Author contributions: X.X. and S.T.W.: wrote the manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS June July 20, 2012; available online without subscription through the open access option.
High-throughput screening (HTS) is a technology widely used for early stages of drug discovery in pharmaceutical and biotechnology industries. Recent hardware and software improvements have enabled HTS to be used in combination with subcellular resolution microscopy, resulting in cell image-based HTS, called high-content screening (HCS). HCS allows the acquisition of deeper knowledge at a single-cell level such that more complex biological systems can be studied in a high-throughput manner. The technique is particularly well-suited for stem cell research and drug discovery, which almost inevitably require single-cell resolutions for the detection of rare phenotypes in heterogeneous cultures. With growing availability of facilities, instruments, and reagent libraries, small-to-moderate scale HCS can now be carried out in regular academic labs. We envision that the HCS technique will play an increasing role in both basic mechanism study and early-stage drug discovery on stem cells. Here, we review the development of HCS technique and its past application on stem cells and discuss possible future developments. Stem Cells2012;30:1800–1807
High-throughput screening (HTS) is technology developed and widely used in the pharmaceutical and biotechnology industry. Using robotics and computers, HTS achieves the full automation of biological or pharmacological experiments so that vast amounts of reagents can be rapidly evaluated to identify biologically active candidates that can potentially be used as drugs. In industry, HTS workstations are now capable of screening tens of thousands of compounds daily, providing an irreplaceable tool for lead discovery [1, 2]. Recently, with the growing availability of instruments and reagents, HTS is increasingly used in academia because of its emerging potential in basic biomedical research.
HTS is made feasible by the latest technologies including robotics, liquid handling, high-sensitivity detectors, and high-performance computing. Since its establishment, its capability continues to expand with the advances in these technologies. HTS technology was initially spurred by the concept of target-based drug discovery. In the 1970s, HTS successfully demonstrated its power in identifying compounds against isolated molecular targets. From the late 1980s to the mid-1990s, great advances in combinatorial chemistry and genomics fuelled a rapid development stage of HTS: large numbers of new compounds were synthesized for screening, and potential new drug targets were discovered at an unprecedented rate. The primary goal of HTS development during this stage was to increase its capacity to allow the screening of massive amounts of new compounds against a rapidly increasing number of new targets. To maximize throughput, assays were carried out with purified biochemical molecules in homogeneous solutions so that the reaction could be readily miniaturized and the signal quickly detected by a plate reader. This effort led to ultra-HTS capable of testing millions of compounds. However, to researchers' and the public's disappointment, this “quantity” improvement did not result in the expected increase in new drug discovery. Instead, the productivity of the whole pharmaceutical industry decreased over the past decades [3, 4]. We now realize that a “quality” improvement to HTS is desired, despite the higher upfront cost, to reduce the high attrition rate in downstream pipelines. Apparently, one of the causes of inapplicability of screening hits in vivo is that biochemical screening assays do not reflect the complexity of living systems. Novel higher order biological assays are needed for a more systematic and accurate evaluation of the reagent effect. To address this challenge, cell-based assays have been increasingly used in HTS since the 1990s as a more physiological alternative to biochemical assays. In vitro cultured cells provide a first-level living complexity, in which not only the reaction with the molecular target itself is considered but also the interaction with the complete regulatory network and feedback control mechanisms . This technique also provides evaluation of membrane permeability, off-target effects, and cytotoxicity, which are critical for the eventual clinical application. Moreover, in contrast to biochemical assays, cell-based assays do not require the purification of biological molecules, which is often challenging. In fact, it can even be carried out without the prior knowledge of a defined molecular target; instead, compounds can be screened with a completely unknown mechanism using phenotypic assays. These advantages have led to increasing consideration that cell-based assays are a mainstream HTS technology, and today, cell-based assays represent approximately half of all HTS performed .
