Team:Dundee/Modeling

From 2014.igem.org

Revision as of 13:05, 11 October 2014 by Rshuttleworth (Talk | contribs)

Dundee 2014

Modeling

Maths.. maths is fun!

Modeling and Analysis of Signaling Pathways


Introduction

In order to help analyze, construct and optimise the biochemical pathways in the Lung Ranger, we used a variety of mathematical tools to create algorithms and simulations.This allowed us to accelerate the development and testing of various hypotheses.

Methodology
As shown in figure 1 each of the approaches; ordinary differential equations (ODEs), stochastic simulation algorithm (SSA) and NetLogo, provided a different understanding of each system.


PQS

Objectives
We wanted to find a method of increasing the expression of mCherry in our engineered PQS system.

Results
Sigmoidal Expression of mCherry in PQS System
When our PQS system was induced with synthetic PQS, no mCherry was expressed. We attempted to use mathematical modelling to find out why and how the situation could be resolved.

After constructing a series of ordinary differential equations (full derivation can be found in the appendix) we established the following relationship between PQS (Se) and mCherry:

Equation (1) was then analysed in MAPLE for varying PQS concentrations using the parameters in table 1.

Fig 1. shows how the concentration of mCherry increases over time. The general trend is that the rate of expression of mCherry is proportional to PQS signal strength. Most striking is the rapid switch from low expression to high expression as the signal is increased.

When the gradients of mCherry were plotted against PQS concentration, Fig 2, the result was a sigmoidal curve with sharp transition located between signal values of 0.1µM and 1µM. Below this, expression was effectively zero, and above it, expression was uniformly high.

These results suggest that the levels of synthetic PQS investigated in our engineered pathway were too low and that the ten fold increases used in the experiments would yield no measurable change in mCherry expression provided the signal level was below the threshold predicted by the model. To investigate this further we went on to carry out stochastic simulations of the system.

Stochastic Simulations Elucidates the “Switch” Behaviour
The lab experiments resulted in no measurable mCherry expression in the PQS system. However, our ODE model predicts that even for very low levels some mCherry is expressed. These levels could represent below-measurable output, thus are essentially zero. We investigated this further at the “single cell” level.

We constructed a stochastic simulation algorithm (SSA) in order to visualise signal-response at the single cell level. Using the SSA we were able to simulate mCherry expression of 1000 cells to varying PQS signals over a period of one cell cycle.

At first sight, the stochastic mean in A appeared too low in comparison with the example realisations. However, further analysis revealed that this low mean is caused by many realisations in which the expression of mCherry was zero (i.e. the promoter never fired). As shown in Figure 4, in 77% of the realisations for 0.02μM PQS, no mCherry was produced. Increasing the PQS concentration through the previously predicted threshold increased the expression rate to 90%. For PQS concentrations greater than 2μM, all realisations produced mCherry in the given time frame.

What is interesting to note is that once expression started, mCherry was produced at essentially the same rate, as shown by the parallel trajectories in Fig 3. We can conclude that a major rate limiting step in the system is the probability of the initiation of mCherry expression.

These stochastic simulations enhanced the deterministic results. The switch behaviour between low and high levels mCherry expression resulted in predicted corresponding molecule numbers shown in Table 2.

Increasing promoter numbers increases mCherry expression
We concluded from the SSA that firing of the mCherry promoter was a major rate limiting step, but wanted to provide the wet team with a more understanding of how to increase mCherry production. Subsequently, after testing several model l hypotheses, it was found that increasing the number of promoters had the greatest impact on the expression of mCherry (varying other experimentally controllable parameters were predicted to have essentially no effect on expression level).

When the number of promoters in the model was increased by 10-fold (30 to 300) there was a corresponding 10-fold increase in the “high” mCherry expression, as shown in Fig 5. We suggested that increasing the number of promoters would be the only practical method of significantly increasing mCherry expression. However it was subsequently discovered that the number of promoters in our chassis could not be increased significantly as high numbers proved toxic.

Conclusions
Our models revealed a distinct switch behaviour in the signal-response curve for mCherry expression. Using this analysis we put forward the hypothesis that the synthetic PQS concentrations used in our experiments were not sufficiently high and although these experiments were conducted using a range of synthetic PQS levels, this range corresponded to fold changes in the low-response region shown in Fig. 2.

The predicted switch behaviour has important and positive consequences for the L.A.S.S.O.(link to lasso page) - we predict that the mCherry output level would be distinctly bimodal, avoiding the necessity for sensitive calibration of the detecting device - low signals would result in approximately no response, whereas supercritical signals would induce maximal response from the sample.


BDSF

Objectives
To investigate why GFP expression was high in the absence of signal in our engineered BDSF system.

Results
BDSF-Induced Phosphorylation of BCAM0228[P] Mediates GFP Expression
We constructed models similar to those used in the PQS system to investigate the signal-response behaviour of the BDSF system. Our results verified what we had expected to happen. Phosphorylation of BCAM0228 is induced by BDSF binding to a cell receptor. BCAM0228[P] then activated the expression of GFP through the engineered cblD promoter. In the absence of a BDSF signal, BCAM0228 remained in its unphosphorylated form and GFP expression was repressed. Increasing the signal induced an increased GFP response (Fig 1).



BCAM0228 is Phosphorylated by Unknown Compound
Our experimental results revealed that in the absence of BDSF, the cblD promoter was active in our engineered system and hence GFP expression upregulated.



From our system of equations we know that the expression of GFP is dependant on the levels of BCAM0228[P]. We hypothesised that one possible mechanism by which our cell could be expressing GFP in the absence of signal was by the presence of BCAM0228[P] independent of BDSF-binding mediated phosphorylation. This suggests that an unknown compound could be phosphorylating BCAM0228 in our engineered cells.

Comparing Fig 2 to Fig 1, it is clear that the model predicts that the expression of GFP is higher when BCAM0228 is being constitutively phosphorylated than in the “normal” case. Moreover, we are also able to conclude that the model predicts that the presence of BDSF has very little effect on the overall expression of GFP when the constitutive phosphorylation process is operating.

Conclusions
Our models predict that an as yet unknown agent phosphorylates BCAM0228 in the absence of a BDSF signal and that in fact the level of the BDSF signal has little quantifiable effect on GFP expression in our engineered cells.


DSF

Objectives

References