Team:Toulouse/Results
Results
Content
Attract
Tests on varroas
Aim: to prove again varroa attraction for butyrate
In the US patent US 8647615 B1, the concentration of butyric acid that attracts varroa mites is 4 % (V/V), but the final description specified the concentration is efficient from 0.00001 %.
In order to verify the results presented in this patent and to make sure we are able to perform such experiment, we designed an attraction test for varroas. Champollion University in Albi welcomed us in their lab to get access to varroas. At the time of this test, there were not a lot of varroas available, so we add to make only one test in order to hope a significant result. Hence, we used the mentioned 4 % butyric acid concentration, like the one first mentioned in the patent.
|
|
Figure 1: Butyric acid test pie chart and statistical test
This test demonstrated that a solution of 4 % butyric acid attracts varroas. However, in the Cytotoxicity part of the results, we also showed that this 4% concentration is lethal for bacteria. We therefore decided to aim for a minimal concentration of 0.00001 % butyric acid, even if further experimentation will have to demonstrate the efficiency of this lower threeshold.
Cloning butyrate genes
Aim: to create the genetic tool to implement a butyrate pathway in E. coli
We designed and ordered synthetic genes to express the full pathway minus the ccr gene. The latter was already placed behind lacI to be ready for circadian circle regulation. Thus, to build a biobrick with the full pathway, we cloned the ccr gene with all the genes necessary for butyrate production, as presented below:
Figure 2: assembly of the synthetic pathway leading to butyrate production (5192 Kb). The first arrow represents the promoter, the others represent genes. The green circles are for the RBS and the red circle is for the terminator. Purple genes are originated from Streptomyces collinus, blue genes from Clostridium acetobutylicum and yellow genes from Escherichia coli.
This construction (BBa_K1587004) was inserted in the plasmid pSB1C3 to be provided to the iGEM in the biobrick format.
Figure 3: Gel electrophoresis of the digestion of the (BBa_K1587004). The size of the first DNA fragment matches 5192 pb (butyrate construction) and the size of the second one (2070 pb) matches linearized pSB1C3.
Test of butyrate production
Aim: to produce butyrate in E. coli
In order to test butyrate production, we cultivated ApiColi under micro-aerobic conditions to simulate the growth conditions in the device. After having harvested the supernatant of the culture, we filtrated it prior to do NMR analysis.
We tested our genetic construction in the E. coli strain BW25113 but we did not detect butyrate production. To improve the production of butyrate, we repeated the experiment in a strain with a deletion of the phosphate acetyltransferase gene. This strain produces lower amount of acetate, a fermentation product which production is likely to compete with butyrate.
Figure 4: Test of butyrate production in E. coli strain deleted for pta gene. Culture under micro-aerobic conditions in 10 mL falcon, result obtained after 28.5 hours of culture.
Unfortunately we could not detect butyric acid in our NMR analysis, but we observed significant differences for other fermentation products, proving that the synthetic pathway is actually interfering with the fermentation products metabolism. Additional experiments will be needed to go further (test of our enzymes expression, measurements of intracellular pools to identify metabolic locks, growth at other pH, deletions to prevent the accumulation of other fermentation products as explained in [1]...).
Conclusion: butyrate production has to be improved in our strain and under our conditions
In a nutshell:
Butyrate is an effective attractant for varroas. Its synthetic pathway has been implemented in E. coli. Butyrate production from this pathway remains to be demonstrated prior to testing the strain with varroas.
Eradicate
Tests on varroas
Aim: determinate the sufficient formate concentration to kill Varroa destructor
In order to determinate the amount of formic acid required to kill the mite, we tested different concentrations of formic acid on varroas as explained in the Protocol part.
Figure 5: Mortality of varroas as a function of time for different formic acid concentrations
Figure 6: Histogram representing mortality of varroas after 2 hours and after 7 hours
Figure 7 presents a dose-dependency of formic acid on varroa mortality. At 10mM formic acid, all varroas died before three hours but as we explained in the Protocol part, varroas also stop moving at lesser concentrations. Figure 6 shows that even with 50 µM of formate, around 30 % varroas died after 7 hours (so close to a night duration). We therefore set our production goal around 50 µmol.L-1.
