m |
m |
||
| Line 6: | Line 6: | ||
<!-- ADD YOUR ARTICLE HERE --> | <!-- ADD YOUR ARTICLE HERE --> | ||
<html><div id='Container'> | <html><div id='Container'> | ||
| - | </html>{{:Team:ETH_Zurich/labblog/20140909hum}} | + | </html> |
| + | {{:Team:ETH Zurich/labblog/20140915hum}} | ||
| + | {{:Team:ETH_Zurich/labblog/20140909hum}} | ||
{{:Team:ETH_Zurich/labblog/20140908meet}} | {{:Team:ETH_Zurich/labblog/20140908meet}} | ||
{{:Team:ETH Zurich/labblog/20140906}} | {{:Team:ETH Zurich/labblog/20140906}} | ||
We are participating to a project of weekly newsletter launched by Paris Bettencourt. We are giving updates on our project every week and other articles such as description of one member of the team. The idea is also to ask questions to other teams on their projects in order to get feedback.
In the last days Stefanie met an architect and a priest to discuss complexity in their occupational field. Both conversations were highly interesting and at the same very different. It was fascinating and inspiring to once change the point of view and to approach complexity from new positions. For the priest, Josef Fuisz, the struggle of men with complexity might be represented by the following psalm (42.5):
Why, my soul, are you downcast?
Why so disturbed within me?
Put your hope in God,
for I will yet praise him,
my Savior and my God.
Other interesting interviews were done in various fields. For example a philosophy teacher, a physicist, a nano-technologist and a theoretical computer scientist were partners for talks. From these interviews we expect to gain insight into the diversity of complexity. It became even more clear than before that complexity can be found in many different fields.
Visit our interview page for the details of each conversation.
Until now we thought Lux was crosstalking too much with las and tried to improve the pair Las-Rhl. However we realized today that Rhl is difficult to improve and that our induction curves of Las and Lux are good enough, since Las is never responding to lux-HSL and we only have to be in a range of las-HSL concentrations that is below 10 nM to avoid Lux responding to las-HSL.
Cross-talk experiments setup is described in this article .
In order to predict accurately the experiments we designed, we defined two different strategies to retrieve parameters for the integrase system.
As we did not find these parameters on the web, we used the data given by Bonnet et al.'s paper "Amplifying genetic logic gates" [9] . The figure S4 is particularly interesting, even if it does not give any information on the dynamics of the system. We modeled and wrote the equations corresponding to Bonnet's experiment.
At first, we try to fit the dynamic parameters of the system with only the steady states values. To fit the parameters with the data, we first try to use a minimum search for the difference between the data and the parametric function. Then, using these first approximations, we tried to do a Markov Chain Monte Carlo approach to find a more suited curve. As these two methods are strongly prior dependent, we planned another step of screening in order to have meaningful results. However, our knowledge on the parameters, the number of data points compared to the high number of parameters did not give any meaningful results.
We then decided to simplify our system :
As we tried to fit the steady state, we obtained two different models, depending on which assumption we made. As these two models fit the Bonnet's curve well, we used them both to derive more precisely the dissociation constant of the dimerization reaction and the one of the DNA binding reaction. Each of them appear in one of the lumped parameters we fitted.
By considering reasonable order of magnitude for other parameters present in the lumped parameters, we find that the value of the dissociation constants are both of the order of 10-4. This order of magnitude is not surprising at all, as we know that the binding of integrases to DNA is very specific (which is underlined by the low value of the dissociation constant of this reaction) and that integrases are more stable as dimer than as monomer.
Now that we tested the different constructs for leakiness, dose response and cross talk, we are investigating in the diffusion rate of AHL. Therefore we designed a chip with chambers connected by channels of different length (see figure). In one chamber we added a strain containing a regulator and a sensor plasmid, to the second chamber we added AHL or an AHL producer strain. As soon as AHL is diffused to the former chamber it induces sfGFP production on the sensor plasmid. The time it takes until we can measure the sfGFP signal and the length of the channel gives us information about the velocity of AHL diffusion.
Three main experiment designs were tested and modified: the agar-design, the liquid-design and the beads-design.
Check the corresponding article for more details.
