---

Summary

The process of Engineering Success requires the iteration through the Design-Build-Test-Learn Cycle. In this page, we present 3 examples of the team's Engineering Success

1. Design of the operon MegaT to overcome design constraints of our project
• Design RBS, reporter and terminator together as a part
• Build MegaT into a negative feedback circuit
• Test the effect of MegaT through microscopy
• Learn the results from the microscopy
• Redesign a family of MegaT suitable for other circuits


2. Design and Characterisation of an autoregulatory negative feedback circuit
• Design an autoregulatory negative feedback circuit
• Build the circuit with JUMP Golden Gate Assembly
• Test the circuit in the Mother Machine
• Learn the results from the Mother Machine Experiment
• Redesign the autoregulatory negative feedback loop to improve system dynamics


3. Insertion of the perturbation circuit into the downstream site of a JUMP acceptor vector
   • Design adapters to allow insertion into downstream site of JUMP vectors
   • Build the linear transcription unit with linear Golden Gate Assembly Pipeline
   • Test through colony PCR
   • Learn the result of successful integration of downstream site through sequencing
   • Redesign the transcription unit allowing easier selection for the next iteration

These Engineering processes are important in the process of our main project of creating an antithetic integral controller. The design of the operon MegaT allows us to measure the amount of output we have, thus showing adaptation. The negative feedback circuit as an example, utilises the operon MegaT to visualise the amount of output and serve as a comparison to the antithetic integral controller. To allow us to perturb the antithetic integral controller, we require both MegaT, showing the increase of production rate of output through perturbation and a downstream site that turns the perturbation on and off. The success of the perturbation circuit is determined by the design of the transcription unit that integrates into the downstream site of the Level 2 pJUMP47-2A acceptor.

Design of MegaT

In our project, we hope to use a reporter protein to visualise the amount of feedback species X we have in the circuit. To avoid tagging the reporter protein directly onto X, we have decided to put in an RBS and a reporter (mVenus) right after X to visualise the amount of X. We are implementing JUMP assembly. However, JUMP assembly requires us to clone in 4 parts (usually the promoter, RBS, CDS and terminator) at the same time and does not allow polycistronic assemblies as is the case we have in our main project. Therefore, we need a way to design an operon to allow polycistronic assembly. For more details on JUMP Golden Gate Assembly, please refer to the Rationale page.

Design (cycle 1)

To allow the reporter to be expressed under the same promoter as X, we have designed MegaT such that the RBS, reporter and terminator are attached together as a fragment. The MegaT part we have designed consists of RBS B0032, CDS mVenus and L3S2P21 terminator (Chen et al.., 2013) and is placed as an order with IDT.

We have chosen to use B0032 because we want to start with a medium strength of RBS. If the RBS is too strong, the reporter may form aggregates when existing in high amounts. If the RBS is too weak however, the amount of mVenus produced may not be detectable. In terms of the terminator, the L3S2P21 is a synthetic high efficiency terminator. A strong terminator is important to terminate transcription.

Build (cycle 1)

The MegaT part itself will not be expressing the reporter itself without the promoter part before it. Therefore, we have built a genetic circuit with a single transcription unit and use MegaT to visualise the relative amount of CDS we have. The circuit we have decided to characterise MegaT in is the autoregulatory negative feedback circuit (nFc). The purpose of this circuit is to characterise the behaviour of autoregulatory feedback circuit and compare it with other adaptation strategies. For more details please look at the part on negative feedback in this page.

Test (cycle 1)

To test for the intensity of the reporter protein in the circuit and show that it is reporting the relative amount of VanR, we have cultured the cells with the plasmid containing the plasmid in 0µM and 100µM Vanillic acid overnight. The following table shows the strain of cells and culture conditions overnight we have used for the characterisation of MegaT. The strain SB7 is a strain with mCherry integrated into chromosomal DNA.

