iGEM Bielefeld 2015

Application of our model

Guiding experiments in the lab.

Our model accurately describes the expression of sfGFP in our cell-free protein synthesis and is able to predict the effects of the competition of two plasmids for resources. Furthermore, it can simulate repression and induction and therefore includes all relevant aspects of our CFPS biosensors. While developing the model and performing modeling-related experiments, we gained new insights into CFPS. However, our model was not only supposed to help us in understanding our biological system. We also wanted to use it to optimize our biosensors. Therefore, we investigated several crucial design aspects with our model.

Two plasmids vs. pre-existing repressor

Comparison of CFPS operating modes — Simulated sfGFP production when 8 nM reporter plasmid and 8 nM repressor plasmid are used (blue lines) or when 8 nM reporter plasmid and 300 nM repressor dimer are present (green lines).

A CFPS biosensor is an open system, which means that the components can be manipulated more easily than in a whole-cell biosensor. One crucial component is the repressor, because its amount is decisive for the sensitivity and background signal of the biosensor. There are two fundamentally different operating modes with regard to the repressor: It can either be encoded on a plasmid and be co-expressed with the reporter, or the protein is already present in the reaction mixture. To analyze the differences, we modelled a CFPS reaction with equimolar amounts of reporter and repressor gene in the non-induced and fully induced state. We then compared the results to a simulation in which the same amount of repressor dimer was present from the beginning. As can be seen from the figure, there is a weak background signal when a repressor plasmid is used, because it takes some time until enough repressor has been expressed to suppress the expression of the reporter. Moreover, the signal in the presence of an analyte is weaker due to the competition of reporter and repressor mRNA for resources. In contrast, when the repressor protein is already present in a sufficient amount, there is no background signal and the signal in the induced state is twice as high.

In view of these results, a pre-existing repressor protein is the better option for a CFPS biosensor. We chose this operating mode for the characterization of our biosensors and used a cell extract that had been prepared from cells expressing the necessary repressor or activator. This is a convenient method, however, it would be an even better option to add a purified protein to the reaction. This would allow for the amount of repressor to be precisely controlled. And as you will see, this is highly desirable.

Choosing the best plasmid ratio

Although the output signal is stronger when working with a purified repressor protein, it is a convenient approach to use two plasmids for reporter and repressor. This makes it possible to control the amount of repressor without the need to purify the protein. Therefore, we looked more closely at the importance of the plasmid ratio. The same analyses could be performed in order to optimize a reaction with a purified repressor.

Importance of plasmid ratio — Importance of the reporter:repressor plasmid ratio. The sfGFP production for three plasmid ratios and analyte concentrations spanning 6 orders of magnitude were simulated.

We simulated the expression of sfGFP for several plasmid and analyte concentrations. The results showed us that the ratio of the plasmid concentrations is very important for a functioning biosensor. When the reporter:repressor ratio is too high, there is a strong background signal and the relative increase in the presence of an analyte is small and would be difficult to discern (orange lines in the figure). In contrast, when the ratio is too low, an induction is not possible because of the strong repression and because most resources are used for the expression of the repressor (blue). Given the parameters of our model, a ratio of 1:1 yields the best compromise between low background and strong induced signal (green).

We tested this prediction with a plasmid that encodes the arsenic repressor and a second plasmid that contains the arsenic operator in front of sfGFP (BBa_K1758300). According to the model, the fluorescence in the presence of the repressor plasmid should be significantly lower than in the presence of a "neutral" plasmid, which competes for resources but does not encode a repressor. However, in the experiment we observed only a slight decrease compared to a reaction that contained an RFP plasmid. Our control reactions suggest that a slight repression does indeed take place, however, an induction with arsenic was not successful.

Model predictions for reactions with two plasmids.

Experimental results for CFPS reactions with two plasmids. Plasmids encoding sfGFP, either with the arsenic operator (arsO) or without, were combined with plasmids encoding the arsenic repressor (arsR) or mRFP1. 8 nM of each plasmid were used.

