Team:BostonU/Experiments
Characterization
Strain Screening
The first step in characterizing LOV2 and PhiReX was validating that our yeast strains had functionally integrated. In order to minimize resource consumption, we screened our strains using a reusable culture block and a red or blue photography lamp in our shaking incubator.
Figure 1: mRuby expression driven by LOV2 exposed to white light in the shaking incubator.
Figure 1 is from a LOV2 induction experiment using white light instead of blue (the preferred induction spectrum). It was necessary to efficiently validate our transformed strains before running long-term eVOLVER experiments. Reviewing relevant literature indicated that white light activates LOV2, albeit to lower levels than blue. . We determined that lower activation levels were sufficient for screening colonies and confirmed white light activation of LOV2 in the shaker experiment in Figure 1.
We later ordered a red light bulb for use in testing our PhiReX strains and blue to further test LOV2.
Figure 2: mRuby expression driven by PhiReX under red light in the shaking incubator.
Our red light shaker experiment revealed unexpected fluorescence from several negative controls. We determined that the measured fluorescence in negative control strains was due to a mutation in our strains which forms a pink pigment in the presence of adenine in media. This pink color was interpreted as fluorescence during flow cytometry measurements, leading us to refine our experimental protocol to dilute cells every hour, halting the accumulation of pink color. Nonetheless, we observed an increase in mRuby expression in the PhiReX strain that indicated successful induction.
Figure 1: mRuby expression driven by LOV2 exposed to white light in the shaking incubator.
Figure 1 is from a LOV2 induction experiment using white light instead of blue (the preferred induction spectrum). It was necessary to efficiently validate our transformed strains before running long-term eVOLVER experiments. Reviewing relevant literature indicated that white light activates LOV2, albeit to lower levels than blue. . We determined that lower activation levels were sufficient for screening colonies and confirmed white light activation of LOV2 in the shaker experiment in Figure 1.
We later ordered a red light bulb for use in testing our PhiReX strains and blue to further test LOV2.
Figure 2: mRuby expression driven by PhiReX under red light in the shaking incubator.
eVOLVER Experiments
Using the eVOLVER, we were able to rapidly obtain data on induction of our two systems. By characterizing the induction capabilities of our systems across integration into different yeast loci, we found that the most reliable and active configurations were LOV2 integrated in the HO locus, and a PhiReX design with chromophores in the HO locus and PhyBNT and PIF3 in the URA3 locus.
Figure 3: mRuby expression driven by LOV2 in response to a unit step blue light input in eVOLVER, plotted with a first order model for the step response of LOV2.
The above graph shows the response of LOV2 to a unit step input of blue light in the eVOLVER. Using control theory, we were able to fit a first order control system model to this dataThe first order model is overlaid on top of the step data in Figure 3. This first order model allowed us to predict how LOV2 would react under different light inputs Experimental analysis of OD and temperature of vials during our initial LOV2 eVOLVER experiments indicated these were not parameters that heavily influenced the functionality of the promoters. Further, light programs are easier to manipulate on the eVOLVER; thus we decided to focus our characterization of LOV2 and PhiRex over different pulsatile light inputs.We used our model to predict the response of LOV2 to different blue light pulsing programs, and then performed an eVOLVER experiment to test our model. The results are shown below.
Figure 4: mRuby expression vs. time under varying pulsatile blue light inputs, with the first order model overlaid on top.
The first order model was not within one standard deviation of the data; we revisited our step data and changed our assumption that the system could be represented by a first order model to a second order model. The model is fitted to the step data as seen below (see Model for more details):
Figure 5: mRuby expression vs. time under varying pulsatile blue light inputs, with the second order model overlaid on top.
We reapplied the second order model onto data from our pulsatile input experiment. While it does not track the data perfectly, the second order model estimates the data better than the first order model, particularly as the duty cycle approaches 100%. The
results of this second order model overlay are displayed below:
Figure 6: LOV2 blue light pulse data with second order model overlaid.
Concurrently with our experiments with LOV2, we applied pulsatile inputs to PhiReX to see if it behaved similarly to LOV2.
Figure 7: mRuby expression driven by PhiReX under several different pulsatile red light inputs.
Figure 7 outlines the output of PhiReX in response to pulsatile red light inputs of different duty cycles. Light pulse programs can be tuned and adjusted to achieve a desired expression level output. The second order model does not fit PhiReX as well as LOV2 because the model was trained on PhiReX step input data; in our pulsatile experiment, PhiReX did not induce as highly as we anticipated, and therefore did not fit the model well. In the future, we plan to run this experiment again, and either alter the model to fit the data or determine that this data is anomalous. Further, we found LOV2 is induced more reliably and generated data that was better suited to our modeling.
Figure 8: Steady State mRuby expression vs. Pulse duty cycle is plotted, showing a linear relationship in which longer duty cycles increase reporter expression.
