Team:Uppsala/Modeling CellCellInteraction
From 2014.igem.org
(Difference between revisions)
Line 5: | Line 5: | ||
document.getElementById("tab1").innerHTML = '<h2>Introduction</h2><p>To facilitate the understanding of our multi-part system, a simulation for cell-cell interaction was developed in Java. Realistic data on movement and reproduction were used, so that our new introduced functions could be evaluated. This resulted in a two dimensional program where bacteria can interact with each other in real time, as if looked at in a microscope. In our scenario we consider the probiotic moving around on the small intestinal wall and reaching the outer surface of a Yersinia enterocolitica colony. By sensing its environment the bacteria can invoke changes in the movement pattern or kill other species in their vicinity. This provides us with a tool to predict how effective our introduced systems will be and add insight to how the systems can be improved. The program can easily be changed and applied to similar systems, hopefully making it useful for understanding and evaluating future projects.</p><h2>Java Design</h2><p>In order to keep track of many bacteria with individual properties we thought it convenient to use an object-oriented language. Since Java also has the advantage of being easily distributed and made executable through a web browser environment, we found it a perfect candidate. Most of the data that has to be visualized are dependent on the programs ability to measure distance. This was made possible by deciding upon a convenient scale and assigning the pixels a suitable size. In our simulation every pixel corresponds to 0.1µm, which makes it easy to visualize the 2x1 µm large bacteria and its movement.</p><h2>Introduced Data</h2><h3>Generation time & Flux</h3><p>Generation time of our bacteria range between 20-40 min and the simulation considers a very limited window of time and space. To work around this issue a flux of bacteria to the visible area has been introduced. Though this will only affect the probiotic since Y. enterocolitica are immobile at gut temperature (De Berardis et al., 2004). In other word, we assume that there are bacteria moving around outside our screen and that they “walk in” at a constant rate. They will still reproduce based on their generation time but to depend solely on this process to drive the simulation would make it very time consuming.<br><br>The generation time for the probiotic is based on E. coli data, once every 20 min. The corresponding value for Y. enterocolitica was a bit more tricky. It has been measured to 34 min at 30°C, which is assumed to be their optimal growth temperature (Aswathy Sreedharan, 2012). Since our simulation will take place at 37°C we decided to use 40min as an approximation. However, the simulation is not meant to run for more than a couple of minutes so the generation times have been translated into a chance of occurrence. The reproductions are also limited by available space so that bacteria are not able to stack on top of each other. </p><h3>Movement & Size</h3><p>Data regarding movement of bacteria other than E. coli is hard to come by so we decided to let the probiotic mirror that pattern. Hopefully motion of E. coli are not too divergent to render the simulation invalid for other species.<br><br>The swimming speed of E. coli is generally estimated to be 10-20 µm/s with tumbling events occurring every 0.5-1s. In the simulation we have tried to mimic this by setting average speed to 10µm/s, maximum speed to 20µm/s and added a stochastic variable allowing it to tumble about once every second.</p><h3>Activation</h3><p>If the movement of our probiotic brings it close to a pathogen, a switch will trigger. This will bring the bacteria into a state where it continuously tumbles and produce a, for Y. enterocolitica, toxic substance. The distance at which the state activates is dependent on the concentration of Acyl-Homoserine-Lactones (AHLs), produced by the pathogen, as well as the logistic function to account for leakage. This was done using a Markov model with a transition probability that increases rapidly around a specified threshold. The probability is calculated as 1/(1 + e^-x) where x is an activation distance yet unknown in our system. While active it uses the same transition probability to return to its inactive stage, which will increase as surrounding pathogens die. An emission probability that oscillates between 100% and 0%, depending on state, will decide if the bacteria produces its toxin or not.