Sequencing mRNA from cryosliced Drosophila embryos to screen genome-wide patterning changes
A complex, spatially and temporally dynamic network of gene expression underlies proper metazoan development, yet methods have not previously existed to assay this network in an efficient, systematic manner. Brute-force application of in situ imaging has been used to painstakingly assay the network, but is impractical to apply to assay the complete gene expression network in multiple mutant conditions. Sequencing, in contrast, is fast and genome-wide, but is generally applied to homogenized tissues, discarding any spatial information.
In the first phase of my research, I have developed methods for performing mRNAseq to globally profile gene expression in dissected subsets of single Drosophila embryos. The patterns I measure are consistent between replicates, and also faithfully reproduce patterns already known from in situ hybridization experiments. However, the global nature of this experiment also reveals spatial patterning in many genes that have not been successfully assayed, even in relatively large-scale projects. Furthermore, I demonstrate that this can be used across samples from different developmental time points to produce a genome-wide developmental atlas of zygotic gene activation.
Because the small sample size of slices of single embryos limited the ability to use standard cDNA library preparation protocols, I next assayed a number of previously published protocols that were specifically designed for minute samples. Previous literature had not addressed whether these protocols would be suitable for reconstructing spatial patterns, and I was particularly concerned that pre-amplification steps would introduce unacceptable non-linearities in the data. Upon analyzing the results, however, I determined that all the protocols I tried were acceptable, and all approximately equally good. I also investigated a few modifications to one of the protocols that reduces the cost of library preparation such that it is no longer the primary limiting factor in terms of number of samples to be sequenced.
Finally, with these methods well refined, I have sliced and sequenced embryos with severe genetic perturbations to maternally provided factors at the head of the patterning network. Comparing embryos deficient in the key pioneer factor zelda, I have revealed that this lethal mutant is nevertheless able to correctly produce the majority of gene expression patterns in the wild-type embryo. Embryos lacking the maternally provided activator bicoid show a greater loss of proper patterning, including the surprising result of ubiquitous over-expression of a number of genes. Simple models of local action of bicoid cannot easily account for this change. Surveying bicoid over-expression mutants and hunchback knock-downs has also yielded a large number of unexpected patterning changes. All of these changes recapitulate previously measured patterning changes, but also highlight new avenues to investigate.