Team:William and Mary/Results

Results

Results


Adhesin Library

Functional Amyloid in Pseudomonas (FAP)

BBa_K3059640 (strong constitutive fap operon) and BBa_K3059641 (weak constitutive fap operon) were characterized with OD600 and congo red assays 20 hours after inoculation. Using a two-tailed paired T-test, a statistically significance difference was found between the absorbance of congo red stained JS006 cells and BBa_K3059640 (p = 0.00017, n = 5), JS006 and BBa_K3059641 (p = 1.59 x 10-5, n = 5) and BBa_K3059640 and BBa_K3059641 (p = 0.0125, n = 5). BBa_K3059641 on average retained slightly more congo red dye than BBa_K3059640. One explanation for the results between the strong and weak circuits may be the high production rate of a large part like BBa_K3059637 (3G compatible fap operon) results in slower growth, allowing cells with lower levels of expression to double faster and produce overall more fiber. However, when considering the average OD600 for the two groups, BBa_K3059640 actually had a higher average absorbance than its weaker counterpart BBa_K3059641, indicating more growth. Overall, the hypothesis that these circuits would express more amyloid (fap) nanofibers, leading to more congo red binding, is supported.

26 = BBa_K3059640, 36 = BBa_K3059641, CG = Congo red + PBS
OD600 of BBa_K3059640 and BBa_K3059641.
Decrease in A490 of BBa_K3059640 and BBa_K3059641

BBa_K3059634 (IPTG-inducible fap operon) was characterized with OD600 and congo red assays 20 hours after inoculation. There were four test groups: induced BBa_K3059634, uninduced BBa_K3059634, induced JS006, and uninduced JS006. Using a two-tailed paired T-test, a statistically significant difference was found between the absorbance of induced BBa_K3059634 and both JS006 groups (p = 0.0039, n = 10; p = 0.0011, n = 10), as well as between uninduced BBa_K3059634 and both JS006 groups (p = 0.0010, n = 10; p = 0.0013, n = 10). There was no significant difference between the induced and uninduced BBa_K3059634 groups (p = 0.1937, n = 10). This seems to indicate leaky repression of pLac, possibly since LacI expression is under the control of the weak promoter J23107. As this part was only functionally confirmed, it is also possible that there is a mutation in pLac or other regulatory gene of the fap operon.

Induced BBa_K3059634
Uninduced BBa_K3059634
Induced JS006
Uninduced JS006
OD600 of induced and uninduced BBa_K3059634.
Decrease in A490 of induced and uninduced BBa_K3059634

Synthetic Curli Operon

We assembled and characterized four curli fiber-producing circuits:

[1] WM19_023, or “circuit 23”: IPTG-inducible curli

[2] WM19_024, or “circuit 24”: Weak constitutive curli

[3] WM19_045, or “circuit 45”: pBlind + curli

[4] WM19_046, or “circuit 46”: pBlind + AG43 + curli


Curli Amyloid Fibers

Of the four curli fiber circuits we constructed, WM19_023 and WM19_024 were constructed first. Before experiments began, we performed a diagnostic restriction digest with enzyme EcoRI. Though Type IIS-compatible and thus legal, the native curli operon contains an EcoRI cutsite, so two cuts were expected: EcoRI within the Biobrick prefix and EcoRI within csgE (internal cut). Analyzing the digests with these cutsites in mind, all minipreps of circuit 23 and 24 appeared correct.

restriction digest on circuit 23. Expected bands at ~2 kb at ~5 kb appear.
restriction digest on circuit 24. Expected bands at ~1.9 kb and ~3.5 kb appear.