</p><h3>Colicin Production</h3><p>In order to kill pathogens, our probiotic produce a bacteriocin called colicin. In this process we assume that colicin will diffuse in three dimensions, as a half-sphere, with a rate that correlate to its diffusion coefficient, 4.2E-7 cm2/s (Schwartz and Helinski, 1971). The strength of the signal when reaching a pathogen is proportional to the production rate and the time it takes for the bacteriocin to reach its destination. This time can be approximated by inserting the diffusion constant in the formula for Brownian motion in three dimensions.<br><br>Since our Killing system will not stack colicin within its cell wall, but produce and export it at constant rate depending on the Targeting system, data of production rate had to be estimated. In a previous study on colicin production in mutated E. coli it was observed that colicin activity could be increased by a factor 10 in 3-6 hours. By this time the average cell contained about 250.000 copies, which meant that 225.00 copies of colicin had been produced within this time window. If we assume a quite optimistic four-hour window this translates to a production rate of 15.625 copies/s, which we used in the simulation (Spangler et al., 1985). </p><h3>Killing</h3><p>Whether or not a pathogen is killed is decided by measuring the combined signal acting upon it. If this signal exceeds a threshold the bacteria will be declared as dead and removed from the simulation. This threshold has been decided based on a unit, called kill unit, defined as the lowest colicin concentration inhibiting growth (Spangler et al., 1985). Initially kill units were approximated from solutions of bacteria in mid exponentiation phase, where each ml corresponded to a specific value. Since mid-exponentiation phase gives us an OD600 of 0.4 we can estimate that the solution contained about 2.4-3.2E8 cells/ml. From section 3.4 we obtained that these mutant bacteria produce an average of 250.000 colicin/cell, which correspond to 5000 kill units/ml. When combining these data we get that 1 kill unit can be converted to 1.4E10 colicin/ml, which we can use in the simulation.</p><h2>Results</h2><h3>Final Product</h3><p>The final software with the introduced data above has been of great value to evaluate our systems and make people understand our design. By running the simulation a couple of times you quickly gain insight about the different systems. It has allowed us to easily detect flaws and make notes about possible future improvements. A screenshot from the program can be seen in fig. x and the results from a number of runs with different parameters is shown in table y.</p><h3>Improvements</h3><p>In order to fully complete the simulation there is still some data that have to be acquired. One of the biggest issues right now is that we have not managed to derive a value for the distance at which the Targeting system will be invoked. We would need to perform further lab experiments on our probiotic to find which concentration of AHLs that is required for activation. Until then we cannot say for sure if we will get close enough to send a lethal signal, in which case the Targeting system could be detrimental to our system since it would hinder the probiotic to get closer to the pathogen. Another data necessary to mirror our real systems is the real colicin production rate, which also could be derived from lab experiments. However, finding these data from experiments require our systems to be fully functional.</p><ul class="reference"><li>[1]Aswathy Sreedharan, C.J., 2012. Preventing Foodborne Illness: Yersiniosis [WWW Document]. URL http://edis.ifas.ufl.edu/fs193 (accessed 8.24.14).</li><li>[2]De Berardis, B., Torresini, G., Brucchi, M., Marinelli, S., Mattucci, S., Schietroma, M., Vecchio, L., Carlei, F., 2004. Yersinia enterocolitica intestinal infection with ileum perforation: report of a clinical observation. Acta Biomed 75, 77–81.</li><li>[3]77–81. H. C. Berg, 2004. E. coli in motion. Biological and medical physics series. (Springer, NewYork)</li><li>[4]Schwartz, S.A., Helinski, D.R., 1971. Purification and characterization of colicin E1. J. Biol. Chem. 246, 6318–6327.</li><li>[5]Spangler, R., Zhang, S.P., Krueger, J., Zubay, G., 1985. Colicin synthesis and cell death. J Bacteriol 163, 167–173.