By Eberhard Krausz
Head, HT-Technology Development Studio
Max Planck Institute of Molecular Cell Biology & Genetics
Dresden, Germany

High-throughput screening is no longer the sole privilege or duty of the pharmaceutical industry. Mainly in Northern America, but increasingly in other parts of the world as well, a remarkable number of academic screening centers have been established  primarily to screen chemical drug-like molecule libraries as molecular modulators of protein functions. In parallel, in the past few years, high-content screening (HCS) based on automated microscopy—particularly in combination with RNA interference for discovery of gene functions in model organisms or in human cells—has been evolving at a number of leading institutes in Europe.

TDS at MPI-CBG
The High-Throughput Technology Development Studio (TDS; http://tds.mpi-cbg.de/webtds/), which is the central high-content screening unit at the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, Germany (MPI-CBG; www.mpi-cbg.de), was founded in 2003 based on a concept elaborated by institute directors Professor Marino Zerial and Dr. Ivan Baines,  with seed funding by the institute. The aim is to develop complex phenotypic cell-based assays in close collaboration with internal and external research groups, and to perform HCS by applying automated microscopy and pattern recognition for image analysis in cultured mammalian cells and in model organisms such as the worm C.elegans. One of the primary goals is to provide an infrastructure to identify functions of genes on a genome-wide scale, applying RNA interference that allows silencing of any targeted genes selectively.

The TDS is also involved in the multi-center RNA Interference Initiative to discover gene functions and the MPG Chemical Genomics Center (CGC; www.cgc.mpg.de) to screen for chemical probes that specifically modulate biological processes. Both of these programs are funded by the Max Planck Society. We are also grateful for two years of support by the Federal Ministry of Education and Research, through its InnoRegio concept of promoting regional key expertise, especially in joint commercial and academic partnerships.

Our network of biological research partners at the MPI-CBG and the Technical University, Dresden was founded on the academic side to develop complex phenotypic assays, and with its partners on the industrial side to develop technology in automated microscopy (Evotec Technologies GmbH, Hamburg & Berlin, Germany); automated pattern recognition and image analysis (Definiens AG, Munich, Germany); and RNA Interference (Cenix BioScience GmbH, Dresden). Additional partnerships were entered into subsequently for lab automation (TECAN AG, Männedorf, Switzerland) and automated microscopy at lower throughput (Cellomics Inc., Pittsburgh, PA).

At the TDS, 12 specialists in cell biology, biochemistry, microscopy, automation, image analysis and information technology work hand in hand in an interdisciplinary fashion to address the challenge of establishing an automated work environment and to deal with the very large data sets generated by large-scale high-throughput automated microscopy-based screening.

Robust, Flexible Infrastructure
In setting up the TDS, we tried to standardize the work flow using automation as much as possible, but still remaining flexible. This was made possible by building integrated conveyor belt-like systems. Cells are counted after trypsination by a CASY cell counter TT (Schaerfe System GmbH, Reutlingen, Germany) and seeded subsequently by dispensers. To pack and deliver the RNAi molecules into the cells, a custom-designed transfection robot was assembled (Freedom Evo 200/8 platform, TECAN) that is completely incorporated into a laminar flow at biosafety level S2. The TDS is probably the first academic laboratory in the world that has established high-throughput siRNA transfection at 55,000 samples per 24 hours in a 384-well format. This was accomplished by applying the Te-Mo 384-steel needle head (TECAN) without using plastic tips, thereby saving up to 15,000 euros in consumables per genome-wide run.

Two additional robotic systems of the same type assist in cell processing and all other in-house projects that require automated liquid handling. To acquire data of biochemical/homogeneous assays, we use a highly sensitive reader for electrochemiluminescence (SECTOR Imager, Meso Scale Discovery) as well as two readers for colorimetric, luminescent and fluorescent signals (Genios Plus and Genios Pro, TECAN). For automated image acquisition at a throughput of up to 100,000 three-color images per day, we use the OPERA QEHS (Evotec Technologies). This confocal high-throughput fluorescent microscope is powered by solid-state lasers at three to four different wavelengths, and is therefore capable of taking pictures by three CCD cameras in parallel. Images at resolutions up to 60-fold magnification are achieved.

To complement OPERA’s confocal optics, we have two epifluorescent ArrayScan VTI systems (Cellomics) that generally work around the clock with little to no supervision to perform large-scale screens where confocal imaging is not required (as is frequently the case for culture cells). The images acquired by the OPERA are applied to image analysis with commercial software packages (Acapella by Evotec Technologies or Cellenger by Definiens), or are analyzed with in-house software solutions such as MotionTracking II (designed and programmed by Yannis Kalaidzidis) to identify desired phenotypical patterns. The ArrayScan VTI provides its own image-analysis software packages called BioApplications.

All data are fed into a web-based front end laboratory information management system that was developed at the TDS, installed on freeware server components, and is conceptually designed for the special needs of HCS, with millions of images linked to other important HCS data such as image processing and analysis results; siRNA molecule design and validation data; assay protocols, plate formats; and more.

