BMC Methods

SignificanceStandard methods to characterize peripheral blood mononuclear cells (PBMCs) are often destructive, lack metabolic information, or do not provide single-cell resolution. Label-free tools that non-destructively measure single-cell metabolism within PBMCs can provide new layers of information to characterize disease state and cell therapy potential. AimDetermine whether non-destructive fluorescence lifetime imaging microscopy (FLIM) of endogenous metabolic co-factors NAD(P)H and FAD, or optical metabolic imaging (OMI), can identify immune cell subsets and activation state within heterogeneous PBMC cultures. ApproachOMI measured single-cell metabolism of PBMCs from 3 different human donors in the quiescent or activated (phorbol 12-myristate 13-acetate and ionomycin) state. Fluorescent antibodies were used as ground truth labels for single-cell classifiers of immune cell subtypes. ResultsOMI identified quiescent vs. activated PBMCs with 93% accuracy at only 2 hours post-stimulation, identified monocytes within quiescent and activated PBMCs with 96% and 88% accuracy, respectively, and identified NK cells within quiescent and activated PBMCs with 74% accuracy. ConclusionOMI identifies activation state and immune cell subpopulations within PBMCs, enabling single-cell and label-free measurements of metabolic heterogeneity within complex PBMC samples. Therefore, OMI could enhance PBMC immunophenotyping for diagnostic and therapeutic applications. Statement of DiscoveryWe demonstrate that autofluorescence lifetime imaging can resolve functional and phenotypic metabolic subpopulations within a mixed culture of immune cells from human blood. This provides a new technique to characterize metabolic activity within immune cells from the peripheral blood of patients, which could improve disease diagnostics and the production of cell therapies.

Manual microscopy is the gold standard for phytoplankton monitoring in diverse engineered and natural environments. However, it is both labor-intensive and requires specialized training for accuracy and consistency, and therefore difficult to implement on a routine basis without significant time investment. Automation can reduce this burden by simplifying the measurement to a single indicator (e.g., chlorophyll fluorescence) measurable by a probe, or by processing samples on an automated cytometer for more granular information. The cost of commercially available flow imaging cytometers, however, poses a steep financial barrier to adoption. To overcome these labor and cost barriers, we developed ARTiMiS: the Autonomous Real-Time Microbial Scope. The ARTiMiS is a low-cost flow imaging microscopy-based platform with onboard software capable of providing species-level quantitation of phytoplankton communities in real-time. ARTiMiS leverages novel multi-modal imaging and onboard machine learning-based data processing that is currently optimized for a curated and expandable database of industrially relevant microalgae. We demonstrate its operational limits, performance in identification of laboratory-cultivated microalgae, and potential for continuous monitoring of complex microalgal communities in full-scale cultivation systems. SynopsisWe introduce a platform for low-cost real-time imaging monitoring of phytoplankton and demonstrate its utility in real-time monitoring of laboratory- and full-scale microalgal cultivation systems.

Escherichia coli is a commonly used host for recombinant protein production. It is advantageous to direct many recombinant proteins, especially those that require disulphide bonding for function, such as antibody fragments, to the periplasm of E. coli. This requires N-terminal fusion of a signal peptide that directs the polypeptide chain through the relevant translocation apparatus. Signal peptides cannot be selected on the basis of recombinant gene sequence, so screening is required to select the optimal signal peptide for each product, typically using subcellular fractionation, a time-intensive process. Fusion of a fluorescent protein such as GFP to the C-terminal of recombinant proteins has previously been used to accelerate cytoplasmic protein production process development, but most GFP proteins are not active in the periplasm. Previous studies have developed GFP derivatives that fold rapidly (such as superfolder GFP, sfGFP) and have been reported to be periplasmically active. Here, we tested the applicability of sfGFP as a periplasmic screening tool using single-cell analysis and structured illumination microscopy. We discovered that sfGFP is very poorly tolerated in the periplasm, causing deleterious effects on E. coli physiology, manifesting as poor growth, cell death, and loss of recombinant protein productivity. A further reason for poor GFP functionality in the periplasm is errant disulphide bonding, so we tested a cysteine-free GFP, which cannot form disulphide bonds; results were similar to sfGFP. In conclusion, currently-available GFP variants are poor fusion partners for screening production and translocation of recombinant proteins to the E. coli periplasm due to their negative impact on physiology. HighlightsO_LIGFP is a useful screening tool recombinant protein production. C_LIO_LIWe tested periplasmic expression of GFP derivatives sfGFP and cfSGFP2. C_LIO_LIWe used structured illumination microscopy to visualise GFP accumulation. C_LIO_LIPeriplasmic GFP derivatives have significant negative effects on bacterial physiology C_LI