</li></ul>'; | document.getElementById("tab1").innerHTML = '<h2>Introduction</h2><p>To facilitate the understanding of our multi-part system, a simulation for cell-cell interaction was developed in Java. Realistic data on movement and reproduction were used, so that our new introduced functions could be evaluated. This resulted in a two dimensional program where bacteria can interact with each other in real time, as if looked at in a microscope. In our scenario we consider the probiotic moving around on the small intestinal wall and reaching the outer surface of a Yersinia enterocolitica colony. By sensing its environment the bacteria can invoke changes in the movement pattern or kill other species in their vicinity. This provides us with a tool to predict how effective our introduced systems will be and add insight to how the systems can be improved. The program can easily be changed and applied to similar systems, hopefully making it useful for understanding and evaluating future projects.</p><h2>Java Design</h2><p>In order to keep track of many bacteria with individual properties we thought it convenient to use an object-oriented language. Since Java also has the advantage of being easily distributed and made executable through a web browser environment, we found it a perfect candidate. Most of the data that has to be visualized are dependent on the programs ability to measure distance. This was made possible by deciding upon a convenient scale and assigning the pixels a suitable size. In our simulation every pixel corresponds to 0.1µm, which makes it easy to visualize the 2x1 µm large bacteria and its movement.</p><h2>Introduced Data</h2><h3>Generation time & Flux</h3><p>Generation time of our bacteria range between 20-40 min and the simulation considers a very limited window of time and space. To work around this issue a flux of bacteria to the visible area has been introduced. Though this will only affect the probiotic since Y. enterocolitica are immobile at gut temperature (De Berardis et al., 2004). In other word, we assume that there are bacteria moving around outside our screen and that they “walk in” at a constant rate. They will still reproduce based on their generation time but to depend solely on this process to drive the simulation would make it very time consuming.<br><br>The generation time for the probiotic is based on E. coli data, once every 20 min. The corresponding value for Y. enterocolitica was a bit more tricky. It has been measured to 34 min at 30°C, which is assumed to be their optimal growth temperature (Aswathy Sreedharan, 2012). Since our simulation will take place at 37°C we decided to use 40min as an approximation. However, the simulation is not meant to run for more than a couple of minutes so the generation times have been translated into a chance of occurrence. The reproductions are also limited by available space so that bacteria are not able to stack on top of each other. </p><h3>Movement & Size</h3><p>Data regarding movement of bacteria other than E. coli is hard to come by so we decided to let the probiotic mirror that pattern. Hopefully motion of E. coli are not too divergent to render the simulation invalid for other species.<br><br>The swimming speed of E. coli is generally estimated to be 10-20 µm/s with tumbling events occurring every 0.5-1s. In the simulation we have tried to mimic this by setting average speed to 10µm/s, maximum speed to 20µm/s and added a stochastic variable allowing it to tumble about once every second.</p><h3>Activation</h3><p>If the movement of our probiotic brings it close to a pathogen, a switch will trigger. This will bring the bacteria into a state where it continuously tumbles and produce a, for Y. enterocolitica, toxic substance. The distance at which the state activates is dependent on the concentration of Acyl-Homoserine-Lactones (AHLs), produced by the pathogen, as well as the logistic function to account for leakage. This was done using a Markov model with a transition probability that increases rapidly around a specified threshold. The probability is calculated as 1/(1 + e^-x) where x is an activation distance yet unknown in our system. While active it uses the same transition probability to return to its inactive stage, which will increase as surrounding pathogens die. An emission probability that oscillates between 100% and 0%, depending on state, will decide if the bacteria produces its toxin or not.</p><h3>Colicin Production</h3><p>In order to kill pathogens, our probiotic produce a bacteriocin called colicin. In this process we assume that colicin will diffuse in three dimensions, as a half-sphere, with a rate that correlate to its diffusion coefficient, 4.2E-7 cm2/s (Schwartz and Helinski, 1971). The strength of the signal when reaching a pathogen is proportional to the production rate and the time it takes for the bacteriocin to reach its destination. This time can be approximated by inserting the diffusion constant in the formula for Brownian motion in three dimensions.