We have gathered numerous substance collections centrally that can be applied in individual projects. Preliminary priority is given to various RNAi libraries—for example, chemically synthesized siRNA libraries covering human and mouse kinases; the whole human genome; and enzymatically derived genome-wide esiRNA libraries for mouse and human. The latter were actually produced at the TDS in close collaboration with Frank Buchholz (MPI-CBG Group Leader), a co-inventor with Dun Yang in the laboratory of Michael C. Bishop at the University of California San Francisco of  esiRNA technology.

Further, we have a genome-wide collection of clones in E.coli  that express RNAi molecules upon feeding C.elegans  worms over numerous generations of progeny. Currently, we are expanding by acquiring chemical drug-like small molecule libraries. Professor Herbert Waldmann of the Max Planck Institute in Dortmund and the MPG Chemical Genomics Center are sources for innovative and proprietary molecule selections, partially natural products, and derivatives thereof. We are actively searching for new partners in order to expand the collections.

Assay Development & Screening Projects
The focus of our current activities is on developing novel complex cell-based assays for automated microscopy and, as a second priority, performing screening of molecular libraries at high-throughput. Usually, individual bench-scale experiments are taken out of the collaborating research labs (most frequently, but not exclusively, the research groups at the MPI-CBG) and developed into screenable assays in the labs of the TDS. This service is provided on a cost-recovery basis. We also have a user facility that provides open access after initial training and prior booking to a remarkable range of equipment and instrumentation, such as two liquid handling platforms; a homogeneous reader; and an ArrayScan VTI that everyone in the institute or in collaborating labs can use to run their own experiments and projects.

Criteria for taking collaborative projects on board include, among others, positive assessment on feasibility; strong data demonstrating reproducibility along with a comfortably big screening window, such that the induced phenotype by a number of selected positive controls is clearly distinct from the negative control treated cells exhibiting the normal phenotype—i.e., the variance of both signals must not overlap. In order to transfer an assay into the TDS labs, it must already exhibit reproducibility and practicability as well as the specificity of the controls and distinctiveness of the specific signal over background—all standard criteria for reducing an assay to practice.

First, the assay is miniaturized down to at least a 96-well format, and the protocols are optimized particularly with consideration for future automation and cost minimization. Further, transfection of the siRNA molecules is optimized, and image analysis scripts are elaborated. Finally, the whole process is adapted to lab automation. Subsequently, during the validation process, several independent runs are performed of plates with negative and positive controls. Variation within a plate, among replicas of identical plates, and among the three independent runs are determined and assessed.  If the process passes these quality control criteria, a small pilot study is started, applying a few plates from a reference library. For a proper screening project, at least two—and ideally, three—independent runs are performed. Eventually, hits must be independently verified in secondary assays.

Since the TDS went operational in the summer of 2004, we have developed a number of innovative multi-parametric phenotypic assays and implemented various large-scale screening projects in close collaboration with internal and external research groups. siRNA libraries covering all human or mouse kinases were applied to cell viability/apoptosis, membrane lipid composition, three viral infection pathways, phagocytosis of bacteria, Alzheimer, and osteoclast differentiation screens. Using the initial cell viability/apoptosis screens at 48 and 96 hours exposure time, we determined a temporal screening window for future screens that consists of an acceptable compromise between maximizing incubation time to achieve the optimal silencing effect on the one hand, and minimizing the number of targets whose depletion significantly affects cell viability on the other hand.

Our first functionality showcase was a multi-assay screening project applying two primary viral infection assays and six secondary endocytosis assays to dissect the viral infection routes. Screening the whole human kinome, we identified a complex regulative network for two endocytotic routes (Pelkmans et al., Nature 2005, 436:78-86). The results formed the basis for the discovery of a new mechanism of caveolae-mediated endocytosis (Pelkmans and Zerial, Nature 2005, 436:128-133).

This project has now evolved dramatically. Replacing the viruses with various fluorescently labeled endocytosed cargoes (e.g., growth factor receptors, other receptors) has made the screens far more robust. A new version of the automated microscope OPERA has allowed us to add an additional fluorescent marker. A novel image analysis software concept programmed by a visiting professor at the institute, Yannis Kalaidzidis, allows extraction of more than 40 parameters pertaining to the movement of the internalized cargoes, while access to the super computer at the Technical University, Dresden, allows processing of such complex image analysis at the required scale of 100,000s of images. A genome-wide screen has now been initiated.

Services provided to the research group by Frank Buchholz have enabled the performance of a number of large-scale screens, as a result of which we published our first report on a screen to identify new genes involved in cell proliferation and division (Kittler et al., Nature 2004, 432:1036-40). Further, a genome-wide four channel screen in dual-stable fluorescent C.elegans worms has been carried out at the TDS and is currently being evaluated, along with the performance of subsequent hit verification studies.

Looking Ahead
In the very near future, parallel chemical screens will be performed under the umbrella of the Chemical Genomics Center of the Max Planck Society. Functional genomics and chemical biology will be linked not only to identify new gene functions, but also to search for chemical modulators of the corresponding protein activities. We are inspired by the belief that multi-parametric cell-based analysis provide extraordinary potential to achieve scientific progress and accelerate opportunities for therapeutic applications, principally by allowing better identification during early discovery of those compounds that are most likely to yield successful drugs.