The Discovery Channel: microfluidics and microengineered systems in drug screening

Abstract

Approaches to drug development have changed significantly over the last decade in response to the continued decline in productivity of the current discovery model. Here, we highlight exciting findings and promising approaches in the recent literature in which researchers integrate advanced micro-engineering, design, and analytical strategies to improve the relevance and utility of high-throughput screening in the drug discovery pipeline.

An often cited reason for the application of micro-engineered technologies towards studies of biology is the ability to increase experimental throughput via the miniaturization of biological assays. This high-throughput screening (HTS) approach was popularized by its initially successful implementation in the drug discovery industry, and has now been prevalent for decades. Today, the integration between 384-, 1536-, and 3456-microtiter well-plates with robotic technologies enables screening of over 100 000 compounds per day. While the early successes of this heavily automated approach to biology is undeniable, the rate of discovery of novel drugs today is rapidly falling, and the HTS paradigm is often criticized for the expenses associated with investigating the number of ‘false positives’ generated during a screen.¹ Since it seems that the ‘low-hanging fruit’ of high-throughput discovery have been picked, researchers are now taking a strongly integrative and creative approach to improving the capabilities of these high-throughput schemes, and utilizing HTS techniques for novel applications. Here, I highlight a small selection of recent strategies that improve and apply HTS approaches, particularly in the areas of designing realistic microenvironments, improved readouts of functional cell activity, and a transition from reading out non-specific drug effects to phenotypic analysis.

It is now well established that the cellular microenvironment plays a pivotal role in driving cell function. Hence, it is perhaps not surprising that cells cultured on hard, flat, tissue-culture plastic well-plates may not adequately recreate in vivo functionality in response to therapies, leading to benchtop discoveries that fail to translate to the bedside. Engineering these microenvironments to be both high-throughput and realistic is the focus of a number of research programs. For example, Takayama and co-workers recently demonstrated the ability to print cancer cell-laden 3D microgels into well-plates, using existing HTS infrastructure. This approach is only slightly more expensive than 2D screens, but better represents the in vivo situation, circumvents transport limitations associated with typical 3D cultures, and is shown to dramatically alter cell response to chemotherapies.² Likewise, Janmey and co-workers developed a microenvironmental screen to study the combined effects of fluid shear stress and substrate stiffness on the inflammatory response of endothelial cells, and demonstrate that substrate stiffness plays a critical role in endothelial function.³

However, replicating every biophysical nuance of the natural environment from the ground up is likely impossible, and in some cases, failure to do so might prevent even the most basic culture of biological material. This is a particular challenge in the growth of bacteria: 99% of naturally occurring bacterial species do not grow in standard nutrient culture plates. Since bacteria are a promising source of antibiotics, and antibiotic resistance is spreading faster than the discovery of new anti-microbials, developing the right culture techniques to leverage this potential source is of critical importance. To circumvent the issue of designing bacteria-specific culture systems, the Epstein lab developed an ‘isolation Chip’ or iChip in 2010,⁴ an array of miniaturized diffusion chambers in a thin polymer plate, compartmentalized from the surrounding environment by two porous membranes (Fig. 1). Single bacterial cells from samples of wet soil or water are loaded into each diffusion chamber by the limited dilution method. The device is then assembled and placed back in situ where the porous membranes allow transfer of natural environmental factors into and out of the device, while keeping the cells within a single confined location. The researchers found that nearly 50% of the isolated bacteria formed colonies, as opposed to 1% on nutrient plates, illustrating how important the culture environment can be for growth. In a recent report in Nature, Ling et al., used this high-throughput methodology to identify a novel anti-bacterial compound, by growing approximately 10 000 previously unculturable colonies from bacterial cells found in a ground soil sample.⁵ The 10 000 isolates were then screened for anti-bacterial activity against Staphylococcus aureus, and an extract from a species of β-proteobacteria showed unusually strong activity. A compound, designated Teixobactin, was isolated and analyzed from this colony, and was found to have potent activity against Mycobacterium tuberculosis, Clostridium difficule, and Bacillus anthracis. Though ineffective against Gram-negative bacteria, Teixobactin was then shown to be non-toxic to mammalian cells in culture, and moreover, was shown to maintain bactericidal activity against antibiotic-resistant strains of infection. Furthermore, unlike in culture with conventional antibiotics, such as oxflacin, S. aureus did not acquire any resistance to Teixobactin over serial passages, suggesting an unusual method of activity. This was subsequently found to be an inhibition of cell wall synthesis by binding to lipid motifs that are highly conserved across species, a mechanism that the researchers theorize will require ∼30 years before immunity can be evolved. This compound was then tested in a mouse septicemia model against methicillin-resistant S. aureus (MRSA), the most common ‘superbug’ responsible for infections picked up in hospitals. The experiments showed excellent activity at doses lower than equivalent treatments using vancomycin, the current standard treatment for MRSA infections (Fig. 2).