<br><br>Since our Killing system will not stack colicin within its cell wall, but produce and export it at constant rate depending on the Targeting system, data of production rate had to be estimated. In a previous study on colicin production in mutated E. coli it was observed that colicin activity could be increased by a factor 10 in 3-6 hours. By this time the average cell contained about 250.000 copies, which meant that 225.00 copies of colicin had been produced within this time window. If we assume a quite optimistic four-hour window this translates to a production rate of 15.625 copies/s, which we used in the simulation (Spangler et al., 1985). </p><h3>Killing</h3><p>Whether or not a pathogen is killed is decided by measuring the combined signal acting upon it. If this signal exceeds a threshold the bacteria will be declared as dead and removed from the simulation. This threshold has been decided based on a unit, called kill unit, defined as the lowest colicin concentration inhibiting growth (Spangler et al., 1985). Initially kill units were approximated from solutions of bacteria in mid exponentiation phase, where each ml corresponded to a specific value. Since mid-exponentiation phase gives us an OD600 of 0.4 we can estimate that the solution contained about 2.4-3.2E8 cells/ml. From section 3.4 we obtained that these mutant bacteria produce an average of 250.000 colicin/cell, which correspond to 5000 kill units/ml. When combining these data we get that 1 kill unit can be converted to 1.4E10 colicin/ml, which we can use in the simulation.</p><h2>Results</h2><h3>Final Product</h3><p>The final software with the introduced data above has been of great value to evaluate our systems and make people understand our design. By running the simulation a couple of times you quickly gain insight about the different systems. It has allowed us to easily detect flaws and make notes about possible future improvements. A screenshot from the program can be seen in fig. x and the results from a number of runs with different parameters is shown in table y.</p><h3>Improvements</h3><p>In order to fully complete the simulation there is still some data that have to be acquired. One of the biggest issues right now is that we have not managed to derive a value for the distance at which the Targeting system will be invoked. We would need to perform further lab experiments on our probiotic to find which concentration of AHLs that is required for activation. Until then we cannot say for sure if we will get close enough to send a lethal signal, in which case the Targeting system could be detrimental to our system since it would hinder the probiotic to get closer to the pathogen. Another data necessary to mirror our real systems is the real colicin production rate, which also could be derived from lab experiments. However, finding these data from experiments require our systems to be fully functional.</p><ul class="reference"><li>[1]Aswathy Sreedharan, C.J., 2012. Preventing Foodborne Illness: Yersiniosis [WWW Document]. URL http://edis.ifas.ufl.edu/fs193 (accessed 8.24.14).</li><li>[2]De Berardis, B., Torresini, G., Brucchi, M., Marinelli, S., Mattucci, S., Schietroma, M., Vecchio, L., Carlei, F., 2004. Yersinia enterocolitica intestinal infection with ileum perforation: report of a clinical observation. Acta Biomed 75, 77–81.</li><li>[3]77–81. H. C. Berg, 2004. E. coli in motion. Biological and medical physics series. (Springer, NewYork)</li><li>[4]Schwartz, S.A., Helinski, D.R., 1971. Purification and characterization of colicin E1. J. Biol. Chem. 246, 6318–6327.</li><li>[5]Spangler, R., Zhang, S.P., Krueger, J., Zubay, G., 1985. Colicin synthesis and cell death. J Bacteriol 163, 167–173.</li></ul>'; | ||