Table 2. Overnight culture for characterisation of MegaT

Strain	Plasmid	Medium (Overnight culture)	Antibiotics	Volume	Purpose
E. coli (SB7)	nFc	EZRDM - 0.1% pluronic	Kanamycin	2mL	Test strain
E. coli (SB7)	nFc	EZRDM - 0.1% pluronic - 100µM Vanillic Acid	Kanamycin	2mL	Test strain
E. coli (SB7)	No plasmid	EZRDM - 0.1% pluronic	Kanamycin	2mL	Negative control

Table 1. Microscope settings for the characterisation of MegaT

Channel	Reporter	Purpose
phase contrast	NA	show the presence of cells
mCherry	mCherry	Show the presence of SB7 - all SB7 cells have mCherry integrated into their chromosome For cell segmentation
YFP	mVenus	To show the concentration of the reporter

We are using the Nikon TI2 Eclipse microscope and have taken 9 images of each cell strain at different fields of views separated from each other to prevent imaging the same cell twice.

Learn (cycle 1)

Processing the image data

Importing images

The microscope images we took came in .nd2 file where the images of separate channels are extracted into .tiff files using NIS Element Viewer. The tiff files are then imported into Python using the scikit image and pillow library and being converted into a 2D NumPy array.

Cell Segmentation

In image analysis, regions where cells are present will have a different value to the background. When analysing the images taken with fluorescence, the region where cells are present will be at a higher intensity value compared to the background. As all the cells we have imaged is the SB7 strain that has mCherry integrated into their genome, we make use of the mCherry images to segment cells.

To segment cells, we use the Omnipose library in Python, putting in the mCherry fluorescent images, to create masks in .png format to segment individual cells into separate regions.

Finding properties of individual cells

After segmentation, the mask has to be labelled i.e. to give each region of interest an associated label and the property of each labelled region can be found. This is done by the function label() from the scikit-image package in Python. After labelling the image, we can find the properties of individual regions which are then exported as Pandas DataFrame with the function regionprops_table, also from the scikit-image package. In this context, area, mean intensity and major axis length are the properties that we are interested in.

Analysis of results

Filtering the data

Looking at the area of individual labelled regions, one can see that some regions have very small areas and some regions have very big areas. The small areas are background noise while the big areas are more than 1 cell being clumped together, both of which are data that we do not want to include in our analysis. Therefore, to filter these outliers, we find the mean of the area and only include the areas that are within 2 standard deviations of the mean.

plotting

To visualise the distribution of the fluorescent signal of mVenus and mCherry of the samples and its respective area, we plotted a scatter plot of fluorescent intensity against area for both 0μM and 100μM vanillic acid in the test strain. This allows us to visualise the relationship between area and fluorescent signal and visualise the area and fluorescent intensity of individual cells. As we can see from the scatter plot that the fluorescent signal and the area is independent of each other.

To visualise the distribution of fluorescent signal of mVenus and mCherry based on counts, we have decided to plot histograms for the counts of fluorescent intensities. We have plotted 2 types of histograms - mCherry and mVenus on the same graph for the same cell type and mVenus of different cell type plotted on the same graph. The first allows us to see the difference between mCherry and mVenus expression while the second allows us to visually compare the effect of Vanillic acid on the system.

The above analysis were all performed at a single timepoint during the experiment. To further characterise MegaT, we did the same analysis on an overnight time-lapse imaging to see where the fluorescent intensity of the cells changes with time. As the cells dried up overnight and died, we splice the time period of the experiment within the period where the cells are not dried up and plotted a time series data of how the mean of the fluorescent intensity change with time. As we can see from the graph, the mean intensity of the cells with nFc in 100μM of vanillic acid is always higher than that of the cells with nFc in 0 μM of vanillic acid.

Statistical Testing

As we are interested in comparing the mean intensities of mVenus expression between 0µM and 100µM Vanillic acid, we perform a t-test between the 2 samples and confirm that the 2 means are statistically significantly different from each other.