Machine learning algorithms hold the promise of greatly improving live cell image analysis by way of (1) analyzing far more imagery than can be achieved by more traditional manual approaches and (2) by eliminating the subjective nature of researchers and diagnosticians selecting the cells or cell features to be included in the analyzed data set. Currently, however, even the most sophisticated model based or machine learning algorithms require user supervision, meaning the subjectivity problem is not removed but rather incorporated into the algorithms initial training steps and then repeatedly applied to the imagery. To address this roadblock, we have developed a self-supervised machine learning algorithm that recursively trains itself directly from the live cell imagery data, thus providing objective segmentation and quantification. The approach incorporates an optical flow algorithm component to self-label cell and background pixels for training, followed by the extraction of additional feature vectors for the automated generation of a cell/background classification model. Because it is self-trained, the software has no user-adjustable parameters and does not require curated training imagery. The algorithm was applied to automatically segment cells from their background for a variety of cell types and five commonly used imaging modalities - fluorescence, phase contrast, differential interference contrast (DIC), transmitted light and interference reflection microscopy (IRM). The approach is broadly applicable in that it enables completely automated cell segmentation for long-term live cell phenotyping applications, regardless of the input imagerys optical modality, magnification or cell type.

Deep learning provides an opportunity to automatically segment and extract cellular features from high-throughput microscopy images. Many labeling strategies have been developed for this purpose, ranging from the use of fluorescent markers to label-free approaches. However, differences in the channels available to each respective training dataset make it difficult to directly compare the effectiveness of these strategies across studies. Here we explore training models using subimage stacks composed of channels sampled from larger, hyper-labeled, image stacks. This allows us to directly compare a variety of labeling strategies and training approaches on identical cells. This approach revealed that fluorescence-based strategies generally provide higher segmentation accuracies but were less accurate than label-free models when labeling was inconsistent. The relative strengths of label and label-free techniques could be combined through the use of merging fluorescence channels and out-of-focus brightfield images. Beyond comparing labeling strategies, using subimage stacks for training was also found to provide a method of simulating a wide range of labeling conditions, increasing the ability of the final model to accommodate a greater range of experimental setups.

MotivationMeasuring fluorescence by flow cytometry is fundamental for characterising single-cell performance. While it is known that fluorescence and scattering values tend to positively correlate, the impact of cell volume on fluorescence is typically overlooked. This makes of fluorescence values alone an inaccurate measurement for high-precision characterisations. ResultsWe developed FlowScatt, an open-source software package that removes volume-dependency in the fluorescence channel. Using FlowScatt, flourescence values are re-calculated based on the unified volume per cell that arises from scattering decomposition. AvailabilityFlowScatt is openly available as a Python package on https://github.com/rstoof/FlowScatt. Experimental data for validation is available online. Contactangel.goni-moreno@newcastle.ac.uk