Complete evaluation of a complex cellular network is beyond the scope of the current technique. In early applications, the complexity had to be reduced using reporter cell lines so that the target-related signal could be readily detected from the whole well; in many studies, the cells were also genetically engineered to overexpress the protein of interest to narrow the target space. This approach is compatible with a traditional plate reader. However, with such compromises, the screening clearly did not fully use the systems biology readout of cell-based assays. Signals detected are average of the population of all the cells in the well, such that the heterogeneity of cellular response is overlooked. Also, it usually only measures the status of a single target, wasting the rich information available. Among all current techniques, microscopy provides the best solution to this problem with its spatial, temporal, and spectral resolution. Combination of microscopy and HTS has led to a new technology: high-content screening (HCS). Broadly speaking, HCS refers to any HTS that uses multiple measurements as readout. However, HCS is now often a specific name for cellular image-based HTS (Fig. 1) [7, 8]. Notably, although microscopes have been used by biologists for over a hundred years, their high-throughput application has only recently become feasible through innovations in both microscope hardware and especially automated image analysis software. The first HCS system was created by the founders of Cellomics in 1996 ; today, it has been widely used in all the different stages of drug development. While it retains high-throughput capacity, HCS permits the quantitative observation of comprehensive phenotypes by providing a spatial resolution at subcellular level, a temporal resolution ranging from milliseconds to days, and a spectral resolution for the observation of multiple targets. Since its introduction, HCS has undergone rapid development and its capacity has been more and more explored. In its simplest application, cells expressing different biomarkers can be counted and analyzed separately in a mixed culture, enabling heterogeneity studies, which are unreachable with traditional ensemble measurements . Novel computer programs have been developed for the automated analysis of cell morphology, enabling phenotypic screening on many important physiological functions, such as cell migration, transformation, and neurite outgrowth [10–12]. Programs were also developed with the attempt to fully use the capacity of HCS through comprehensive phenotype profiling, encompassing the intensity and distribution of all detectable signals in the cellular image. Such techniques may enable screening with unclear biological mechanisms, and therefore greatly expanding the application of HCS [13, 14] (Fig. 1). Today, with more than a decade of development, HCS has become a mature technique that is not only used for lead discovery in industry but also for mechanism research in academia.
STEM CELLS AS AN ASSAY VEHICLE FOR HCS
In order for HCS results to be able to be faithfully transferred to downstream applications in vivo, screenings should ideally be carried out on cells from the final tissue of interest. However, under most circumstances, such cells cannot be obtained and alternatives have to be used. Several advantages make immortalized cell lines currently the most commonly used alternative: first, they are cheap and can be amplified in virtually unlimited quantities, providing a consistent and homogeneous vehicle source; second, they can be easily modified genetically to generate pure clones of reporter lines, allowing straightforward readout that can be conveniently automated. Despite these advantages, the genetic backgrounds of cell lines differ considerably from those of native cells. The marked difference often makes the screening result inapplicable in vivo . Based on this concern, some HCS applications have used primary mammalian cells, which can be derived from the tissue of interest and therefore more closely resemble the endogenous state. However, a major limitation that prevents use of primary cells as a standard for HCS is their relative scarcity—it is often difficult to obtain large amount of cells reproducibly. Moreover, while several specialized primary cultures, such as hepatocytes, umbilical endothelial cells, and keratinocytes, can be established from adult human, many primary cells need to be cultured from embryonic tissues so that they have to be obtained from non-human vertebrates. The biochemical differences between species certainly pose an obstacle for the final application of the screening results . In addition, transfection of primary cells is challenging, often preventing incorporation of reporter genes.
Limitations to immortalized cell lines and primary cells can be addressed, at least to a certain extent, using stem cells . Stem cells are nontransformed cells derived from primary tissue. Certain stem cell lines, such as embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs), can be amplified through self-renewal so that unlimited supplies may be obtained. As we will show in the later section, these cell sources have been successfully used either directly for HCS or to produce large amount of definitive stem cells for HCS. Various techniques have been developed for the highly efficient genetic modification of stem cells, with many reporter lines established that can be used for high-throughput applications . Stem cells can be generated from human tissue and, more importantly, they can be subsequently differentiated in vitro into any desired tissue type, providing a unique cell source for personalized drug discovery [18–21]. Numerous studies have confirmed that stem cell derivatives highly resemble cells from the native tissue, both phenotypically and transcriptionally [22, 23]. All these properties make stem cells ideal vehicles for high-throughput drug screening. Conversely, stem cells can potentially be used as regenerative medicine directly, raising great interest in screening for reagents that can regulate their self-renewal and differentiation. HCS is particularly well-suited for such screening, which almost inevitably requires single-cell resolution for the analysis of morphology and biomarker expression in highly heterogeneous cultures. Application of HCS on stem cells is also propelled by the lack of understanding of stem cell biology. The capability of phenotypic screening enables HCS to be carried out when a defined molecular target is unavailable, and in return the identified hits may be used as probes for dissecting the underlying mechanism. With such positive feedback, HCS provides a powerful tool for stem cell research and is expected to eventually lead to more novel medicine discovery.