We could use quorum sensing dynamic curves to fit degradation rates of external, internal AHL and GFP. We fitted the experimental results for lux with riboregulator and thus didn't consider leakiness (we showed leakiness was low enough in that case when fitting quorum sensing Hill function parameters ). By writing GFP production rate and replacing AHL in this equation by $$[AHL]=[AHL_i]-[AHL_{bound}]$$ we find $$\frac{d[GFP]}{dt}=\frac{k_{45}[P_{tot}]\Big(-1+\sqrt{1+8\frac{AHL_i}{K_{d3}}}\Big)^2}{16\frac{K_{d4}}{K_{d3}}+\Big(-1+\sqrt{1+8\frac{AHL_i}{K_{d3}}}\Big)^2}-d_{GFP}[GFP]$$. For very high initial concentrations AHLi, we have $$\frac{d[GFP]}{dt}=k_{45}[P_{tot}]-d_{GFP}[GFP]$$
so by taking
$$t_{1/2}=\frac{ln(2)}{d_{GFP}}$$
from experimental curves, we are able to retrieve
$$d_{GFP} = 4.9 . 10^{-3} min^{-1}$$. For very low initial concentrations of initial AHL and considering degradation, we have
$$ \frac{d[GFP]}{dt}=\frac{k_{45}[P_{tot}][AHLi]^2e^{-2d_{AHL}t}}{K_{d4}K_{d3}}-d_{GFP}[GFP].$$
We find
$$ [GFP]=\frac{k_{45}[P_{tot}][AHLi]^2}{K_{d4}K_{d3}(d_{GFP}-2d_{AHL})}(e^{-2d_{AHL}t}-e^{-d_{GFP}t}).$$
This curve has a maximum at
$$ t_{max}=\frac{1}{d_{GFP}-2d_{AHL}}ln\big(\frac{d_{GFP}}{2d_{AHL}}\big).$$
This way we can find from experimental curves $$d_{AHL}=4,0.10^{-3} min^{-1}$$
This AHL degradation rate is alumped parameter between internal and external degradation rates, equivalent to
$$d_{AHLext}+\alpha d_{AHLint}$$
where α is the fraction of the volume occupied by cells in the whole culture. From that on we could find reasonable values for internal and external AHL degradation rates.
Finally the curves simulated by our model with these parameters fit the experiments quite well:
We have a Comsol simulation for diffusion experiments that explains why at 1mM AHL concentration in sender wells, all 4 wells glow at the same time irrespectively of channel length. On the other hand if we use μ , the wells should start glowing at different times. The channels are 1 to 4mm long. We used a Hill function to model activation of GFP in the receiver wells.
The design of punched agar doesn't seem to be the best but at least we can see something. Diffusion experiments with different designs are running now.
The plasmids for quorum sensing experiments concerning the Lux system are made and operational to be used for experiments. We have three different plasmids:
The third plasmid corresponds to a riboregulating system. The second plasmid allowed us to validate the crucial role of taRNA in this riboregulating system.
On the modeling point of view, we only tried to fit the data coming from experiments done with plasmid one (with no riboregulating system) and plasmid three (with an effective riboregulating system). They were the only one to show an answer to a change of input that was not on the noise level.
The experimental setting was to inject different concentration of the signaling molecule AHL at the same time point into different bacterial colonies. The factor that we want to observe is the strength of the activation of the Lux promoter. The read-out is made by the fluorescence. The data was gathered over several time points in order to be able to observe the dynamic behavior. More precisely, the read-out of the promoter activation is made by the relative fluorescence, which corresponds to the ratio of the absolute fluorescence over the Optical Density (OD). We consider that the concentration in GFP per cell is proportional to the relative fluorescence. As we try to avoid fitting the saturation events, we chose to normalize with respect to the highest value before a certain amount of time, which corresponds for each experiment to the apparition of the saturation.
As a first approach, we only try to fit the steady state value. As the signaling molecule is inducing the production of GFP, we chose to use an Hill function to fit the function [GFP] = f([AHL]) at steady state.
If Hill functions fit these two curves good, the parameters used are surprising. For our system, we suppose that AHL binds to the regulator protein LuxR. This complex called Rlux dimerizes before binding to the promoter to induce its expression. Thus, the dimerization event makes cooperativity emerge from AHL molecules. It is translated for a Hill function by n > 1. In this particular case, one could expect to find n=2 (as it is a dimerization event). Nevertheless, the n found are on the order of magnitude of 1 (0.89 for the non-riboregulated promoter plasmid and 1.22 for the ribo-regulated promoter plasmid). The data points were collected in triplicates and this allow us to have a look at the variability (namely standard deviation) around each value. Where the switch happens, the variability is higher (it is expected because the output GFP is more sensitive to variation of the input [AHL] around the switching point). Therefore, the inflexion points could need some improvment, by taking more points around this identified sensitive region. Moreover, this critical concentration of AHL was the one that was expected given the litterature and our system. We observe that it needs a higher [AHL] for the riboregulated system to be activated than for the non-riboregulated system.