Light pulse duty cycle is a readily tunable parameter, so we sampled a range of inputs for optimal activation levels with PhiReX and LOV2. Thus, modulation of light pulse dynamics as inputs to LOV2 and PhiReX allows precise control of transcription without having to test inducer dose responses as small molecule inducers require. Further, modeling of light inducible system responses provides an opportunity to predict outputs generated in response to tunable pulsatile inputs. We have demonstrated that steady-state gene expression increases linearly as duty cycle approaches a step function. Our results indicate that steady-state gene expression under the control of LOV2 and PhiReX in a bioreactor can be controlled using tunble light programs.
Figure 3: mRuby expression driven by LOV2 in response to a unit step blue light input in eVOLVER, plotted with a first order model for the step response of LOV2.
The above graph shows the response of LOV2 to a unit step input of blue light in the eVOLVER. Using control theory, we were able to fit a first order control system model to this dataThe first order model is overlaid on top of the step data in Figure 3. This first order model allowed us to predict how LOV2 would react under different light inputs Experimental analysis of OD and temperature of vials during our initial LOV2 eVOLVER experiments indicated these were not parameters that heavily influenced the functionality of the promoters. Further, light programs are easier to manipulate on the eVOLVER; thus we decided to focus our characterization of LOV2 and PhiRex over different pulsatile light inputs.We used our model to predict the response of LOV2 to different blue light pulsing programs, and then performed an eVOLVER experiment to test our model. The results are shown below.
Figure 4: mRuby expression vs. time under varying pulsatile blue light inputs, with the first order model overlaid on top.
The first order model was not within one standard deviation of the data; we revisited our step data and changed our assumption that the system could be represented by a first order model to a second order model. The model is fitted to the step data as seen below (see Model for more details):
Figure 5: mRuby expression vs. time under varying pulsatile blue light inputs, with the second order model overlaid on top.
Figure 6: LOV2 blue light pulse data with second order model overlaid.
Figure 7: mRuby expression driven by PhiReX under several different pulsatile red light inputs.
Figure 7 outlines the output of PhiReX in response to pulsatile red light inputs of different duty cycles. Light pulse programs can be tuned and adjusted to achieve a desired expression level output. The second order model does not fit PhiReX as well as LOV2 because the model was trained on PhiReX step input data; in our pulsatile experiment, PhiReX did not induce as highly as we anticipated, and therefore did not fit the model well. In the future, we plan to run this experiment again, and either alter the model to fit the data or determine that this data is anomalous. Further, we found LOV2 is induced more reliably and generated data that was better suited to our modeling.
Figure 8: Steady State mRuby expression vs. Pulse duty cycle is plotted, showing a linear relationship in which longer duty cycles increase reporter expression.
Light pulse duty cycle is a readily tunable parameter, so we sampled a range of inputs for optimal activation levels with PhiReX and LOV2. Thus, modulation of light pulse dynamics as inputs to LOV2 and PhiReX allows precise control of transcription without having to test inducer dose responses as small molecule inducers require. Further, modeling of light inducible system responses provides an opportunity to predict outputs generated in response to tunable pulsatile inputs. We have demonstrated that steady-state gene expression increases linearly as duty cycle approaches a step function. Our results indicate that steady-state gene expression under the control of LOV2 and PhiReX in a bioreactor can be controlled using tunble light programs.
Orthogonality of LOV2 & PhiReX
In order to exact layered transcriptional control, we aimed to demonstrate that our blue- and red-light inducible systems could be used simultaneously in the same cells to multiplex control. We therefore needed to demonstrate that their activation spectra were functionally orthogonal when co-expressed in vivo.
Figure 9: Absorbance spectra of LOV2 and PhiReX. LOV2 peaks at 450-490 nm and PhiReX peaks at 640-670 nm.
Figure 10: Average mRuby expression in a.u. of PhiReX and LOV2 under different experimental conditions.
Figure 11: mRuby expression under PhiReX and LOV2 in a blue light eVOLVER experiment for orthogonality, as well as fold change from negative control strain.
Figure 9: Absorbance spectra of LOV2 and PhiReX. LOV2 peaks at 450-490 nm and PhiReX peaks at 640-670 nm.
As the peak absorbance wavelengths of PhiReX and LOV2 lie on opposite ends of the visible spectrum, the systems are theoretically orthogonal. To confirm this hypothesis, we tested PhiReX and LOV2 orthogonality with both batch culture shaking incubator experiments and eVOLVER experiments.
Figure 10: Average mRuby expression in a.u. of PhiReX and LOV2 under different experimental conditions.
The data in Figure 10 confirms that LOV2 is not activated by red light, and PhiReX is not activated by blue light. Thus, the systems are indeed orthogonal and could be used to implement layered control over gene expression.
Figure 11: mRuby expression under PhiReX and LOV2 in a blue light eVOLVER experiment for orthogonality, as well as fold change from negative control strain.
Examining the fold change from negative control for each system in orthogonality experiments, we found that LOV2-driven mRuby expression is significantly activated under blue light while PhiReX-driven expression mirrors control expression values. Conversely, PhiReX-driven mRuby expression is responsive to red light while LOV2-driven expression remains at baseline levels. Although PhiReX expression is slightly elevated over the negative control in the above figure 11, this less than 2-fold activation can be reconciled as likely due to either transient activation of PhiReX by ambient white light in the lab, or due to a low-level basal "leakiness" in the promoter system.