Conclusion

The result shows that MegaT can respond to the relative changes of VanR, it shows us that it is a desirable candidate to report the relative amount of the species it is in the same transcription unit with. As we are trying out a combinatorial approach to test for the antithetic integral controller with the best performance, we have decided to design MegaT with a different ribosome binding sites, matching that of the ribosome binding site of species X in our project.

Design 2

Due to what we have learnt above, we have designed 2 other MegaT, one with RBS B0033 (weak) and the other with RBS B0034 (strong). The following figure shows the design of all the MegaT, showing that this can be put into a package to allow users of the antithetic circuit to choose the MegaT based on the RBS of species X.

Autoregulatory Negative feedback

Negative feedback is a type of circuit design motif that exists in nature to improve system response and reject disturbances. To know more about how a negative feedback motif acts to reject disturbances, please refer to the modelling page. The purpose of the experiment is to characterise the autonegative feedback circuit by seeing how it adapts to constant disturbance as a perturbation. Using the mother machine, a microfluidic device, we are able to control the medium the cells are cultured in. We image the cells with time-lapse microscopy at 1 µM Vanilic acid for 3 hours, making sure that the steady state value is reached. We then perturb the system by switching the culture medium to 10 µM vanillic acid for another 3 hours to see how the system adapts.

Design (cycle 1)

We have designed an autonegative feedback loop (nFc) to study the effect of adaptation. An autoregulatory negative feedback refers to a motif that involves a species feeding back to downregulate itself. The feedback species we have chosen is VanR and the promoter we have chosen is P_VanCC, the effect of VanR on P_VanCC at different concentration of Vanillic acid pf inducer is well characterised by Meyer et al., (2019). Essentially, VanR will act on P_VanCC, repressing its production, following a Hill Function dynamics. In order to characterise the amount of vanR, we have created an operon (MegaT) with a RBS, mVenus and Terminator after VanR coding sequence (CDS). The detailed design on MegaT is discussed above.

Build (cycle 1)

We perform JUMP assembly to clone the transcription unit into pJUMP27-1A using the enzyme BsaI to clone the transcription unit into the main cloning site. For the protocol we used for JUMP assembly, please refer to the Experiments page. For details of how JUMP Golden Gate Assembly works, please refer to the Rationale page.

After Golden Gate Assembly, the construct is then chemically transformed into DH5a cloning strain and plated out. For details of the protocol, please refer to the Chemical transformation protocol on the Experiments page.

After overnight incubation on Kanamycin selection plates, we want to confirm the construct we want exists in the colony. To do this, we perform colony PCR and gel electrophoresis on a non-fluorescent colony that does not contain superfolder GFP (sfGFP) and check whether it has the correct length of amplicon after colony PCR (cPCR). The reason we are choosing non-fluorescent colonies is because sfGFP is the drop out cassette of the JUMP vector. Therefore, if the transcription unit is successfully inserted into the JUMP vector, the sfGFP will be lost, thus the colony will not be fluorescent. For the details of the protocol, please refer to the colony PCR and the gel electrophoresis protocol on the Experiments page.

For the colony with the correct cPCR amplicon length on gel electrophoresis, the colony grown overnight and glycerol stocks and Miniprep are done the next day. For details of the protocol, please refer to the glycerol stocks and Miniprep protocol on the Experiments page.

For the construction of the autoregulatory negative feedback circuit, colony PCR is a good indication that cloning is successful. The band length of the autoregulatory negative feedback circuit is about 2000bp while for superfolder GFP is about 1300bp. Moreover, to make sure that there is no mutation within the construct, the sequence has to be verified through sequencing. Sequence verification in this case is very important because MegaT contains a fluorescent protein, mVenus, which may fail to report on the amount of VanR if it contains mutations.

Test (cycle 1)

To test the effect of the negative feedback circuit in response to perturbation, we have decided to conduct the experiment in a mother machine.

Fabrication of the Microfluidic Device

A mother machine is a microfluidic device that is made up of a feeding lane and trenches on its side. The design of the mother machine allows cells to be captured within the trenches. The first cell that gets captured in the trench is the "Mother" cell. The feeding lane allows fresh medium to flow through it constantly, providing nutrients to the cells inside the trenches of the mother machine. The following figure shows a mother machine under the microscope where we can see the trenches on the side of the feeding lane.

