Abstract
Introduction
SLAS Europe presented a High-Content Screening Conference at the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, Germany, June 27-29, 2016. The conference was part of a series of conferences organized by SLAS Europe on various topics as SLAS expands its educational and networking activities in Europe.
The conference brought together speakers from academia and industry to present their latest results and innovations in the field of high-content screening (HCS). HCS is a vibrant, expanding field that is evolving rapidly, and the conference offered a glimpse of the current state of the art and future perspectives. The achievements of four conference participants comprise this JBS Special Collection on High-Content Screening.
The conference and resulting special collection reflect some of the aspects where HCS is likely to develop in the near future. These fields can be summed up in three broad categories: (1) complex cell biological systems for screening, (2) novel screening modalities and readouts, and (3) data analysis.
Complex Cell Biological Systems for Screening
A search in PubMed with the term high-content screening 3D yields 163 publications with the first dating back to 2002 ( Fig. 1A ). A similar analysis with the term high-content screening stem cells yields 461 publications dating back to 2000 ( Fig. 1B ), and last, a search with the term high-content screening primary cells yields 752 publications starting in 1975 ( Fig. 1C ). All three searches show increasing numbers of publications per year. As control, a search with Drosophila does not show this trend ( Fig. 1D ) and zebrafish only a more modest increase (data not shown).
Number of publications by year found in PubMed with the search terms (
This simplistic analysis reveals the trend of applying more complex biology in HCS. The purpose of embracing more complex cell culture is to study more physiologically relevant biology. For drug discovery, the hope is that fewer drug candidates will fail in the clinic as the assays will be more predictive and animal testing will be reduced. For basic research, many cell biological phenomena cannot be studied in cancer cells, as they do not try to maintain a homeostatic balance within a body but are solely focused on growth. Thus, developmental processes and homeostatic regulation need to be studied in physiologically relevant cellular systems.
This trend toward more advanced cell culture is also evident at specialized HCS or phenotypic screening conferences like the SLAS Europe conference in Dresden. SLAS invited its conference speakers to submit manuscripts based on their presentations for this special collection. The four speakers whose articles are showcased in this collection discuss methods to either cultivate or analyze advanced cell culture systems, which again demonstrates the trend toward sophisticated cell culture systems in HCS.
The articles discuss the use of primary cells, induced pluripotent stem cells (iPSCs), and 3D cell culture. Shana et al. 1 share an RNA interference screen to determine factors secreted by fibroblasts to maintain the liver function of primary human hepatocytes. Primary hepatocytes lose rapidly their liver-specific functions when cultivated ex vivo and require extracellular matrix components and support from stromal cells to maintain function. If isolated hepatocytes are to be used for therapy, it becomes necessary to determine the factors that maintain liver function, and the described assay aims at discovering the ones secreted by fibroblasts.
Kerz et al. 2 present an analysis of phase contrast images of live iPSCs. iPSCs are increasingly used in HCS either in their pluripotent state or as differentiated cells.3,4 They replace immortalized cell lines that have suffered genetic drift over many decades in cell culture. HCS is also beneficial in optimizing differentiation protocols as it is an optimal method for carrying out multifactorial analysis and thus monitoring multiple cellular phenotypes in parallel.5,6
The other two articles describe the use of 3D cell culture in HCS. Booij et al. 7 describe the implementation of a 3D cancer cell culture in hydrogel to study the inhibition of metastasis by small chemical compounds. This assay is an excellent example of the direction in which HCS is developing. It involves 3D cell culture, complex imaging in 3D, and complex analysis to extract the crucial information. Das et al. 8 use 3D cancer spheroids grown in liquid overlay on agarose-coated plates to study growth. They find that plate edge effects affected their readout and diminished the power of their assay. In this article, they describe the optimization of their cell culture conditions to improve the homogeneity of their spheroid size.
These four reports exemplify the trend in HCS to study more complex cellular assays ranging from stem cells and primary cells to cocultures and 3D cell culture. One further advanced biological system used in HCS is small model organisms. In particular, zebrafish are amenable to microscopy-based HCS, and several posters at the conference presented assays using zebrafish. In particular for drug discovery, zebrafish are an attractive vertebrate screening model as they recapitulate many physiological traits of humans. Translatability of isolated drug candidates into humans is a worry, but the current lack of any whole-body physiology, even in advanced cell culture systems, is also a cause for worry about translatability of drug candidates. There is therefore a case to study model organisms, especially organisms with powerful genetic tools. A presentation by Erdem Bangi 9 about Drosophila in profiling cancers of patients and predicting the best therapeutic approaches is an excellent example of the power of small organisms in drug discovery.
Novel Screening Modalities and Readouts
As more complex cellular systems are used in HCS, their analysis becomes more challenging. The SLAS Europe HCS conference was heavily focused on imaging, and several examples of advanced imaging applied at high throughput were presented.
One of the emerging themes was smart microscopy, where the microscope acquisition is controlled by feedback from image analysis. Smart microscopy has been developed to allow acquiring images where something “interesting” has or is happening and avoid taking images with no object of interest, thereby reducing data quantity. Typically, the microscope surveys the plate for an event of interest and then switches to an acquisition mode that is generally more data intensive, such as high-resolution 3D stacks, fast live-cell imaging, super-resolution microscopy, total internal reflectance (TIRF), or fluorescence recovery after photobleaching (FRAP).10,11 Currently, smart microscopy on commercial systems is restricted to low-resolution scans of a plate followed by high-resolution imaging of objects of interest (prescan/rescan), and the other modalities are currently not commercially available.
