Teams’ protocols.

3rd Iteration

Design

Evaluating the reflection of the past iteration we found the following options to optimize the efficiency and time of the process. By increasing the available gas exchange surface we could shorten the period to reach optimal density, thanks to its effect on accelerating cell duplication. While it is possible to increase the efficiency of the transformation by considering the concentration of cells plated on the agar medium, where it is possible to increase the total number of cells by harvesting the pellet of transforming cells, and resuspending it in a smaller volume.

Build

To perform our third iteration we modified the chemocompetent protocol described in the Experiments Page of the Tec-Chihuahua 2022 Team, applying the changes described in the second iteration, and using a 250 mL Erlenmeyer flask instead a 100 mL one. In addition, its iteration includes the harvest of the total transforming cells incubated for two hours, and resuspending it into 100 uL SOC medium, to plated the total volume into preheated plates.

Test

Figure 2. Third iteration’s transformation results.

Our results were up to the NEB cell competent efficiency, where the positive pUC-19 control reached uncountable colony forming units. At the same time, the lapse to achieve the optical density required for the competent cell protocol was shortened to ninety minutes.

Learn

We reached the NEB cell competent efficiency by incorporating all the modifications which optimized the process, it was good enough to transform our Golden Gate Assembly constructs. Our final protocols to the chemocompetent cell’s production, and heat-shock transformation are available in our Experiments Page.

Success

We actually achieved good enough results, and obtained the standardization of our protocols to continue with the process!

Second sub-cycle: Golden Gate Assembly

Three fusion-endolysins are the nucleus of AureBos, our non-antibiotic treatment against the mainly pathogenic bacteria which causes bovine mastitis: Staphylococcus aureus, Streptococcus agalactie, Streptococcus uberis, and Escherichia coli. The biologic principle behind our product has basis in the composition of the endolysins in which we conducted an in-deep analysis to optimize the original capacities of these enzymes; eventually we synthesized our new codifying sequences. All the endolysins principle is acutely described in our Parts Page. As a brief description each fusion-endolysin is composed by three mainly domains: CHAP, in which lies the catalytic activity that makes possible the lysate of specific pathogenic bacteria; the albumin binding domain (ABD) which increase up to 30 hours the endolysin’s lifespan; and finally the SH3 domain, that gave to AureoBos the biological safety through the ability to recognize specific bonds in the peptidoglycan layer of the previously described bacterium. Attending the stakeholder’s suggestions, we expanded the effect spectrum by the addition of a third endolysin which incorporates a polycationic nonapeptide, and the Cecropin A peptide, both have abilities to destabilize the gram-negative bacterial cell wall. You could know more about this incorporation in our Human Practices Page.

Click on each accordion to see details

1st Iteration

Design

To obtain our constructs we designed the codifying sequences for each endolysin as a result of an extensive bibliographic search. After an in-depth research, we designed the endolysins that will be expressed, LysCSA13 and LysK, with some modifications: an ABD and the cell-wall binding domain (CBD) of SH3B30. The ABD increases the in vivo half-life from 20 minutes to 30 hours² and the CBD makes the endolysin specific for Staphylococcus aureus, Streptococcus agalactiae and Streptococcus uberis.^{3, 4} A third endolysin (PCNP-CecA-LysSS) was added as the way to incorporate the stakeholders’ suggestions. Both enzymes were modeled in AlphaFold 2, to verify the proper folding of the protein. In the case of the LysK endolysin, the cleavage site was not necessary since the addition of the histidine tag did not interfere with the folding of the enzyme. The sequences were optimized for E. coli and synthesized as a transcriptional unit with a sponsorship from TWIST. The biobricks were selected and assembled in silico using Snapgene. For the constructions of both enzymes, the same parts were used: promoter (BBa_J435350), RBS (BBa_Z0262), each coding sequence, histidine tag, terminator (BBa_J435371) and backbone (BBa_J435330) were assembled by Golden Gate (Figure 3).
Figure 3. Expressions cassettes for the endolysins LysCSA13-ABD (a), LysK-ABD-SH3B30 (b). and PCNP-CecA-LysSS (c)

Build

To perform the Golden Gate Assembly cloning method we followed the NEB’s protocol, attending all the advice from the reagent’s trading company. All the process is described in our Experiments Page. We resuspended the DNA of the backbone from the plate with nuclease-free water. The DNA was transformed into E. coli DH5α from NEB and extracted with the Miniprep protocol. When we received the DNA of the transcriptional units from TWIST, we resuspended and stored it. The Golden Gate assembly was performed with enzyme BsaI, the backbone, and the synthesized transcriptional unit. Three mixes of the reaction were made, one for each transcriptional unit, LysK, LysCSA13, and LysSS. A negative control was made for each reaction. Posterior to the transformation with our constructs, we planed the plasmid DNA extraction resulting of a Miniprep reaction, and finally, evaluating the plasmid concentration, plasmid topology, and weight to compare with our in silico digestions, to assured that we had our endolysin constructs.

