Project development P5 (2015)

The biological systems

The characteristics of cancer biology and its clinical consequences render a strong case for studying cell division and migration at the systems level. Cell division and cell migration are dynamic processes, in which the dramatic spatiotemporal dynamics of cells, sub-cellular machineries and molecules dictates the outcome. The molecular complexity of these biological processes and their deregulation in cancer cells need to be addressed by powerful imaging tools and combined with systems-scale perturbation experiments. The study of cell division and cell migration form the basis for the activities in WP1 and WP2 respectively.

During the fifth year of the project, in WP1, we continued the work on the functional characterization and modeling of the earlier identified candidate Ki-67 which provided novel insights into the biomechanics and regulation of mitotic chromosomes. We also finalized the work on the validation of protein complex functions in live cell imaging-based structure function assays using the novel human mitotic gene ARHGEF17 identified in a genome wide RNAi screen previously during the project. We have generated and validated several cell lines and staining protocols using photoactivable markers for super-resolution imaging of key proteins in mitosis. Further, we have continued our work on the recently developed technologies based on CRISPR/Cas9 gene editing and the technology is now readily available and can be applied to a vast number of biological questions. Using the newly developed Cellular Microscopy Phenotype Ontology (CMPO), we investigated the relationships between cellular phenotypes and cellular functions using CMPO- and GO (gene ontology)-annotated data from large microscopy-based screens, thus coming closer to not only test hypotheses about gene functions, but to provide as well a method to organize and search a repository of RNAi screen data. In addition, we have built a comprehensive dynamic and quantitative 4D model of a human dividing cell which provides a dynamic standard cellular spatio-temporal reference system that can be used to integrate quantitative information on protein distributions during mitosis sampled in thousands of different experiments. The model allows building an interactive atlas of the mitotic cell, thus allowing the exploration of multiple protein dynamics with spatial and temporal continuity that are currently impossible experimentally. Knowledge gained through such exploration and data mining can then be used to formulate new mechanistic hypotheses about the function of the proteins in the atlas.

In WP2, we continued our work on the relationships between dynamic features of cell migration and the organization of the related cytoskeletal networks. To this end, we have developed an integrative imaging approach followed by an automated image analysis procedure of time-lapse image sequences in order to gain understanding of the relationship between cancer cell migration and the dynamics of matrix adhesions. Migration regulators were found to be associated with specific GO annotations, such as cell adhesion, cytoskeletal regulators, focal adhesion, the hippo pathway, mechanical regulation of the nucleus, oncogenic signaling, mechanotransduction, cell proliferation regulators, and Wnt receptor signaling. In addition, we have used computational approaches to define the genes for which levels of expression are correlated with the breast tumor cell migration and invasion. Furthermore, we identified and validated the most promising hits of a migratory gene screen in clinical samples for subsequent screens by using bioinformatic in silico analysis.

The consortium has also been dedicated to leverage the powerful systems microscopy tools developed within the project in specific translational applications, such as exploration and diagnosis of the dependency of cancer on specific targets, or reactivity towards specific drugs (WP7). Thus, our work demonstrate the potential of systems microscopy, combined with high throughput (drug) screening and other –omics technologies, as a method to diagnose dependency of solid tumours, such as ovarian cancer, on specific gene targets and pathways and to determine the patient’s drug responses in combination with other readouts. In addition, the earlier pipeline, created to link the systems microscopy findings within the consortium with bioinformatics data was updated and used for further analysis of the hit genes from migratory screens.

Improving throughput

Within the last years we have got ahead in establishing the workflow for automated acquisition and analysis of FC(C)S data acquired on both Zeiss LSM780 ConfoCor3 and Leica TCS SP5 SMD imaging systems to extract quantitative protein interaction and mobility parameters. The automated high-throughput imaging and FC(C)S workflow (HT-FCS) has been developed to integrate a screening, a time-lapse, a population, and a calibrated imaging mode, thus  a useful tool to characterize proteins biochemically and biophysically in living cells. Applied to different proteins, it enables high-throughput protein interaction studies that complement established in vitro assays. We have also focused on further development of the data acquisition software for semi-automated, high-throughput super-resolution microscopy. The DSRT platform has been extended from five 384-well plates to a set of eight 384-well plates with over 527 drugs and have optimized tens of antibodies for high-content imaging -based DSRT.

Maximizing content

Available high-throughput image analysis software offer efficient algorithms for analysis of single time-point assays while the existing tools for the analysis of cellular dynamics in multi-dimensional large-scale imaging are very limited. To enable systems biology analyses of the cell division and cell migration, the network started to develop image analysis tools (WP4) aimed at increasing the data content that can be extracted in parallel through microscopy. During our fifth year, the open source software platform CellCognition has been extended by an improved user interface that enables the automated sorting of cell objects based on phenotypic distances, and iterative interactive learning. The new user interface is linked to the high-performance batch-processing module of CellCognition via the standardized exchange data format CellH5. Computational methods were developed for intracellular motion analysis with a new software module aimed at inferring general parameters describing intracellular motility, which can be used for phenotype comparison and for modeling of cellular dynamics. The motion analysis module is implemented in the open source software package R. Integration with other software packages of the Systems Microscopy is achieved through support of the common open source file format CellH5, which was developed in previous years of the project.