Conclusion: 50 µmol.L-1 could be sufficient to kill varroas
Cloning formate genes
Aim: to create the genetic tool to implement a formate pathway in E. coli
We designed and ordered synthetic genes to express the full pathway. Thus, we only had to clone the synthesized genes into pSB1C3 to be sent to the iGEM registry.
Figure 8: assembly of the synthetic pathway leading to formate production (3090 Kb). The first arrow represents the promoter, the others represent genes. The green circles are for the RBS and the red circle is for the terminator.
This construction (BBa_K1587007) was inserted in the plasmid pSB1C3 to be provided to the iGEM in the biobrick format.
Test of formate production
Aim: to increase the production of formate in E. coli
For formate production, we designed a synthetic polycistron coding for pyruvate formate lyase and its activator protein. The polycistron was cloned in a pUC57 plasmid to be tested for formate production. We then made a biobrick by subcloning the polycistron in the pSB1C3 plasmid (BBa_K1587007).
Figure 9: extracellular substrate and product concentrations with or without formate polycistron overexpression under micro-aerobic conditions.
Figure 9 shows that the only difference between ApiColi and the control is in the formate accumulation, so we plotted the specific histogram for formate after 3 days (Figure 10). This highlights that formate production is increased significantly by 10%.
Figure 10: Formate production tests after 3 days cultivation under micro-aerobic conditions.
As mentioned earlier, our goal was to produce 50µM of formic acid in 7 hours. To reach this quantity, we needed to produce 77mM of formate. We observed a maximal formate concentration of 25mM (Figure 9). Therefore our production level is not far from the target. By improving the strain and/or conditions (as for Test of butyrate production), we can reasonably hope to reach this goal.
Conclusion: formate production was increased with our synthetic construction
In a nutshell:
formate is effective to kill varroas from as low as 50 µmol.L-1. Overexpressing its biosynthetic genes enables obtention of a 10 % increase in production. Optimizing this production will have to be achieved before testing the strain capacity to kill varroas.
Regulate
NOT logic gate
Aim: NOT logic gate construction
We want to express the butyrate synthetic pathway by day and the formate synthetic pathway by night. As a first step to develop our regulation module, we replaced the butyrate pathway by the RFP reporter gene and the formate pathway by GFP reporter gene. Doing so, it should be easy to assess the functioning of the light sensor and the NOT logic gate since the colonies should be red by day and green by night.
To do so, we synthesized the pOMPC-cI gene (new biobrick BBa_K1587006)Through subcloning the GFP reporter gene (BBa_E0240) was placed under control of this promoter. Likewise, the RFP reporter gene (BBa_K081014) was cloned behind the PL-lacI synthetic gene. Both constructions were placed together on the same plasmid but unfortunately sequencing revealed that there was some incoherence in the RFP sequence, preventing further testing.
Figure 11: NOT logic gate test construction
Conclusion: the strain can grow and survive over 10 days in the TPX® bag.
Light sensor
Aim: Light sensor system construction.
For our project we need to implement a light sensor system in our strain. To that end, we needed to clone the three genes (cph8, hoI, pcyA) of this system behind a strong promoter. We received the strain containing the three genes on different plasmids as courtesy from Dr Clark Lagarias, Christopher Voigt, and Nico Claassens [2]. These genes were amplified by PCR and subcloned in pSB1C3 (at this step, a RBS was added to sequence during the PCR). The new biobricks for hoI and pcYA were sent to the registry under the accession number BBa_K1587000 and BBa_K1587002, respectively. The cph8 biobrick was not obtained but we managed to build a biobrick (BBa_K1587008) with the cph8 encoding sequence under the control of a strong promoter originated from the biobrick (BBa_J23119).
Figure 12: Targeted light sensor construction
Figure 13: :cp biobrick
We did not succeed in associating the three genes together behind a strong promoter in pSB1C3 using the In Fusion cloning technique before the iGEM deadline. Consequently, we were not able to test our light sensor system soon enough for this project.
Conclusion: the light sensor system is not yet achieved.
In a nutshell:
About 70% of the clonings for the light sensor and the NOT logic gate has been done, but the last cloning steps remain to be finished before the tests.
ApiColi containment and culture
Growth tests in TPX bag
Aim: to perform the growth of the modified strain in a sealed container
We investigated if the bacteria could grow inside a small bag of TPX®. Thus, the strain E. coli BW 25113 has been inoculated in a small bag that was sealed. The bag was placed inside a tube and incubated at 37 °C. Growth was visualy assessed in the bags after 17 hours at 37°C (figure 14), or with a culture in tube with agitation as a control over the same period (figure 15). This showed that the cells were able to proliferate in the plastic bag.