We modeled leakiness as an offset in the Hill function : [GFP] = g([AHL]) = f([AHL]) + aLeakiness. We obtain the expected information that the signal over noise ratio is greater by one order of magnitude for the riboregulated system than for the non-riboregulated system. A drawback from the riboregulated system is that the amplitude of the signal corresponds to the amplitude of the noise in the non-riboregulated system.
The experimental design also allowed us to detect eventual cross-talk with other signaling molecules, the one corresponding to the las system and the one corresponding to the rhl system. No cross-talk was observed with the signaling molecule of the rhl system. However, the las system signaling molecule activated the Lux promoter. We chose to model this cross-talk, which corresponds to the induction of a promoter, by an Hill function. This fitting allowed us to verify that the Hill function was adapted to describe a cross-talk phenomena. The derived parameters indicated that the concentration in the las signaling molecule had to be high in order to have an effect on our system (It was two orders of magnitude away from the needed concentration of AHL to activate the system). Nevertheless, once activated, the effects of these crosstalk are not to be neglected. Check the following article for more information on the experimental setting.
These results are promising and we look forward to fit the dynamic data of this system.
After having constructed the necessary regulator and sensor plasmids we were able to start first cross talk experiments. Therefore we transformed competent cells with different combinations of regulator and sensor plasmids. One example would be the strain siG0024 which contains the regulator plasmid piG0041 producing LuxR under a constitutive promoter and the sensor plasmid piG0051 that encodes for sfGFP. The latter is under control of a pluxR promoter, hence it will be transcribed in the presence of LuxR and Lux. To assess crosstalk and leakiness we conducted multiple experiments: after adding one of the three AHLs to the cultures we analyzed the raise in sfGFP production using a microtiterplate reader. In a first step we measured the dose response of constructs with different ribosomal binding sites (RBSs) and with or without riboregulator. We could observe differences not only in sensitivity, but also in leakiness; some strains expressed GFP even without AHL induction. To give an example: we evaluated the sensitivity of the siG0024 strain to different concentrations of 3OC6-HSL (LuxI product) and compared its characteristics to those of other constructs. Additionally, we measured the sensitivity of siG0024 to 3OC12-HSL (LasI product) and C4-HSL (RhlI product). Indeed, we could observe the induction of sfGFP transcription by comparably low levels of 3OC12-HSL. Crosstalk can be observed between the Las and the Lux system. To simulate a situation most similar to the final experiment, we added the diluted supernatant of AHL producer cells to the test cultures. However, we could not detect any GFP production as reaction to the supernatant. We assume that the AHL concentration is not high enough to induce transcription. By constructing stronger RBSs on the producer plasmids we aim at increasing the AHL production. All these informations we will use to design a system tailored to our plans.
The figure shows the dose response of xxx to 3OC12-HSL, 3OC6-HSL and C4-HSL. Both 3OC12-HSL and 3OC6-HSL induce production of GFP, although at different concentrations. This is an example for cross talk between the lux and the las system.
We performed double transformations with different combinations of regulator + sensor plasmids for cross-talk experiments. We have results for our first quorum sensing experiment : fluorescence under control of wildtype lux sensor (s24) over time for different concentrations of AHL. s1 is a negative control.
We need an efficient way to simulate diffusion experiments. A receiver cell at the end of a channel receives AHL produced by the well at the other end of the channel, and starts to produce GFP. By using quasi steady state aproximations (considering that binding is much faster than transcription and diffusion), we have a simple equation for GFP production as a function of AHL :
By using producer cells on one side of the channel and receiver cells in the other side, and putting cells in a punched cylinder in agar, we should see fluorescence appear at different times depending on the channel length (1 to 4mm) :
Here is our first millifluidic chip for diffusion experiments ! This time we used a layer of PDMS thin enough so that we could peel it without breaking it. In some wells will be added producer cells or pure AHL, and in the wells on the other side of the channels, receiver cells. There are several lengths of channels, this should enable us to retrieve diffusion parameters by recording when receiver cells start glowing.