We assume that there are two main reasons for this result:

The position of the arsenic operator relative to the T7 promoter in the P_T7-arsO-UTR-sfGFP plasmid is not ideal, as the distance needs to be very small to efficiently block the T7 RNA polymerase (Karig et al. 2012).
An optimization is often necessary for protein expression in vitro, which we had not carried out for the arsenic repressor, ArsR. Therefore, the amount of active ArsR was likely lower than predicted by the model.

These results show that the model predictions are only valid for optimized genetic constructs and CFPS conditions. Otherwise, further tests and adjustments to the model parameters are necessary in advance.

Adjusting the detection limit

The amount of repressor is not only important for obtaining a reasonable output signal. It also determines the detection limit of the biosensor. For whole-cell biosensors, it has been demonstrated that the detection limit can be adjusted by changing the strength of the promoter or RBS of the sensing device (Wackwitz et al. 2008). This means that the sensor can be tuned to give a signal when a certain threshold concentration of the analyte is exceeded. This allows the user to easily decide whether the concentration is above the safety limit or not. While whole-cell biosensors require time-consuming genetic manipulations in order to achieve this, our cell-free biosensors can be tuned by simply adjusting the amount of repressor. Our model further facilitates this process because it can predict the necessary amount of proteins or plasmids. The figure illustrates this for a hypothetical safety limit of 100 µM analyte. The horizontal line marks the amount of sfGFP which we were able to detect with a smartphone camera. By keeping the analyte concentration constant and varying the repressor concentration it is possible to determine the conditions that result in a positive test result for concentrations above 100 µM analyte and a negative result for concentrations below this value.

Using this procedure, it is also possible to generate test strips that include several spots that detect the same analyte, but give a signal at different analyte concentrations. This can serve as an internal calibration of the test strip (Wackwitz et al. 2008). Furthermore, when the intensity of the fluorescence is to be used for quantification, it is possible to obtain a larger linear range using several spots with different detection limits.

Future applications

Currently our model uses parameters for repression and induction which were determined for the lac operon. This is sufficient for making qualitative predictions about the behaviour of our biosensors, but it does not enable us to predict exact concentrations for the different biosensors we used. For future applications it would therefore be desirable to determine the specific parameters of the repressor proteins we worked with. A sensitivity analysis of the model showed that in the absence of an analyte, the translation rate constants are most important for the output signal. When an analyte is added, however, the derepression parameter k_dr2 becomes crucial. Consequently this and, if possible, other parameters should be determined for every repressor protein and analyte. This would be possible by means of techniques such as surface plasmon resonance or by measuring the sfGFP production as a result of known analyte concentrations and fitting the parameters. The second approach would have the advantage that it takes into account the efficiency of the genetic devices, because our experiments have shown that aspects such as operator placement and translation efficiency have a great impact on the functionality of the biosensor and are difficult to predict.

For the future, we envision that a model with optimized parameters could be used to predict a range of repressor concentrations for an optimal output signal and the desired sensitivity. These concentrations could then be tested in the lab. Due to the batch-to-batch variations between cell extracts, such a process would need to be performed regularly. Therefore, the model could significantly reduce the number of necessary experiments. The information from the model could also be integrated into our app for fluorescence detection.

References

Karig, David K.; Iyer, Sukanya; Simpson, Michael L.; Doktycz, Mitchel J. (2012): Expression optimization and synthetic gene networks in cell-free systems. In Nucleic acids research 40 (8), pp. 3763–3774. DOI: 10.1093/nar/gkr1191.

Wackwitz, Anke; Harms, Hauke; Chatzinotas, Antonis; Breuer, Uta; Vogne, Christelle; Van Der Meer, Jan Roelof (2008): Internal arsenite bioassay calibration using multiple bioreporter cell lines. In Microbial biotechnology 1 (2), pp. 149–157. DOI: 10.1111/j.1751-7915.2007.00011.x.