Figure 14: Growth test of E. coli BW 25113 inside a small bag of TPX®.
(t = 17 hours, 37 °C).
Figure 15: Control of figure 14 with a growth test of E. coli BW 25113 in a culture tube.
The culture tube contains bacteria growing in parallel of the biological sample tubes shown above (t = 17 hours, 37 °C, 130 rpm).
We then assessed the biomass evolution during long term growth in the TPX® bag (7 to 10 days). Figure 16 displays the result. The growth tendency is toward a continous increase over the experiment time.
Figure 16: Growth test in TPX® bag by monitoring OD at 600 nm over 7 days to 10 days.
We then checked the viability of the cells after 10 days of growth in the TPX® bag (tube 1, 1' and 2) or in the control tube (tube 3), by spreading diluted volume of the cultures on Petri plates. Similar numbers of colonies were obtained, indicating that the cells survival is the same in both culture conditions.
Figure 17: Colonies of E. coli BW 25113 on Petri dishes after an overnight incubation at 37°C to check survivability.
Conclusion: the strain can grow and survive over 10 days in the TPX® bag.
Gas diffusion tests
Aim: ensuring that formate and butyrate permeate through the TPX® bag
For our system to be efficient, it was necessary to check that butyrate and formate could permeate through the TPX® bag. This was verified as described in the Protocol section. Shortly, a solution of butyrate or formate was placed in the TPX® bag. The molecules exchanged between the bag and the surrounding air were desorbed in a solution of sodium bicarbonate. This resulting solution was analysed by NMR.
On a first assay, we did not detect any permeation of butyric acid through the TPX® bag. The same experiment was therefore performed by placing directly the bag in the solution of sodium bicarbonate. The control is an injection of a 4% (V/V) solution of butyric acid in water.
Figure 18: NMR Spectrum of butyric acid liquid control in red and butyric acid liquid which passed through TPX bag in blue. * Blue curve is zoomed 1340 times more than red curve. Each condition was tested in two replicates.
Table 1: Concentrations of butyric acid calculated from the NMR spectrum
From these results, we can conclude that the TPX® enables butyric acid
to weakly pass outside the bag. We detected only a small quantity but
an optimization of the device could be made with a plastic containing
bigger pores.
For the formic acid, we were able to detect it in the gas, probably because its pKa is lower than the butyric acid one.
Figure 19: NMR spectrum of formic acid gas control in red and formic acid gas which passed through TPX bag in blue.
Table 2: Concentrations of formic acid corresponding to NMR spectrum
According to these results, TPX® lets 56% of formic acid to pass outside the bag in the gas phase. Therefore, formic acid permeats through the TPX plastic. Using a more porous plastic as proposed for the butyrate, we supposed this percentage to further increase.
Conclusion: formate permeates through the TPX® while butyrate exchange will need to be improved
Safety tests
Aim: demonstration of TPX® capacity to retain the bacteria
The bacteria impermeability of the TPX® was tested through inoculation of the strain
E. coli BW 25113 in M9 defined medium complemented with glucose inside the bag. Then, the inoculated bag was immersed in a glass measuring cylinder containing M9 medium with glucose. OD600 nm served to monitor a putative growth in the external medium.
Figure 20: Measuring cylinders used for the safety test of the TPX® polymer.
The cylinder on the left contains the TPX® bag with E. coli BW 25113 immersed in M9 medium after 27 hours of growth at 37 °C. On the right, the negative control cylinder contains a bag of TPX® without bacteria after 27 hours of growth at 37 °C.
Over this time, no growth was detected in the medium surrounding the plastic bag. We concluded that the bacteria are contained in the bag.
Conclusion: bacteria cannot permeate through the TPX® material
In a nutshell:
The TPX® material successfully contains growing bacteria while allowing the diffusion of formate and, in a lesser extend, of butyrate. These tests will profit to all Igem teams looking for a solution to contain their strains .
Growth tests
Our final objective was to prepare a bag containing bacteria producing alternatively butyric acid (during the day)
and formic acid (during the night) and for a period of time of at least ten days (so that beekeepers don't have to change them every morning...).
