High-Content Analysis (HCA 2013): 3-D Reduces the Gap Between In Vitro and In Vivo

The 10th High-Content Analysis meeting attracted about 300 scientists to San Francisco’s Fairmont Hotel (January 8–11, 2013). A Venn diagram of attendees would certainly include overlapping circles for specialists in biology, microscopy, informatics, statistics, and biochemistry. Major themes included screening in 1536-well plates; label-free screening, especially of live cells; and imaging zebrafish embryos plus 3- or 4-D imaging of cells and spheroids. Most applications are related to improving health care, but the field of view is expanding to include nano materials.

Initially, 10 years ago, the assays were relatively simple. Tissue or cells were challenged with elements of a compound library, incubated for an arbitrary fixed time, and then stained. Often, in vitro results did not correlate with in vivo. Over the years, assays have evolved to be more biologically relevant with assays more similar to the mode of action of the biological system. Other factors that are essential for success are: Each project needs a power user to drive and motivate the individuals involved in the screening project. Data files are huge, and most data are not used since they are difficult to comprehend. Red/green heat maps are not working for displaying complex relationships. Data integration from various sources is still a common pain point.

Session on 1536-well technology

Dr. Debra Nickischer (Bristol-Myers Squibb, New York, NY) led off a session focused on 1536-well plates. Reagent cost, particularly antibodies, is a major driver favoring higher well density and reduced well size. Other drivers include the need to screen larger libraries and more sophisticated cell models. Assembling all the pieces‒including washer, robot for plate moving, informatics, and two imagers (Thermo Scientific CellInsight™, Thermo Fisher Scientific, Waltham, MA)‒took two years. However, in late 2012, the vast majority of plates were 1536 wells. The last 96-well plates were run in Q3 of 2011. Dr. Nickischer advised that plate vendors need to be qualified. Only a few offer consistent performance. She anticipated problems with scale-down such as cell washing, but these were less than anticipated. Cell population in each well was one key factor‒too high, and the cells clumped, which was recorded as more than 100% inhibition.

In her experience, informatics is a complex problem since the modules were from different vendors. Drivers for most modules are available, with effort. But the structure of the output files and databases is different, which inhibits data integration. Data storage needs to be scaleable, at least into the mid-TB range. Raw data, including images, need to be stored indefinitely. Attempts to save time by screening initially with a limited parameter set were counterproductive, since the derived models were too simple or misleading.

Image searching and processing

HCA is all about image processing. Each well, or z slice, can lead to more than 100 descriptors and data points. One session focused on processing of images. Prof. Robert F. Murphy (Carnegie Mellon University, Pittsburgh, PA) discussed the problem of comparing images with a computer. Key questions are“What is this? Has anyone reported this before?” He described development of Omero searcher, which sits on top of Omero for searching image files from private and public databases. (Omero is an open-source software for acquisition and curation of images.) The search can be content- and context-based. Content-based searches find images that show similar features to one or more query images. The search can be iterative as the search criteria are refined. Murphy pointed out that image searching of early images from screens is effective in predicting side effects. Image searching is an ideal candidate for active learning. Prof. Murphy closed by soliciting participation of data-rich early adopters for retrospective drug discovery studies based on image analysis.

Imagers come with a range of capabilities, speed, and cost, according to Prof. Tiao Xie (Harvard Medical School, Cambridge, MA). Imagers include the Opera® Confocal Imager from PerkinElmer (Waltham, MA) and the ImageXpress Velos Laser Scanning Cytometer from Molecular Devices (Sunnyvale, CA). The images obtained are more than adequate for assays based on classifying and counting. Prof. Xie advised that the imager should be selected appropriately because each high-content screening (HCS) assay has its own set of idiosyncrasies.

Study of zebrafish embryos

Zebrafish embryos are studied as a way to identify toxicity. Dr. Jyotshna Kanungo of the FDA’s National Center for Toxicological Research (Bethesda, MD) described how ethanol and ketamine affect neurogenesis, particularly axon development. The fish are genetically labeled with green fluorescent protein. This facilitates imaging of the motor axons. The biggest problem appears to be random orientation of the fish. Since axon growth is perpendicular to the major axis of the fish, the side view is required.