We are participating to a project of weekly newsletter launched by Paris Bettencourt. We are giving updates on our project every week and other articles such as description of one member of the team. The idea is also to ask questions to other teams on their projects in order to get feedback.
We will have a chip with these four different channels 1 to 4 mm long. The model still has to be refined because at the moment it doesn't take into account internal/external concentrations of AHL.
We are trying several strategies : minimization of the error function and Markov Chain Monte Carlo.
Competent cells were transformed with piG0028 (K553003, lasR) and selected on chloramphenicol-LB-plates. Plasmid DNA was extracted by performing a miniprep. The relevant plasmid sequence was sequenced by Microsynth using the primers oiG0001 and oiG0002. Amplification of lasR by PCR using oiG0003 and oiG0004 resulted in the fragment fiG0004 (1.0 kb). The vector piG0034 (pSEVA 181) was digested with the restriction enzymes HindIII and PacI to fiG0001 (3.0 kb). The two fragments, fiG0001 and fiG0004, were assembled using Gibbson assembly. Thus we used the plasmid backbone of piG0034 and lasR of piG0028 to construct the lasR regulator plasmid piG0040. Competent cells were transformed with piG0040 and selected on ampicillin-LB-plates. Colony PCR using the primers oiG0035 and oiG0036 was conducted to check the size of the inserted fragment (expected bands at 1.0 and 1.2 kb). The sequence was verified by Microsynth using the same primers.
Competent cells were transformed with piG0008 (F2620, luxR) and selected on chloramphenicol-LB-plates. Plasmid DNA was extracted by performing a miniprep. The relevant plasmid sequence was sequenced by Microsynth using the primers oiG0001 and oiG0002. Amplification of luxR by PCR using oiG0005 and oiG0006 resulted in the fragment fiG0005 (0.8 kb). The backbone of the vector piG0040 (lasR in pSEVA 181) was amplified by PCR using oiG0007 and oiG0008. This fragment, fiG0011 (3.2 kb), and fiG0005 were assembled using Gibbson assembly. Thus we formally replaced lasR by luxR to construct the luxR regulator plasmid piG0041. Competent cells were transformed with piG0041 and selected on ampicillin-LB-plates. Colony PCR using the primers oiG0035 and oiG0036 was conducted to check the size of the inserted fragment (expected bands at 1.1 and 1.2 kb). The sequence was verified by Microsynth using the same primers.
Competent cells were transformed with piG0023 (C0171, rhlR) and selected on chloramphenicol-LB-plates. Plasmid DNA was extracted by performing a miniprep. The relevant plasmid sequence was sequenced by Microsynth using the primers oiG0001 and oiG0002. Amplification of rhlR by PCR using oiG0009 and oiG0010 resulted in the fragment fiG0006 (0.8 kb). The backbone of the vector piG0040 (lasR in pSEVA 181) was amplified by PCR using oiG0011 and oiG0012. This fragment, fiG0012 (3.2 kb), and fiG0006 were assembled using Gibbson assembly. Thus we formally replaced lasR by rhlR to construct the rhlR regulator plasmid piG0042. Competent cells were transformed with piG0042 and selected on ampicillin-LB-plates. Colony PCR using the primers oiG0035 and oiG0036 was conducted to check the size of the inserted fragment (expected bands at 1.1 and 1.2 kb). The sequence was verified by Microsynth using the same primers.
The promoter site of the luxR regulator plasmid piG0041 was mutated by QuikChange site-specific mutagenesis to produce promoters of different strength. The primers oiG0031 and oiG0032 were used to establish piG0047, a plasmid with a promoter of intermediate strength (J23111). Analoguos the primers oiG0033 and oiG0034 were used to construct piG0046, a plasmid with a weak promoter (J23109). Competent cells were transformed with piG0046 and piG0047 and selected on ampicillin-LB-plates. The sequence was verified by Microsynth using the primers oig0035 and oiG0036.
We want to optimize quorum sensing to have non-leaky non-cross-talking constructs. We want to prevent cross-talk between integrases and check their dynamics with different quorum sensing promoters. Then we will test the XOR gate without production of LuxI, to check how it works without the loop, and we will finally test our final construct with the loop, hoping that the delay between GFP production and the possible second switching due to AHL production is long enough to have enough fluorescence.