WM19_023

Preliminary experiments with circuit 23 aimed to determine a reasonable amount of IPTG to add for curli fiber expression. We initially followed the Harvard OptiPoly protocol, hoping that the instructions for arabinose induction of pBbB8K-csgBACEFG would work similarly for IPTG induction of WM19_023. However, 0.75 uL of 1M IPTG into 3 mL LB caused no observable curli fiber expression or statistically significant difference in supernatant color. A second experiment with 3 uL 1M IPTG into 3 mL LB (~1 mM final concentration) proved similarly ineffective. To verify that the circuit was at all functional, we induced with 30 uL 1M IPTG into 3 mL LB (~10 mM final concentration), and finally saw results (both to the naked eye and in plate reader data). Though we needed more replicates to draw any conclusions regarding the effectiveness of circuit 23, this experiment provided us with a starting point for IPTG induction.

When induced with 30 uL 1M IPTG, cell pellets were visibly redder than uninduced counterparts. The redness of these pellets should not be compared to that of pellets in future experiments, however; 150 uL of 0.015% Congo Red solution was mistakenly used instead of 100 uL.

Believing the full 30 uL to be excessive, we tested induction with 10, 20, and 30 uL of 1M IPTG. From visual inspection and plate reader data, we determined that no significant difference in curli production/supernatant color occurred past 10 uL 1M IPTG, and quickly devised a follow-up experiment to confirm the effectiveness of 10 uL. This experiment showed that induced samples consistently formed visible aggregates in glass culture tubes overnight, and consistently produced redder pellets/clearer supernatants (confirmed by plate reader) after staining with Congo Red solution.

Pellets visible on the bottom of glass culture tubes with induced samples. No such aggregation was visible for negative control samples of untransformed JS006.
Uninduced cultures, like the one above, showed no aggregation after overnight growth.
From left to right: WM19_023 1C3 #1 uninduced, WM19_023 1C3 #1 induced, WM19_023 1C3 #2 uninduced, WM19_023 1C3 #2 induced, WM19_023 1C3 #3 uninduced, WM19_023 1C3 #3 induced, WM19_023 1C3 #4 uninduced, WM19_023 1C3 #4 induced, untransformed JS006 #1, untransformed JS006 #2, uninduced pBbB8K-csgBACEFG, induced pBbB8K-csgBACEFG. pBbB8K-csgBACEFG is a gift from Dr. Neel S. Joshi at Harvard’s Wyss Institute. We discuss this plasmid later.

As seen above, for each pair of samples, the induced/right of the two samples has a redder pellet. Uninduced samples (the left of each pair) resemble the negative control, untransformed JS006. A student’s t-test (n=4, p=0.0196) found statistically-significant difference between uninduced and induced samples with a confidence level of 98.0%.

Box and whisker graph comparing Congo Red spin down assay results for uninduced and induced samples. For each sample, A490 of supernatant (provided by plate reader) was subtracted from A490 of a positive control (pure 1X PBS mixed with 100 uL 0.015% Congo Red solution). Samples that produced curli fibers bound more Congo Red dye, resulting in a greater difference between pure PBS/Congo Red solution and supernatant. As seen in the graph above, induced samples bound more Congo Red dye, resulting in a greater difference between supernatant A490 and control than uninduced samples. Blue lines represent averages, while red dots represent individual samples.

However, only four colonies were tested in this preliminary assessment. A final experiment tested ten colonies, producing 22 tubes total (ten induced, ten uninduced, two untransformed/negative control). To ensure that uninduced/induced sample pairs contained genetically identical bacteria, each inoculated with the same colony solution.

Again, results were apparent even before qualitative assessments: aggregates formed overnight in induced cultures, and cell pellets for induced samples were redder than their counterparts. Plate reader absorption measurements at 490 nm showed a consistent decrease in supernatant redness (Congo Red bound in the cell pellet decreases supernatant redness) from uninduced to induced samples. This difference was verified by a student’s t-test as statistically significant (n=10, p=0.00298) with a confidence value of 99.7%.

Box and whisker plot formatted identically to the one above. Once again, induced samples bound more Congo Red dye, resulting in a greater difference between supernatant A490 and control than uninduced samples. Blue lines represent averages, while red dots represent individual samples.