PROGRESS OF HCS ON STEM CELLS
While the potential of stem cells for high-throughput drug screening, especially for toxicity assessment and personalized medicine discovery [24, 25], is well-recognized and major pharmaceutical companies are already investing in it, it may still take years for the value to be fully realized due to the lack of understanding on stem cell biology, especially on ESCs and iPSCs. Nevertheless, pioneering work on HCS with stem cells has already been reported in the literature, and its power for identifying stem cell management reagents as well as dissecting the underlying mechanism has already emerged . One of the obstacles to the clinical application of human ESCs (hESCs) is that their maintenance requires feeder cells as well as medium containing serum derivatives. Moreover, hESCs survive poorly under single-cell enzymatic dissociation, adding to the complexity of in vitro amplification and manipulation of the cells. In order to develop techniques for highly reproducible large-scale amplification of hESCs under chemically defined conditions, several HCSs have been carried out for compounds that maintain the survival and self-renewal of ESCs, especially hESCs, in feeder-free conditions [27–32]. These screens used fluorescence images generated from Oct4 expression to identify and count the pluripotent cells and evaluate the compound effect on stem cell self-renewal. Along with a number of marketed drugs and natural compounds, several kinase inhibitors were identified through these studies to control hESC self-renewal. These include rho-associated protein kinase (ROCK) inhibitors and RasGAP/ERK1 dual inhibitor. While ROCK inhibition is previously known to be important for ESC survival, dual inhibition of RasGAP and ERK1 revealed novel mechanism regulating self-renewal. More recently, the mechanism on ERK1 was confirmed and identified to be through interaction with Klf4 . These screening studies not only provided a novel ESC culture technique but also deepened our understanding of the mechanism of ESC self-renewal. Using a library of purified recombinant proteins, a similar HCS effort also identified pigment epithelium-derived factor to promote long-term pluripotent growth of hESCs without basic fibroblast growth factor or TGFbeta/Activin/Nodal ligand supplementation , providing direct evidence for the involvement of the protein in hESC self-renewal.
HCS was used to identify reagents that direct stem cells to differentiate into various lineages. One such application is the endodermal differentiation of ESCs, with particular interest in pancreatic differentiation for diabetes treatment. Using the expression of endoderm marker Sox17 as an indicator, HCSs identified compounds that can induce more than 80% of ESCs to an endodermal lineage [35, 36]. A subsequent study using the compound stauprimide as probe identified NME2 as the protein target, to regulate ESC self-renewal and endodermal differentiation through nuclear localization . This study showed for the first time a role of NME2 in ESC pluripotency, demonstrating the power of HCS in studying stem cell biology. To further induce ESC-derived endodermal cells to a pancreatic lineage, HCS was also carried out using PdX1 fluorescence images as readout. The screening identified a small molecule, (−)-indolactam V, that induces the differentiation of a substantial number of pancreatic progenitors from hESCs . Further study indicated that the biological effect is achieved at least in part through the activation of protein kinase C signaling. HCS was also used in search for reagents that induce stem cells into neuronal cells that can potentially be used as transplantation medicine for neurodegenerative diseases. In one study, a screening of a combinatorial heterocycle library of 50,000 compounds identified a new compound, neuropathiazol, able to significantly drive neural differentiation from rat neural progenitor cells, as indicated by the βIII tubulin fluorescence images . In a more basic mechanism-oriented study, a genome-wide screening of approximately 6,500 genes identified eight known and seven new genes that can specify mouse ESCs into an ectodermal neural lineage, as indicated by the number of Tα1-eGFP expressing cells . Mesodermal differentiation of ESCs was also explored with the HCS technique. Using a transgenic cell line that expresses mCherry under a MYH6 promoter, compound IWR-1 was identified that increases the number of cardiomyocytes differentiated from hESC mesodermal derivatives . Transcriptional profiling of the inhibitor-treated samples confirmed Wnt inhibition to be critical for cardiogenesis.