Test

The constructs were transformed into E. coli 5-alpha by heat shock and inoculated in kanamycin plates. We obtained individual colonies in every plate, including the negative controls, in which nothing should have grown.

Learn

When analyzing our results, we remember that (Figure 4) the chosen backbone contains two resistances to antibiotics (kanamycin and ampicillin), after being cut in the Golden Gate reaction, it only retains the resistance to kanamycin.1 However, We realized that there was no way to have 100% efficiency in the cloning stage, and considering that we didn't add any fluorescence or color reporter genes, we were in a hard situation. In other words, we could not differentiate which colonies were transformed with our endolysin constructs, and which colonies were transformed with the original backbone with both resistances.

Figure 4. Backbone after Golden Gate assembly.

Reflection

Evaluating the situation, we planned to identify our constructs by using all the information about our current plasmids. The new challenge was to find the way to counteract the reporter gene absence, with the knowledge of having two resistance genes in the undigested acceptor backbone, and only one resistance in our interest plasmids. We also confirmed that the transformation efficiency was very positive.

2nd Iteration

Design

After reflecting on our mistakes, we realized that it was possible to identify endolysin constructs by inoculating the same colony onto kanamycin and ampicillin plates. We reached almost 90 transformed colonies and considered 25 colonies to be significant to increase the chances of choosing the right transformant.

Build

We chose 25 different colonies from every endolysin to inoculate in kanamycin and ampicillin plates. Later, we designed the work plan in which we divided each plate into four quadrants, in the same way, all the labeling was made to duplicate, one by each antibiotic. After all the experiment was correctly labeled, we performed four different cross-striations by each plate, identifying every colony by a number, which was the same for each antibiotic. Finally, we incubated all the plates at 37°C for 18 hours. To confirm that our endolysin constructions were in our plates, we expected that the growth would appear only in the kanamycin plates, but nothing in the ampicillin ones. While the growth on both of the plates would reveal an acceptor backbone transformant.

Test

We showed bacterial growth in both plates for the majority of the colonies, but some of the colony samples appeared only in the kanamycin plates (Figure 5).

Figure 5. LB agar plates with kanamycin (50 g/mL) from the transformation of the LysK construct. A. Transformed E. coli 5-alpha with the plasmid of LysK (quadrant 16). B. Transformed E. coli 5-alpha with the plasmid of LysK (quadrant 18).

Figure 6 LB agar plates from the transformation of the LysK construct. A. LB agar with ampicillin (100 g/mL) inoculated with transformed E. coli 5-alpha with the plasmid of LysK (quadrant 18). B. LB agar with kanamycin (50 g/mL) inoculated with transformed E. coli 5-alpha with the plasmid of LysK (quadrant 18).

Learn

We found our endolysin constructs! The antibiotic screening made possible the identification of our endolysin constructions, confirmed by the growth of the transformed bacteria only in the kanamycin plates. We also have growth of some of the colonies in both of the plates, which corresponded to the acceptor backbone transformants. Any of the colonies grow only in the ampicillin plates, which coincided with our expectations.

Reflection

We assumed the Golden Gate reaction was done correctly, due to the fact that growth was only present on the kanamycin agar plates for some of the replated colonies. Otherwise, it would have grown in the presence of both antibiotics, kanamycin and ampicillin, meaning that the backbone was still intact. When performing a ligation reaction of any kind, the backbone can get auto-ligated. This occurs because of the ligases that are not specific to a site, unlike enzymes, which make them very prone to ligate the same backbone without the insert. In a first instance, there is nothing to tell us that this did not happen without doing other types of tests 5. We inferred that some colonies grew on the negative control because the backbone ligated itself without the transcriptional unit; this was verified by the antibiotic selection mentioned above, confirming that the colonies had the DNA of the backbone with both antibiotic resistances. It’s very important to consider the addition of reporter genes for the construction of expression plasmids, however, it is more important to evaluate the challenging situations by the Engineering cycle, taking into account all the information that already exists. In the same way, we conclude that the antibiotic screening is not enough to ensure the presence of our endolysins constructs. At this point, our next step is to perform a Miniprep reaction of the supposed transformants to obtain its plasmids and analyze them by an enzymatic digestion on an agarose gel.