Preparation of Overnight Culture

The following table shows the strain of cells and culture conditions overnight we have used for the characterisation of the negative feedback circuit. The test strain we have used is called SB7, a strain with mCherry integrated into chromosomal DNA. We have used E. coli SB6, a strain with mVenus integrated into the chromosomal DNA as a positive control and E. coli SB5, a strain with CFP integrated into the chromosomal DNA as a positive control.

Table 3. Overnight culture for characterisation of negative feedback circuit (nFc)

Strain	Plasmid	Medium (Overnight culture)	Antibiotics	Volume	Purpose
E. coli K-12 MG1655 7740 (ΔmotA, glmS::PRNA1-mCherry-mKate2) [aka. SB7]	nFc	EZRDM - 0.1% pluronic	Kanamycin	2mL	Test strain
E. coli (SB7)	nFc	EZRDM - 0.1% pluronic - 100µM Vanillic Acid	Kanamycin	2mL	Pre-test sample to check full expression fluorescence level before experiment start under the microscope to test required microscope settings
E. coli (E. coli MG1655 7740 ΔmotA attTn7::PrpsL-SCFP3A) [aka. SB5]	No plasmid	EZRDM - 0.1% pluronic	Kanamycin	2mL	Negative control
E. coli K-12 MG1655 7740 (ΔmotA, attTn7::PrpsL-Venus) [aka. SB6]	No plasmid	EZRDM - 0.1% pluronic	Kanamycin	2mL	Positive control

Microscope settings

As we need to visualise the cells and the mother machine under the microscope (Nikon TI2 Eclipse), we have to have appropriate microscope settings to image the cells.

Table 4. Microscope settings for the characterisation of negative feedback circuit (nFc)

Channel	Reporter	Wavelength	Purpose
phase contrast	NA	705nm	show the presence of cells
mCherry	mCherry	610nm	Show the presence of SB7 - all SB7 cells have mCherry integrated into their chromosome For cell segmentation
YFP	mVenus	530nm	To show the concentration of the reporter To show positive control For cell segmentation
CFP	CFP	475nm	Show negative control For cell segmentation

Note: Segmenting cells with phase contrast alone can be difficult due to low contrast. Therefore, we require a fluorescent signal image for cell segmentation.

Microscope images were taken every 3 minutes to allow for good time resolution at 9 different field of views along the trenches of the mother machine.

Below is a video showing the microscope in action, imaging the cells in different channels!

Testing in the Mother Machine

We have prepared 2 media which was put into the mother machine in the first 3 hours and last 3 hours of the experiment respectively. The first media we have made consists of EZRDM, pluronic and 1μM Vanillic acid. The second media we have made consists of EZRDM, pluronic and 10μM Vanillic acid. The explanation of the choice of the media is explained below.

Explanation

With 1μM vanillic acid added, vanR produced represses on p_vanCC, inhibiting mVenus production. After the circuit reaches its steady state, we introduce the perturbation as 10μM vanillic acid. As this is a negative feedback loop but not robust perfect adaptation, 10μM of vanillic acid should act to inhibit the repression of vanR on p_vanCC so the signal of mVenus is expected to increase. The concentration of vanillic acid chosen for the experiment is based on the characterisation from Meyer et al., (2019). To allow the system to have a large dynamic range, we have chosen concentrations of vanillic acid that lie on the lower end and the higher end of the slope of the Hill curve. As we are interested in a positive perturbation, we started the system off with a vanillic acid concentration of 1μM then use 10μM vanillic acid as a perturbation.

