"
Team:Johns Hopkins-Wetware/etohproject
From 2012.igem.org
Revision as of 03:53, 4 October 2012 by Mrjiaruiwang (Talk | contribs)
Cost effective production of high value compounds, either through chemical synthesis or extraction procedures, is often unattainable using traditional agricultural or chemical processes. Industrial fermentations using microorganisms, such as yeast, is quickly becoming an important alternative and has been employed for the synthesis of compounds ranging from pharmaceuticals to human nutrients. During yeast fermentation, the major chemical stress that impedes optimal production of such compounds is ethanol toxicity (Birch et al. 2000). The presence of ethanol, which yeast cells generate as a by-product of fermentation, activate the natural stress response of the cell, leading to denaturation of intracellular proteins and glycolytic enzymes, decreased membrane integrity, and ultimately cell death. Further, cellular resources devoted to combating ethanol stress result in lost productivity given that resources are diverted from biosynthesis of the desired compound.
The current solutions for ethanol stress are inadequate. Directed evolution and systematic overexpression are the most common means by which engineers select strains that demonstrate increased ethanol tolerance. These solutions are slow, unpredictable, and aimed towards building ethanol resistance rather than eliminating the stressor. The current solution is well suited for the biofuel industry, but ethanol is not the only fermented compound. If we re-frame the problem and look at the entire spectrum of valuable compounds, we begin to see a need for an ethanol control mechanism that targets the source of ethanol accumulation.
To address this problem, we have constructed an ethanol control system in yeast. Central to this system is the human cytochrome p450 CYP2E1 gene, whose encoded protein converts ethanol to acetaldehyde with high efficiency. We have developed and tested a series of constructs in which CYP2E1 expression is driven by a native yeast promoter sequences that are activated by the presence of ethanol. Thus, CYP2E1 expression is triggered when ethanol concentration reaches the threshold level associated with the upstream promoter, resulting in the enzymatic conversion of ethanol to acetaldehyde. Engineering a solution to the problem of ethanol toxicity represents a paradigm shift to the slow and random approaches of traditional lab evolution experiments to isolate ethanol resistant strains.
We have designed, built, and tested a control system to monitor ethanol concentration in yeast. The human cytochrome P450 2E1 (CYP2E1) is a membrane-bound protein that converts ethanol into acetaldehyde. The goal of CYP2E1 expression is to reduce ethanol level in the cell thereby reducing ethanol toxicity. This may seem counterintuitive, given the major push from the biofuel industry to increase ethanol production by yeast cells. However, advances in synthetic biology are enabling us to use yeast fermentation to produce many other interesting compounds, and in this setting, ethanol toxicity is indeed a major hurdle.
The reaction catalyzed by CYP2E1 is:
ethanol + NADP+ -> acetaldehyde + NADPH
In our control system, CYP2E1 expression is driven by an ethanol-inducible promoter derived from yeast (see below). We hypothesized the yeast genome, which has evolved over years to contain a wealth of pre-existing stress responsive promoters, could be "hijacked" for the purposes of expressing CYP2E1. In our synthetic system, a variety of ethanol responses can be obtained by modifying promoter parameters such as strength or percent ethanol of induction. This means that the response can be tailored according to the engineering specifications required for optimizing the synthesis of interest.
The main control mechanism is built from a library of 27 ethanol responsive promoters, the human CYP2E1, and yeast terminator from the MFA2 gene. These parts were constructed using yeast Golden Gate (yGG) Assembly (RFC88), which is virtually seamless and amenable to high throughput construct assembly. We chose yGG to eliminate the possibility of restriction enzyme site "scars" interfering with the native promoter induction system. The Golden Gate standard also modularizes the parts, a requirement for the advancement of synthetic biology. The promoter library was built from two sources. Gene descriptions from the hand-curated Saccharomyces Genome Database were mined for ORFs including ethanol in their functional description. Additionally, publications from microarray studies were compared and condensed. We selected genes that showed higher mRNA levels in the presence of ethanol across multiple studies.
We have characterized ethanol inducible promoters which turn on when specific ethanol parameters are met. With a promoter library characterized by ethanol threshold, our genes can be activated at controlled levels.
24 candidates for ethanol inducible activity were screened by inserting the promoter in front of GFP. We first tested the promoters in various ethanol concentrations to establish base conditions for cell viability and promoter activity. The cells were put into SC -Leu media with ethanol concentrations ranging from 0-14% in 2% increments. The screening results were promising since we saw a significant increase in GFP fluorescence in 8% ethanol media.
Strains containing 24 ethanol-inducible promoters and 3 constitutive control promoters with EGFP were constructed by genomic integration in the Leu2 domain. These strains need to be grown at a temperature-controlled environment under mild shaking in order to characterize the promoter library. A plate reader device can automate this process and combine high-throughput capabilites. Both absorbance at 600nm and fluorescence of EGFP were monitored over 24 hours at 15 minute intervals, and the ratio of fluorescence per OD were plotted. The figure shown is only for 1 out of 27 promoters in our library.
**To see the characterization results from all 27 promoters, click here
The data from all 27 promoters can be better visualized by a histogram of each parameter. The left histogram is induction threshold. This parameter is defined as the % ethanol at which the promoter is turned on, or when GFP fluorescence / OD is maximum, compared with all other concentrations, after 4 hours of induction. The right histogram is of the ratio of induction. This parameter is defined as how much the promoter is induced, or the ratio of fluorescence / OD between the final induced state after three hours over the background signal of 0% ethanol. These plots show that we have an adequately flexible toolbox to meet multiple control specifications.