So we faced some biological questions:
- Can bacteria live during ten days under micro-aerobic conditions?
- Which carbon source is suitable to produce continuously acids?
- Are formic and butyric acids toxic for E. coli?
In order to characterize the E. coli strain growth in the conditions we plan to use in our device, we used a culture under aerobic and micro-aerobic conditions, measured the growth rate (OD) and analyzed the supernatant to measure the concentration of fermentation products.
Micro-aerobic conditions were obtained with culture in specific Falcon tubes with holes and covered with a membrane that let the oxygen pass through. They were incubated at 37 °C without agitation (mimicking the aerobic and lack of agitation condition present in our device).
Aerobic conditions were obtained with a classical Erlenmeyer incubated at 37 °C with agitation.
We used a minimal M9 medium to identify formic or butyric acid production by NMR.
Biomass, substrate and products
Aim: to measure formate production of wild-type strains in microaerobic conditions
In order to plot biomass concentration we converted the OD values with the following equation:
$$ X=OD_{600nm}\times 0,4325 $$
Where X is the cell concentration (g.L-1)
Substrate and products concentration was inferred from the peak area of each molecule on the NMR spectrum.
Concentrations were calculated with the following equation:
$$[A]=\frac{Area_{molecule}}{Area_TSP} \times [TSP] \times \frac{\textrm{TSP proton number}}{\textrm{A proton number}} \times DF $$
- [A] = concentration of molecule in our solution in mM
- AreaTSP = 1
- [TSP] = 1.075 mM
concentration of Trimethylsilyl propanoic acid in NMR tube, internal reference for quantification - TSP proton number = 9
- DF = Dilution Factor = 1.25
Biomass and concentration of various molecules were issued from fermentation under aerobic and microaerobic conditions.
Figure 21: Results of micro-aerobic and aerobic culture after 5 days. Culture of E. coli BW25113 in M9 medium with [glucose] = 15 mM, at 37 °C
Glucose was consumed approximately at the same rate for both conditions but was converted differently depending on the condition. Under aerobic conditions the biomass reached 3 g/L whereas under micro-aerobic conditions there was six times less biomass. On the contrary, there were far less products in aerobic conditions.
Our results were in line with our objective to produce acids in a microporous bag because we demonstrated that with microaerobic conditions, bacteria grew slowly and produced fermentation products.
To convert the concentration of formate into formic acid we used the famous Henderson Hasselbalch equation:
$$ pH=pKa+log \left(\frac{C_{b}}{C_{a}} \right) $$
- pH = 7 (pH of the M9 culture).
- pKa: 3.7 for formic acid and 4.8 for butyric acid
- Cb: base concentration
- Ca: acid concentration
As mentioned in the Eradicate part, our goal was to produce 50 µM of formic acid to kill varroa, and this corresponds to 77,7 mM of formate.
Under our microaerobic conditions, the wild type E. coli strain produced 32 mM of formate. It was therefore necessary to further improve the metabolic production by adding genes involved in formate production in order to increase it by a factor of 240%. For a perfect regulation it would have been necessary to delete the chromosomic version of pflB in E. coli genome to avoid formate production during the day.
Conclusion: the wild type E. coli strain produces formate in microaerobic conditions but production has to be further improved
Bacteria survival
Aim: to verify how E. coli survives in long culture conditions
As explained here calculation of the survival of bacteria was perfomed via colony counting after plating on solid agar medium. Wild-type bacteria can easily survive at least 15 days in aerobic or microaerobic conditions without a carbon source. We hypothesized that in presence of a carbon source, we may be able to extend even further their survival time period.
Figure 22: Bacteria survival test performed with the BW25113 strain on M9 with 15 mM of glucose during 15 days.
In a nutshell:
E. coli can produce important amount of fermentation products, among which formate, when cultured in microaerobiosis. The strain can survive up to two weeks
in normal culture conditions. This validates the possible use of long lasting cultures in our device.
Choice of the carbon source to sustain formic and butyric acids production during 10 days
Characteristics of the Biosilta kit
Aim: optimizing the release of glucose using Biosilta kit
Biosilta kit is a technology which enables production of a lot of recombinant proteins owing to a slow carbon source delivery during 24 hours. This technology is based on polymer degradation by an depolymerisation enzyme leading to a perfect control of the quantity of substrate released in the medium for a given period of time. We wanted to use this technology to cultivate our cells during one or two weeks in conditions allowing the production of butyrate and formate. The medium with the polymer was solid and contained in separate bags. In order to quantify the production of butyrate and formate in these conditions, we first performed a test with a high concentration of the enzyme (50 U/L) and measured the kinetics of the polymer degradation.