Learn (cycle 1)

Through looking at the data from the microscope as a quick analysis, we have realised that only the cells in the middle of the field of view survived but those on the side are all dead at the end of the experiment (Figure 4). Moreover, this happens consistently in all the field of views we have captured. We proceed on looking at the time-lapse images and realise that the cells on the periphery of the field of views are slowly dying out as the experiment proceeds. As the experiment requires using high energy photons to detect CFP emission from SB6 strain, we suspect that this effect we saw from such quick-analysis is due to the effect of phototoxicity. As the different field of views are close together and the area that the excitation light shines on is wider than the field of view itself, the cells on the periphery are excited for more than once during every cycle of taking images, leading to their death.

As the cells in the middle of the field of view are largely unaffected, we looked at the time-lapse images to see whether we see the effect we expect after the addition of vanillic acid. As we are going through the images, we see that the cells increases in intensity of mVenus as expected. However, towards the end of the experiment, we see signs that the cells are “struggling” and getting dimmer, probably also due to the phototoxicity brought about by the UV light that excites CFP.

To show that the fluorescent intensities of cells in the presence of perturbation of vanillic acid, we plotted a time series data of the cells that survived the imaging. However, as it is evident that some cells are dying as the imaging goes on and lost their fluorescent, we have limited the time range we analyse the time series data. The following figure shows the change in intensity of 3 cells as an example to see how the intensity of the cells have increased with the change of medium from 1µM of vanillic acid to 10µM of vanillic acid.

To visualise the effect of vanillic acid on the fluorescent intensity, the distribution of fold change of fluorescent intensites of all the cells is plotted below. The fold change is calculated by the ratio of the intensity of the end point and the intensity of the starting point. Apart from the fold change, the halfway point, the time point where half of the maximum intensity is reached, is also being plotted out as a distribution to look at the response time of the circuit.

Design (cycle 2)

Learning from the data obtained from the mother machine experiment, we have decided to iterate through another cycle by doing simulations of the autoregulatory negative feedback system to further improve system response. Through doing the simulation, we find that increasing the strength of the production rate of VanR reduces the system deviation from the setpoint in response to the same perturbation. Therefore, we have designed the next negative feedback circuit, replacing the RBS with B0034, a stronger RBS than B0032 to investigate whether it can improve system dynamics. Below is the figure showing the changes we are trying to implement for this new cycle.

We have also noted that the data we have obtained from the first cycle may not be reliable as the phototoxicity damaged the cells. Therefore, for the next iteration, we plan to perform the experiment by manipulating the fieldstop, ensuring that the excitation light only illuminate the field of view itself but not the area out of the field of view, avoid illuminating cells in the periphery of the field of view more than once in each cycle.

Downstream site insertion

In the main project, we want to create a perturbation circuit to test our whether the antithetic integral controller can robustly adapt. As building the antithetic integral controller together with the perturbation circuit requires 5 transcription units and a level 2 plasmid can accommodate 4 transcription units at the same time, we have decided to insert the remaining transcription unit of the perturbation circuit into the downstream site of pJUMP47-2A.

Design (Cycle 1)

At the first instance, we have decided to look at the protocol given in the original paper of JUMP Golden Gate Assembly (Valenzuela-Ortega et al., 2021) where the authors make use of a 2-step assembly method to perform a downstream site integration. In the first step, basic parts required for the linear transcription unit for insertion is assembled and at the time the destination vector is digested separately with BbsI and dephosphorylated. After this, in the second step, the ligated mixture and the digested and dephosphorylated backbone is put together in the same reaction mixture and is ligated.

As we believe the protocol used in the paper was a bit too complicated for us, we have decided to design our own protocol to insert the linear assembly into the downstream site. We have designed a linear Golden Gate Assembly pipeline to create an insert that has adapter on both the 5’ and the 3’ end with the recognition site of BbsI.

The 5’ adapter is divided into 3 parts - the primer amplification region, the BbsI recognition site with the correct overhangs and the BsaI recognition site with the correct overhangs. The primer amplification region allows the same primer design to be able to be compatible with the amplification of all linear Golden Gate assembly products. The BbsI recognition site allows the Golden Gate assembled product to be inserted into the downstream site. The BsaI recognition site allows linear Golde Gate assembly to occur, forming a linear ligated product of 5’ adapter, promoter, RBS, CDS, Terminator and 3’ adapter.