Third sub-cycle: Golden Gate Assembly

The next sub-cycle had the objective to demonstrate our constructs identity to be completely assured that we were obtaining the endolysins expression constructs, before performing the protein inductions. In the same way, we had to extract the plasmids to their posterior heat shock transformation in the E. coli BL21 (DE3) strain, which is feasible after the standardization of the protocols as we did in the previous semi-cycles.

Click on each accordion to see details

1st Iteration

Design

To continue with the experimentation, we planned to perform a plasmid DNA extraction for each of the three endolysins’ constructs that compounds the AureBos treatment. Performing a plasmid DNA extraction was possible after following the Miniprep PureLinkTm Quick protocol, which is also included in the Experiments Page. In the same way, an agarose electrophoresis was required for its semi-cycle.

Build

The building up of this stage was prepared considering all the steps described above, but now taking special care of all the temperature steps, as we reflect in our previous semi-cycle iterations. Our samples were the E. coli 5-alpha apparently transformed with each of the fusion-endolysins constructs inoculums, incubated at 37°C at the maximum shaking parameters, for 24 h. For the plasmid analysis we performed an in silico digestion on the SnapGene software, taking into account our fusion-ensolysins constructs, and the specific enzyme cuts: SmaI. Our plasmids have a length of: 2997 bp (LysK), 3159 bp (LysCSA13), 2892 bp (LysSS). The in silico results demonstrated the expected length for each linearization on a 1% agarose electrophoresis. In Figure 7 is shown the expected results for LysK and LysCSA13, in which were also included the acceptor backbone control, and the undigested samples for both, acceptor backbone, and LyK construct.

Figure 7. In silico agarose gel (1%) of the results from the enzymatic digestion with SmaI. MW. 1 kb DNA Ladder 1. Digested backbone (3893 bp). Ladder 2. Negative control reaction of the backbone (3893 bp). Ladder 3. Digested plasmid with LysK (2997 bp). Ladder 4. Negative control reaction of the plasmid with LysK (2997 bp). Ladder 5. Digested plasmid with LysCSA13 (3159 bp). Ladder 6. Negative control reaction of the plasmid with LysCSA13 (3159 bp).

Test

After the plasmids digestion, we performed an 1% agarose gel electrophoresis as is described in our Experiments Page, to analyze the plasmid identity of the LysK and LysCSA13 fusion-endolysin constructs. Our samples were performed as Figure 8 reveals.

Figure 8. In silico agarose gel (1%) of the results from the enzymatic digestion with SmaI. Ladder 1. Quick-Load Purple 1 kb Plus DNA. Ladder 2. Digested backbone above 4000 bp (expected band size: 3893 bp). Ladder 3. Negative control reaction of the backbone (3893 bp). Ladder 4. Digested plasmid with LysK (2997 bp). Ladder 5. Negative control reaction of the plasmid with LysK, between 2000 and 3000 bp (expected band size: 2997bp). Ladder 6. Digested plasmid with LysCSA13, two bands at 1200 bp and between 1500 and 2000 bp (expected band size: 3159 bp). Ladder 7. Negative control reaction of the plasmid with LysCSA13, between 2000 and 3000 bp (expected band size: 3159 bp).

Learn

We realized that the gel's appearance was not the best, not only because of the undesired spots, but also because the sample bands were not aligned to the DNA marker. In addition, an unexpected double digestion was revealed for Ladder 6. After analyzing the entire experimentation, we found that everything was according to the protocol, the DNA sequences were correctly reviewed, and no problem was found in the in silico results. Evaluating the situation, we concluded that there was a problem with the reagents because of their expiration date. To confirm our hypothesis, we repeated the experiment, but nothing changed.