Figure 23: Kinetics of Biosilta polymer degradation measured with glucose release.
In order to have a global idea of the release of glucose per unit of time we then calculated the average rate of release with the following formula:
$$ v_{glucose1}=\frac{[glucose]}{time}=\frac{11.1}{3.97}=2.80 g.L^{-1}.h^{-1} , for [E]_{1}=50 U.L^{-1} (4) $$
With a final glucose concentration of 13 g/L for one bag of polymer, the rate of glucose release necessary to have glucose during 13 days can be calculated with the following formula.
$$ v_{glucose2}=\frac{13}{13 days}=\frac{13}{322 hours}=0.0403 g.L^{-1}.h^{-1} (5) $$
The reduction factor was calculated:
$$ RF=\frac{v_{glucose1}}{v_{glucose2}}=\frac{2.8}{0.0403}=69.44 (6)$$
So, the concentration of enzyme that we had to use should have been:
$$ [E]_{2}=\frac{[E]_{1}}{RF}=\frac{50}{69.44}=0.72 U.L^{-1} (7)$$
Growth assessment using the Biosilta kit
As we do not know any growth characteristics of bacteria in the Biosilta medium we tested different enzyme concentrations and not only the one calculated previously. Acquisition of the growth stopped due to a problem with the plate reading system where the bacteria where growing. Hence, there is a break at 5 days.
Figure 24: Bacteria growth as a function of different enzyme concentrations in Biosilta medium. Growth tests were performed in 48 wells plates attached to an OD reader.
Except for the 1.5 U/L enzyme concentration, there was an increase of OD in all conditions during 12 days. This means that glucose releasing worked as expected. At the beginning there is an exponential growth because some glucose is directly available in the medium. This phase was therefore not the one to analyse. After 2 days and until the end, the growth rate changes gradually and becomes constant, at the highest between 2 and 4 days and decreasing thereafter.
We therefore knew that it was possible to have continuous growth during at least 12 days. However, in our control without enzyme, there was also some bacterial growth so either another substrate was available or bacteria could realy degrade the polymer. More tests will be needed to verify if and how E. coli can use this polymer.
In order to answer these questions we performed some cultures in Falcon tubes to analyse the products and the evolution of the polymer quantity. Culture without enzyme and with a concentration of enzyme of 0.72 U/L were performed.
Figure 25: Results of BW25113 culture on Biosilta medium after 4 days. Test of two conditions: without enzyme and with [Enzyme] = 0.72 U/L
Figure 25 (right panel) shows that the polymer is not degraded by E. coli which is probably capable of finding another carbon source within the Biosilta medium. The concentration of products (left panel) is clearly different in the two conditions: with the enzyme, more fermentation products are formed.
Fermentation products have high concentrations in comparison to the cultures in performed in M9 (15 mM glucose), around 20 times more for lactate, 3 times more for acetate and 2 times more for ethanol.
We hypothesized that by deleting the various production pathways for lactate, acetate and ethanol and the degradation pathway for
formate we should be able to produce enough formate.
Modelling of formic and butyric acid production
Aim: modeling the production of formic and butyric acids as a function of the glucose release rates.
With the rate of glucose release calculated above, an FBA and FVA simulation were launched as explained in Modeling part.
Some conversion between the model and the real condition are necessary and they are also explained there.
In order to model production in the most similar conditions to fit real experiment we chose a glucose rate of 0.0403 g.L-1.h-1 that correspond to 0.72 U/L of enzyme.
To convert formate production into formic acid concentration we used equation (3).
Figure 26: Modeling of formic acid production as a function of different growth rates for a glucose release of 0.0403 g.L-1.h-1.
Our goal was to produce at least 50 µmol/L. The graph clearly shows a maximum production of 6 µmol/L. To reach our goal we then have to change the growth rate.
Figure 27: Modelling formic acid production as a function of glucose release for different growth rates (in h-1).