Extracellular electrical potentials have been observed in a number of filamentous mycelial species with incommensurable and non-stereotypic features. In Basidiomycetes, detecting these signals reliably is dependent on the properties of the cell wall and plasma membrane and requires implementation of microelectrode array hardware, filtering and spike sorting methods. In this paper, we present recording methods for detection of discrete unit extracellular spikes in biofilm forming liquid cultures of Hericium erinaceus. We utilised custom designed microelectrode arrays (MEAs) with passive planar hard gold microelectrodes and individual radius of 100 {micro}m in recordings at a sample rate of 30 kHz. Triplicate recordings of mycelial samples in a double shielded electromagnetic and RF shield box were conducted for wild-type, ionophore assays and fungicidal assays. The recordings were analysed offline using the Kilosort4 sorting algorithm resulting in detection of discrete unit spikes with milliseconds durations. The clustered spike waveforms for the wild-type triplicates were estimated to have a mean trough-to-peak-time of 2.68 {+/-} 0.087 ms and width at half maximum of 0.8 {+/-} 0.031 ms across a combined total of 418 spiking units. Ionophore assays using nystatin solution (10,000 units/ml) exhibited significant statistical differences including a reduction of total units to 97. A decrease in the trough-to-peak time of the mean waveform (1.97 {+/-} 0.32 ms) and an increase in the width at half maximum (2.7 {+/-} 2.45 ms) were also observed. Nystatin was found reduce the mean extracellular spike amplitude from 173.06 {micro}V in the wild-type to 25.76 {micro}V in the assay. Physiological disruption of the cell wall and plasma membrane was confirmed by environmental scanning microscopy comparison of triplicates at 90 % humidity. In comparison, a fungicidal assay utilising 12% w/v H2O2 solution resulted in zero spiking units and absence of discrete unit activity across all channels in triplicate recordings.

Fluorescence miniature microscopy in vivo has recently proven a major advance, enabling cellular imaging in freely behaving animals. However, fluorescence imaging suffers from autofluorescence, phototoxicity, photobleaching and non-homogeneous illumination artifacts. These factors limit the quality and time course of data collection. Bioluminescence provides an alternative kind of activity-dependent light indicator. Bioluminescent calcium indicators do not require light input, instead generating photons through chemiluminescence. As such, limitations inherent to the requirement for light presentation are eliminated. Further, bioluminescent indicators also do not require excitation light optics: the removal of this component should make lighter and lower cost microscope with fewer assembly parts. While there has been significant recent progress in making brighter and faster bioluminescence indicators, parallel advances in imaging hardware have not yet been realized. A hardware challenge is that despite potentially higher signal-to-noise of bioluminescence, the signal strength is lower than that of fluorescence. An open question we address in this report is whether fluorescent miniature microscopes can be rendered sensitive enough to detect bioluminescence. We demonstrate this possibility in vitro and in vivo by implementing optimizations of the UCLA fluorescent miniscope. These optimizations yielded a miniscope (BLmini) which is 22% lighter in weight, has 45% fewer components, is up to 58% less expensive, offers up to 15 times stronger signal (as dichroic filtering is not required) and is sensitive enough to capture spatiotemporal dynamics of bioluminescence in the brain with a signal-to-noise ratio of 34 dB.

Background53BP1 foci are reflective of DNA double-strand break formation and have been used as radiation markers. Manual focus counting, while prone to bias and time constraints, remains the most accurate mode of detecting 53BP1 foci. Several studies have pursued automated focus detection to replace manual methods. Deep learning, spatial 3D images, and segmentation techniques are main components of the highest performing automated methods. While these approaches have achieved promising results regarding accurate focus detection and cell classification, they are not compatible with time-sensitive large-scale applications due to their demand for long run times, advanced microscopy, and computational resources. Further, segmentation of overlapping foci in 2D images has the potential to represent focus morphologies inaccurately. ResultsTo overcome these limitations, we developed a novel method to classify 2D fluorescence microscopy images of 53BP1 foci. Our approach consisted of three key features: (1) general 53BP1 focus classes, (2) varied parameter space composed of properties from individual foci and their respective Fourier transform, and (3) widely-available machine learning classifiers. We identified four main focus classes, which consisted of blurred foci and three levels of overlapping foci. Our parameter space for the training focus library, composed of foci formed by fluorescently-tagged BP1-2, showed a wide correlation range between variables which was validated using a publicly-available library of immunostained 53BP1 foci. Random forest achieved one of the highest and most stable performances for binary and multiclass problems, followed by a support vector machine and k-nearest neighbors. Specific metrics impacted the classification of blurred and low overlap foci for both train and test sets. ConclusionsOur method classified 53BP1 foci across separate fluorescent markers, resolutions, and damage-inducing methods, using off-the-shelf machine learning classifiers, a diverse parameter space, and well-defined focus classes.