Reflection

To apply the learning from this iteration, we realized that some of our lab reagents were not in good enough condition to be used for experimentation. Initially, we thought that this situation would apply to the agarose, the staining gel’s SYBR Safe, but after looking for troubleshooting sheets, we realized that the unexpected gel stains could be an incorrect manipulation while preparing it. In the same way, we evaluated the possibility of having star activity as a consequence of rough manipulation, but it could only occur during long incubation periods, which was not the case. Our new question was related to finding a way to correctly evaluate the plasmids' identity with a high level of confidence. To solve this problem, we consider the option of using other enzymes and take into account the troubleshooting data sheets for obtaining better agarose gels.

2nd Iteration

Design

After reflecting on our learnings we intended an analysis of the most common troubleshooting data sheets related to the plasmid identification techniques, in which we realized that we could improve our results by taking care of the agarose dissolving steps. In the same way, we recognized another enzyme currently in our lab, which also produces plasmid linearizations, hence we decided to repeat the experiment by using EcoRI for LysCSA13 and LysSS instead of SmaI, which corresponded to the failed results. Finally, our new ladder arrived just in time to try again, considering that some of the possible causes of the non alignment between the DNA ladder and our samples could be because of the previous uses of this shared lab reagent.

Build

To perform this iteration we followed the current protocol described in our Experiments Page, in which we consider a long time and less temperature to the dissolution process of the agarose. Besides, we chose the EcoRI enzyme because it was also useful for our plasmids in producing pDNA linearized. To be assured that it was a correct selection, we performed an in silico digestion through SnapGene as is shown in Figure 9. At the same time, we incorporated the new Quick-Load Purple 1 kb Plus DNA ladder.

Figure 9. in silico pDNA agarose gel electrophoresis (1%). (a) Lane 1. Quick-Load® Purple 1 kb Plus DNA Ladder. Lane 2. LysK digested with single-cut enzyme SmaI (2997 bp) pDNA.Lane 3. Negative control LysK digestion (2997 bp) pDNA. (b) Lane 4.LysCSA13 digested with single-cut enzyme EcoRI. Lane 5. Negative control LysCSA13 digestion (3159 bp) pDNA. (c) Lane 6. LysSS digested with single-cut enzyme EcoRI (2892 bp).Lane 7. Negative control LysSS digestion (2892 bp) pDNA.

Test

After running our 1% agarose electrophoresis we showed that our digestions coincided with the in silico results expected (Figure 10).

Figure 10. pDNA agarose gel electrophoresis (1%). (a) Lane 1. Quick-Load® Purple 1 kb Plus DNA Ladder. Lane 2. LysK digested with single-cut enzyme SmaI (2997 bp) pDNA.Lane 3. Negative control LysK digestion (2997 bp) pDNA. (b) Lane 4.. Quick-Load® Purple 1 kb Plus DNA Ladder. Lane 5. LysCSA13 digested with single-cut enzyme EcoRI. Lane 6. Negative control LysCSA13 digestion (3159 bp) pDNA. (c) Lane 7. LysSS digested with single-cut enzyme EcoRI (2892 bp).Lane 8. Negative control LysSS digestion (2892 bp) pDNA. Lane 9. Quick-Load® Purple 1 kb Plus DNA Ladder

Learn

Once we analyzed our results, we discovered that using the enzyme EcoRI, it was possible to linearize our plasmids, which gave us the possibility to compare our results with the expected bands. In fact, we verified that our constructs were correctly assembled, besides, we improved the agarose gel’s quality so that unwanted spots didn’t appear.

Reflection

The overall knowledge acquired for this iteration was related to the reagents' usage. Through our first iteration, we faced a challenge because some of our plasmids weren’t correctly digested, instead, a double digestion occurred, which had no relation to our designing or assembly method; We confirmed it once our constructions were digested with another enzyme. The only variation between our constructions was the codifying sequences, and considering that the restriction site was in the acceptor backbone, no change should have shown between the plasmids’ digestions, but only the LysK plasmid aligned its results with the expected ones. After performing the same experiment but using the EcoRI enzyme, all the results coincided with the in silico simulation. Evaluating the situation and taking into account that the 1 kb DNA ladder condition was not the best, we concluded that the obstacle was a consequence of non-ideal storage conditions.

First sub-cycle:

Second sub-cycle

Protein Induction

First sub-cycle: Induction Math Model

1st Iteration

Design

Once we finalized the first cycle, we aimed to establish the conditions that would be most favorable for protein induction. The gene expression system in E. coli BL21(DE3) is inducible by isopropyl β-D-1-thiogalactopyranoside (IPTG), so our objective was to predict the protein yield based on different initial IPTG concentrations and analyze how the system behaves while accounting for stochasticity.