The following figure shows the design of the adapters

Build (Cycle 1)

As the adapter sequence is less than 50bp, we have ordered each adapter sequence as 2 strands of oligos. The adapters are first annealed together into double-stranded DNA with the oligo annealing protocol. Please refer to the Experiments page for the protocol of oligo annealing.

After obtaining the adapter in the double-stranded DNA form, the adapter is molaritised into 25nM. By diluting the necessary parts of the linear transcription unit (J23106, B0032, TetR, ECK120029600) into 15nM, we then perform linear Golden Gate Assembly using the linear Golden Gate Assembly pipeline specified in the Experiments page.

After linear Golden Gate assembly, the product is then amplified with Q5 PCR then validated through gel electrophoresis whether it shows a correct length on the gel. If it does, we purify the linear Golden Gate product. If not, we have to repeat the above steps.

After purification, the linear transcription unit is ready to be integrated into the downstream site. For this, we make use of the downstream site integration protocol in the experiments page. Subsequent steps of JUMP assembly is the same for downstream site integration compared to cloning into the main site. For details, please refer to the cloning workflow described in the Experiments page.

Test (Cycle 1)

To make sure that the downstream site is inserted, we have to formulate ways to test and confirm that the integration is successful. Normally, there are 3 ways to validate integration into the cloning site - visual inspection through fluorescence, colony PCR and sequencing. While we can perform visual inspection on inserts into the main cloning site due to the presence of a superfolder GFP dropout cassette, the downstream site integration does not displace any dropout cassette so when looking at the colonies, we cannot use visual inspection to decide which colonies will have the downstream site integrated thus which ones to pick. As a result, the second way to confirm this is through colony PCR. To increase the chance of success, we have decided to pick about 12 colonies, a lot more than the normal cloning workflow into the mainstream site. Figure x shows an example of the gel with 3 lanes being the downstream site integrated into the Level 2 acceptor. As we can see, the band length shown in the gel is about 2000, an expected band length produced through colony PCR. After confirming the band length on the gel, we confirm the construct through sequence verification with Sanger sequencing.

Learn (Cycle 1)

After submitting the plasmid for sequencing, we have confirmed that the downstream site has been inserted successfully and have proved that inserting a linear transcription unit into the downstream site through the design of adapters is possible. The following figure shows the sequence alignment of the sequencing results and the design of our linear Golden Gate Assembly into the downstream site.

We have also submitted an entry of the Downstream site perturbation circuit onto the Registry! Registry No. BBa_K4491145

Design (Cycle 2)

Through doing the downstream site cloning, we have also realised how inconvenient it is to be unable to select for colonies through visual inspection because we have to pick a larger number of colonies for colony PCR. As a result, we think of replacing the current terminator (ECK120029600) with MegaT_RFP, a variant of MegaT that we have decided to allow for modularity of our system (see Implementation page for more details). As MegaT_RFP contains a whole transcription unit and appear red by the naked eye, it can act as a selection marker that reduces the number of colonies we need to pick for colony PCR.

The following figure shows the selection plate with colonies containing MegaT_RFP on it, as we can see, the colonies are visibly red, allowing red-green selection.

References

1. Meyer, A.J., Segall-Shapiro, T.H., Glassey, E. et al. Escherichia coli “Marionette” strains with 12 highly optimized small-molecule sensors. Nat Chem Biol 15, 196–204 (2019). https://doi.org/10.1038/s41589-018-0168-3


2. Chen, YJ., Liu, P., Nielsen, A. et al. Characterization of 582 natural and synthetic terminators and quantification of their design constraints. Nat Methods 10, 659–664 (2013). https://doi.org/10.1038/nmeth.2515


3. Marcos Valenzuela-Ortega, Christopher French, Joint universal modular plasmids (JUMP): a flexible vector platform for synthetic biology, Synthetic Biology, Volume 6, Issue 1, 2021, ysab003, https://doi.org/10.1093/synbio/ysab003