The red line is our critical timeline. Numbers of days are very optimistic,therefore we multiply the whole timeline by 2. The stars stand for modeling inputs.
Sensor plasmids will be approx. 9.3kb long, and logic plasmids will be approx. 5,2kb long.
The idea of our system is to associate a quorum sensing part, which should be able to react to a certain concentration of input, with a integrase-base logic part, which should be able to derive a computation given the transformed inputs.
From a modeling point of view, it seemed obvious that we had to consider these two entities separately. That is why we chose to have a modularized approach, where we regrouped, on one hand, the equations corresponding to quorum sensing and, on the other hand, the ones corresponding to the integrase logic phenomena.
As the integrase module is not that well known, we further decomposed it into two successive parts. The first one only models the path integrases go through in order to bind to DNA. The second one models the switch behavior of the integrases (after they are bound to DNA) and describe the XOR logic of our system.
As we see, these modules are successive. After the logic module, we added a production module, which is responsible for the production of our outputs (GFP and an Homoserine Lactone that depends on the type of cell modeled). As one of our output is a messenger molecule, it has an influence on the quorum sensing module, which detects those molecules. Therefore, we have a feedback loop. This feedback loop comes from our choice to put only two different strains in our grid.
ΦC31 and Bxb1 are integrases. LuxR an RhlR are quorum sensing repressors. A riboswitch construct is placed around quorum sensing constructs to prevent leakiness of lux and rhl promoters. Type p colonies produce AHL by expressing the enzyme LuxI. Type q colonies produce Rhl by expressing the enzyme RhlI. In fact, we don't know yet which quorum sensing systems we will use. We will have to perform cross-talk experimetns in order to choose the ones that are the most orthogonal.
This week, we have talked with people from the microfluidics group and it came out that we should be able to use beads to encapsulate our cells and use a microfluidic chip. Therefore we will try to develop this possibility in parallel with the 3D-printed agar chip.
We have set up a modeling pipeline. We divided the modeling project into 3 parts :
We know that the modeling challenges will be to model how integrases work, and to model the delay between reception of a quorum sensing signal and production of a quorum sensing signal by the receiver cell. Indeed, a colony could switch itself OFF when it should be ON, if it receives QS1 and produces QS2, for example. If a delay is present between reception of QS1 and production of QS2, the colony will produce GFP before it is switched OFF, and as GFP is stable enough, it will stay visible during a long time and the self-switching OFF won't be observed. This delay should be modeled.
After more than one month of endless meetings and passionate debates, we finally chose the project that will keep us occupied in the next five months. From the beautiful pattern made by the Sierpinski triangles, we will focus on cellular automata and try to implement one.
Sierpinski triangles appear when the rule 90 is followed by every cell on the grid :
Ideally we will use a microfluidic chip. We could also use a 3D printed agar plate like this one to load the colonies. On this grid we can implement the rule 6, which is the simplification of rule 90 considered as a rule with 2 inputs : each cell computes a simple XOR gate of its two parents.
The logic part will be built with integrases and the colony-to-colony communication will use quorum sensing.
Every colony will receive two quorum sensing signals (QSp and QSq) from the two cells above it. These two signals trigger the production of two different integrases r and s in the colony. Integrases enable to build biological XOR logic gates by switching twice a terminator. Indeed, every integrase can switch the terminator only once. Thus if the colony produces only r or only s, the terminator is switched only once, so the terminator is OFF, and GFP and QS1 or QS2 are produced (depending on the colony). If the colony produces r and s, the terminator is switched twice, so it is ON and it blocks expression of GFP and of the quorum sensing molecule.
We need to :
On May 10th the public in Basel had the unique chance to get an insight into many different scientific laboratories and the work done there. It was the joint open house day of D-BSSE of ETH (Department of Biosystems Science and Engineering) and the Biozentrum of the University of Basel. The many different labs opened their doors to the public and many scientists were present to give interested people some details about their daily work. So did the ETH iGEM team 2014. The team was present with a poster showing the history of iGEM, the previous ETH iGEM teams with their projects and general information about synthetic biology. Additionally there was a slideshow giving a best-of photo collection of last years jamboree. The goal of this day was to inform the public about synthetic biology in general and specifically about the spirit and the many different projects of iGEM. Many people showed strong interest in truly student driven projects and are curious to follow our team wiki for the next months.
"