During these experiments, we also tested pBbB8K-csgBACEFG, a synthetic curli plasmid from Neel S. Joshi. This plasmid showed dramatic curli expression when induced, compared to our IPTG-inducible circuit. However, cells with pBbB8K showed impressive curli production even when uninduced. Uninduced samples of the arabinose-inducible circuit resembled the induced samples of our IPTG-inducible circuits, rather than the negative control (untransformed JS006).

From left to right: untransformed JS006, uninduced IPTG-inducible curli circuit, induced IPTG-inducible curli circuit, uninduced pBbB8K-csgBACEFG, induced csgBACEFG. The uninduced sample of pBbB8K-csgBACEFG resembled the induced sample of our IPTG-inducible circuit, rather than the negative control. The data suggests that, though our circuit maintains weaker expression of curli fibers, expression is less leaky.

WM19_024

WM19_024 placed our synthetic curli operon under medium-weak constitutive promoter J23107. No visible pellets formed in the glass culture tubes during overnight growth, and no difference in redness was observed between control and experimental pellets after Congo Red staining. Furthermore, the average A490 measurement for experimental samples’ supernatants was slightly higher than the average for control circuits (if curli fibers were expressed, the supernatant should be lighter, resulting in a lower A490 measurement). Future constitutive circuits may require a stronger promoter such as J23100.

WM19_045

WM19_024 places our synthetic curli operon under the promoter pBlind, which is activated by blue-light sensitive transcription factor EL222 (Jayaraman et al., 2016). Both pBlind and EL222 were obtained through DNA synthesis, and sequence-confirmed in multiple circuits.

Though considered an optogenetic circuit, pBlind-EL22 had resulted in constitutive gene expression in circuit 46, which was miniprepped before circuit 45. With this in mind, we tested the function of circuit 45 when exposed to ambient light as well as blue light.

Our first experiment to test WM19_045 did not involve blue light specifically. Five experimental tubes were exposed to ambient light while growing in the shaking incubator overnight, while experimental tubes were covered with foil (foiled and unfoiled sample pairs were inoculated using the same colony solution, allowing for direct comparison). JS006 cultures were grown as a negative control. Some experimental tubes, both foiled and unfoiled, showed small aggregates at the bottom. However, no obvious differences in pellet color were visible to the naked eye.

A sample that had been exposed to ambient light (left) as well as a sample that had been covered with foil (right). Both have small aggregates at the bottom.

The average A490 value for each group reflected expectations: average A490 was the lowest for experimental samples exposed to ambient light, followed by the average for foiled samples, followed by the average for the negative controls. However, these averages were hardly different (0.2438, 0.2440, 0.2482), and no statistical significant difference was found between ambient light and foiled samples (n=5, calculated t=-0.0673). Statistical significance was found between ambient light and control samples (n=5, calculated t=-2.475, 2.475 > 2.132) but only with a confidence level of 90%.

A second experiment exposed samples of circuit 45 to blue light. Two six-well plates were inoculated with three colonies of circuit 45 and three control (untransformed JS006) colonies. One plate was projected with blue light while the other was grown within a dark box. Both plates were grown in a stationary incubator and filled with M63 media rather than LB broth.

A clear difference between experimental and control samples was visible the next day; aggregation was observed within experimental wells whereas the turbidity within control wells appeared more “smooth.” This trend was apparent for both the plate grown under blue light and the plate grown in darkness.

For both the plate exposed to blue light (left image) and the plate grown in the dark (right image), experimental wells (left of two wells) showed aggregation whereas control wells (right of two wells) were turbid but uniformly so. Furthermore, experimental pellets (both those grown in darkness and those exposed to blue light) stained a darker red than negative control pellets. 

Two stained samples of circuit 45 between two negative control samples. Note the redder pellets of the middle samples. Both are similarly red, redder than the negative controls, although one was exposed to blue light and the other was grown in darkness.

No statistically significant difference was found between WM19_045 samples grown in the dark and samples exposed to blue light. Though curli expression (and thus A490) appeared to decrease from experimental samples to negative control samples, it was difficult to verify this relationship statistically with only n=3. The inability of the projector to cover many samples with intense blue light prevented the use of large sample sizes. Though future experiments were designed, such experiments were unfeasible with the time constraints.