Issues in 3-D imaging

Graduating from 2-D to 3-D or higher imaging was a major topic at HCA 2013. The initial lecture in the session was from Prof. Anthony M. Davies of the University of Dublin (Ireland). He explained that the transition is not simple since the cells need to be kept healthy and happy. (Tools for 3-D cell culture are described at http://www.biocompare.com/Editorial-Articles/117891-3D-Cell-Culture/.)

One technology uses cells suspended in hanging drops below a plate (InSphero AG, Schlieren, Switzerland). Several others have engineered scaffolds or matrixes, but these interfere with imaging. Others recommend hydrogels, but these are loaded with toxic concentration of calcium ion. How do you get nutrients to the cells and waste products out? Then there is labeling or staining. Micropatterned structures are not ideal since the pattern can interfere with autofocus on the upper end imagers.

Prof. Davies described the trinity cell suspension media for 3-D culture. The media has a low viscosity, which makes it compatible with liquid handlers and fluidics-based imagers. Cells can be washed and labeled. Secreted substances as well as the cells or microtissues are available for sampling. Best yet, cells tend to self-organize and grow. He presented one study of formation of prostate cancer spheroids during a 72-hr incubation. The yield was more than twice that using 2-D conditions.

Spheroids for in vitro studies of solid cancer tumors

Carsten Wenzel of Bayer’s Lead Discovery Facility in Berlin, Germany introduced spheroids for in vitro studies of solid cancer tumors. At Bayer, spheroids are regarded as a reliable tool for HCA, particularly for identification of compounds targeting the tumor’s interior regions. Many cancer therapeutics only access peripheral cells. Some hits were identified that failed to induce cell death in 2-D screens. These hits showed improved induction of cell death when coadministered with commonly prescribed cytostatics.

In more detail, cells in spheroids experience strong concentration gradients, which control cell activity. Cells near the surface of the spheroid have access to nutrients including oxygen. In contrast, interior cells are starved, leading to quiescence. Chemotoxics often affect only the peripheral cells. These are killed, but they can still protect the interior cells. Eventually the dead cells are degraded, exposing the quiescent cells to the nutrients, which stimulates them to become more active. The hypothesis is that targeting quiescent cells could improve efficacy of cancer treatment.

Bayer invested in developing technology to screen spheroids by culturing spheroids in 384- and recently 1536-well plates. The spheroids are one/well. Morphology appears uniform. Imaging was with a confocal Opera from PerkinElmer. Some spheroids are studied further off-line with light sheet microscopy, which provides image slices every 2.6 μm. Files for these images are about 1 GB or TB/plate.

3-D + time = 4-D

Later in the session, Dr. Bonnie Sloane (Wayne State University, Detroit, MI) described live cell imaging of the attack of breast cancer tumor cells on normal tissue. The images showed the time course of the invasion (i.e., 4-D) and optimistic images of effective compound screening.

I was especially impressed with a lecture on heart failure by Mark Mercola of Sanford-Burnham Medical Research Institute (La Jolla, CA). Dr. Mercola was the first to show films of cardiac myocytes beating due to Ca++ induced contraction and relaxation. Heart attacks often produce fibronetic cells in the heart tissue, which weakens the heartbeat. His team performed a screen to find small molecules that apparently induced apoptosis of fibronetic cells. Seven compounds (hits) were found out of several hundred thousand. To evaluate their potential in reversing cardiac fibrolysis, three cohorts of a particular mouse strain were selected. One cohort was untreated (i.e., control), and two cohorts were surgically treated to reduce blood flow to the heart, which induced fibrolysis. Half of the cohort received no treatment; many died. The other half was treated with injections of the hits; none died. Plus, the calcium-regulated pumping cycle of each cell in the heart tissue returned to near normal, returning cardiac function to near normal. These studies were based upon measuring the pumping action of thousands of single functioning cells. The results were dosage dependent and dramatic.

Credits

Cambridge Healthech Institute’s team on HCA deserves special recognition for organizing another fine meeting in the series. I think that their efforts have been responsible for reducing the discrepancy between in vitro and in vivo by about one log per year. Having a meeting in a famous, nostalgic venue year after year is a real attraction. Of course, the views of San Francisco from the rooms cannot be beat. Please monitor the CHI website (http://www.healthtech.com/) for news about HCA 2014.

Robert L. Stevenson, Ph.D., is a Consultant and Editor of Separation Science for American Laboratory/Labcompare; e-mail: rlsteven@comcast.net.

Comments