Build

The model was constructed by integrating the biochemical reactions within the cell, encompassing the introduction of IPTG via simple diffusion. A representation of each binding interaction was diagrammed, culminating in the development of a systematic set of Ordinary Differential Equations (ODEs). The equations were derived in accordance with the Law of Mass Action, ensuring a foundation for the subsequent analyses conducted in the model. Additionally, we incorporated a range of parameters obtained from literature, accompanied by the necessary assumptions to ensure model´s tractability.

Test

In addressing the system ́s behavior, we utilized a two-fold approach: solving the ODEs set and performing a Monte Carlo simulation. The simulation was executed by incorporating a defined normal distribution of known parameters, enabling an assessment of the stochasticity of biomolecular mechanisms. However, it was not feasible to simultaneously plot each scenario.

Learn

During the analysis of this conflict, we identified a key factor to the numerical solver we employed in MATLAB, the ODE15s. This solver dynamically adjusted steps to accurately approximate the solutions based on the system´s evolving requirements. However, this led to varying matrix sizes for each scenario, presenting a challenge in attempting to graph all scenarios simultaneously due to these differing dimensions.

Reflection

This insight shed light on a significant aspect influencing our visualization approach. In essence, this realization prompted us to reevaluate and fine-tune our strategy to accommodate this variability effectively while ensuring the integrity and clarity of our results.

2nd Iteration

Design

The primary objective was to optimize computer memory usage without compromising data integrity. We employed strategies to streamline storage while dynamically adapting to changing matrix sizes, by creating n copies of a matrix while adjusting the row and column dimensions accodingly.

Build

Incorporated into the Monte Carlo code, we developed an algorithm to address this challenge. The algorithm compared matrices, identifying the required adjustments for each one without the need for data discarding. When discrepancies in sizes were detected, the algorithm seamlessly utilized the last obtained data from each scenario and repeated it as necessary until all matrices were equal in size.

Test

After implementing the algorithm, we were able to successfully run the code and achieve our primary objective of plotting each scenario. Nevertheless, upon comparison, the simulation results did not align cohesively with the experimental results, which prompted an examination of the model´s parameters.

Learn

The unexpected and unusual behaviors observed in the system led us to a critical realization - certain parameters were inadequate. We couldn't find these parameters in the literature, and upon reviewing our approach, we identified that the established ranges for the equations constants were not sufficiently robust. This insight caused us to reevaluate and refine them to ensure accuracy and reliability in our model, so our instructor, Ph.D. Segio Medina, advised for parameter estimation.

Reflection

We emphasize the iterative nature of the DBTL cycle and the importance of adaptability. Uncovered complexities and a disconnect between simulation and experimental results, reveals the crucial role of accurate parameters in modeling. Though, we only had qualitative data of the SDS-PAGE gel electrophoresis images, depicting inductions at different concentrations and varying times for each sample.

3rd Iteration

Design

To analyze protein distribution comprehensively, we integrated quasi-quatitative data extracted from ImageJ, an image analysis program. This data was coupled with a heuristic algorithm designed to estimate the unknown kinetic parameters.

Build

The process consisted of defining protein bands and extracting concentration data across the different gel lanes. Each gel lane corresponded to the same initial concentration, differing only in terms of induction time points within the same sample. Additionally, assumptions were established to streamline the calculation of endolysin concentration from total protein. The data analyzed formed the basis for creating an objective function in a Genetic Algorithm (GA) to calculate and minimize the error between experimental results and our model. The GA iteratively optimized kinetic parameters of the ODEs system by comparing simulated protein concentrations with experimental data.

Test

Utilizing the ODE15 solver in MATLAB to conduct the simulation based on the established model. The simulation was run for each initial IPTG concentration. The primary evaluation criterion was the convergence of the algorithm to a stable point, indicating that the GA effectively minimized the error between simulated protein concentrations and the experimental data. Another critical measure of success was the consistency in the converged point across several generations of iterations

Learn

We observed consistent patterns in the graphs generated during the simulation, consistently displaying three lines corresponding to the different induction concentrations. Notably, each graph consistently stabilized at a non-zero point, suggesting that our solution space might be overly complex or potentially suffering from overfitting. However, this was not a major concern as the coefficients effectively described protein production under varying conditions.

