Technical considerations

Extraembryonic development in M. abdita is a dynamic process that lasts for 10 to 14 hours and takes place during almost the entire embryogenesis from gastru- lation to dorsal closure. The serosa is specified at the blastoderm stage on the dorsal side of the embryo, expands and eventually fuses ventrally. Accordingly, the most important requirement for an in vivo analysis of this developmental process was to image the embryo in its entirety and as gentle as possible. A

microscope that allows long imaging, in toto and at high temporal resolution with low photo-toxicity is a Multiview Selective-Plane Illumination Microscope (MuVi-SPIM).

The use of a MuVi-SPIM for in vivo time-lapse recordings requires fluorescently labeled samples [Keller, 2013]. For model organisms, transgenic lines with ubiq- uitously expressed markers outlining cell membrane, nuclei, or other cellular components and structures are readily available [Edwards et al., 1997]; [Martin and Wood, 2002]; [Riedl et al., 2008]. However, such tools have not yet been established for M. abdita. During the course of my PhD, I therefore established a protocol for stable transgenesis in M. abdita and generated a transgenic line that expressed Mab-His2Av-mCherry as nuclear marker that could be used for MuVi-SPIM time-lapse recordings ( [Caroti et al., 2015] and Appendix 2). Although the transgenic M. abdita retained the insertion, it was not possi- ble to maintain the ubiquitous expression of Mab-His2Av-mCherry over more than ten generations ( [Caroti et al., 2015] and Appendix 2), exposing one of several challenges in the usage of transgenic lines to introduce fluorescent markers. Aside from a loss or selective silencing of transgene expression, it has been difficult to identify M. abdita enhancer sequences that could be used for ubiquitous transgene expression. Until very recently [Vicoso and Bachtrog, 2015], available genome assemblies of M. abdita had been sufficient to identify genes, but not good enough to identify several kilobases of putative regulatory sequence upstream or downstream of the putative coding sequence. As pos- sible alternative to ubiquitous transgenic expression, the injection of mRNA encoding, e.g., a His2Av-GFP fusion, have been successfully used for ubiqui- tous marking cell structures and outlines [Benton et al., 2013]; [Benton et al., 2016]. Unfortunately, such mRNA constructs did not seem to provide ubiqui- tous labeling in D. melanogaster and M. abdita, possibly because the protein did not diffuse ubiquitously and/or its expression was not high enough to allow

in vivo analysis [Danai, 2012]. To analyze the shape and the morphology of the

serosa at cellular and tissue level, I therefore decided to inject recombinantly expressed protein. To label cell membrane, actin is a possible candidate and has been used before, e.g. in D. melanogaster [Verkhusha et al., 1999]. Direct

emerged as widely used alternative. Lifeact is a 17-amino-acids peptide that links the filamentous actin (F-actin) in eukaryotic cells and tissues without interfere with actin dynamics [Riedl et al., 2008]. The peptide Lifeact in frame with the fluorescent protein mCherry was recombinantely expressed in E. coli and injected in M. abdita embryos at blastoderm stage. After injection, the embryos were mounted for the MuVi-SPIM as described in Appendix 1. Using this method I was able to ubiquitously label the cell boundaries in fly embryos during the entire embryonic development.

Analysis of extraembryonic development in M. abdita required to observe the dorsal side of the embryo – where the serosa and amnion anlage are specified – and the ventral side of the embryo – where the serosa eventually fuses – all at the same time. To achieve this, the 3D reconstructed embryo (called fusion in Fig. 2.10.1) had to be transformed into a 2D image using a cylindrical projec- tion that simultaneously displayed the dorsal as well as the ventral side of the embryo. To this end, I collaborated with Everado Gonzalez, who developed a dedicated image-processing algorithm for this task. In brief, the algorithm was programmed to process a recorded 3D volume such that (1) it recognized the shape of the embryo, (2) it removed the volume around the embryo, (3) eroded the outermost layer of image information from the embryo, and (4) subtracted the eroded image from the full embryo to end up with a thin layer that was just a few micrometers in thickness and contained all relevant image information of cells at the surface of the embryo minus the yolk and other cells that were deeper in the embryo. This process reduced the data to a quasi two- dimensional image that was projected onto a flat surface and thus generated a 2D map of the embryo surface, which contained ventral as well as dorsal views in one image simultaneously (Fig. 2.10.5). In the rest of the thesis I will refer to this process as “unrolling” and at the 2D movie as “unrolled movie”.

3D

2D

Figure 2.10: Processing of the data from the microscope to the analysis. The four

views obtained with the MuVi-SPIM were fused together to form a complete 3D embryo (1). To be able to transform this 3D information into 2D, the volume around the embryo was removed by the thresholding (2). The surface of the embryo was eroded until only the volume inside it was left (3). The eroded embryo was subtracted from the thresholded embryo therefore only the surface was left (4). The cylindrical surface was cut on the ventral side and unrolled (5) giving rise to an entire embryo in 2D, where dorsal and ventral sides were visible at the same time. Black is the volume around the embryo; purple is the surface of the embryo; pink is the yolk; grey dashed line indicates the position of the cut. The 3D and 2D embryos are oriented with the anterior side to the left.