Our available data suggests that, if pBlind is truly light-sensitive, the level of induction by blue light is somewhat weak (especially compared to the pDawn system). Instead, pBlind seems to express downstream genes constitutively, even moreso than our constitutive circuit J23107-curli (WM19_045).

WM19_046

Like WM19_045, WM19_046 utilizes the pBlind-EL222 system. However, circuit 46 includes both AG43 and curli fibers. As with circuit 45, pBlind-EL222 appears to result in constitutive, uninduced expression. Inoculations for miniprep, though covered completely in foil and grown in a shaking incubator, exhibited massive aggregates at the bottom of glass culture tubes and minimal cells in the LB supernatant.

Aggregates at the bottom of inoculations for miniprep, resulting in clearer LB broth supernatant.



This behavior persisted in JS006 samples inoculated for experiment. Though no statistically significant difference was found between foiled and unfoiled experimental samples, a statistically significant difference was found between unfoiled experimental samples and the negative control (n=5, calculated t=-3.8069, 3.8069 > 3.747) with a confidence interval of 98%. If one assumes that light plays little to no role in adhesin production and combines foiled and unfoiled groups, the confidence interval rises to 99.9% (n=10, p=0.000389).

Box and whisker graph comparing circuit 46 samples to untransformed JS006, the negative control. Samples expressing curli fibers bound more Congo Red dye, resulting in lighter supernatants and lower A490 values. As shown in the graph above, circuit 46 samples had generally lighter supernatants than JS006 samples. Blue lines represent averages while red circles represent individual samples.

We also completed a blue light experiment identical to the one used for circuit 45. However, instead of performing a Congo Red assay on the cultures, the wells were crystal violet stained. Before crystal violet staining, aggregation was visible in experimental wells (as opposed to uniformly turbid negative control wells). Unfortunately, aspirating the media and subsequently washing with 1X PBS removed most of the aggregation produced by experimental samples. This suggests that AG43 and curli fibers were expressed constitutively by cells floating freely within the liquid media. Had the cells expressed AG43 as a result of adsorbing onto the region of the plate exposed to blue light (as proposed by the original PNAS paper by Jin and Riedel-Kruse), a purple film would have been observed after aspiration and washing. Instead, it appears that most of the adhesin was suspended in the liquid media.

Though optogenetics was the intended use for circuits 45 and 46, we now have constitutive circuits to replace the ineffective circuit 24.

SaSuhB

IPTG-Inducible Circuits

To assess amyloid fiber formation for each of the SaSuhB circuits, a congo red assay was performed for each of the circuits. Colonies of 033 and 040 were picked, and inoculated into a colony solution. Half of the colony solution was added to 4mL of LB + chloramphenicol, and the other half of the colony solution was added to 4mL of LB + chloramphenicol + 10mM IPTG to induce expression of SaSuhB. After being grown for 18 hours, the solutions were Congo Red stained using our standardized protocol.

As can be seen in the graph above, the induced circuit 033 has a significantly higher decrease in A490 as compared to uninduced circuit 033 (p = 0.013678551, n = 4). This result is confirmed qualitatively by looking at the congo red spin down solutions. The tube on the left is uninduced, while the tube on the right is induced with 10mM IPTG.

Circuit 040 was grown for 18 hours as well. As can be seen, both uninduced circuit 040 ( p=2.08 X 10^-06, n = 8) , and induced circuit 040(p = 0.0003, n = 8) have a significantly higher decrease in A490 as compared to uninduced circuit 033. Furthermore, the induced circuit 040 also has a significantly higher decrease in A490 as compared to uninduced circuit 040 (p = 0.038, n = 8).

EL222-Circuits

Circuit 039 was inoculated directly into 4mL of LB + chloramphenicol. To induce expression of circuit 039, we found that ambient lighting was sufficient. After being grown overnight at 37C with 250rpm shaking, visible SaSuhB fibers were isolated from several of the culture tubes.