Reflection

Across some of the graphs, several generations of iterations appeared as though the optimization did not show apparent improvement. This observation hinted at the possibility that the coordinates leading to the optimum for each line, specifically the minimum value, remained consistent. The convergence to the same point in parameter estimation suggested a stable and optimal solution for each line, indicating a robust and reliable estimation process.

Second sub-cycle: Protein Induction Kinetics

The next step in our process was to standardize the optimal conditions to induce our fusion-proteins.

Click on each accordion to see details

1st Iteration

Design

We planned to perform the protein synthesis by inducing our transformed E. coli BL21(DE3) strains which contain the three plasmid constructions previously identified. For this to be possible, we incorporated the biobrick BBa_J435350 that codifies for the T7 promoter and also contains LacO regulations, inducible by the addition of IPTG. Our proteins have a molecular weight of: 28.437 kDa, 35.098 kDa, and 24.359 kDa, for LysK-ABD-SH3B30, LysCSA13-ABD, and PCNP-CecA-LysSS, respectively.

Build

The experiment was performed as described on our Experiments Page. Through this protocol, we induced each endolysin once the inoculums reached 0.4–0.5 OD (600 nm) at 0.2 mM, 0.5 mM, and 1 mM IPTG. Two controls were added: the E. coli BL21(DE3) non-modified, and the E. coli BL21(DE3) transformed non-induced. All of them were incubated for 6 hours. To determine the optimal IPTG concentration to reach the highest level of protein production, each hour's samples were harvested and lysed for their posterior SDS-PAGE analysis. In the same way, we separated the soluble and insoluble fractions by centrifugation. Taking into account the proteins’ weight, our polyacrylamide gels were prepared at 6% and 12% for the concentrator and separator phases, respectively.

Test

Our results didn't show the LysCSA13-ABD (Figure 11) and the PCNP-CecA-LysSS(Figure 12) proteins , but the LysK appeared successfully (Figure 13), and the protein soluble fraction was identified (Figure 14).

Figure 11. LysCSA13-ABD 0.5 mM IPTG protein induction kinetic, 3th hour. A) ColorBurst protein ladder. B) E. coli BL21 (DE3)-LysCSA13 non-inducted, zero hour. C) E. coli BL21 (DE3)- LysCSA13 inducted at 0.5 mM, zero hour. D)E. coli BL21 (DE3)-LysCSA13 non-inducted, first hour. E) E. coli BL21 (DE3)- LysCSA13 inducted at 0.5 mM, first hour. F)E. coli BL21 (DE3)-LysCSA13 non-inducted, second hour. G) E. coli BL21 (DE3)- LysCSA13 inducted at 0.5 mM, second hour. H) E. coli BL21 (DE3)-LysCSA13 non-inducted, third hour. I) E. coli BL21 (DE3)- LysCSA13 inducted at 0.5 mM, third hour.

Figure 12. PCNP-CecA-LysSS 0.2 mM IPTG protein induction kinetic, 4th to 6th hour. A) Precision Plus Protein Dual Color Standards (10–250 kD) ladder. B) E. coli BL21 (DE3). C) E. coli BL21 (DE3)- PCNP-CecA-LysSS non-inducted, fourth hour. D) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 0.2 mM IPTG, fourth hour. E) E. coli BL21 (DE3)- PCNP-CecA-LysSS non-inducted, fifth hour. F) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 0.2 mM IPTG, fifth hour. G) E. coli BL21 (DE3)- PCNP-CecA-LysSS non-inducted, sixth hour. H) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 0.2 mM IPTG, sixth hour. I) E. coli BL21 (DE3), first hour.

Figure 13. LysK-ABD-SH3B30 0.2 mM IPTG protein induction kinetic, 4th to 6th hour. A) ColorBurst protein ladder. B) E. coli BL21 (DE3). C) E. coli BL21 (DE3)- LysK-ABD-SH3B30 non-inducted, fourth hour. D) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, fourth hour. E) E. coli BL21 (DE3)- LysK-ABD-SH3B30 non-inducted, fifth hour. F) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, fifth hour. G) E. coli BL21 (DE3)- LysK-ABD-SH3B30 non-inducted, sixth hour. H) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, sixth hour. I) E. coli BL21 (DE3), first hour.

Figure 14. LysK-ABD-SH3B30 soluble (A) and insoluble (B) fractions. Visualization through a polyacrilamide gel 12% (SDS-PAGE) results. Both gels present the same distribution of wells, each one corresponding to the fraction (soluble or insoluble) assigned.

