Quantifying transcriptional decisions in living Drosophila embryos

Hernan Garcia, Mariella Petkova & Thomas Gregor

Biological problem

Over the last few years new technologies have allowed us to query the dynamics of transcriptional decisions in single cells. As a result, a picture of noisy transcriptional processes including, for example, transcriptional bursting has emerged. Dynamics of transcriptional decisions in development are key to the timely and precise establishment of developmental programs. In some organisms, such as the fly embryo, these transcriptional decisions orchestrate the underpinnings of biological pattern formation. Yet, so far our intuition about the dynamics of development come either from live experiments in single cells or from the examination of patterns of expression in fixed tissue.

We have developed a new technique to quantify transcriptional decisions in living Drosophila embryos at the single cell level. Implementation of a well-established reporter system allows us to visualize sites of nascent transcript formation in individual nuclei whose fluorescence is proportional to the number of RNA polymerase molecules that are actively transcribing. In no other multicellular organism do we currently possess such exquisite access to transcription dynamics in vivo. This technique together with single molecule counting of transcriptional output (i.e. mRNA molecules), which gives us an absolute measure, brings the Drosophila embryo into a position where it lends itself to study the most general aspects of transcriptional regulation, a position that is typically occupied by single celled organisms only.

In this course we will use this system to explore the dynamics of the establishment of an expression boundary that forms a step-function along the embryo's anterior-posterior axis about 2h after the egg is laid and fertilized. The challenges to be addressed in the course fall in three categories: experimental, data analysis and modeling.

Experimental challenges:
o From previous analyses we have a good idea about the transcription dynamics when averaging over multiple neighboring nuclei. The next step is to extend the analysis to single-cell dynamics of transcription and determine the nucleus-to-nucleus variability. Quantifying this variability is meaningful for two reasons. First, it allows us to ask how the fly copes with noise in order to produce precise patterns of gene expression. Second, the variability allows us to constrain molecular models of transcription. We can access these single nucleus dynamics of transcription by tagging either the 5' or 3' end of the gene. Each construct will report on different aspects of the process: 5' reports on the rate of initiation while 3' also reports on the rate of elongation along the gene.
o In previous analyses we have extracted transcription dynamics from both the 5' or the 3' end, but in different embryos. However, imaging these constructs in different flies misses valuable dynamic information when reporting on the whole process of transcription at the single cell level is desired. For example, is the rate of elongation reproducible between different nuclei/different cell cycles/different embryos? We have extended our mRNA dynamics approach to multiple colors to monitor the activity of RNA polymerase at different points along the gene. We will image constructs were the 5' and 3' ends of the gene can be simultaneously visualized using two different colors.
o The variability of transcriptional activity between nuclei is about 40% at any time point in development. However, we don't know if this variability is intrinsic to the transcription process or the result of extrinsic sources such as variability in the number of RNA polymerase molecules in each nucleus. In order to disentangle the intrinsic and extrinsic contributions to variability we will quantify the dynamics of expression of two alleles of the same gene in the same nucleus and compare the results among different nuclei.
o Technical constrains have limited our characterization of dynamics to only 20% of the embryo at a time. The extend of the boundary region of our reporter pattern is on that same spatial order, making it difficult to image the emergence of the whole boundary in one embryo. During the course we will explore options to increase both the size of the imaged spatial domain in the embryo as well as the time resolution in order to obtain a dynamical view of the establishment of the whole pattern during development.


Data analysis challenges:
o Averaging nuclear time traces over regions along the anterior-posterior (AP) axis of the embryo provides valuable information about the mean transcription dynamics, but information about individual traces is lost. We would like to expand this analysis by looking more closely at single transcription site time traces. What are their noise characteristics? Are there spatial or temporal correlations that could provide insights into the underlying patterning processes?
o The behavior of individual nuclear traces changes drastically between nuclear cycles 13 and 14. In nc14 the promoter seems to change between discrete values of the rate of transcription initiation. These are single molecule, non steady-state experiments for which new data analysis techniques will have to be developed in parallel with the modeling approaches outlined below.
o More generally, what is the rate of transcription initiation changing as a function of time? How variable is it?
o Our code is setup right now to analyze one transcription locus per nucleus as we use heterozygote flies for our reporter. Measuring the intrinsic and extrinsic contributions to noise will require extending our analysis code to be able to deal with two transcription loci per nucleus.


Modeling challenges:
o Models of bursty transcription where the promoter switches slowly between two states, on and off, have been widely used to describe observations in single-celled organisms. Recent results suggest that these models are insufficient to describe the dynamics of gene expression in the early embryo. We will explore alternative models of transcription to determine whether a larger number of transcriptional states need to be considered or whether other sources of noise such as RNA polymerase stalling, enhancer looping or chromatin state modification are necessary. We will develop Gillespie simulations in order to compare the quantitative consequences of different molecular pictures of transcription with the experimental data.
o Gillespie simulations are useful to determine the intrinsic variability of a biochemical process. Our measurements will also yield light on the extrinsic variability of transcription, which we will incorporate in our modeling efforts.
o Every locus of transcription we measure corresponds to two sister chromatids that have just replicated. We will explore limiting cases in our models where these sister chromatids are completely correlated or uncorrelated.
o Different pieces of evidence point at the role of spatial averaging over noisy transcriptional decisions in order to create smooth patterns of gene expression. We will explore models of mRNA diffusion and reconcile the expected evolution of the mRNA pattern with our mRNA production measurements in space and time as well as measurements of mRNA and protein profiles in fixed tissue.