Fig. 1 (A) Schematic of the iChip (isolation Chip) design and operation. (B) Each plate consists of machined microcompartments, which are loaded with a single bacterium (on average, based on the limited dilution of a sample), and (C) sandwiched between porous membranes in a clamp system with aligned wells. The device is then placed back in the original environment from which the sample originated. The original ‘uncultured’ environment permits a substantially greater fraction of the bacterial isolates to form colonies. The figure is reused with permission from Nichols et al.⁴

Fig. 2 (a, b) Activity of Teixobactin at minimal inhibitory concentrations (MIC) is comparable to other antibiotics in killing pathogens at the (a) early and (b) late stages of growth. (c) Use of Teixobactin results in lysis of the bacterial cells, and (d) no resistance is acquired during serial passaging in the presence of sub-MIC levels of Teixobactin, as compared to treatment with the antibiotic Ofloxacin. The figure is reused with permission from Ling et al.⁵

These results are particularly exciting in that they demonstrate both a new candidate anti-bacterial agent, and a new mechanism of activity, at a time at which both discoveries are critical to address the challenges posed by antibiotic-resistant infections. This study is also particularly striking to the Integrative Biology community: using simple technologies, it is a clear demonstration that the combination of high-throughput screening and design of appropriate culture environments can play an effective role in the discovery and design of next-generation therapies.

While this first example demonstrates the Goldilocks principle of designing the environment to be “Just right”, a second key technological approach to the development of effective HTS is the ability to read-out useful and relevant assay results. In high-throughput applications, the quality of biological information obtained is typically inversely proportional to the time, cost and difficulty of the analysis. For example, a simple study of viability as measured by fluorescence imaging can be performed relatively rapidly on a robotic imaging platform, and analyzed using automated software. More subtle analyses of biological activity, such as cell morphology or structure of expressed proteins, may require careful fluorescence staining protocols and more advanced or manual image analysis techniques. Robust biological assays for gene and protein expression are best performed with a cherry-picking automated flow cytometer system, which greatly adds to the time, cost and complexity of the assay. While these are each feasible on the small scale, expanding these assays to the current rates of 100 000 experiments per day is prohibitive.

Novel engineering strategies could be employed to improve our ability to read-out biological function from a high-throughput experiment. In a recent example in Nature Medicine, Mei et al. apply microfabrication techniques to improve the quality, analyzability and relevance of a fluorescence readout, for the specific application of screening drugs to address multiple sclerosis (MS).⁶ Multiple sclerosis is caused by the loss of the myelin sheath surrounding neural axons, limiting the conductivity of action potentials in the neuron. To date, there are no therapies for oligodendrocyte remyelination. Here, the researchers recognized that identifying potential drugs for MS requires time-consuming, high-resolution confocal imaging and analysis to identify culture conditions in which ‘rings’ of fluorescently-labelled myelin membrane can be induced to form around axon-like structures, such as nanofibers. Imaging and analysis at this size scale is not amenable to automation, as they are various confounding issues arising from the varying lengths of myelin sheaths, and a convoluted spatial structure of oligodendrocyte processes and cell bodies. Rather than try and improve the imaging and analysis procedures, they designed the culture system itself to clearly display biological activity relevant to MS. Recognizing that myelin-membrane wrapping can be observed on pseudoaxonal substrates, they microengineered an array of cone-shaped structures into an HTS-compatible 96-well plate (Fig. 3). By observing the de novo formation of concentric myelin membrane strands (stained positive for myelin bound protein, red) wrapped around the conical micropillars in a co-culture of oligodendrocytes and oligodendrocyte precursor cells (OPCs; labelled green), they assessed the ability of those molecules to induce remyelination activity. These micropillars can be readily imaged using a compressed z-stack without concerns regarding spatial overlap of the structures. The data can be easily analyzed using a count of the number of myelinated pillars as a proxy for myelination activity. Following the validation of this analysis, the researchers used HTS techniques to screen 1000 bioactive molecules. They identified a new cluster of compounds that enhance differentiation and remyelination, which are now being studied further.