More recently, iPSC has become attractive for its potential to overcome the ethical concerns of ESCs as well as the immunorejection problem . However, the current technique to generate iPSCs uses a transgenic method, raising concerns regarding its future clinical utility. In order to identify small molecules that can replace the transgenes, several HCSs have been carried out, using either Oct4-GFP  expression or alkaline phosphatase (ALP) staining combined with morphology recognition  as indicators for pluripotent clones. These efforts have led to the identification of several compounds that can successfully replace some of the transgenes (e.g., Sox2), and it is expected that the effort may eventually lead to a transgene-free approach for iPSC generation.
In addition to regeneration medicine study, HCS has also attracted interest in cancer research. Increasing evidence suggests that the capability of a tumor to grow and propagate resides in a very small population of cancer stem cells (CSC) [44, 45]. In order to achieve the effective treatment of cancer, it is the CSC subpopulation that needs to be eliminated instead of the bulk tumor cells. Traditional HTS is very limited in such drug discovery as it only measures the ensemble effect. Besides, therapeutic agents may work through induction of differentiation rather than apoptosis, making them indistinguishable by viability assays. HCS is ideal for such application. By subpopulation analysis and morphological examination, a recent work screened the family of cancer testis antigens, on their effect on glioblastoma tumor stem cell differentiation, and differential effects of gene knockdown were observed . Using dye-efflux as an indicator, compounds that can specifically eliminate the stem-like cancer cells were also identified through HCS . The resolution that HCS brings to a screen at the subpopulation level makes it an excellent tool for CSC drug discovery.
Applications of HCS on stem cells have just emerged; while its power has already been demonstrated, its capability has not been fully explored. Most current studies have only applied the simplest image analysis function of HCS, that is, nuclei counting. More sophisticated functions of HCS will soon be used and expectedly will promote significant advance in stem cell research considering the past success on other cells [11, 47, 48]. It is worth noting that using cellular images as readout, HCS may also help to reduce cost and allow regular academic labs to carry out screening studies that used to be unaffordable. In a recent report, bright-field images were used for result evaluation in a screening of more than 1,000 compounds. Through image analysis, the cell confluence was automatically calculated to assess the growth rate of stem cells, leading to the discovery of compounds that can specifically inhibit glioma stem cells but not normal neural stem cells . In this approach, only minimal reagent is required and the cost of screening is greatly reduced, facilitating its application in academia. A summary of the reviewed HCSs with stem cells is listed in Table 1.
Table 1. Recent work on high-content screening with stem cells
HTS is a technology traditionally used in industry. However, there is a trend for HTS, and especially HCS, to be increasingly used in academia. Many universities and research institutes have set up HCS facilities that used to be found only in industry. Microscopes and associated image analysis software specifically designed for HCS are commercially available from several vendors (for reviews, please read [8, 50]). More and more academic labs are now well-equipped to carry out small-to-moderate scale HCS research. In such academic screenings, most labs used libraries of compounds with known substrate-protein interactions. Each compound may serve as a probe for a specific pathway, so that the screening may not only identify potential drug candidate but also represents a “chemical genetics” approach to investigate the complete repertoire of signaling pathways underlying stem cell biology [51, 52]. Moreover, by screening compounds that have already been assessed toxicologically and pharmacokinetically, the “drug repositioning” strategy also greatly reduces the cost of drug development, enables faster to market production, and quickly translates the scientific discovery to the patient bedside . To facilitate such screening, a number of compound libraries are now available either commercially or through nonprofit academic institutes. These libraries consist of hundreds to thousands of compounds that can be readily screened with academic setting. A list of such libraries that have been previously used in stem cell research is shown in Table 2. The potential of HCS for stem cell mechanism research has already been demonstrated by the works reviewed above, with growing access HCS is expected to be used by more and more labs for basic biomedical research on stem cells.
Table 2. Commercial and academic compound libraries commonly used for stem cell screening
FUTURE DEVELOPMENT OF HCS ON STEM CELLS
Previous work on stem cell HCS has only used its minimal function and the full capacity is yet to be explored. It is expected that HCS techniques previously developed for other applications will gradually be adopted for stem cell research. Meanwhile, new techniques including novel software, reagents, and devices are being developed for stem cell HCS.