Saccharomyces cerevisiae is a powerful system for the expression of genome-wide or combinatorial libraries for diverse types of screening. However, expressing large libraries in yeast requires high-efficiency transformation and controlled expression. Transformation of yeast using electroporation methods is more efficient than chemical methods; however, protocols described for electroporation require large amounts of linearized plasmid DNA and often yield about 106 cfu/{micro}g of plasmid DNA. We optimized the electroporation of yeast cells for the expression of whole-genome libraries to yield up to 108 cfu/{micro}g plasmid DNA. The protocol generates sufficient transformants for 10-100x coverage of diverse genome libraries with small amounts of genomic libraries (0.1{micro}g of DNA per reaction) and provides guidance on calculations to estimate library size coverage and transformation efficiency. It describes the preparation of electrocompetent yeast cells with lithium acetate and dithiothreitol conditioning step and the transformation of cells by electroporation with carrier DNA. We validated the protocol using three yeast surface display libraries and demonstrated using nanopore sequencing that libraries size and diversity are preserved. Moreover, expression analysis confirmed library functionality and the methods efficacy. Hence, this protocol yields a sufficient representation of the genome of interest for downstream screening purposes while limiting the amount of the genomic library required.

Determining the relationship between biomarkers via fluorescence microscopy is a key step in the characterisation of cellular phenotypes. We define a simple distance-based measurement termed a perimeter distance mean (PDmean) which quantifies the relative proximity of objects in one fluorescent channel to objects in a second fluorescent channel in 2D or 3D microscopy datasets. PDmean measurements were able to accurately identify known changes in colocalisation in computer-generated and real-world microscopy datasets. We argue that this approach provides substantial advantages over currently used distance-based colocalisation analysis methods. We also introduce PyBioProx, an extensible open-source Python module and graphical user interface that produces PDmean measurements.

Highly multiplexed tissue imaging and in situ spatial profiling aim to extract single-cell data from specimens containing closely packed cells of diverse morphology. This is challenging due to the difficulty of accurately assigning boundaries between cells (segmentation) and then generating per-cell staining intensities. Existing methods use gating to convert per-cell intensity data to positive and negative scores; this is a common approach in flow cytometry, but one that is problematic in imaging. In contrast, human experts identify cells in crowded environments using morphological, neighborhood, and intensity information. Here we describe a computational approach (Cell Spotter or CSPOT) that uses supervised machine learning in combination with classical segmentation to perform automated cell type calling. CSPOT is robust to artifacts that commonly afflict tissue imaging and can replace conventional gating. The end-to-end Python implementation of CSPOT can be integrated into cloud-based image processing pipelines to substantially improve the speed, accuracy, and reproducibility of single-cell spatial data.