We were successful in introducing functional CYP2E1 into yeast. After quantitatively characterizing our library of ethanol-inducible yeast promoters, we created 27 CYP2E1 control systems using each promoter in the library. Top 3 most promising strains were picked based on characterization results as candidates for demonstrating control function. Our CYP2E1 strain reduced ethanol better than wild type yeast in a small-scale fermentation experiment (see figure 1). The initial conditions were: 35 mL working volume, 10% dextrose YPD, 30 degrees C, 120 rpm shaking, starting strain BY362.
Ethanol concentration in the media decreased by almost half in strains constitutively expressing CYP2E1. Strains with ethanol-inducible promoters were slightly less effective than the constitutive promoter, but still performed better than the wild type yeast. These results are expected since a constitutive promoter would express CYP2E1 all the time and would be constantly breaking down ethanol, while an ethanol-inducible promoter would only activate the gene when ethanol content reaches a critical level, which is not all of the time.
The data above shows both CYP2E1 and native ethanol-induction function. Additionally, the effect of expressing CYP2E1 on growth rate of these same strains seems to be marginal (see figure 2).
The highly similar growth curves of these strains indicate that synthetic addition of the human CYP2E1 cost the cell very little resources. There is much benefit of ethanol reduction gained at very little cost to the cell. In a resource-competitive environment where metabolic trade-offs must be made for cell survival, we think it is more valuable to invest resources into our synthetic control system rather than on optimizing native ethanol stress response genes. Our solution directly reduces the amount of ethanol in the media instead of tolerating the problem and allowing ethanol concentration to increase with time. This is a promising value proposition that may be applied in an industrial setting.
We designed a mathematical model to explain our results and to predict future behavior of our ethanol control system. Data from the fermentation test and library characterization were used to fit a mathematical model. The control system was built from a classic closed-loop P control system, where the Kp is the promoter strength. Ethanol induction threshold is the input, since the lowest level of ethanol needed to induce the system would be the level that it is trying to control to. Both of these parameters can be tuned by choosing the desired promoter from our toolbox. The transfer function is a black box. It represents the native protein expression mechanism of the cell and all of the variability that comes with biology. We fit the parameters of this transfer function to the protein expression kinetics observed in the characterization results. Some of the mystery of the transfer function can be reduced by thinking of it as taking ethanol concentration as input and giving ethanol consumption rate out. In this way, we assume that the rate of ethanol consumption is directly proportional to the level of CYP2E1 expression. An integrator is added at the end to make sure the feedback is again concentration of ethanol, not rate.
The model, as simulated in MathWorks Simulink:
The base ethanol production of the cell can be thought of holistically as a disturbance. The model for this signal was fitted to the ethanol output of the control strain during fermentation. In order to feed this ethanol concentration disturbance into the CYP2E1 output signal from the transfer function, a derivative is needed to convert the signal to rate of ethanol change. Below is the result for the first 24 hours of fermentation, fit to ethanol concentration data from the actual experiment.
The manipulated parameters used for this model are in the format: {Name, Kp, induction threshold or input}. {Control, 0, 0};{Ethanol-inducible promoter, -0.5, 2};{Constitutive promoter, -0.5, 0};{Weak promoter, -0.3, 2} . The constant parameters are as follows: a = 10, starting OD = 0.5 step at time 3, max %EtOH yield = 4.6%, fermentation rate a = 0.095.
There are limitations to this control model, since we are assuming a linear time-invariant system. We know that yeast population growth is not a linear system, however the relationship between number of yeast cells and ethanol concentration can be approximately related by the transfer function. This way, we can condense all of the parameters that rely on a lot of assumptions into one component, which will make future improvements to the model easy, and the rest of the model predictable.
We were interested to see what the ethanol response would look like given a longer simulation time. According to our hypothesis, the ethanol level should go down if no more ethanol is being produced. This is a safe assumption since we had designed the fermentation conditions to consume all of the dextrose at around 24 hours. Below is the simulation result after 48 hours.
The plateau in the middle may be caused by a time delay in protein expression. There were too many assumptions made to tell exactly where this behavior is originating from. However, we know that using a PI or a PID controller will help fix the problem. We are looking into biological analogs of these controllers and see if they can be implemented with our current CYP2E1 control system. We are confident that with more information about protein expression mechanisms and kinetics, we can come up with a more predictable model.
This model is a great tool for future users of our CYP2E1 control system to simulate fermentation before the actual experiment to see which parameters might yield the best results. The user can then take these parameters and find the closest match in our promoter toolbox. This has the potential to save the user a lot of time and money by eliminating initial screening. We are working to further validate our model with this purpose in mind.
The next step of this project is to design a yeast strain with and without our control system that also expresses the biosynthetic pathway of an interesting compound, such as beta-carotene or artemisinin. We can use model predictions to select which promoter out of our library would be best suited for optimum control, and compare the titers between strains. We have already shown that our control system removes ethanol from the cell without affecting growth rate, so we would expect to see an increase in titer over wild type yeast. We think this will be the case because a lower ethanol concentration in the cell means fewer stress response proteins will be expressed, thus freeing more usable resources for the pathway of interest.
Rosslyn M. Birch, Graeme M. Walker, Influence of magnesium ions on heat shock and ethanol stress responses of Saccharomyces cerevisiae, Enzyme and Microbial Technology, Volume 26, Issues 9D10, June 2000, Pages 678-687, ISSN 0141-0229, 10.1016/S0141-0229(00)00159-9.