Qualitatively, we found that the fibers were incredibly sticky, and we had difficulty removing the fibers from the side of the glass culture tubes if they touched the sides.

Engineering Mycobacterium

Result Summary:

  • 8 Function-confirmed Mycobacterium smegmatis compatible 3G parts (3 promoters, 2 RBSs, 2 terminators)
  • 1 3G compatible shuttle vector with dual origin of replication for both E. coli and M. smegmatis
  • 2 Constitutive gene expression circuits (mScarlet-I for testing)
  • Validated Electroporation and Electrocompetent cell protocol

Electroporation Protocol Validation

Mycobacterium smegmatis is difficult to transform using traditional chemically-competent method. Thus, we adapted the electrocompetent cell protocol from Dr. Broussard lab (Broussard, 2009)and electroporation protocol from Cirillo et al. (1993) and Goude et al. (2008). To test out these protocols, we electroporated pSUM36 plasmid with gfp inserted (acquired from addgene). We imaged the transformed M. smegmatis under inverted fluorescence microscope. The validation test showed a positive result, which confirmed that the adapted protocols work well for our strain of Mycobacteria.

Magnification: 1000X. Filter: FITC
Magnification: 1000X. Filter: Transmission

Test Circuits Characterization

In order to confirm that BBa_K3059423 and BBa_K3059424 function as we expected, we imaged Mycobacterium smegmatis transformed with these two constructs under fluorescence microscope.



M. smegmatis transformed with BBa_K3059423. Magnification: 1000X. Filter: RGB-TRITC -> Transmission -> RGB


M. smegmatis transformed with BBa_K3059424. Magnification: 1000X. Filter: RGB-TRITC -> Transmission -> RGB


Untransformed M. smegmatis. Magnification: 1000X. Filter: RGB-TRITC -> Transmission -> RGB


To further quantify the gene expression level, we measured the optical density and fluorescence intensity for both transformed strains with plate reader experiments. Experimental data with BBa_K3059423 and BBa_K3059424 is plotted in the following graph. BBa_K3059423 showed a 10-fold to 50-fold increase in F.I., while BBa_K3059424 increased F.I. by up to 1500X fold.

Fluorescence intensity normalized with respect to OD600 plotted on a log scale. Each dot represents a distinct biological replicate (colonies). Negative control measures untransformed, nonfluorescent M.smeg mc2155 while blank measures 7H9 media with no inoculation.


Patterning

Engineered Quorum Sensing


We tested our sender and receiver quorum sensing circuits at multiple stages of the development and assembly process (see design page for more information on the interaction between sender and receiver strains). First, we assessed sender and receiver circuits separately, rather than immediately growing the strains together and hoping that one would induce the other. These individual experiments allowed us to confirm (using plate reader measurements taken over time) that prototype sender bacteria were inducible by IPTG and that receiver bacteria were inducible by HSL. Both circuits showed increase in fluorescence when induced by the proper molecule. With the circuits confirmed separately, we continued to experiments where sender bacteria (induced by IPTG) produced HSL and induce mScarlet production in receiver bacteria. Increased fluorescence over time after addition of IPTG (again quantified by plate reader data) functionally confirmed our final circuits. Results were further visualized via microscopy, and final sender and receiver circuits were further confirmed via Sanger Sequencing on both strands (Epoch Life Sciences, Inc.)

After functional and sequence confirmation of our final sender and receiver circuits, we set out to create a gradient of fluorescence via “ring experiments” (see experiment page for more information). Such a gradient would exemplify distance-dependent patterning. Visualization of our ring experiment under a dissection scope showed, as expected, no fluorescence within the inner circle of sender cells (sender cells produce HSL, not mScarlet). Fluorescence was observed in the outer ring of receiver cells, showing proper induction by HSL from the sender strain. However, no gradient of fluorescence was observed; the outer ring of receiver cells was uniformly fluorescent. We attributed this observation to over-induction of sender cells by IPTG, and thus over-induction of receiver cells by HSL from senders. We consulted the math team, asking them to scale their simulations so that the size of their simulated biofilms matched the size of the biofilms we were inducing (there was previously a discrepancy). We also asked them to model various concentrations of IPTG and various lengths of IPTG incubation. Based on their recommendations, we revised our parameters for future ring experiments.