Wisoft analysis software packages, Athena and Minerva, were enriched with variety of algorithms and applications. Analysis of FRAP and FRET experiments is now available, where internal calculations are applied to allow the accurate extraction of the diffusion factor in the FRAP experiment and the FRET efficiency in the FRET application.

We have further enhanced the methods for segmentation of nuclei in histopathology data. In particular, we have investigated the use of Generalized Fast Radial Symmetry Transform (GFRST) for finding cell centers. We also have increased the accuracy of the segmentation algorithm by applying learning methods to the separation lines between candidate regions, which proved to be both computationally efficient and beneficial for segmentation performance. Importantly, we have integrated this final version of nuclei detection in H&E stained images into CellCognition, the open-source software for cellular phenotyping developed by members of this consortium.

Data processing, modeling and query

The large, multi-dimensional image-based data sets that are generated from systems microscopy studies pose high demands on the tools used for statistical analysis. Network members have continued to improve software (WP5) that performs primary statistical analysis of complex data along with quality assessment and significance analysis.

A tremendous amount of information exist in the literature, in bioinformatics databases and in published experimental datasets on the function of gene products but whole gene products can usually not be tested in a functional assay, thus identifying the genes most likely to be of interest is of critical importance to avoid ignoring relevant experimental data. Therefore, we developed FUN-L (Functional Lists) to help in selecting target genes for experiments (http://funl.org). During this fifth year, we have further worked on application and validation of this tool to better understand different biological systems as the mitotic chromosome condensation or cell migration.

Development and application of modelling methods for systems microscopy

Data derived from a number of different experimental set-ups have been used to commence bold modeling projects that will ultimately serve to describe the dynamic processes of cell migration and division. Close collaboration between experimental and modeling partners, which are actively exchanging modelling methodology and know-how, has generated extremely promising results (WP6). During the fifth period, we focused on developing new approaches for mining large image data sets generated by Image-based RNAi screens by developing a data integration strategy to combine different data sets and a mining strategy based on a semi-supervised approach. Through tertiary RNAi screening, as described in the reports of the previous years we established quantitative data for phenotypic modules, including the control of the mitotic chromosome periphery and the progression through mitotic exit. We used cellular phenotypes from large scale RNAi screens annotated with the cellular microscopy phenotype ontology to explore how cellular phenotypes relate to GO cellular functions. Further, we have developed a generic method that can predict the localization of proteins in networks of filaments. This method is applicable to arbitrary networks containing overlaps or crossing filaments, for example the mitotic spindle or the branched networks of F-actin found in motile cells.

Standardization measures

Systems biology aims to produce comprehensive models of biological systems, requiring integration of multiple experimental techniques, analytical methods and modeling approaches. A central task for this NoE is therefore the development of standards that enable exchange, interoperability and integration of data from different laboratories (WP8). We continued our work on development of standardized nomenclature of quantitative parameters such as “statistical geometric features” or “convex hull features” extracted form cellular images. The definition of a feature groups is currently generated in CellCognition and persistently stored as output in the CellH5 format. In addition, we developed a CellH5 module with an interface to the open source software CellProfiler, enabling exchange of feature names, parameters, and identifiers. The RBioFormats (https://github.com/aoles/RBioFormats) software package which was first released in 2014, it now provides a fully functional and tested interface between the powerful statistical and visualization environment R and the Bio-Formats microscopy image formats reader. To increase the performance and to enhance user experience, core package functions were refactored. In comparison to the initial implementation, the image data and metadata readout is now 2-3 times faster than before.

Database Development

A prototype database for systems microscopy data has been developed within the framework of WP9 to facilitate standardization efforts toward a common platform for data sharing. The Cellular Phenotype Database (CPD) is running on a production server since April 2013 http://www.ebi.ac.uk/fg/sym and it now contains 10 datasets associated with scientific publications. Phenotypes associated with these 10 studies have been mapped to Cellular Microscopy Phenotype Ontology (CMPO) terms and ontology-enabled query has been implemented through the CPD user interface for phenotype search and browsing. Additionally, over the course of the project period, a lot of focus was given to improve the database performance.

Training the next generation

The network was pleased to welcome a high number of junior participants to its fourth Annual Consortium meeting, many of them presenting their work either by giving an oral presentation (10) or presenting a poster (23). During the fifth project period, several courses supported by the NoE were organized and attended by both NoE external and NoE associated participants, thus providing training opportunities in systems microscopy methodology. In training the next generation of scientists, both events contributed to the durability objective of the training work package (WP10).  A large number of postdocs and students have further been enrolled within the NoE and started their training by research, which constitutes the main training method applied in this NoE.

Dissemination

Structures for internal as well as external communication (WP11) that have been set up during P1 were followed up during the whole duration of the project: the project website (both internal and external) was updated continuously, the brochure (reaching both the scientific community as well as the general public) that was produced earlier in the project was distributed at our symposia, and meetings. The partners were disseminating the project extensively at conferences, meetings, public lectures, by publications in scientific journal but also in local newspapers. In addition, the management office actively disseminated the consortium and its achievements by organizing a mini-symposium during P5 but also attending several conferences and workshops (by distributing our brochure and assigning the Systems Microscopy logo as well as EU-FP7 logo.

Seventh Framework Programme

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer