Deep learning guided image-based droplet sorting for on-demand selection and analysis of single cells and 3D cell cultures.

Uncovering the heterogeneity of cellular populations and multicellular constructs is a long-standing goal in fields ranging from antimicrobial resistance to cancer research. Emerging technology platforms such as droplet microfluidics hold the promise to decipher such heterogeneities at ultra-high-throughput. However, there is a lack of methods able to rapidly identify and isolate single cells or 3D cell cultures. Here we demonstrate that deep neural networks can accurately classify single droplet images in real-time based on the presence and number of micro-objects including single mammalian cells and multicellular spheroids. This approach also enables the identification of specific objects within mixtures of objects of different types and sizes. The training sets for the neural networks consisted of a few hundred images manually picked and augmented to up to thousands of images per training class. Training required less than 10 minutes using a single GPU, and yielded accuracies of over 90% for single mammalian cell identification. Crucially, the same model could be used to classify different types of objects such as polystyrene spheres, polyacrylamide beads and MCF-7 cells. We applied the developed method for the selection of 3D cell cultures generated with Hek293FT cells encapsulated in agarose gel beads, highlighting the potential of the technology for the selection of objects with a high diversity of visual appearances. The real-time sorting of single droplets was in-line with droplet generation and occurred at rates up to 40 per second independently of image size up to 480 × 480 pixels. The presented microfluidic device also enabled storage of sorted droplets to allow for downstream analyses.

High-Throughput, Lysis-Free Screening for Sulfatase Activity Using Escherichia coli Autodisplay in Microdroplets.

Directed evolution of enzymes toward improved catalytic performance has become a powerful tool in protein engineering. To be effective, a directed evolution campaign requires the use of high-throughput screening. In this study we describe the development of an ultra high-throughput lysis-free procedure to screen for improved sulfatase activity by combining microdroplet-based single-variant activity sorting with E. coli autodisplay. For the first step in a 4-step screening procedure, we quantitatively screened >105 variants of the homodimeric arylsulfatase from Silicibacter pomeroyi (SpAS1), displayed on the E. coli cell surface, for improved sulfatase activity using fluorescence activated droplet sorting. Compartmentalization of the fluorescent reaction product with living E. coli cells autodisplaying the sulfatase variants ensured the continuous linkage of genotype and phenotype during droplet sorting and allowed for direct recovery by simple regrowth of the sorted cells. The use of autodisplay on living cells simplified and reduced the degree of liquid handling during all steps in the screening procedure to the single event of simply mixing substrate and cells. The percentage of apparent improved variants was enriched >10-fold as a result of droplet sorting. We ultimately identified 25 SpAS1 variants with improved performance toward 4-nitrophenyl sulfate (up to 6.2-fold) and/or fluorescein disulfate (up to 30-fold). In SpAS1 variants with improved performance toward the bulky fluorescein disulfate, many of the beneficial mutations occur in residues that form hydrogen bonds between α-helices in the C-terminal oligomerization region, suggesting a previously unknown role for the dimer interface in shaping the substrate binding site of SpAS1.

Ultrahigh-Throughput Screening of Single-Cell Lysates for Directed Evolution and Functional Metagenomics.

The success of ultrahigh-throughput screening experiments in directed evolution or functional metagenomics strongly depends on the availability of efficient technologies for the quantitative testing of a large number of variants. With advanced robotics, libraries of up to 105 clones can be screened per day as colonies on agar plates or cell lysates in microwell plates, albeit at high cost of capital, manpower and consumables. These cost considerations and the general need for high-throughput make miniaturization of assay volumes attractive. To provide a general solution to maintain genotype-phenotype linkage, biochemical assays have been compartmentalized into water-in-oil droplets. This chapter presents a microfluidic workflow that translates a frequently used screening procedure consisting of cytoplasmic/periplasmic protein expression and cell lysis to the single cell level in water-in-oil droplet compartments. These droplets are sorted based on reaction progress by fluorescence measurements at the picoliter scale.

Exploring sequence space in search of functional enzymes using microfluidic droplets.

Screening of enzyme mutants in monodisperse picoliter compartments, generated at kilohertz speed in microfluidic devices, is coming of age. After a decade of proof-of-principle experiments, workflows have emerged that combine existing microfluidic modules to assay reaction progress quantitatively and yield improved enzymes. Recent examples of the screening of libraries of randomised proteins and from metagenomic sources suggest that this approach is not only faster and cheaper, but solves problems beyond the feasibility scope of current methodologies. The establishment of new assays in this format - so far covering hydrolases, aldolases, polymerases and dehydrogenases - will enable the exploration of sequence space for new catalysts of natural and non-natural chemical transformations.

Development of an open technology sensor suite for assisted living: a student-led research project.

Many countries have a rapidly ageing population, placing strain on health services and creating a growing market for assistive technology for older people. We have, through a student-led, 12-week project for 10 students from a variety of science and engineering backgrounds, developed an integrated sensor system to enable older people, or those at risk, to live independently in their own homes for longer, while providing reassurance for their family and carers. We provide details on the design procedure and performance of our sensor system and the management and execution of a short-term, student-led research project. Detailed information on the design and use of our devices, including a door sensor, power monitor, fall detector, general in-house sensor unit and easy-to-use location-aware communications device, is given, with our open designs being contrasted with closed proprietary systems. A case study is presented for the use of our devices in a real-world context, along with a comparison with commercially available systems. We discuss how the system could lead to improvements in the quality of life of older users and increase the effectiveness of their associated care network. We reflect on how recent developments in open source technology and rapid prototyping increase the scope and potential for the development of powerful sensor systems and, finally, conclude with a student perspective on this team effort and highlight learning outcomes, arguing that open technologies will revolutionize the way in which technology will be deployed in academic research in the future.

Programming Light-Harvesting Efficiency Using DNA Origami.

The remarkable performance and quantum efficiency of biological light-harvesting complexes has prompted a multidisciplinary interest in engineering biologically inspired antenna systems as a possible route to novel solar cell technologies. Key to the effectiveness of biological "nanomachines" in light capture and energy transport is their highly ordered nanoscale architecture of photoactive molecules. Recently, DNA origami has emerged as a powerful tool for organizing multiple chromophores with base-pair accuracy and full geometric freedom. Here, we present a programmable antenna array on a DNA origami platform that enables the implementation of rationally designed antenna structures. We systematically analyze the light-harvesting efficiency with respect to number of donors and interdye distances of a ring-like antenna using ensemble and single-molecule fluorescence spectroscopy and detailed Förster modeling. This comprehensive study demonstrates exquisite and reliable structural control over multichromophoric geometries and points to DNA origami as highly versatile platform for testing design concepts in artificial light-harvesting networks.