The prescan/rescan protocol is very useful when acquiring 3D objects, tissue slices, or small organisms. For homogeneously growing immortalized cell lines, randomly acquired images always have cells in the field of view (except for toxic conditions). For 3D objects, most images do not contain any objects. Several possibilities exist to circumvent this problem. One possibility is to acquire low-resolution images of the entire well. This approach is used in the articles by Das et al. 8 and Booij et al. 7 to study cancer spheroid growth and cancer spheroid metastasis, respectively. This approach is also typically used for zebrafish screening.12,13 The low resolution precludes measuring subcellular features but allows measuring cell numbers, colony shape, and texture features.
For high-resolution imaging, two other solutions exist: either the objects need to be brought where the lens is or the lens is brought to the object. For centering objects in the middle of the well, round-bottom plates or agarose-coated plates can be used. Greiner Bio-One proposes a method using magnetic manipulation by coating the objects with magnetic beads and attracting the object in the center of the well with a magnet.
Last, smart microscopy is available in a number of commercially available microscopes such as the CV7000 of Yokogawa, the Opera Phenix of PerkinElmer, and the ImageXpress Micro of Molecular Devices. Users can perform a low-resolution scan of a plate, determine objects of interest by image analysis, and use the coordinates of the objects to acquire high-resolution images. The use of smart microscopy will presumably grow in HCS, as it allows acquiring high-quality data while maintaining the amount of data within reasonable bounds. It is also to be expected that it will allow the introduction of advanced fully automated imaging modalities such as super resolution, FRAP, and TIRF. This, in turn, will allow the development of novel, informative assays to interrogate and study biology in the future.
At the SLAS Europe HCS conference, Jan Huisken presented his automated single-plane illumination microscope (SPIM) for use in screening zebrafish embryos and other 3D objects. His group has developed software solutions to project the data on a 2D map, thereby preserving the 3D information but discarding the voxels that do not contain information. 14 With this approach, it is possible to exploit 3D information and SPIM without running into a computational bottleneck.
A further innovative technology was presented by Oliver Otto from Zellmechanik who developed a high-throughput instrument to measure the stiffness of cells. This technology will be coupled in the future with image cytometry to yield a multiparametric profile including biophysical parameters such as cellular deformability (www.zellmechanik.com).
Data Analysis
Obviously, the development of more complex cell culture and more advanced imaging has an impact on data analysis. Here, high-content screeners have two main challenges: first, more complex and larger data that require larger computing resources and, second, communicating assay quality metrics and screening results to people not used to multiparametric analysis to enable them to make decisions and fully comprehend the implications of the analysis. For the first problem, it will be interesting to follow how commercial providers of analytical software adapt their licensing model to large computer clusters. Commercial software installed on a workstation controlling a high-content imager will not be sufficient in the future for large-scale data analysis. The model of paying annual licensing fees for each CPU will not be applicable to the cloud due to the number of CPUs used and the short duration of use. It is to be expected that a model where licensing fees are paid for the usage of the software will appear.
An alternative is the use of open-source software such as CellProfiler, KNIME, CellCognition, ImageJ, and other platforms. Pharmaceutical companies will need to settle upon a strategy to keep their software costs under control as the size of data increases. AstraZeneca, Janssen Pharmaceuticals, and Novartis have come up with three different solutions to the problem. AstraZeneca opted for purchasing Definiens, the company producing its favorite image-processing software (www.astrazeneca.com/media-centre/press-releases/2014/medimmune-definiens-aquisition-tissue-phenomics-26112014.html). Janssen Pharmaceuticals developed its own image analysis solution (phaedra), and Novartis uses open-source software such as CellProfiler and KNIME. 15 Thus, different companies have come up with different solutions to address the issue of analyzing large data sets on computer clusters, whether in the cloud or on in-house computer farms.
The second problem of reporting complex analytical results to colleagues or management who are not used to multiparametric analysis is trickier. After all, you have spent an awful lot of time obtaining a multiparametric profile that yields a fine-grained analysis, and you would like to be able to show it off. Kerz et al. 2 in this collection use the first component of their principal component analysis of time series to follow the evolution of their cells over time. A multiparametric extension of the Z′ quality control criteria has also been proposed by Kümmel et al., 16 and the Mahalanobis distance has been used for quantifying phenotypic strength. 17 The problem with these tools is that they do not capture the diversity of phenotypes found in the assays. Other fields are also facing the same challenge such as mass cytometry (CyTOF), which analyzes the expression of up to 60 proteins. Self-organizing maps have been used for visualization, and there will likely be more solutions in the future. 18
In summary, HCS is clearly a vigorously developing field driven by innovation in cell culture techniques, imaging technologies, and computing methods. These advances are fueled by the goal to obtain physiological relevant readouts from high-throughput assays. HCS is a demanding, multidisciplinary field requiring knowledge in biology, image processing, statistics, and programming, and the future is likely to demand more of these skills rather than less. Biologists are often not well trained in mathematics and computation, but it is important that they become familiar with these fields to be able to tackle the challenges of HCS in the future.