Endogenous tagging makes it possible to study a proteins localization, dynamics, and function within its native regulatory context. This is typically accomplished via CRISPR, which involves inserting a sequence encoding a functional tag into the reading frame of a gene. However, this process is often inefficient. Here, we introduce the "quickTAG," or qTAG system, a versatile collection of optimized repair cassettes designed to make CRISPR-mediated tagging more accessible. By including a desired tag sequence linked to a selectable marker in the cassette, integrations can be quickly isolated post-editing. The core sequence scaffold within these constructs incorporates several key features that enhance flexibility and ease of use, such as: specific cassette designs for N- and C-terminus tagging; standardized cloning sequences to simplify the incorporation of homology arms for HDR or MMEJ-based repairs; restriction sites next to each genetic element within the cassette for easy modification of tags and selectable markers; and the inclusion of lox sites flanking the selectable marker to allow for marker gene removal following integration. We showcase the versatility of these cassettes with a diverse range of tags, demonstrating their applications in fluorescence imaging, proximity-dependent biotinylation, epitope tagging, and targeted protein degradation. The adaptability of this scaffold is also exhibited by incorporating novel tags such as mStayGold, which offer enhanced brightness and photostability, reconciling prolonged live-cell imaging of proteins at their endogenous levels. Finally, by leveraging the restriction sites, entirely distinct cassette structures and editing schemes were developed. These enabled scenarios that included conditional expression tagging, selectable knockout tagging, and safe-harbor expression. Our existing and forthcoming collection of plasmids will be accessible through Addgene. It includes ready-to-use constructs targeting common subcellular marker genes, as well as an assortment of tagging cassettes for the tagging of genes of interest. The qTAG system offers an accessible framework to streamline endogenous tagging and will serve as an open resource for researchers to adapt and tailor for their own experiments.

Red blood cells (RBCs) must be highly deformable to transit through the microvasculature to deliver oxygen to tissues. The loss of RBC deformability resulting from pathology, natural aging, or storage in blood bags can impede the proper function of these cells. A variety of methods have been developed to measure RBC deformability, but these methods require specialized equipment, long measurement time, and highly skilled personnel. To address this challenge, we investigated whether a machine learning approach could be applied to determine donor RBC deformability using single cell microscope images. We used the microfluidic ratchet device to sort RBCs based on deformability. Sorted cells are then imaged and used to train a deep learning model to classify RBCs based on deformability. This model correctly predicted deformability of individual RBCs with 84 {+/-} 11% accuracy averaged across ten donors. Using this model to score the deformability of RBC samples were accurate to within 4.4 {+/-} 2.5% of the value obtained using the microfluidic ratchet device. While machine learning methods are frequently developed to automate human image analysis, our study is remarkable in showing that deep learning of single cell microscopy images could be used to measure RBC deformability, a property not normally measurable by imaging. Measuring RBC deformability by imaging is also desirable because it can be performed rapidly using a standard microscopy system, potentially enabling RBC deformability studies to be performed as part of routine clinical assessments.

Our goal is to identify and understand cellular behaviors using 3D live imaging of cell organization. To do this, we image human inducible pluripotent stem cell (hiPSC) lines expressing fluorescently tagged protein representing specific cellular organelles and structures. To produce large numbers of standardized cell images, we developed an automated hiPSC culture procedure, to maintain, passage and Matrigel coat 6-well plastic plates and 96-well glass plates compatible with high-resolution 3D microscopy. Here we describe this system including optimization procedures and specific values for plate movement, angle of tips, speed of aspiration and dispense, seeding strategies and timing of every step. We validated this approach through a side-by-side comparison of quality control results obtained from manual and automated methods. Additionally, we developed an automated image-based colony segmentation and feature extraction pipeline to predict cell count and select wells with consistent morphology for high resolution 3D microscopy.

Proper reporting of metadata is essential to reproduce microscopy experiments, interpret results and share images. Experimental scientists can report details about sample preparation and imaging conditions while imaging scientists have the expertise required to collect and report the image acquisition, hardware and software metadata information. MethodsJ2 is an ImageJ/Fiji based software tool that gathers metadata and automatically generates text for the methods section of publications.

Identifying nuclei is a standard first step to analysing cells in microscopy images. The traditional approach relies on signal from a DNA stain, or fluorescent transgene expression localised to the nucleus. However, imaging techniques that do not use fluorescence can also carry useful information. Here, we demonstrate that it is possible to accurately segment nuclei directly from brightfield images using deep learning. We confirmed that three convolutional neural network architectures can be adapted for this task, with U-Net achieving the best overall performance, Mask R-CNN providing an additional benefit of instance segmentation, and DeepCell proving too slow for practical application. We found that accurate segmentation is possible using as few as 16 training images and that models trained on images from similar cell lines can extrapolate well. Acquiring data from multiple focal planes further helps distinguish nuclei in the samples. Overall, our work liberates a fluorescence channel reserved for nuclear staining, thus providing more information from the specimen, and reducing reagents and time required for preparing imaging experiments.