A) MWM (for gel A: Precision Plus Protein Dual Color Standards (10–250 kD), for gel B: PageRuler™ Plus Prestained Protein Ladder, 10 to 250 kDa). B) E. coli BL21 (DE3) C) E. coli BL21 (DE3)- LysK-ABD-SH3B30 non-inducted. D) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 2 μL. E) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 4 μL. F) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 6 μL. G) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 8 μL. H) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 10 μL. I) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 15 μL. J) E. coli BL21 (DE3)- LysK-ABD-SH3B30 inducted at 0.2 mM IPTG, 20 μL.

Learn

As shown in the test stage, the experimentation revealed that some of the samples didn't appear in the SDS-PAGE, while others did. All the processes were prepared and performed in the same way. In the first place, we could see that, for LysK-ABD-SH3B30 the proper bands appeared, in which we could identify that the best IPTG concentration was 0.2 mM for 5 hours and the protein was in the soluble fraction. Roughly speaking, we confirmed that our construction was correctly designed. On the other hand, LysCSA13-ABD didn’t show bands for any of the samples, but we know the gel was correctly run because of the protein ladder appearance. In addition, the PCNP-CecA-LysSS only showed results in the negative controls (non-induced, and non-transformed strains).

Reflection

While it is true that the results for this iteration included good news in the development of the product, the obstacles were really challenging. To analyze the situation, we actually know that all the processes were the same for each of the samples, but the results were completely different for each one. Knowing that LysK-ABD-SH3B30 achieved the expected results, we can be assured that our protocols work correctly. Taking into account that for LysCSA13-ABD none of the samples showed bands, we assumed that the problem was on the lysis step since we are completely sure that the cell growth was the same for each of the samples, as was verified through the optical density. In that case, our first question for the second iteration was: "Was the lysis exactly the same for each sample? How can we improve the lysis efficiency?" Considering that the process depends completely on temperature, we know how to counteract this kind of issue. To solve this problem, we needed to prolong the incubation time. In the case of the PCNP-CecA-LysSS, the challenge looked harder. Evaluating that only the transformed strains showed no bands, we were completely confident that something about our protein was affecting cell growth, but it also confirmed that our construction worked. Our main hypothesis was that our protein, which lyses E. coli strains, was currently toxic for our host. In the background research, we didn’t find anything about killing the specific E. coli BL21(DE3) strain by expressing each of the parts that compound our fusion-endolysin, but results demonstrated that it could be occurring. The main question to solve is: "Was our protein lysing the bacteria? What can we do to prevent it?" To be assured that our results were right, we planned to repeat the experiment before assuming that our hypothesis could be true.

2nd Iteration

Design

Before evaluating the first iteration of this semi-cycle, we decided to check our hypothesis by changing the lysis incubation. Considering that the LysK protein revealed that the process worked, and as a consequence of researching other ways to lysate without using additional reagents, we concluded that extending the incubation time from 5 min up to 10 min and adding a posterior step in boiling water for 5 min, there was no way of having unlysed cells in the SDS-PAGE wells.

Build

We performed the protocol for the protein induction kinetics as described in our Experiments Page, taking into account the lysis incubation changes. We induced both endolysins once the inoculums reached 0.4–0.5 OD (600 nm) at 0.2 mM, 0.5 mM, and 1 mM IPTG. Two controls were added: the E. coli BL21(DE3) non-modified, and the E. coli BL21(DE3) transformed non-induced. All of them were incubated for 6 hours. To determine the optimal IPTG concentration to reach the highest level of protein production, each hour's samples were harvested and heat-lysed for their posterior SDS-PAGE analysis. In the same way, we separated the soluble and insoluble fractions by centrifugation. Taking into account the proteins’ weight, our polyacrylamide gels were prepared at 6% and 12% for the concentrator and separator phases, respectively.

Test

After the induction and SDS-PAGE electrophoresis we identified the expected LysCSA13-ABD bands, as well as its optimal conditions at 0.2 mM for 6 hours (Figure 15). In addition, we identified that this protein was in the insoluble fraction (Figure 16). On the other hand, PCNP-CecA-LysSS notice better bands, but only for the negative controls.