To produce 50 µmol/L of formic acid different strategies are possible. If we choose a low growth rate then a low glucose rate is be necessary and vice-versa. As bacteria have to live during at least ten days it was better to have a continuous slow growth rate. Moreover, it would consume less glucose per hour so we would need a lower polymer concentration in our bag at the beginning. Thus, we chose a growth rate of 0.2h-1, and we could determinate the glucose release value needed.
[Formic acid] (μmol.L-1 )=166.88 ×[Glucose] (g.L-1.h-1 ) (9) $$ [Glucose] = \frac{50}{166.88}=0.3 g.L^{-1}.h^{-1} (10) $$
Now, we will see which butyric acid concentration we can theoretically produce.
Figure 28: Modelling of butyric acid production as a function of glucose release for different growth rates.
According to the modelling results described in figure 28, we have to produce around 100µmol/L of butyric acid
that corresponds to 0.0092% (V/V). As explained in the "Results" part, "attract" section,
our objective was to produce at least 0.00001%, so the modelling indicates that we can theoretically reach it.
Nevertheless, in order to have this glucose release production, it was necessary to calculate
how much polymer and enzyme concentration are needed.
With the same equations as we used in Characteristics of Biosilta kit
we determined which total quantity of glucose is needed for a fortnight.
$$[Glucose]=v_{glucose}\times time=0.3\times 322=96.6 g.L^{-1} (11)$$
Knowing that one Biosilta kit contains the equivalence of 13 g/L of glucose, we should therefore concentrate the medium 7 times. To obtain a glucose release of 0.3 g.L-1.h-1 value, we nee 5 U/L of enzyme. Thus, we tested different concentrations of Biosilta medium with different enzyme concentrations.
Testing different concentrations of Biosilta medium
Aim: to determine if we can concentrate Biosilta medium to improve the concentration of fermentation products.
We do not have the exact composition of Biosilta medium and cannot ascertain innocuity of higher titers of the polymer on growth. The best approach was to actually test it directly.
Figure 29: Bacteria growth as a function of time for different concentrations of Biosilta medium. Culture with BW25113 on 48 wells plate and optical reader
From these three figures it is evident that we cannot reach the growth level that is expected and that concentrated Biosilta is actually detrimental to growth.
In a nutshell:
The Biosilta medium was successfully used to have a continuous release of glucose over two weeks. Optimizing the protocol settings allows improving the production of fermentation product. The modeling suggests that concentration the Biosilta medium should even improve formate and butyrate productions but the experimental validation of this production has shown that this is not compatible with E. coli growth.
Testing acids toxicity
Aim: to test butyric and formic acid effects on E. coli growth.
In order to measure BW25113 resistance to different acids concentrations we tested two medium: LB and M9 with 15 mM glucose.
Figure 31: Optical density in function of time for different formic acid concentrations. LB medium is represented with green curves and M9 medium with blue curves. Each condition is tested in three replicates and standard deviation is represented in orange.
Figure 32: Optical density in function of time for different butyric acid concentrations and two medium. LB medium is represented with green curves and M9 medium with blue curves. Each condition is tested in three replicates and standard deviation is represented in orange.
In M9 medium, the growth is slower at the beginning of the growth curve, however the maximum value of OD reached is almost the same for both medium.
There is a visible difference between the two culture media. Growth in LB is clearly less tolerant to either formic or butyric acid. As M9 (and not LB) is a buffered culture medium we measured the pH of both media with different acids concentrations.
Figure 33: pH of the two media in function of the concentration of formic acid and butyric acid. LB medium is represented with green curves and M9 medium with blue curves. pH was measured with pH paper. Each condition was tested three times.
It is clear that in M9 medium the pH remains at pH 7 at high acid concentrations but not in the LB medium. Moreover we now know that bacteria do not grow when the pH is around 5. We will now see if buffering our medium has a positive action on growth.
In a nutshell:
Butyric and formic acids production does not stop the growth within our targeted ranges of concentrations.
READ MORE
References
- [1] Saini M, Wang ZW, Chiang C-J & Chao Y-P (2014) Metabolic Engineering of Escherichia coli for Production of Butyric Acid. J. Agric. Food Chem. 62: 4342–4348
- [2] Levskaya A, Chevalier AA, Tabor JJ, Simpson ZB, Lavery LA, Levy M, Davidson EA, Scouras A, Ellington AD, Marcotte EM & Voigt CA (2005) Synthetic biology: Engineering Escherichia coli to see light. Nature 438: 441–442
