Global Transcriptional Profiling Analyses

      

One area of emphasis in our work is the analysis of pancreas tumorogenesis. Pancreatic cancer is the fifth most common cause of cancer related deaths in industrialised countries, with a dismal prognosis, an increasing incidence and no or only ineffective means of treatment. The development of new treatment modalities and diagnostic and preventive approaches requires the understanding of the molecular mechanisms of tumorogenesis in pancreas. Since they are only poorly understood, it is of major importance to identify pathways and/or individual genes or proteins relevant to carcinogenesis.

In an EU-funded Concerted Action, a group of nine laboratories coordinated by Thomas Gress at the University of Ulm is working at the identification of disease genes relevant to pancreatic cancer.

   

A complex Cancer DNA-Microarray has been established that contains known cancer-associated genes in addition to gene fragments which were isolated by Representational Difference Analyses (RDA) studies on various cancer tissues. The gene set is steadily extended with new genes of interest. In addition to this complex microarray, more specific arrays are being used for analyses on particular cancer types such as pancreatic carcinomas, colon cancer and others.

Analyses are usually performed by standard experimental procedures followed by data analysis using the M-CHiPS software package.
We currently pursue several new projects on various cancer types and tissues. By direct integration of relevant clinical data and other annotations, such as Gene Ontology information, in the clustering and by a purely statistical analysis of the relevance and significance of annotated factors for the occurrence of a cluster or differences in patterns, corresponding factors can be identified. This is a feature which is unique to the M-CHiPS software package (and is also patent protected). Thereby, particularly relevant pathways can be deduced, for example. Apart from a better diagnosis and prognosis, new therapeutic sites can be indicated with much higher accuracy and probability of being real.

In several cases, we were not able to obtain from samples any data or data of sufficient quality because only small amounts of material were available. While amplification prior to hybridisation was helpful, we obtained even better results by using radioactive label rather than fluorescent dyes. Not only could we increase sensitivity by accummulating signal during longer exposure times, also the dynamic range of detection was substantially improved by this process and thus the accuracy and significance of the results. For the analyses of microdissected material, for example, total RNA was frequently analysed by labelling with ³³P. Hybridisation was nevertheless performed on standard DNA-microarray containing many thousand PCR-products or oligonucleotides. Detection of the results was performed with a phosporimager of appropriate resolution (5 µm pixel size). The figure to the right shows a typical result of a radioactive hybridisation with a sample made of total RNA from microdissected cells. Only part of a microarray grid is shown.

Recently, we substituted this process by a newly developed method for detecting even few binding events with fluorescence. Still, radioactive labelling provides a means to obtain data from samples that would be beyond analysis by standard procedures.

The complete genomic sequence of several metazoan organisms such as Drosophila melanogaster and Caenorhabditis elegans are available. The next task arising from this sequence data is the deciphering of the role and function of the identified genes and their corresponding protein products in the context of an entire organism. As often the expression pattern of a gene provides clues to its function, we produced a DNA-microarray that enabled us to monitor gene expression in the context of the entire Drosophila genome. Such a system should enable us to identify genes whose activities are required for the execution of complex developmental gene networks and signal transduction pathways. As such networks and pathways are evolutionary highly conserved among metazoans, the analyses of gene and protein function in Drosophila should also provide valuable clues for a better knowledge of corresponding pathways in vertebrates.

While the genome sequences for a variety of organisms are now available, the precise number of genes encoded is still a matter of debate. We based our whole-transcriptome microarray, the Heidelberg FlyArray, on the combination of the BDGP annotation and a novel ab initio gene prediction of lower stringency using the Fgenesh software. A microarray was established with altogether some 24,000 different features, each actually present in duplicate. The primer set used to produce the PCR-products is available from Eurogentec.


Apart from being used for the production of the microarray, the very primer set was also applied to the generation of a genome-wide dsRNA library, the actual work being performed in the laboratory of Norbert Perrimon at Harvard Medical School in Boston (USA). This molecule set allows the identification of gene functions by cell-based RNAi-screens.

To assess the overall quality of our array design as well as to validate the novel predictions, we performed developmental profiling of the Drosophila lifecycle using 9 different stages. We were able to provide evidence for the transcription of ~2,600 additional genes predicted by Fgenesh. Validation of the developmental profiling data by RT-PCR and in situ hybridization indicates a lower limit of 2,000 novel annotations, thus substantially raising the number of genes that make a fly. The successful design and application of this Drosophila microarray confirms our expectation that mere in silico approaches will always tend to be incomplete. The identification of at least 2,000 novel genes highlights the importance of gathering experimental evidence to discover all genes within a genome. Moreover, as such an approach is independent of homology criteria, it will allow the discovery of novel genes unrelated to known protein families or which have not been strictly conserved between species.

We are participating in the International Drosophila Array Consortium (INDAC; www.indac.net), which aims at establishing a common, standardised microarray and a corresponding set of controls for the entire Drosophila community.

On the basis of shotgun subclone libraries used in the sequencing of the Drosophila melanogaster genome, a minimal tiling path of subclones across much of the genome was determined. About 320,000 shotgun clones for chromosomes X(12-20), 2R, 2L, 3R and 4 were available from the Berkeley Drosophila Genome Project. The clone inserts have an average length of 3.4 kb and are amenable to standard PCR-amplification. The resulting tiling path covers 86.2% of chromosome X(12-20), 86.2% of chromosomal arm 2R, 79.0% of 2L, 89.6% of 3R and 80.5% of chromosome 4. In total, the 25,135 clones represent 76.7 Mb – equivalent to about 67% of the genome – and would be suitable for producing a microarray on a single slide.