Turing Patterns

Part Sources: Sequences for pLas-OR1, pRhl-LacO, RBSII and RBSH were found in the Supplementary Information section of the paper “Stochastic Turing patterns in a synthetic bacterial population.” Sequences for pRhl-lacO, pLacIq, RBSG, GFP(LVA), dsRed-exp, and RhlI were found in the NCBI sequence for plasmids PFNK-512 and PFNK-806. Sequences for LasI(LVA) and cI(LVA) were found on the iGEM Parts Registry. Two silent mutations were introduced into the sequence for GFP(LVA) to remove illegal cutsites.

Wetlab: To create this system, pLas-OR1, RBSII, pLaqIq, RBSG, RhlR mutant, RhlI, pRhl-lacO, and RBSH were ordered from IDT in 3G compatible format with sticky ends, and BsaI cut sites. dsRed-exp and RhlR were ordered from Twist, also in 3G compatible format with sticky ends, and BsaI cut sites. All of the parts ordered from IDT and Twist were also ordered with Pad1 and Pad2 sequences, except for the RBSs which were ordered as oligos and therefore did not include the Pad1 and Pad2 sequences. All parts were ordered with the standard sticky ends (see 3G Protocol). LasR, the spy terminator, LacI, WM18_DE_002, and the B0015 terminator were all retrieved from the WM iGEM 2018 Inventory. LasI(LVA) and cI(LVA) were retrieved from the 2019 iGEM distribution kit. To make LasI(LVA) and cI(LVA) 3G compatible, designer primers were used to add sticky ends, BsaI cut sites, and Pad1 and 2 sequences to the parts. However, new sticky ends were added to RhlI, GFP(LVA), dsRed-express, RBSII, and the spy and B0015 terminators using designer primers. These parts were then placed into WM Pad 1C3 backbone through the use of Gibson Assembly and were transformed into NEB 5 alpha cells. Then, a colony PCR was run using primers that anneal to the WM Pad1 and Pad2 sequences and the correct size bands were inoculated for minipreps the following day. The minipreps were used in the Golden Gate process to create transcriptional units, which were then made into a circuit using Gibson Assembly (see 3G and Gibson Assembly Protocols).

Applications: What advantages do Turing patterns have over other methods of patterning? Why use Turing patterns in our engineered living materials toolkit? Stochastic Turing patterns are a unique part of our toolbox because they are a form of cellular self-patterning. Most alternative methods of patterning have many limitations that make them inefficient or impractical for use in real life applications. Some methods, such as printing, require direct access to the substrate and an open environment. Others, such as patterning substrates, rely on surface pretreatment. In some cases, such as with microfluidics, expenses become an obstacle. Turing patterns are not limited by any of these requirements. Under IPTG induction, the cells are able to produce these patterns without the need for external management. Turing patterns strengthen our toolkit in countless ways. Unlike the other patterning methods included in the toolkit, Turing patterns do not require any manual programming such as continuous access to light. Our engineered quorum sensing circuits rely on blue light for placement of the sender and receiver strains. In contrast, when engineered with the Turing pattern system, cells are able to form patterns independently starting from a lawn of cells. Therefore, they require very little equipment and maintenance, especially when being formed over large areas. In addition, the expression level of certain genes can be altered based on the concentration of IPTG used for induction. As shown by researchers Karig et al., induction with high concentrations of IPTG causes more green fluorescence, while induction with low concentrations of IPTG allows for more red fluorescence (Karig). Since it consists of two non-interfering signalling pathways, the Turing pattern system can allow for the control of two genes in place of GFP(LVA) and dsRed-express within the same culture. Furthermore, the level of expression of each gene can be controlled. The minimal need for external interference in pattern formation and the ability to alter the level of gene expression through IPTG concentration levels be useful for many sophisticated applications. How can Turing patterns be incorporated into biofilms? What kinds of engineered biomaterials can this lead to? Along with all of the advantages that Turing patterns have over other patterning methods, there are several benefits to incorporating Turing patterns into biofilms for the creation of engineered biomaterials. Engineering E. coli to create these types of patterns allows for greater control over the spatial placement of the bacteria within the biofilm and could increase the complexity of the gene expression. For example, by replacing GFP(LVA) and dsRed-express with genes that code for lysins that are specific to certain virulent strains of bacteria, this system could be used to form an antimicrobial surface coating. By changing the concentration of IPTG, the expression of these lysins can be altered depending on which virulent strains are more prevalent or dangerous. Another possible application of Turing patterns is to increase the efficiency of water filtration. In 2018, researchers Tan et al. created polyamide membranes for water filtration and found that membranes formed by Turing patterns made desalination more efficient due to the “bumps, voids, and islands” they create within them (Tan). One possible application for Turing patterns formed in E. coli is to create biofilms that express Turing patterns within them for the purpose of water filtration. The instability caused by the Turing patterns could allow for more efficient water filtration and help to reduce costs of filtration in the long term.