Figure 15 LysCSA13-ABD 0.2 mM IPTG protein induction kinetic. A) ColorBurst protein ladder. B) E. coli BL21 (DE3). C) E. coli BL21 (DE3)- LysCSA13-ABD non-inducted, fourth hour. D) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, fourth hour. E) E. coli BL21 (DE3)- LysCSA13-ABD non-inducted, fifth hour. F) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, fifth hour. G) E. coli BL21 (DE3)- LysCSA13-ABD non-inducted, sixth hour. H) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, sixth hour. I) E. coli BL21 (DE3), first hour.

Figure 5. LysCSA13-ABD soluble (A) and insoluble (B) fractions. Visualization through a polyacrilamide gel 12% (SDS-PAGE) results. Both gels present the same distribution of wells, each one corresponding to the fraction (soluble or insoluble) assigned. A) MWM (for gel A: Precision Plus Protein Dual Color Standards (10–250 kD), for gel B: PageRuler™ Plus Prestained Protein Ladder, 10 to 250 kDa). B) E. coli BL21 (DE3). C) E. coli BL21 (DE3)- LysCSA13-ABD non-inducted. D) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 2 μL. E) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 4 μL. F) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 6 μL. G) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 8 μL. H) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 10 μL. I) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 15 μL. J) E. coli BL21 (DE3)- LysCSA13-ABD inducted at 0.2 mM IPTG, 20 μL. K)

Learn

As shown in the test stage, increasing the incubation time and incorporating a posterior phase in boiling water improved the lysis method. For this iteration, the protein LysCSA13-ABD appeared and matched the expected band weight. We identified that the optimal conditions were induction at 0.2 mM IPTG with an incubation period of 6 hours. In the same way, we confirmed that it’s an insoluble protein. Unfortunately, the modification only improved the lysis phase, and for the PCNP-CecA-LysSS protein, the problem still appears.

Reflection

By using the engineering cycle, one of our current problems was solved. In the same way, we verified that the PCNP-CecA-LysSS issues were not related to the lysis phase. Our hypothesis about the toxicity of our protein was more accurate. In fact, it confirms that our protein was correctly designed and possesses functional enzymatic activity. The new question was about how we could prevent the early death of our cell host and effectively produce our third endolysin.

3rd Iteration

Design

After reflecting on our drawbacks, we realized that our PCNP-CecA-LysSS protein lyses the cell host because of the differences on the culture turbidity between the inducted samples and the non-inducted ones. To identify that our hypothesis was right we considered running an SDS electrophoresis, in that case, we evaluated the possibility of increasing the acrylamide concentration up to 15% and 8% for the separator and concentrator phases respectively. In the same way, we planned to take a sample from our final crude protein extract, to be assured of the presence of our protein.

Build

We performed massive protein inductions at 0.2 mM, 0.5 mM, and 1.0 mM IPTG concentrations of our PCNP-CecA-LysSS at 37°C for 6 hours, once our inoculums reach 0.4 - 0.6 OD (600 nm). We also included BL21 and non-inducted negative controls. To lyse our inducted cells, we use the B-PER reagent as is described in the Experiments page. After the protein extraction, a SDS-PAGE at 15% and 8% polyacrylamide concentrations (separator and concentrator phases).

Test

We performed an SDS-PAGE to identify the expected protein bands for the PCNP-CecA-LysSS, which has a molecular weight of 24 kDa. The results can be observed at Figure 17.

Figure 17. PCNP-CecA-LysSS expression. Visualization through a polyacrilamide gel 15% (SDS-PAGE) results. A) MWM (iBright). B) E. coli BL21 (DE3). C) E. coli BL21 (DE3)- PCNP-CecA-LysSS non-inducted. D) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 0.2 mM IPTG. E) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 0.5 mM IPTG. F) E. coli BL21 (DE3)- PCNP-CecA-LysSS inducted at 1.0 mM IPTG

Learn

The protein expression of our third endolysin was identified. Considering the principles of catalytic activity of this endolysin, we realized that the toxicity of the host cell hinders adequate protein production, in this way we addressed the problem by increasing the final concentration of proteins. This was possible by analyzing the crude protein extracts at different concentrations of IPTG.

Reflection

Reflecting on the Engineering cycle gave us the opportunity to affront and resolve one of our biggest problems. In that case, we finally identified the presence of our protein in our IPTG inductions by increasing the final concentration of protein for the sample. Generally, we only took samples of 200 uL of bacterium, but for this case, 20 uL of the final crude protein of 20 mL was lysed and charged in our SDS-PAGE.