Quantifying synthetic gene transcription in plants

The article we write about today, “Quantitative Characterization of Genetic Parts and Circuits for Plant Synthetic Biology“, was published online in Nature Methods about a year ago. But its importance is such that we still thought it worth describing and to point out to readers other sites that have also provided excellent overviews of this paper (see The New Leaf, the GARNet Community Blog and Science Daily‘s write ups).

The Study

The researchers behind the study were looking to address one of the biggest problems in plant synthetic biology (and synthetic biology generally), being the ability to design gene circuits with a solid understanding of the rate of transcription of each of the genes within the circuit. Being able to predict and ‘tune’ the amount that a gene is transcribed will be a great step forward in allowing researchers and biological engineers to design and test circuits on computers, saving the time and expense of having to synthesise and test each component in the wet lab.

But biology, particularly multicellular biology, is messy and noisy and affected by so many different factors that we are unlikely to know and control each and every one of them. This study showed as much, but also showed that with careful testing we can develop the underlying knowledge required to develop general rules that will assist higher throughput development and research.

Using protoplasts derived from Arabidopsis cells to express the synthetic constructs, the researchers developed three small circuits. One circuit was an inducible promoter which, when induced by an externally applied inducer to begin transcribing the circuit, would result in the production of the protein firefly luciferase (F-luc), a fluorescent protein that can be detected. The level of fluorescence detected is a function of how much the gene is being transcribed. The second circuit was also controlled by an inducible promoter which, when induced by the same inducer, would result in the transcription of a protein with DNA-binding domains which would bind to specific DNA sequences, in turn repressing the transcription of those genes. The idea behind the two circuits using the same inducer is that the amount of fluorescence from the first would act as a proxy reporter for the amount of repressor being transcribed.

The third circuit synthesised contained a gene for Renilla luciferase (R-luc), another protein which fluoresces but does so at a shorter wavelength than the firefly luciferase and is therefore distinguishable. This gene was linked to a constitutive (constantly active) promoter that contained a number of DNA sequences that could be bound by the repressor. The repressible promoter contained DNA binding sites at different points to test the ability to fine tune repression rates with different repressor/binding site combinations.

nature-methods-fig-1-protoplasts-inducers

Figure 1 from article. (a) is the repressible promoter with DNA binding sites at different points. (b) is the three

The result were protoplasts that (at least theoretically) would fluoresce at the shorter, R-luc wavelength when there was no inducer added to the cell as the gene circuit containing the repressor would not have been induced for transcription. When the inducer is added and increased, the R-luc fluorescence would reduce proportional to the amount of inducer added (as more and more of the repressor is transcribed) and simultaneously the amount of F-luc florescence would increase.

In a less messy environment, the input-output response would predictable somewhat like an electronic circuit – an increase in the input would result in an equivalent (whether that be linear, exponential, logarithmic etc) change in the output. The aim of the research was to quantify the input-output ratio and see if a mathematical formula could be applied to allow predictions of input and outputs to be made.

The Results

But biology is messy and the input to output ratio over a number of reproductions varied considerably even though the generally expected increase in F-luc and reduction in R-luc as more inducer was added was observed.

So the researchers looked at where the variability could be coming from and how it may be accounted for so that predictable quantification of the transcription rates could be derived. To test this they examined the amount of F-luc fluorescence in the protoplasts when no inducer was added which theoretically should result in no fluorescence or, if induced by something other than the external inducer, should be fluorescing at the same level. What they found was that there was an unexpected variability in the fluorescence levels even when there was not external inducer added. When sorted by different batches of protoplasts which were prepared on different days, it was apparent that some variation in the preparation of each batch (which they weren’t able to explicitly identify) was causing a variation in the F-luc levels that was not controlled for (Figure 2(b) below).

noise-in-protoplast-data

Figure 2 from article showing noise in the protoplast fluorescence data grouped by batch and inducer type.

Given the noise wasn’t completely random, the researchers looked to mathematics to attempt a solution to their quest to help predict output levels despite the noise.

The Mathematics

It was assumed that the input to output relationship would function in accordance with the repressing Hill function. But analysing their data compared to that expected according to the Hill function showed that the output amount was the Hill function multiplied by some factor that was related to whatever was affecting each batch. What they hypothesised what that the ratio between average of fluorescence of all batches to that of each batch, with no inducer applied, should normalise the fluorescence levels and eliminate the batch effect observed. Calculating the normalisation factor in this way resulted in the reduction in noise between batches.

normalisation

Figure 4 from article showing the effect of normalising the input-output data using the calculated normalisation factor.

Reproducibility and Usability

To test the circuits and normalisation of output quantification in a monocot, sorghum protoplasts were transiently transformed with the same result – normalisation of the fluorescence values reduced or eliminated the batch effects.

Further testing was done to compare the use of transient expression in protoplasts to expression of the same circuits in stably transformed plants to determine whether one is reliably indicative of the other. Again, the purpose of confirming this comparison is to allow faster testing of circuits using transient expression with confidence that similar results will flow from the same circuits being installed stably into transgenic lines. Comparing the two in Arabidopsis protoplasts, the researchers found that luciferase expression was lower in the stably transformed protoplasts compared to the transient expression but the difference could again be normalised and fitted with the repressing Hill function.

Thus, it was suggested by the authors of the paper that by using these normalisation techniques, transient expression of gene constructs in protoplasts could be reliably used to predict the expression levels of the same constructs in stably transformed plant cells.

Conclusion

The paper addresses the importance of predictable, reproducible quantification of genetic parts in multicellular organisms that produce a lot of noise when trying to quantify results. It demonstrates the issue that will consistently arise as we attempt to address food security and environment concerns with technology aimed to produce larger yields with lower input and land usage. But it gives an insight into the ability we have to overcome the hurdle and begin designing and testing circuits with larger throughput and greater reproducibility.

And was clearly worth writing about again!

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s