- | document.getElementById("tab2").innerHTML = '<h2>Running the simulation</h2><h3>Mode Selection</h3><p>The program will give you multiple options before the simulations starts. The first popup will require you to choose mode to run the simulation in, graphical or statistical. The graphical mode lets you watch the whole simulation in graphics, which is strongly recommended if you run the program for the first time. Statistical mode has been introduced to give a statistical estimate of the performance and will instead run the simulation in the background as quick as possible. If you choose this option you will also receive a followup question regarding how many times you wish to run the simulation before the mean value is estimated.</p><h3>Movement near <i>Yersinia enterocolitica</i></h3><p>This option will allow you to choose whether to use the targeting system or not. Depending on pathogen resilience the targeting system can be either detrimental or of grave importance for the killing potential.</p><h3>Parameter Selection</h3><p>In this step you will input values for different parameters. These have all been selected due to importance for the overall performance of the probiotic.<br><br><i>Resilience of pathogen</i> - What concentration that is required to kill a pathogen.<br><i>Average activation distance</i> - At what distance there will be a 50% chance of activation (and deactivation).<br><i>Delay of activation</i> - Delay before the bacteria stops and starts producing bacteriocin (it will take a while before the AHLs will take effect in the probiotic).<br><i>Bacteriocin range</i> - Effective range of the bacteriocin before the signal is removed due to degradation or insignificance.<br><i>Runtime</i> - Time that the simulation will run before the final result is shown.</p><h3>Assumptions</h3><p>Some | + | document.getElementById("tab2").innerHTML = '<h2>Running the simulation</h2><h3>Mode Selection</h3><p>The program will give you multiple options before the simulations starts. The first popup will require you to choose mode to run the simulation in, graphical or statistical. The graphical mode lets you watch the whole simulation in graphics, which is strongly recommended if you run the program for the first time. Statistical mode has been introduced to give a statistical estimate of the performance and will instead run the simulation in the background as quick as possible. If you choose this option you will also receive a followup question regarding how many times you wish to run the simulation before the mean value is estimated.</p><h3>Movement near <i>Yersinia enterocolitica</i></h3><p>This option will allow you to choose whether to use the targeting system or not. Depending on pathogen resilience the targeting system can be either detrimental or of grave importance for the killing potential.</p><h3>Parameter Selection</h3><p>In this step you will input values for different parameters. These have all been selected due to importance for the overall performance of the probiotic.<br><br><i>Resilience of pathogen</i> - What concentration that is required to kill a pathogen.<br><i>Average activation distance</i> - At what distance there will be a 50% chance of activation (and deactivation).<br><i>Delay of activation</i> - Delay before the bacteria stops and starts producing bacteriocin (it will take a while before the AHLs will take effect in the probiotic).<br><i>Bacteriocin range</i> - Effective range of the bacteriocin before the signal is removed due to degradation or insignificance.<br><i>Runtime</i> - Time that the simulation will run before the final result is shown.</p><h3>Assumptions</h3><p>Some phenomena have been disregarded in this simulation, mostly due to complex calculations or missing data. First of all, the simulation does not account for external parameter that will impact the scenario. These are for example movement of the gut, available nutritions and other bacterial species and compounds that will compete for the available space. The other big assumption is regarding the bacteriocins killing potential. The signal that is sent out by our probiotic will not be decreased by hitting a cell, dead or alive, and will only be based on distance. We also assume an equal distribution over the considered volume, so distant pathogens are affected by a stronger signal than what is realistic. The considered signal x is calculated as P/V where P is the total number of colicins produced during the time it takes for one colicin molecule to diffuse to the target. V is the volume of the half sphere with radius equal to the distance to the pathogen.</p><ul class="launchlist"><li><a href="#" id="downloadDemo">Download</a></li></ul>'; |
</script> | </script> |
Revision as of 17:44, 14 October 2014
Stephanie Herman
Teresa Reinli
Joakim Hellner
Alexander Virtanen
Jennifer Rosenius
Marcus Hong
Miranda Stiernborg
Tim Hagelby Edström
Viktor Blomkvist
Megha Biradar
Niklas Handin
Jonas Mattisson
Arina Gromov
Nils Anlind
Eric Sandström
Gunta Celma
Oliver Possnert
Martin Friberg
Kira Karlsson
Christoffer Andersson
Laura Pacoste
Andries Willem Boers
Home
Failed to load tracking. JS is probably not enabled