Fig. 3 (top panel) A 96-well culture plate is designed with conical micropillars, around which oligodendrocytes can wrap their processes. (middle panel) High-throughput screening would reveal an increase in the number of oligodendrocytes around the micropillars by staining for myelin-basic protein (red), which wraps around the micropillar in concentric rings. (bottom panel) SEM image of the oligodendrocyte/OPC co-culture, in which processes wrap around the conical projections. The figure is reused with permission from Mei et al.⁶

This second example demonstrates principles of highly specific design: the cell culture platform is uniquely developed to study remyelination around axon-like topographies. The general principle underlying this approach is the concept of utilizing phenotypic screens in which evidence of favourable functional activity is first demonstrated, without prior knowledge of biochemical activity or molecular mechanism of action. In contrast, target-based approaches identify mechanisms of interest first, and then use HTS to develop candidate modifiers of those targets. A recent meta study comparing the use of these approaches in drugs approved between 1999 and 2008⁷ suggests that a smaller, but still sizable, fraction of viable drugs was identified through target-based discovery. While target-based discovery is more likely to fail to translate to clinical applications, they still play a role in informing the HTS process.⁸ In fact, phenotypic assays are rarely performed in complete isolation from any knowledge of the target mechanism. For example, Mei et al. specifically tested antimuscarinic compounds in the design of their MS screening system, based on prior knowledge of mechanism. This integrated approach between functional phenotypic and target-based discovery is termed ‘mechanism-informed phenotypic drug discovery’,⁸ and may provide a substantially more efficient, rational approach towards designing HTS experiments.

The key to integrating these two approaches while minimizing the discovery of molecules that ultimately fail to translate to the clinic, may hence be in identifying drug targets in physiologically-realistic environments. In a recent Integrative Biology paper, Peyton and co-workers develop a high-content, systems biology-based method to screen cell adhesion, motility, and growth factor responses on biomaterial surfaces that are specifically designed to simulate the extracellular matrix in tissues of interest.⁹ Focusing on the three primary sites of breast cancer metastasis, they created a simple biomaterial platform in which the extracellular matrix compositions of the brain, bone and lung are recapitulated, and breast cancer cell lines specific to these sites were assayed for a panel of functional markers. These observations were collated to create a phenotypic ‘fingerprint’, that was shown to predict in vivo metastasis: while a single measurement was not enough to predict behaviour, the combination of markers provided sufficient predictive capability. These functional findings led the researchers to identify a specific group of integrins associated with each of the metastasis sites, which interestingly, did not correlate well with previously-collected gene expression data from secondary tumor sites. This mismatch may explain why integrin-based therapies to date have not been successful in treating this disease, and directs future studies in druggable targets against tissue metastasis.

These studies represent three distinct approaches in applying novel technologies towards high-throughput screening practices for drug discovery. It is likely that some combination of each will be necessary to create an improved drug discovery pipeline, and the successes of these approaches to date suggest an important role for integrative and creative thinkers in this field.

References

1. J. W. Scannell, A. Blanckley, H. Boldon and B. Warrington, Nat. Rev. Drug Discovery, 2012, 11, 191–200.
2. B. M. Leung, C. Moraes, S. P. Cavnar, K. E. Luker, G. D. Luker and S. Takayama, J. Lab. Autom., 2014 DOI:10.1177/2211068214563793.
3. P. A. Galie, A. van Oosten, C. S. Chen and P. A. Janmey, Lab Chip, 2015, 15, 1205.
4. D. Nichols, N. Cahoon, E. M. Trakhtenberg, L. Pham, A. Mehta, A. Belanger, T. Kanigan, K. Lewis and S. S. Epstein, Appl. Environ. Microbiol., 2010, 76, 2445–2450.
5. L. L. Ling, T. Schneider, A. J. Peoples, A. L. Spoering, I. Engels, B. P. Conlon, A. Mueller, T. F. Schäberle, D. E. Hughes, S. Epstein, M. Jones, L. Lazarides, V. A. Steadman, D. R. Cohen, C. R. Felix, K. A. Fetterman, W. P. Millett, A. G. Nitti, A. M. Zullo, C. Chen and K. Lewis, Nature, 2015, 517, 455.
6. F. Mei, S. P. J. Fancy, Y.-A. A. Shen, J. Niu, C. Zhao, B. Presley, E. Miao, S. Lee, S. R. Mayoral, S. A. Redmond, A. Etxeberria, L. Xiao, R. J. M. Franklin, A. Green, S. L. Hauser and J. R. Chan, Nat. Med., 2014, 20, 954–960.
7. D. C. Swinney, Clin. Pharmacol. Ther., 2013, 93, 299–301.
8. J. G. Moffat, J. Rudolph and D. Bailey, Nat. Rev. Drug Discovery, 2014, 13, 588–602.
9. L. E. Barney, E. C. Dandley, L. E. Jansen, N. G. Reich, A. M. Mercurio and S. R. Peyton, Integr. Biol., 2015, 7, 198.

---