This work was performed in collaboration with Susan Celniker (Berkelely), Eric Johnson (University of Oregon) and Eileen Furlong (EMBL).

Collaborators: Johannes Beckers , Martin Hrabe de Angelis, GSF, Munich; Werner Mewes, GSF, Munich; Martin Vingron, MPIMG, Berlin.

The sequencing of the mouse and human genomes has basically been accomplished. The next step in the integration of this genomic information into biological and biomedical research, will be the systematic analysis of gene function. The similarities between man and mouse in their genomes, molecular pathways, physiology and developmental mechanisms make the mouse the most important model organism for the study of inherited diseases in man.
overall picture of mouse arrayTo discover new mutants that serve as model for human diseases or that have developmental defects, mice that have been subjected to ENU mutagenesis are routinely examined for clinical parameters, behaviour and dysmorphologies (Mouse ENU Mutagenesis Screen, Institute of Experimental Genetics, GSF, Neuherberg-Munich, Germany). In the 'MouseExpress' project, we extend phenotypic description to a new molecular level: using DNA microarrays in a high-throughput approach, we compare RNA expression profiles of thousands of genes from tissues or embryos of mutant and wildtype mice. We use the RNA expression profiles to identify molecular pathways that are affected in disease and analyse interdependencies between pathways within a molecular network. Expression profiles of mutant and wildtype mouse strains are filed in a database and will be linked to the phenotype and mutant databases of the Mouse ENU Mutagenesis Screen of the GSF.

   

Global transcriptional profiling in Arabidopsis thaliana was started in the EU-funded PPMdb-network and extended as part of the German ZIGIA consortium. Analyses were initially performed on a set of some 13,000 non-redundant EST-clones combined from the EST-clone collection of the Institut National de la Recherche Agronomique (Versailles, France), kindly provided by Herman Höfte, and the MSU EST-clone collection obtained from the Arabidopsis Biological Resource Center at the Ohio State University (Columbus, USA). Various conditions have been studied. One area of emphasis was the analysis of pathogen responses, done in collaboration with Nikolaus Schlaich and Alan Slusarenko of the RWTH Aachen, work to which the figure shown on the right refers. For more detailed information and comprehensive data sets, please, press the button below.

While in most cases, relatively moderate changes in the transcript levels were identified, a couple of experimental conditions revealed changes of up to 80% of all genes. While difficult to analyse and interpret, such data sets could become invaluable in determining the regulative changes and interrelations that take place on transcriptional level.

The life cycle of Trypanosoma brucei involves adaptation to a variety of conditions in the host and Tsetse fly. The successive changes in morphology, biochemistry and plasma membrane proteins, some of which also involve cell cycle arrest, are still very poorly understood. Many of these changes are likely to be directed by changes in mRNA abundance and translation, and over two decades of effort have now been expended in the identification of stage-specific mRNAs. Because of technical limitations, however, most of the transcripts identified have been rather abundant, and the regulation studied has mainly been restricted to the rapidly-dividing long slender bloodstream and procyclic forms.

Given the relatively small size of the T. brucei genome, there is a good prospect of a complete exploration of its genome by microarray analyses. This will allow identification of lower-abundance regulated transcripts, and (using amplification methods) the study of the transcriptome of the less accessible forms found in the Tsetse fly. One format for the analysis is to perform genome-wide expression studies on genomic instead of gene-specific fragments. Such arrays have several intrinsic advantages. Overall genome representation is usually good in shotgun libraries with comparatively little variation across the genome. Thus, even if randomly selected clone inserts are used as probes, there should be good coverage and relatively little redundancy. Also, not only coding but also intergenic regions can be studied, for example in chromatin immunoprecipitation experiments. Insert amplification can be performed with a single primer pair and functional analyses can actually precede sequencing.
In collaboration with the group of Christine Clayton, we have produced DNA-microarrays containing more than 21,000 PCR-products of 2 to 2.5 kb long genomic fragments of T. brucei strain TREU927/4, which are being used in several projects.

For the identification of stage-specific gene activity, for example, total RNA from in vitro cultures of the human, long slender form and the insect, procyclic form of the parasite was labelled and hybridised to the microarray. Approximately 75% of the genomic fragments produced a signal and about 2% exhibited significant differences between the transcript levels in the bloodstream and procyclic forms. Results were confirmed by Northern blot analysis or reverse-transcription and PCR.

Ever since Tatum and Beadle formulated their one-gene-one-enzyme hypothesis on the basis of studies with Neurospora crassa, this filamentous fungus served as a model organism not only in genetics but also many other fields of basic research. Despite a lot of successful research, only about one tenth of the genes of Neurospora crassa had been described and localised on the seven chromosomes prior to genome initiatives. Genome analysis started by ordering cosmid and BAC clones along individual chromosomes. Based on the physical clone maps of linkage groups II and V, sequencing of the two chromosomes was done as part of the German Neurospora Genome Project. Simultaneously, a whole-genome shotgun approach was taken at the Whitehead Genome Center, Cambridge, USA, recently yielding the complete genomic sequence.

For an initial insight into transcriptional variations in Neurospora crassa, we started with the creation of a microarray prior to sequence assembly and annotation, however. Some 4,700 EST-clones were arrayed on glass slides and used to monitor nutrient-dependent functional phenomena in Neurospora crassa. Upon availability of the sequence, also arrays made by in situ synthesis of oligonucleotides were used in other analyses.