The design of the plasmids, based on the paper by Karig et al., is shown on the “Design” page. Transcriptional units WM19_1_3_12, WM19_3_5_03, and WM19_5_10_05 were assembled to create circuit pFNK-806 using Gibson Assembly. Transcriptional units WM19_1_3_11 and WM19_3_10_08 were not successfully created, although all of the parts that it consists of were placed into 3G compatible format.

Reference

Broussard, G. (2009). Electrocompetent Mycobacterium Cells. [online] Gregory W. Broussard, Ph.D. Available at: https://gregorybroussard.com/2009/04/21/electrocompetent-mycobacterium-cells/ [Accessed 21 Oct. 2019].
Cirillo, J. (1993). Efficient Electro-transformation of Mycobacterium smegmatis. [online] Bio-rad.com. Available at: https://www.bio-rad.com/webroot/web/pdf/lsr/literature/Bulletin_1360.pdf [Accessed 21 Oct. 2019].
Goude, R. and Parish, T. (2008). Electroporation of Mycobacteria. Journal of Visualized Experiments, (15).
Jayaraman, P., Devarajan, K., Chua, T. K., Zhang, H., Gunawan, E., & Poh, C. L. (2016). Blue light-mediated transcriptional activation and repression of gene expression in bacteria. Nucleic acids research, 44(14), 6994–7005. doi:10.1093/nar/gkw548
Jin, X., & Riedel-Kruse, I. H. (2018). Biofilm lithography enables high-resolution cell patterning via optogenetic adhesin expression. PNAS 115, 3698-3703. doi: 10.1073/pnas.1720676115
Karig, D., Martini, K. M., Lu, T., Delateur, N. A., Goldenfeld, N., & Weiss, R. (2018). Stochastic Turing patterns in a synthetic bacterial population. Proceedings of the National Academy of Sciences, 115(26), 6572-6577. doi:10.1073/pnas.1720770115
Karig, D. K. (2007). Engineering multi-signal synthetic biological systems. (Doctoral dissertation). Princeton University, USA.
Tan, Zhe (2018). Polyamide membranes with nanoscale Turing structures for water purification. Science, Vol. 360(6388), 518-521. DOI: 10.1126/science.aar6308
Tay, P. K. R., Nguyen, P. Q., & Joshi, N. S. (2017). A Synthetic Circuit for Mercury Bioremediation Using Self-Assembling Functional Amyloids. ACS Synthetic Biology, 6(10), 1841–1850. doi: 10.1021/acssynbio.7b00137
Turing, A. M. (1952). The Chemical Basis of Morphogenesis. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 237(641), 37-72.
Yang, Xin-She. (2003). Modelling Simul. Mater. Sci. Eng. 11(321).