Critical to HCS is the image processing programs that automatically manage and interpret the complex cellular information. In the past, advanced software tools have been developed to address the most common challenges. A typical challenge for stem cell research is heterogeneity. Stem cell cultures consist of various cell subtypes, including undifferentiated cells and various derivatives. Even within the same subtype, increasing evidence has shown that an individual cell may respond differently to the same perturbation and the cellular heterogeneity serves an important biological function, possibly representing distinct stereotyped signaling states [54–56]. Tools have already been developed for single-cell analysis and the multivariate profiles generated from immunocytochemistry images can be used to systematically characterize the compound effect [13, 14]. It is expected that these tools will be increasingly used for stem cell HCS and lead to more accurate evaluation of compound effects as well as provide deeper insight into the signaling states of stem cells. Also, while we have discussed the many advantages of HCS, it has to be noted that HCS is typically slower than traditional HTS by 2 orders of magnitude so that its screening capacity is limited. As the rate-limiting step of HCS is image processing, it is anticipated that novel computing methods will be kept developed to significantly increase the throughput of HCS.
Novel probing techniques are being developed to enable the fast detection and quantification of specific cellular states. For example, a recent work has shown that small molecules can be used to distinguish distinct states resulting from cellular differentiation from myoblasts to myotubes . Such small molecules represent a significant advance over the traditionally used immunocytochemistry method by offering faster, cheaper, and easier to use reagents to distinguish the differentiation process. Novel microscopy techniques are also emerging for the detection of specific biochemical molecules, providing possible opportunities for cell lineage differentiation. In a recent work, screening was carried out for fat regulating genes using stimulated-Raman scattering microscopy, which can specifically detect the CH2 stretching vibration of fatty acid chains . Such a label-free technique is specific, quantitative, and avoids the complications associated with immunostaining. It can also be applied to live cells or even animals, allowing time-lapse studies, so that the effects can be studied more thoroughly and accurately.
Most current HCS use monolayer cultured stem cells for their convenience and low cost. However, three-dimensional (3D) cell culture technique is receiving increased interest with evidence showing that the cellular phenotype and biological response of 3D cultured cells are closer to cells in vivo . It is probably not surprising as the 3D methods facilitate greater cell-cell and cell-matrix interactions allowing cells to adapt to their native morphology as well as physiological function. Various techniques have been developed for 3D cell culture, including the use of extracellular matrix protein gels, agitation-based bioreactors, porous polymeric materials, and nanoparticle materials [60–62]. 3D culture is not new to stem cells. In fact, several types of stem cells such as neural stem cells and mammary stem cells were originally cultured in 3D in spheroid form [63, 64]. Using 3D culture technique, a recent report even showed that organoid culture can be established from pure stem cells that faithfully resemble the living tissue . Apparently, this will make stem cell more valuable for biomedical research and drug discovery. However, although pioneer high-throughput studies have already been carried out , a number of challenges still remain for the 3D culture to be used in HCS. Culture cells in 3D not only increase the cost but also complicate the result, leading to large standard deviation and poor screening performance. Perhaps more importantly, 3D culture poses a significant challenge for imaging as well as image processing. Since cells are not grown in a single focal plane, confocal sections followed by 3D reconstruction will be needed to evaluate the cell status. With these challenges addressed by further improvement of the hardware and software systems, it is anticipated that HCS on 3D cultured cells will lead to more faithful mechanism and drug discovery.
Another recent innovation is gene screening using RNAi libraries. By facilitating the systematic study of the role of individual genes in any biological process, RNAi screening opens new prospects in biological research [10, 67]. Customized RNAi libraries, either in the form of virus or synthetic oligos, can now be obtained commercially. In addition, siRNA microarrays have been printed on chips and used for genome-wide RNAi screening [47, 68]. These techniques, together with other microfluidic devices that are being quickly developed, will greatly increase the throughput and reduce the cost of screening [69, 70]. With the capability of HCS in combination with RNAi screening to query the role of genes in individual cells at a subceullular resolution, RNAi screening will emerge as a powerful tool for understanding the molecular mechanism that controls the self-renewal and pluripotency of stem cells [71–75]. In return, the insightful knowledge of stem cell biology will eventually translate into novel medicine and clinical applications.
This research was supported by NIH R01AG028928, and Ting Tsung and Wei Fong Chao Center for Bioinformatics Research and Imaging in Neurosciences (BRAIN) to S.T.W.
Disclosure Of Potential Conflicts Of Interest
The authors indicate no potential conflicts of interest.