Correlative light and electron microscopy (CLEM) allows to link light microscopy (LM) of living cells to ultrastructural analyses by electron microscopy (EM). Pre-embedding CLEM often suffers from inaccurate correlation between the LM and EM modalities due to chemical and physical distortions. Post-embedding CLEM enables precise registration of fluorescent structures directly on thin resin sections. However, in-resin CLEM techniques require fluorescent markers withstanding EM sample preparation. Most fluorescent proteins lose their fluorescence during EM sample preparation. Synthetic dyes present an alternative as their photostability and brightness exceed those of fluorescent proteins. Together with self-labeling enzymes (SLE) as protein tags, these fluorophores can be used to precisely label cellular structures of interest. By applying SLE labelling for post-embedding CLEM, we compared Janelia Fluor dyes and TMR to identify most suitable fluorophores. Epithelial cells expressing HaloTag fusion proteins were stained with various ligand-conjugated dyes, and fluorescence preservation was quantified after conventional room temperature sample preparation with embedding in EPON. The results obtained show that only the red dyes TMR, JF549, JFX549 and JFX554 retain their fluorescence in resin, with JFX549 and JFX554 yielding best signal intensity and signal-to-background ratio during in-resin super-resolution microscopy. Since all red dyes possess an oxygen atom within their xanthene structure, our results indicate that this might be a crucial feature making them more tolerant to sample preparation for electron microscopy. Our work reports a rapid in-resin CLEM approach that combines fast and efficient labeling of SLE tags with EM-compatible fluorophores, and serve as benchmarks for experimental planning and future engineering of fluorophores for CLEM.

Image processing methods offer the potential to improve the quality of fluorescence microscopy data, allowing for image acquisition at lower, less phototoxic illumination doses. The training and evaluation of such methods is informed and driven by full-reference image quality metrics (IQMs); however, these metrics derive from applications to natural scene images, not fluorescence microscopy images. Here we investigate the response of IQMs to common properties of fluorescence microscopy data and whether IQMs are capable of reporting the biological information content of images. We find that IQM scores are biased by image content for both raw and processed microscopy data, and that improvements in IQM values reported after processing are not reliably correlated with performance in downstream analysis tasks. As common IQMs are unreliable proxies for guiding image processing developments in biological fluorescence microscopy, image processing performance should be benchmarked according to downstream analysis success.

Microdroplets are compartments made in the laboratory that allow the miniaturisation of chemical and biological experiments to the femto- to picolitre scale, replacing the classical test tube with a droplet. Ideally containment of the contents of individual droplets would be perfect, but in reality this situation rarely occurs. Instead the leaking of molecules even from intact droplets presents a challenge to the success of miniaturisation and must be assessed on a case-by-case basis. We now present a new method for quantitative determination of leakage: a sheath fluid-free flow cytometer (Guava EasyCyte) is used to directly determine the fluorescence of water-in-oil droplets as a function of time. We validate this method by demonstrating that this assessment of leakage provides a framework for experimental improvements that reduce the leakage of two widely used fluorophores. A 40-fold better retention compared to current protocols is achieved for resorufin with an optimized mix (oil: FC-70, surfactant: 0.1% w/w AZ900C, additive: 1% BSA) to maintain useful retention for up to 130 hours. Likewise leakage of the fluorophore methylumbelliferone is reduced by 75-fold. The availability of a method to quantitate leakage quickly for a variety of experimental conditions will facilitate future applications of droplet-based experiments (e.g. in directed evolution or diagnostics), aid miniaturisation of lab-scale assays into this format, and improve the degrees of freedom in setting up such ultrahigh-throughput experiments.